Reviews

The proportion of older adults in the world population is growing rapidly. In the next 10 to 15 years, the population age >60 will grow 3.5 times more rapidly than the general population.¹ As a result, there is an increased urgency in examining benefits vs risks of antipsychotics in older individuals. In a 2010 U.S. nationally representative observational study, antipsychotic use was observed to rise slowly during early and middle adulthood, peaking at approximately age 55, declining slightly between ages 55 and 65, and then rising again after age 65, with >2% of individuals ages 80 to 84 receiving an anti­psychotic.² This is likely due to the chronology of psychotic, mood, and neurocognitive disorders across the life span. In this large national study, long-term antipsychotic treatment was common, and older patients were more likely to receive their prescriptions from non-psychiatrist physicians than from psychiatrists.² Among patients receiving an antipsychotic, the proportion of those receiving it for >120 days was 54% for individuals ages 70 to 74; 49% for individuals ages 75 to 79; and 46% for individuals ages 80 to 84.

This 3-part review summarizes findings and risk–benefit considerations when prescribing antipsychotics to older individuals. Part 1 focused on those with chronic psychotic disorders, such as schizophrenia or bipolar disorder,³ and part 3 will cover patients with dementia. This review (part 2) aims to:

• briefly summarize the results of randomized controlled trials (RCTs) of second-generation antipsychotics (SGAs) and other major studies and analyses in older patients with major depressive disorder (MDD)
• provide a summative opinion on the relative risks and benefits associated with using antipsychotics in older adults with MDD
• highlight the gaps in the evidence base and areas that need additional research.

Summary of benefits, place in treatment armamentarium

The prevalence of MDD and clinically significant depressive symptoms in community­dwelling older adults is 3% to 4% and 15%, respectively, and as high as 16% and 50%, respectively, in nursing home residents.⁴ Because late-life depression is associated with suffering, disability, and excessive mortality, it needs to be recognized and treated aggressively.⁵ Antidepressants are the mainstay of pharmacotherapy for late-life depression. Guidelines and expert opinion informed by the current evidence recommend using selective serotonin reuptake inhibitors, such as escitalopram or sertraline, as a first-line treatment; serotonin norepinephrine reuptake inhibitors, such as duloxetine or venlafaxine, as a second-line treatment; and other antidepressants, such as bupropion or nortriptyline, as a third-line treatment.^5,6 However, antipsychotics also have a role in treating late-life depression.

Over the past decade, several anti­psychotics have been FDA-approved for treating MDD: aripiprazole and brexpiprazole as adjunctive treatment of MDD in adults; olanzapine-fluoxetine combination for acute and maintenance treatment of treatment-resistant depression in adults and geriatric adults; and quetiapine extended-release (XR) as monotherapy for MDD in adults and as adjunctive treatment of MDD in adults and geriatric adults who have had an inadequate response to antidepressants alone (Table 1). However, “black-box” warnings for all first-generation antipsychotics (FGAs) and SGAs alert clinicians that these medications have been associated with serious adverse events in older adults with dementia, including “deaths […] due to heart-related events (eg, heart failure, sudden death) or infections (mostly pneumonia),” with 15 of 17 placebo-controlled trials showing a higher number of deaths with an antipsychotic compared with placebo.⁷ Although similar controlled data on the mortality risk of antipsychotics in older adults with mood disorders do not exist, most experts limit their use to 2 groups of older patients: those with MDD and psychotic features (“psychotic depression”) and those with treatment-resistant depression.

Clinical trials

Olanzapine plus sertraline as first-line pharmacotherapy for MDD with psychotic features. Meyers et al¹¹ reported on a double-blind randomized comparison of olanzapine plus placebo vs olanzapine plus sertraline in 259 patients with MDD with psychotic features. An unusual feature of this trial is that it included a similar number of younger and older participants (ages 18 to 93): 117 participants were age <60 (mean age [standard deviation (SD)]: 41.3 [10.8]) and 142 were age ≥60 (mean age [SD]: 71.7 [7.8]). The same dose titration schedules based on efficacy and tolerability were used in both younger and older participants. At the end of the study, the mean dose (SD) of sertraline (or placebo) did not differ significantly in younger (174.3 mg/d [34.1]) and older participants (165.7 mg/d [43.4]). However, the mean dose (SD) of olanzapine was significantly higher in younger patients (15.7 mg/d [4.7]) than in older participants (13.4 mg/d [5.1]).

In both age groups, olanzapine plus sertraline was more efficacious than olanzapine plus placebo, and there was no statistical interaction between age, time, and treatment group (ie, the trajectories of improvement were similar in older and younger patients receiving either olanzapine or olanzapine plus sertraline). Similarly, drop-out rates because of poor tolerability did not differ significantly in younger (4.3%) and older participants (5.6%). However, in a multinomial regression, older participants were more likely to discontinue treatment because of poor tolerability.²² Older participants were significantly less likely to experience weight gain (mean [SD]: +3.3 [4.9] vs +6.5 [6.6] kg) or an increase in fasting glucose and more likely to experience a fall, pedal edema, or extrapyramidal symptoms.^11,22-24 Cholesterol and triglyceride increased significantly and similarly in both age groups. The incidence of symptoms of tardive dyskinesia (TD) over the 12-week trial was low (<5%) in both younger and older participants, and clinically diagnosed TD was reported in only 1 (older) participant.²⁵

Venlafaxine plus aripiprazole for treatment-resistant MDD. In the largest double-blind randomized study of augmentation pharmacotherapy for late-life treatment-resistant depression published to date, Lenze et al²¹ compared venlafaxine plus aripiprazole vs venlafaxine plus placebo in 181 patients age >60 (mean age 66, with 49 participants age >70) with MDD who did not remit after 12 weeks of treatment with venlafaxine (up to 300 mg/d). After 12 weeks of augmentation, remission rates were significantly higher with aripiprazole than with placebo: 40 (44%) vs 26 (29%); odds ratio (95% confidence interval [CI]): 2.0 (1.1 to 3.7). The median final aripiprazole dose was 7 mg/d (range 2 to 15 mg/d) in remitters and 10 mg/d (range 2 to 15 mg/d) in nonremitters.

Five of 90 participants (5%) discontinued aripiprazole (1 each: suicide, jitteriness/akathisia, worsening parkinsonism; and 2 withdrew consent); 8 of 90 (9%) discontinued placebo (2 each: lack of efficacy, headache; 1: worsening parkinsonism; and 3 withdrew consent). The completed suicide occurred after 5 weeks of treatment with aripiprazole and was judged to be “neither due to emergent suicidal ideation nor to aripiprazole side-effects, but was concluded by investigators to be a result of the individual’s persisting and long-standing suicidal ideation.”²¹ Including the suicide, there were 4 serious adverse events (5%) in those receiving aripiprazole (1 each: suicide, congestive heart failure, mild stroke, and diverticulitis) and 2 (2%) in those receiving placebo (1 each: myocardial infarction, hospitalized for vomiting due to accidentally taking extra venlafaxine). In 86 participants receiving aripiprazole and 87 receiving placebo, the most frequently reported adverse effects were increased dream activity (aripiprazole: 23 [27%] vs placebo: 12 [14%]), weight gain (17 [20%] vs 8 [9%]), and tremor (5 [6%] vs 0). Akathisia and parkinsonism were observed more frequently with aripiprazole than with placebo (akathisia: 24 [26%] of 91 vs 11 [12%] of 90; parkinsonism: 15 [17%] of 86 vs 2 [2%] of 81). Akathisia was generally mild and resolved with dose adjustment; however, it was associated with a transient increase in suicidality in 3 (3%) participants receiving aripiprazole vs 0 receiving placebo and persisted at the end of the trial in 5 (5%) participants receiving aripiprazole vs 2 (2%) receiving placebo. Participants receiving aripiprazole had a significantly larger increase in weight (mean [SD]: +1.93 [3.00] vs +0.01 [3.15] kg), but there were no differences between aripiprazole and placebo in changes in body fat, total cholesterol, high-density lipoprotein, low-density lipoprotein, triglycerides, glucose, insulin concentration, or QTc.

Citalopram plus risperidone for treatment-resistant MDD. Alexopoulos et al²⁶ reported an analysis of data from 110 patients age ≥55 years (mean age [SD]: 63.4 [4.8]), among 489 mixed-age patients with MDD. Participants (n = 110) who did not respond to 1 to 3 antidepressants (venlafaxine, sertraline, mirtazapine, fluoxetine, paroxetine, or bupropion in >90%) during their current depressive episode completed 4 to 6 weeks of treatment with citalopram up to 40 mg/d; 93 did not respond and were treated with open-label risperidone (0.25 to 1 mg/d) augmentation for 4 to 6 weeks. Sixty-three (68%) of these 93 patients remitted and were randomized to 24 weeks of double-blind continuation treatment with citalopram plus risperidone vs citalopram plus placebo. Neither the median times to relapse (105 vs 57 days) nor the relapse rates (risperidone: 18 of 32 [56%] vs placebo: 20 of 31 [65%]) differed significantly. During the open-label risperidone augmentation, the most common adverse events were dizziness and dry mouth (n = 9 each, 9.7% of 93). During the continuation phase, headache (n = 3; 9.1% of 32) was observed with risperidone but not with placebo (n = 0). There was no incident parkinsonism or abnormal movements noted, but risperidone was associated with weight gain during both the open-label risperidone augmentation phase (mean [SD]: +0.9 [2.1] kg) and the continuation phase (risperidone: +0.8 [3.5] vs placebo: −0.3 [2.8] kg).

Quetiapine XR monotherapy for MDD. Katila et al²⁷ reported on a placebo-controlled RCT of quetiapine XR (median dose, 158.7 mg/d; range, 50 to 300 mg/d) in 338 patients age ≥66 years (mean age [SD], 71.3 [7.5]) presenting with MDD and a major depressive episode with a duration <1 year and no history of failed antidepressants trials from 2 classes (more than two-thirds of participants had not received treatment). After 9 weeks, the reduction in depressive symptoms on the Montgomery-Åsberg Depression Rating Scale was significantly larger with quetiapine XR than with placebo (mean [SD]: −16.0 [9.3] vs −9.0 [9.9]). There were congruent, significant differences between quetiapine and placebo in terms of response rate (quetiapine XR: 105 of 164 [64%] vs placebo: 52 of 171 [30.4%]) and remission rate (92 of 164 [56.1%] vs placebo: 40 of 171 [23.4%]). The drop-out rates for all causes were similar, but the drop-out rate attributed to adverse events was higher with quetiapine than placebo (16 of 166 [9.6%] vs 7 of 172 [4.1%]). Most quetiapine drop-outs were attributable to dizziness, headache, and somnolence (n = 4 each), and placebo drop-outs were because of headache (n = 2). Consistent with the profile of quetiapine, adverse events with a rate that was at least 5% higher with quetiapine than with placebo included somnolence (64 of 166 [38.6%] vs 16 of 172 [9.3%]), dry mouth (34 [20.5%] vs 18 [10.5%]), and extrapyramidal symptoms (12 [7.2%] vs 4 [2.3%]). Changes in weight and laboratory test results (eg, glucose, lipid profile) were minimal and not clinically meaningful.

Other clinical data. The efficacy and relatively good tolerability of aripiprazole in older patients with treatment-resistant depression observed in the RCT by Lenze et al²¹ is congruent with the earlier results of 2 small (N = 20 and 24) pilot studies.^18,19 In both studies, the remission rate was 50%, and the most prevalent adverse effects were agitation/restlessness/akathisia or drowsiness/sedation. Similarly, in a post hoc pooled analysis of 409 participants ages 50 to 67 from 3 placebo-controlled randomized trials, the remission rate was significantly higher with aripiprazole than with placebo (32.5% vs 17.1%), and the most common adverse effects were akathisia or restlessness (64 of 210 [30.4%]), somnolence (18 [8.6%]), and insomnia (17 [8.1%]).²⁰

Clinical considerations

When assessing the relative benefits and risks of antipsychotics in older patients, it is important to remember that conclusions and summative opinions are necessarily influenced by the source of the data. Because much of what we know about the use of antipsychotics in geriatric adults is from clinical trials, we know more about their acute efficacy and tolerability than their long-term effectiveness and safety.²⁸ There are similar issues regarding the role of antipsychotics in treating MDD in late life. Based on the results of several RCTs,^8,11 a combination of an antidepressant plus an antipsychotic is the recommended pharmacotherapy for the acute treatment of MDD with psychotic features (Table 2).^{8,11,12,19-21,23-27} However, there are no published data to guide how long the antipsychotic should be continued.²⁹

In older patients with MDD without psychotic features, 1 relatively large placebo-controlled RCT,²¹ 2 smaller open studies,^18,19 and a post hoc analysis of a large placebo-controlled RCT in mixed-age adults²⁰ support the efficacy and relatively good tolerability of aripiprazole augmentation of an antidepressant for treatment-resistant MDD. Similarly, 1 large placebo-controlled RCT supports the efficacy and relatively good tolerability of quetiapine for non–treatment-resistant MDD. However, there are no comparative data assessing the relative merits of using these antipsy­chotics vs other pharmacologic strategies (eg, switching to another antidepressant, lithium augmentation, or combination of 2 antidepressants). Because older patients are more likely to experience adverse effects that may have more serious consequences (Table 3), many prudent clinicians reserve using antipsychotics as a third-line treatment in older patients with MDD without psychotic features and limit the duration of their use to a few months.³⁰

The proportion of older adults in the world population is growing rapidly. In the next 10 to 15 years, the population age >60 will grow 3.5 times more rapidly than the general population.¹ As a result, there is an increased urgency in examining benefits vs risks of antipsychotics in older individuals. In a 2010 U.S. nationally representative observational study, antipsychotic use was observed to rise slowly during early and middle adulthood, peaking at approximately age 55, declining slightly between ages 55 and 65, and then rising again after age 65, with >2% of individuals ages 80 to 84 receiving an anti­psychotic.² This is likely due to the chronology of psychotic, mood, and neurocognitive disorders across the life span. In this large national study, long-term antipsychotic treatment was common, and older patients were more likely to receive their prescriptions from non-psychiatrist physicians than from psychiatrists.² Among patients receiving an antipsychotic, the proportion of those receiving it for >120 days was 54% for individuals ages 70 to 74; 49% for individuals ages 75 to 79; and 46% for individuals ages 80 to 84.

This 3-part review summarizes findings and risk–benefit considerations when prescribing antipsychotics to older individuals. Part 1 focused on those with chronic psychotic disorders, such as schizophrenia or bipolar disorder,³ and part 3 will cover patients with dementia. This review (part 2) aims to:

• briefly summarize the results of randomized controlled trials (RCTs) of second-generation antipsychotics (SGAs) and other major studies and analyses in older patients with major depressive disorder (MDD)
• provide a summative opinion on the relative risks and benefits associated with using antipsychotics in older adults with MDD
• highlight the gaps in the evidence base and areas that need additional research.

Summary of benefits, place in treatment armamentarium

The prevalence of MDD and clinically significant depressive symptoms in community­dwelling older adults is 3% to 4% and 15%, respectively, and as high as 16% and 50%, respectively, in nursing home residents.⁴ Because late-life depression is associated with suffering, disability, and excessive mortality, it needs to be recognized and treated aggressively.⁵ Antidepressants are the mainstay of pharmacotherapy for late-life depression. Guidelines and expert opinion informed by the current evidence recommend using selective serotonin reuptake inhibitors, such as escitalopram or sertraline, as a first-line treatment; serotonin norepinephrine reuptake inhibitors, such as duloxetine or venlafaxine, as a second-line treatment; and other antidepressants, such as bupropion or nortriptyline, as a third-line treatment.^5,6 However, antipsychotics also have a role in treating late-life depression.

Over the past decade, several anti­psychotics have been FDA-approved for treating MDD: aripiprazole and brexpiprazole as adjunctive treatment of MDD in adults; olanzapine-fluoxetine combination for acute and maintenance treatment of treatment-resistant depression in adults and geriatric adults; and quetiapine extended-release (XR) as monotherapy for MDD in adults and as adjunctive treatment of MDD in adults and geriatric adults who have had an inadequate response to antidepressants alone (Table 1). However, “black-box” warnings for all first-generation antipsychotics (FGAs) and SGAs alert clinicians that these medications have been associated with serious adverse events in older adults with dementia, including “deaths […] due to heart-related events (eg, heart failure, sudden death) or infections (mostly pneumonia),” with 15 of 17 placebo-controlled trials showing a higher number of deaths with an antipsychotic compared with placebo.⁷ Although similar controlled data on the mortality risk of antipsychotics in older adults with mood disorders do not exist, most experts limit their use to 2 groups of older patients: those with MDD and psychotic features (“psychotic depression”) and those with treatment-resistant depression.

Clinical trials

Olanzapine plus sertraline as first-line pharmacotherapy for MDD with psychotic features. Meyers et al¹¹ reported on a double-blind randomized comparison of olanzapine plus placebo vs olanzapine plus sertraline in 259 patients with MDD with psychotic features. An unusual feature of this trial is that it included a similar number of younger and older participants (ages 18 to 93): 117 participants were age <60 (mean age [standard deviation (SD)]: 41.3 [10.8]) and 142 were age ≥60 (mean age [SD]: 71.7 [7.8]). The same dose titration schedules based on efficacy and tolerability were used in both younger and older participants. At the end of the study, the mean dose (SD) of sertraline (or placebo) did not differ significantly in younger (174.3 mg/d [34.1]) and older participants (165.7 mg/d [43.4]). However, the mean dose (SD) of olanzapine was significantly higher in younger patients (15.7 mg/d [4.7]) than in older participants (13.4 mg/d [5.1]).

In both age groups, olanzapine plus sertraline was more efficacious than olanzapine plus placebo, and there was no statistical interaction between age, time, and treatment group (ie, the trajectories of improvement were similar in older and younger patients receiving either olanzapine or olanzapine plus sertraline). Similarly, drop-out rates because of poor tolerability did not differ significantly in younger (4.3%) and older participants (5.6%). However, in a multinomial regression, older participants were more likely to discontinue treatment because of poor tolerability.²² Older participants were significantly less likely to experience weight gain (mean [SD]: +3.3 [4.9] vs +6.5 [6.6] kg) or an increase in fasting glucose and more likely to experience a fall, pedal edema, or extrapyramidal symptoms.^11,22-24 Cholesterol and triglyceride increased significantly and similarly in both age groups. The incidence of symptoms of tardive dyskinesia (TD) over the 12-week trial was low (<5%) in both younger and older participants, and clinically diagnosed TD was reported in only 1 (older) participant.²⁵

Venlafaxine plus aripiprazole for treatment-resistant MDD. In the largest double-blind randomized study of augmentation pharmacotherapy for late-life treatment-resistant depression published to date, Lenze et al²¹ compared venlafaxine plus aripiprazole vs venlafaxine plus placebo in 181 patients age >60 (mean age 66, with 49 participants age >70) with MDD who did not remit after 12 weeks of treatment with venlafaxine (up to 300 mg/d). After 12 weeks of augmentation, remission rates were significantly higher with aripiprazole than with placebo: 40 (44%) vs 26 (29%); odds ratio (95% confidence interval [CI]): 2.0 (1.1 to 3.7). The median final aripiprazole dose was 7 mg/d (range 2 to 15 mg/d) in remitters and 10 mg/d (range 2 to 15 mg/d) in nonremitters.

Five of 90 participants (5%) discontinued aripiprazole (1 each: suicide, jitteriness/akathisia, worsening parkinsonism; and 2 withdrew consent); 8 of 90 (9%) discontinued placebo (2 each: lack of efficacy, headache; 1: worsening parkinsonism; and 3 withdrew consent). The completed suicide occurred after 5 weeks of treatment with aripiprazole and was judged to be “neither due to emergent suicidal ideation nor to aripiprazole side-effects, but was concluded by investigators to be a result of the individual’s persisting and long-standing suicidal ideation.”²¹ Including the suicide, there were 4 serious adverse events (5%) in those receiving aripiprazole (1 each: suicide, congestive heart failure, mild stroke, and diverticulitis) and 2 (2%) in those receiving placebo (1 each: myocardial infarction, hospitalized for vomiting due to accidentally taking extra venlafaxine). In 86 participants receiving aripiprazole and 87 receiving placebo, the most frequently reported adverse effects were increased dream activity (aripiprazole: 23 [27%] vs placebo: 12 [14%]), weight gain (17 [20%] vs 8 [9%]), and tremor (5 [6%] vs 0). Akathisia and parkinsonism were observed more frequently with aripiprazole than with placebo (akathisia: 24 [26%] of 91 vs 11 [12%] of 90; parkinsonism: 15 [17%] of 86 vs 2 [2%] of 81). Akathisia was generally mild and resolved with dose adjustment; however, it was associated with a transient increase in suicidality in 3 (3%) participants receiving aripiprazole vs 0 receiving placebo and persisted at the end of the trial in 5 (5%) participants receiving aripiprazole vs 2 (2%) receiving placebo. Participants receiving aripiprazole had a significantly larger increase in weight (mean [SD]: +1.93 [3.00] vs +0.01 [3.15] kg), but there were no differences between aripiprazole and placebo in changes in body fat, total cholesterol, high-density lipoprotein, low-density lipoprotein, triglycerides, glucose, insulin concentration, or QTc.

Citalopram plus risperidone for treatment-resistant MDD. Alexopoulos et al²⁶ reported an analysis of data from 110 patients age ≥55 years (mean age [SD]: 63.4 [4.8]), among 489 mixed-age patients with MDD. Participants (n = 110) who did not respond to 1 to 3 antidepressants (venlafaxine, sertraline, mirtazapine, fluoxetine, paroxetine, or bupropion in >90%) during their current depressive episode completed 4 to 6 weeks of treatment with citalopram up to 40 mg/d; 93 did not respond and were treated with open-label risperidone (0.25 to 1 mg/d) augmentation for 4 to 6 weeks. Sixty-three (68%) of these 93 patients remitted and were randomized to 24 weeks of double-blind continuation treatment with citalopram plus risperidone vs citalopram plus placebo. Neither the median times to relapse (105 vs 57 days) nor the relapse rates (risperidone: 18 of 32 [56%] vs placebo: 20 of 31 [65%]) differed significantly. During the open-label risperidone augmentation, the most common adverse events were dizziness and dry mouth (n = 9 each, 9.7% of 93). During the continuation phase, headache (n = 3; 9.1% of 32) was observed with risperidone but not with placebo (n = 0). There was no incident parkinsonism or abnormal movements noted, but risperidone was associated with weight gain during both the open-label risperidone augmentation phase (mean [SD]: +0.9 [2.1] kg) and the continuation phase (risperidone: +0.8 [3.5] vs placebo: −0.3 [2.8] kg).

Quetiapine XR monotherapy for MDD. Katila et al²⁷ reported on a placebo-controlled RCT of quetiapine XR (median dose, 158.7 mg/d; range, 50 to 300 mg/d) in 338 patients age ≥66 years (mean age [SD], 71.3 [7.5]) presenting with MDD and a major depressive episode with a duration <1 year and no history of failed antidepressants trials from 2 classes (more than two-thirds of participants had not received treatment). After 9 weeks, the reduction in depressive symptoms on the Montgomery-Åsberg Depression Rating Scale was significantly larger with quetiapine XR than with placebo (mean [SD]: −16.0 [9.3] vs −9.0 [9.9]). There were congruent, significant differences between quetiapine and placebo in terms of response rate (quetiapine XR: 105 of 164 [64%] vs placebo: 52 of 171 [30.4%]) and remission rate (92 of 164 [56.1%] vs placebo: 40 of 171 [23.4%]). The drop-out rates for all causes were similar, but the drop-out rate attributed to adverse events was higher with quetiapine than placebo (16 of 166 [9.6%] vs 7 of 172 [4.1%]). Most quetiapine drop-outs were attributable to dizziness, headache, and somnolence (n = 4 each), and placebo drop-outs were because of headache (n = 2). Consistent with the profile of quetiapine, adverse events with a rate that was at least 5% higher with quetiapine than with placebo included somnolence (64 of 166 [38.6%] vs 16 of 172 [9.3%]), dry mouth (34 [20.5%] vs 18 [10.5%]), and extrapyramidal symptoms (12 [7.2%] vs 4 [2.3%]). Changes in weight and laboratory test results (eg, glucose, lipid profile) were minimal and not clinically meaningful.

Other clinical data. The efficacy and relatively good tolerability of aripiprazole in older patients with treatment-resistant depression observed in the RCT by Lenze et al²¹ is congruent with the earlier results of 2 small (N = 20 and 24) pilot studies.^18,19 In both studies, the remission rate was 50%, and the most prevalent adverse effects were agitation/restlessness/akathisia or drowsiness/sedation. Similarly, in a post hoc pooled analysis of 409 participants ages 50 to 67 from 3 placebo-controlled randomized trials, the remission rate was significantly higher with aripiprazole than with placebo (32.5% vs 17.1%), and the most common adverse effects were akathisia or restlessness (64 of 210 [30.4%]), somnolence (18 [8.6%]), and insomnia (17 [8.1%]).²⁰

Clinical considerations

When assessing the relative benefits and risks of antipsychotics in older patients, it is important to remember that conclusions and summative opinions are necessarily influenced by the source of the data. Because much of what we know about the use of antipsychotics in geriatric adults is from clinical trials, we know more about their acute efficacy and tolerability than their long-term effectiveness and safety.²⁸ There are similar issues regarding the role of antipsychotics in treating MDD in late life. Based on the results of several RCTs,^8,11 a combination of an antidepressant plus an antipsychotic is the recommended pharmacotherapy for the acute treatment of MDD with psychotic features (Table 2).^{8,11,12,19-21,23-27} However, there are no published data to guide how long the antipsychotic should be continued.²⁹

In older patients with MDD without psychotic features, 1 relatively large placebo-controlled RCT,²¹ 2 smaller open studies,^18,19 and a post hoc analysis of a large placebo-controlled RCT in mixed-age adults²⁰ support the efficacy and relatively good tolerability of aripiprazole augmentation of an antidepressant for treatment-resistant MDD. Similarly, 1 large placebo-controlled RCT supports the efficacy and relatively good tolerability of quetiapine for non–treatment-resistant MDD. However, there are no comparative data assessing the relative merits of using these antipsy­chotics vs other pharmacologic strategies (eg, switching to another antidepressant, lithium augmentation, or combination of 2 antidepressants). Because older patients are more likely to experience adverse effects that may have more serious consequences (Table 3), many prudent clinicians reserve using antipsychotics as a third-line treatment in older patients with MDD without psychotic features and limit the duration of their use to a few months.³⁰

The proportion of older adults in the world population is growing rapidly. In the next 10 to 15 years, the population age >60 will grow 3.5 times more rapidly than the general population.¹ As a result, there is an increased urgency in examining benefits vs risks of antipsychotics in older individuals. In a 2010 U.S. nationally representative observational study, antipsychotic use was observed to rise slowly during early and middle adulthood, peaking at approximately age 55, declining slightly between ages 55 and 65, and then rising again after age 65, with >2% of individuals ages 80 to 84 receiving an anti­psychotic.² This is likely due to the chronology of psychotic, mood, and neurocognitive disorders across the life span. In this large national study, long-term antipsychotic treatment was common, and older patients were more likely to receive their prescriptions from non-psychiatrist physicians than from psychiatrists.² Among patients receiving an antipsychotic, the proportion of those receiving it for >120 days was 54% for individuals ages 70 to 74; 49% for individuals ages 75 to 79; and 46% for individuals ages 80 to 84.

This 3-part review summarizes findings and risk–benefit considerations when prescribing antipsychotics to older individuals. Part 1 focused on those with chronic psychotic disorders, such as schizophrenia or bipolar disorder,³ and part 3 will cover patients with dementia. This review (part 2) aims to:

• briefly summarize the results of randomized controlled trials (RCTs) of second-generation antipsychotics (SGAs) and other major studies and analyses in older patients with major depressive disorder (MDD)
• provide a summative opinion on the relative risks and benefits associated with using antipsychotics in older adults with MDD
• highlight the gaps in the evidence base and areas that need additional research.

Summary of benefits, place in treatment armamentarium

The prevalence of MDD and clinically significant depressive symptoms in community­dwelling older adults is 3% to 4% and 15%, respectively, and as high as 16% and 50%, respectively, in nursing home residents.⁴ Because late-life depression is associated with suffering, disability, and excessive mortality, it needs to be recognized and treated aggressively.⁵ Antidepressants are the mainstay of pharmacotherapy for late-life depression. Guidelines and expert opinion informed by the current evidence recommend using selective serotonin reuptake inhibitors, such as escitalopram or sertraline, as a first-line treatment; serotonin norepinephrine reuptake inhibitors, such as duloxetine or venlafaxine, as a second-line treatment; and other antidepressants, such as bupropion or nortriptyline, as a third-line treatment.^5,6 However, antipsychotics also have a role in treating late-life depression.

Over the past decade, several anti­psychotics have been FDA-approved for treating MDD: aripiprazole and brexpiprazole as adjunctive treatment of MDD in adults; olanzapine-fluoxetine combination for acute and maintenance treatment of treatment-resistant depression in adults and geriatric adults; and quetiapine extended-release (XR) as monotherapy for MDD in adults and as adjunctive treatment of MDD in adults and geriatric adults who have had an inadequate response to antidepressants alone (Table 1). However, “black-box” warnings for all first-generation antipsychotics (FGAs) and SGAs alert clinicians that these medications have been associated with serious adverse events in older adults with dementia, including “deaths […] due to heart-related events (eg, heart failure, sudden death) or infections (mostly pneumonia),” with 15 of 17 placebo-controlled trials showing a higher number of deaths with an antipsychotic compared with placebo.⁷ Although similar controlled data on the mortality risk of antipsychotics in older adults with mood disorders do not exist, most experts limit their use to 2 groups of older patients: those with MDD and psychotic features (“psychotic depression”) and those with treatment-resistant depression.

Clinical trials

Olanzapine plus sertraline as first-line pharmacotherapy for MDD with psychotic features. Meyers et al¹¹ reported on a double-blind randomized comparison of olanzapine plus placebo vs olanzapine plus sertraline in 259 patients with MDD with psychotic features. An unusual feature of this trial is that it included a similar number of younger and older participants (ages 18 to 93): 117 participants were age <60 (mean age [standard deviation (SD)]: 41.3 [10.8]) and 142 were age ≥60 (mean age [SD]: 71.7 [7.8]). The same dose titration schedules based on efficacy and tolerability were used in both younger and older participants. At the end of the study, the mean dose (SD) of sertraline (or placebo) did not differ significantly in younger (174.3 mg/d [34.1]) and older participants (165.7 mg/d [43.4]). However, the mean dose (SD) of olanzapine was significantly higher in younger patients (15.7 mg/d [4.7]) than in older participants (13.4 mg/d [5.1]).

In both age groups, olanzapine plus sertraline was more efficacious than olanzapine plus placebo, and there was no statistical interaction between age, time, and treatment group (ie, the trajectories of improvement were similar in older and younger patients receiving either olanzapine or olanzapine plus sertraline). Similarly, drop-out rates because of poor tolerability did not differ significantly in younger (4.3%) and older participants (5.6%). However, in a multinomial regression, older participants were more likely to discontinue treatment because of poor tolerability.²² Older participants were significantly less likely to experience weight gain (mean [SD]: +3.3 [4.9] vs +6.5 [6.6] kg) or an increase in fasting glucose and more likely to experience a fall, pedal edema, or extrapyramidal symptoms.^11,22-24 Cholesterol and triglyceride increased significantly and similarly in both age groups. The incidence of symptoms of tardive dyskinesia (TD) over the 12-week trial was low (<5%) in both younger and older participants, and clinically diagnosed TD was reported in only 1 (older) participant.²⁵

Venlafaxine plus aripiprazole for treatment-resistant MDD. In the largest double-blind randomized study of augmentation pharmacotherapy for late-life treatment-resistant depression published to date, Lenze et al²¹ compared venlafaxine plus aripiprazole vs venlafaxine plus placebo in 181 patients age >60 (mean age 66, with 49 participants age >70) with MDD who did not remit after 12 weeks of treatment with venlafaxine (up to 300 mg/d). After 12 weeks of augmentation, remission rates were significantly higher with aripiprazole than with placebo: 40 (44%) vs 26 (29%); odds ratio (95% confidence interval [CI]): 2.0 (1.1 to 3.7). The median final aripiprazole dose was 7 mg/d (range 2 to 15 mg/d) in remitters and 10 mg/d (range 2 to 15 mg/d) in nonremitters.

Five of 90 participants (5%) discontinued aripiprazole (1 each: suicide, jitteriness/akathisia, worsening parkinsonism; and 2 withdrew consent); 8 of 90 (9%) discontinued placebo (2 each: lack of efficacy, headache; 1: worsening parkinsonism; and 3 withdrew consent). The completed suicide occurred after 5 weeks of treatment with aripiprazole and was judged to be “neither due to emergent suicidal ideation nor to aripiprazole side-effects, but was concluded by investigators to be a result of the individual’s persisting and long-standing suicidal ideation.”²¹ Including the suicide, there were 4 serious adverse events (5%) in those receiving aripiprazole (1 each: suicide, congestive heart failure, mild stroke, and diverticulitis) and 2 (2%) in those receiving placebo (1 each: myocardial infarction, hospitalized for vomiting due to accidentally taking extra venlafaxine). In 86 participants receiving aripiprazole and 87 receiving placebo, the most frequently reported adverse effects were increased dream activity (aripiprazole: 23 [27%] vs placebo: 12 [14%]), weight gain (17 [20%] vs 8 [9%]), and tremor (5 [6%] vs 0). Akathisia and parkinsonism were observed more frequently with aripiprazole than with placebo (akathisia: 24 [26%] of 91 vs 11 [12%] of 90; parkinsonism: 15 [17%] of 86 vs 2 [2%] of 81). Akathisia was generally mild and resolved with dose adjustment; however, it was associated with a transient increase in suicidality in 3 (3%) participants receiving aripiprazole vs 0 receiving placebo and persisted at the end of the trial in 5 (5%) participants receiving aripiprazole vs 2 (2%) receiving placebo. Participants receiving aripiprazole had a significantly larger increase in weight (mean [SD]: +1.93 [3.00] vs +0.01 [3.15] kg), but there were no differences between aripiprazole and placebo in changes in body fat, total cholesterol, high-density lipoprotein, low-density lipoprotein, triglycerides, glucose, insulin concentration, or QTc.

Citalopram plus risperidone for treatment-resistant MDD. Alexopoulos et al²⁶ reported an analysis of data from 110 patients age ≥55 years (mean age [SD]: 63.4 [4.8]), among 489 mixed-age patients with MDD. Participants (n = 110) who did not respond to 1 to 3 antidepressants (venlafaxine, sertraline, mirtazapine, fluoxetine, paroxetine, or bupropion in >90%) during their current depressive episode completed 4 to 6 weeks of treatment with citalopram up to 40 mg/d; 93 did not respond and were treated with open-label risperidone (0.25 to 1 mg/d) augmentation for 4 to 6 weeks. Sixty-three (68%) of these 93 patients remitted and were randomized to 24 weeks of double-blind continuation treatment with citalopram plus risperidone vs citalopram plus placebo. Neither the median times to relapse (105 vs 57 days) nor the relapse rates (risperidone: 18 of 32 [56%] vs placebo: 20 of 31 [65%]) differed significantly. During the open-label risperidone augmentation, the most common adverse events were dizziness and dry mouth (n = 9 each, 9.7% of 93). During the continuation phase, headache (n = 3; 9.1% of 32) was observed with risperidone but not with placebo (n = 0). There was no incident parkinsonism or abnormal movements noted, but risperidone was associated with weight gain during both the open-label risperidone augmentation phase (mean [SD]: +0.9 [2.1] kg) and the continuation phase (risperidone: +0.8 [3.5] vs placebo: −0.3 [2.8] kg).

Quetiapine XR monotherapy for MDD. Katila et al²⁷ reported on a placebo-controlled RCT of quetiapine XR (median dose, 158.7 mg/d; range, 50 to 300 mg/d) in 338 patients age ≥66 years (mean age [SD], 71.3 [7.5]) presenting with MDD and a major depressive episode with a duration <1 year and no history of failed antidepressants trials from 2 classes (more than two-thirds of participants had not received treatment). After 9 weeks, the reduction in depressive symptoms on the Montgomery-Åsberg Depression Rating Scale was significantly larger with quetiapine XR than with placebo (mean [SD]: −16.0 [9.3] vs −9.0 [9.9]). There were congruent, significant differences between quetiapine and placebo in terms of response rate (quetiapine XR: 105 of 164 [64%] vs placebo: 52 of 171 [30.4%]) and remission rate (92 of 164 [56.1%] vs placebo: 40 of 171 [23.4%]). The drop-out rates for all causes were similar, but the drop-out rate attributed to adverse events was higher with quetiapine than placebo (16 of 166 [9.6%] vs 7 of 172 [4.1%]). Most quetiapine drop-outs were attributable to dizziness, headache, and somnolence (n = 4 each), and placebo drop-outs were because of headache (n = 2). Consistent with the profile of quetiapine, adverse events with a rate that was at least 5% higher with quetiapine than with placebo included somnolence (64 of 166 [38.6%] vs 16 of 172 [9.3%]), dry mouth (34 [20.5%] vs 18 [10.5%]), and extrapyramidal symptoms (12 [7.2%] vs 4 [2.3%]). Changes in weight and laboratory test results (eg, glucose, lipid profile) were minimal and not clinically meaningful.

Other clinical data. The efficacy and relatively good tolerability of aripiprazole in older patients with treatment-resistant depression observed in the RCT by Lenze et al²¹ is congruent with the earlier results of 2 small (N = 20 and 24) pilot studies.^18,19 In both studies, the remission rate was 50%, and the most prevalent adverse effects were agitation/restlessness/akathisia or drowsiness/sedation. Similarly, in a post hoc pooled analysis of 409 participants ages 50 to 67 from 3 placebo-controlled randomized trials, the remission rate was significantly higher with aripiprazole than with placebo (32.5% vs 17.1%), and the most common adverse effects were akathisia or restlessness (64 of 210 [30.4%]), somnolence (18 [8.6%]), and insomnia (17 [8.1%]).²⁰

Clinical considerations

When assessing the relative benefits and risks of antipsychotics in older patients, it is important to remember that conclusions and summative opinions are necessarily influenced by the source of the data. Because much of what we know about the use of antipsychotics in geriatric adults is from clinical trials, we know more about their acute efficacy and tolerability than their long-term effectiveness and safety.²⁸ There are similar issues regarding the role of antipsychotics in treating MDD in late life. Based on the results of several RCTs,^8,11 a combination of an antidepressant plus an antipsychotic is the recommended pharmacotherapy for the acute treatment of MDD with psychotic features (Table 2).^{8,11,12,19-21,23-27} However, there are no published data to guide how long the antipsychotic should be continued.²⁹

In older patients with MDD without psychotic features, 1 relatively large placebo-controlled RCT,²¹ 2 smaller open studies,^18,19 and a post hoc analysis of a large placebo-controlled RCT in mixed-age adults²⁰ support the efficacy and relatively good tolerability of aripiprazole augmentation of an antidepressant for treatment-resistant MDD. Similarly, 1 large placebo-controlled RCT supports the efficacy and relatively good tolerability of quetiapine for non–treatment-resistant MDD. However, there are no comparative data assessing the relative merits of using these antipsy­chotics vs other pharmacologic strategies (eg, switching to another antidepressant, lithium augmentation, or combination of 2 antidepressants). Because older patients are more likely to experience adverse effects that may have more serious consequences (Table 3), many prudent clinicians reserve using antipsychotics as a third-line treatment in older patients with MDD without psychotic features and limit the duration of their use to a few months.³⁰

“It is a mystery how a ubiquitous treatment used since antiquity was unknown, unnamed, and unidentified until recently. It is even more remarkable because this is the only treatment common to all societies and cultures.”¹

The treatment discussed above is not a specific pill, surgery, plant, or herb. Rather, the authors are referring to placebo. Indeed, the history of medical treatment is largely a chronicle of placebos. When subjected to scientific scrutiny, the overwhelming majority of treatments have turned out to be devoid of intrinsic therapeutic value; they derived their benefits from the placebo effect. Despite these benefits, the term “placebo” comes with unfortunate baggage. Latin for “I shall please,” it is the first word of the Christian vespers for the dead. In the 12th century these vespers were commonly referred to as placebos. By the 1300s, the term had become secular and pejorative, suggesting a flatterer or sycophant. When the word entered medical terminology in the late 18th century, the negative connotation stuck. A placebo was defined as a medicine given to please patients rather than to benefit them. In the modern era, the lack of pharmacologic activity became part of the definition as well.

The word placebo brings with it connotations of deception, fakery, and ineffectiveness. But one of the things about placebos that contribute mightily to the health care community’s aversion toward them is, in fact, their effectiveness. They bring relief across a wide range of medical conditions.² In doing so, placebos impugn the value of our most cherished remedies, hamper the development of new therapeutics, and threaten our livelihoods as health professionals.³

Placebos often are conceptualized as any treatment that lacks intrinsic therapeutic value, such as sugar pills. But looking at what placebo treatment actually entails, both in placebo-controlled treatment trials and in clinical settings, suggests a more comprehensive definition. Placebos encompass all the elements common to any treatment or healing situation. These include a recognized healer, evaluation, diagnosis, prognosis, plausible treatment, and most importantly, the expectation that one will recover. Along these lines, the placebo response can be thought of as the response to the common elements of the treatment or healing situation.³

Research regarding the placebo effect has mushroomed in the past 2 decades. Over this time, we have learned a good deal about both the mechanisms underlying the placebo effect and how the placebo effect can be applied to enhance the benefit of conventional treatment. Brain imaging technology has revealed that when placebo treatment alleviates pain, Parkinson’s disease, and depression, brain changes occur that are similar to those observed with active pharma­cologic treatment.^4,5 Recent studies also show that deliberate, open (nondeceptive) use of placebo can improve the symptoms of several conditions, including depression, pain, and irritable bowel syndrome.⁶ Furthermore, intermittent substitution of placebo pills for pharmacologically active treatment in a conditioning paradigm can be as effective as the “real” treatment.⁷ Also, research over the past decade has verified that certain common features of the treatment situation, particularly the quality of the doctor–patient encounter, contribute to the placebo response and have a demonstrable impact on the outcome of treatment.⁸ Clearly, the placebo effect has gone from being simply a nuisance that interferes with the evaluation of new treatments to a variable worthy of study and application in its own right. Although, for the most part, clinical practice has not kept up with these advances.

Placebos seem to have their greatest impact on the subjective symptoms of disease—pain, distress, and discouragement. It should come as no surprise, then, that placebos are particularly effective in certain psychiatric conditions. In some forms of anxiety and depressive disorders, for example, distress is the illness, and placebos reliably bring relief. Patients with panic disorder, mild to moderate depression, or generalized anxiety disorder get almost as much relief with placebo as they do with conventional treatment (about one-half improve with placebo).^9-11 But <20% of those with obsessive-compulsive disorder improve with placebo, and placebo response rates are also low in patients with schizophrenia or dementia. Mania, attention-deficit/hyperactivity disorder (ADHD), and severe depression fall somewhere in the middle.³

Harnessing the placebo response

There may be a few circumstances in psychiatric practice when it makes sense to intentionally prescribe a placebo as treatment, and we discuss those below. But far more frequently, what we know about the elements that contribute to the placebo effect can be applied to enhance the benefits of any treatment. Patients might be best served if deliberate mobilization of the placebo effect was a standard adjunct to conventional clinical care.

Various components of the treatment situation, collectively referred to as placebo, are a powerful antidote for illness, and some of these healing components exert their influence without special activity on the clinician’s part:

• Simply seeking psychiatric care can bring relief by providing some sense of control over distressing symptoms. The standard trappings of the office or clinic and customary office procedures—from the presentation of one’s insurance card to taking a history—offer reassurance and evoke the expectation that improvement or recovery is around the corner.
• The comfort provided by the psychiatrist’s presence is enhanced when patients feel that they are in the hands of a recognized healer. Psychiatrists inspire confidence when they look like a psychiatrist, or more precisely, like the patient’s idea of what a psychiatrist should look like. In our culture, that means a white coat or business attire.

A thorough evaluation is one of the common treatment elements that does the most to reduce distress and inspire confidence. The quality of an evaluation bears a strong relationship to patients’ satisfaction with the medical encounter, and can influence the amount of disability they suffer.^3,12-15

Although guidelines for conducting effective psychiatric interviews have been around for almost 100 years, psychiatrists vary considerably in the extent to which they elicit complete and accurate information, build rapport, give patients the sense that they are listened to, and provide a thorough assessment. The degree to which patients feel that the clinician is responsive to their concerns depends as much on the style of the interview as on the amount of time devoted to it. Nonverbal behavior can carry the message that the clinician is paying full attention. Something as simple as not answering the phone during an interview (this seems obvious, but a surprising and troubling number of mental health professionals take phone calls during interviews and treatment sessions) conveys an important message about the importance that the clinician places on the patient’s problems.³

The idea that the treatment situation itself provides reassurance and reduces distress, and in doing so, powers a good bit of the placebo effect, is enshrined in such concepts as the importance of good bedside manner. Many feel that the doctor’s thoughtful attention, positive regard, and optimism—so valued by patients—are justified on humanitarian grounds alone; actual evidence that this caring behavior contributes to healing isn’t required. To many, the healing properties of the treatment situation are self-evident. But as the costs of health care snowball and the demands for efficiency and cost-effectiveness rise, the time that psychiatrists can devote to patients has dwindled. Third-party payors demand evidence, beyond intuition and common sense, that diagnostic procedures and treatments have some usefulness, and rightly so.

Is there any evidence that the common components of the treatment situation provide benefit?³ More specifically, does the quality of the doctor–patient relationship and the patient’s feelings about a therapeutic encounter promote healing? Several studies suggest that the doctor–patient relationship has a demonstrable impact on symptom relief.¹⁶ In 1 study, oncologists were randomly assigned to receive a Communication Skills Training (CST) program or not. CST included a 1.5-day face-to-face workshop and 6 hours of monthly videoconferencing that focused on improving communication skills with patients.¹⁷ Lessons included building rapport, engaging in appropriate eye contact, and normalizing difficult experiences. One week after initially consulting with their physician, patients who saw an oncologist in the CST group experienced less anxiety and depression than those who saw an oncologist who did not receive CST. The benefit of CST for patient anxiety mostly persisted at a 3-month follow-up.

A recent meta-analysis pooled the results of 47 studies to examine the relationship between how much trust patients have for their doctors and health outcomes. There was a small to medium association: More trust was associated with greater improvement.¹⁸ It is possible that a good doctor–patient relationship enhances expectancies. However, it is also likely that a positive therapeutic relationship is inherently soothing and reduces distress or dysfunction independent of expectation. Regardless of the precise mechanism, these studies warrant attention. We all understand that it is important on ethical grounds to treat patients with respect and kindness. Research shows that this type of behavior also promotes recovery.

Patient expectations. The idea that expectation of improvement has a major impact on treatment outcome is firmly grounded in research on the placebo effect. Studies have shown that what people expect to experience as an outcome of treatment has a substantial impact on what they actually experience. In a classic study, a doctor told some patients with symptoms of minor illness that they would feel better soon and another group with the same symptoms that he didn’t know what ailed them.¹⁹ Two weeks later, 64% of patients in the “positive expectation” group were improved, compared with only 39% of patients in the “negative” group. In another study, adults were exposed to an allergen that caused a skin reaction.²⁰ Hand lotion (ie, a therapeutically inert substance) was then spread on the skin. Patients were led to believe that the cream would either alleviate or exacerbate the itching. The experimentally-induced wheal-and-flare was measured in both groups a few minutes after the allergen and cream were applied. The wheal-and-flare were worse for participants in the group that expected exacerbation.

Not uncommonly, expectation can have more impact on clinical outcome than a drug’s pharmacologic activity. In a double-blind placebo-controlled study, patients with depression were treated with St. John’s wort, sertraline, or placebo.²¹ They improved to the same extent with all 3 treatments. But when patients were asked to guess the treatment to which they had been assigned, those who thought they had received placebo showed little improvement, irrespective of which intervention they actually received, and those who guessed they had been given St. John’s wort or sertraline showed uniformly large improvement, irrespective of which intervention they actually received (including placebo). The researchers concluded that “Patient beliefs regarding treatment may have a stronger association with clinical outcome than the actual medication received.”

Psychiatrists who wish to use all the therapeutic tools at their disposal must attend to and manage patient expectations. One part of channeling a patient’s expectation is to thoroughly assess the patient’s beliefs regarding the efficacy of various treatments. If a patient’s uncle said that a certain drug is a miracle cure for anxiety, and the patient believes it to be true, then that expectation must be taken into consideration. Many patients prefer alternative treatments to conventional therapies. As long as there is no reason to think an alternative treatment will cause harm, a compromise might be reasonable. For example, if a patient with schizophrenia wants to treat her symptoms with herbal tea, the psychiatrist could say, “In addition to the tea, I recommend that you also take clozapine. The combination is likely to improve your symptoms.”³ More than anything else, the words a psychiatrist uses when recommending treatment shape the patient’s expectations. “You should be feeling a lot less anxious soon after you start taking this” has a different effect than “Try this. It may help.”

Prescribing ‘open-label’ placebo

There may be some limited circumstances where an actual placebo (eg, a sugar pill) might be suitable as a treatment. These include when placebo and conventional treatment provide similar results and a patient is reluctant to take conventional medicine, or when there is no effective conventional treatment. The deceptive prescription of placebo (providing placebo and calling it a drug) has a long history and was considered ethical—and recommended by medical authorities—until the latter half of the 20th century. This practice was deemed unethical in the 1980s, because it was dishonest and violated patient autonomy. Because it was widely believed that placebos given openly would be ineffective, the end of placebo treatment seemed at hand. An intriguing body of evidence, however, suggests that placebos can be effective even when patients know they are taking a placebo. Patients given an “open-label” placebo are told something along the lines of “the pill being prescribed contains no medicine, but some people improve with it, perhaps because the pill stimulates the body’s self-healing.” Open-label placebo has been evaluated for depression,²² low back pain,²³ irritable bowel syndrome,²⁴ neurosis,²⁵ allergic rhinitis,²⁶ and anxiety.²⁷ Most of these studies are small, and some were uncontrolled. Yet they consistently have shown that symptoms improve with a nondeceptive placebo, and improve to a greater extent than with no treatment.

The most recent trial is a promising example of the potential of open-label placebos. In this study, 96 patients with chronic low back pain were randomly assigned to 3 weeks of treatment as usual (TAU) or 3 weeks of TAU plus open-label placebo.²³ Patients who received open-label placebo were educated about the placebo effect and shown a film clip describing promising results of a prior open-label placebo study. They were then given placebo pills to be take once daily, and clearly told the pills contained no active medication. After 3 weeks, patients in the TAU plus placebo group reported less pain and less disability than patients who received TAU without a placebo. Some patients even requested a placebo prescription at the end of the study.

The placebo response provides a rational basis for prescribing innocuous alternative therapies with no intrinsic therapeutic value. Patients who prefer and believe in the effectiveness of alternative remedies—herbal compounds, massage, magnets, homeopathic solutions, etc.—can be recommended these treatments to mobilize a placebo response.

Using a conditioning model. Prescribing a placebo to obtain a conditioned drug response has enormous but untapped clinical potential. Both animal and human research indicates that a wide range of drug responses, from immune suppression to motor stimulation, can be conditioned (a neutral stimulus, such as a pill or injection, associated with drug administration can in itself evoke the drug effect). In many conditioning or dose-extending models, a particular response to real medication (such as pain relief after analgesics) first becomes conditioned due to repeated exposure to the drug given in a particular vehicle. Then, the treatment shifts to some doses comprising of real medicine and some doses comprising of placebo. Because the drug response has been conditioned, it is thought that the response to an identically appearing placebo will mirror the drug response. The active drug often is only replaced by placebo for certain doses under a schedule of partial reinforcement, given the ubiquity of extinction (the conditioned response lessens when the conditioned stimulus is presented alone on repeated trials).

In 1 version of a conditioning study, children with ADHD were randomized to 1 of 3 groups.²⁸ One group (full dose) took the standard dose of medication for 2 months, a second group (reduced dose) took a standard dose during 1 month followed by a half dose during the second month, and children in the third group (reduced dose with placebo) took the standard dose plus a visually distinctive placebo during the first month, followed by a half dose plus the visually distinctive placebo during the second month. Not surprisingly, ADHD symptoms were worse among children in the reduced-dose group. However, there was no difference between those in the reduced-dose with placebo group and those in the full-dose group. It appears as though the symptom reduction associated with a 100% dose was an unconditioned response that could be mimicked with the addition of a placebo pill.

In another study, patients with psoriasis were randomly assigned to receive a full dose of active medication (0.1% triamcinolone cream) twice a day, or a full dose of active medication for 25% to 50% of the doses, with a placebo (moisturizing cream) given for the other 50% to 75% of the doses.²⁹ Relapse rates were not statistically different between groups.

“It is a mystery how a ubiquitous treatment used since antiquity was unknown, unnamed, and unidentified until recently. It is even more remarkable because this is the only treatment common to all societies and cultures.”¹

The treatment discussed above is not a specific pill, surgery, plant, or herb. Rather, the authors are referring to placebo. Indeed, the history of medical treatment is largely a chronicle of placebos. When subjected to scientific scrutiny, the overwhelming majority of treatments have turned out to be devoid of intrinsic therapeutic value; they derived their benefits from the placebo effect. Despite these benefits, the term “placebo” comes with unfortunate baggage. Latin for “I shall please,” it is the first word of the Christian vespers for the dead. In the 12th century these vespers were commonly referred to as placebos. By the 1300s, the term had become secular and pejorative, suggesting a flatterer or sycophant. When the word entered medical terminology in the late 18th century, the negative connotation stuck. A placebo was defined as a medicine given to please patients rather than to benefit them. In the modern era, the lack of pharmacologic activity became part of the definition as well.

The word placebo brings with it connotations of deception, fakery, and ineffectiveness. But one of the things about placebos that contribute mightily to the health care community’s aversion toward them is, in fact, their effectiveness. They bring relief across a wide range of medical conditions.² In doing so, placebos impugn the value of our most cherished remedies, hamper the development of new therapeutics, and threaten our livelihoods as health professionals.³

Placebos often are conceptualized as any treatment that lacks intrinsic therapeutic value, such as sugar pills. But looking at what placebo treatment actually entails, both in placebo-controlled treatment trials and in clinical settings, suggests a more comprehensive definition. Placebos encompass all the elements common to any treatment or healing situation. These include a recognized healer, evaluation, diagnosis, prognosis, plausible treatment, and most importantly, the expectation that one will recover. Along these lines, the placebo response can be thought of as the response to the common elements of the treatment or healing situation.³

Research regarding the placebo effect has mushroomed in the past 2 decades. Over this time, we have learned a good deal about both the mechanisms underlying the placebo effect and how the placebo effect can be applied to enhance the benefit of conventional treatment. Brain imaging technology has revealed that when placebo treatment alleviates pain, Parkinson’s disease, and depression, brain changes occur that are similar to those observed with active pharma­cologic treatment.^4,5 Recent studies also show that deliberate, open (nondeceptive) use of placebo can improve the symptoms of several conditions, including depression, pain, and irritable bowel syndrome.⁶ Furthermore, intermittent substitution of placebo pills for pharmacologically active treatment in a conditioning paradigm can be as effective as the “real” treatment.⁷ Also, research over the past decade has verified that certain common features of the treatment situation, particularly the quality of the doctor–patient encounter, contribute to the placebo response and have a demonstrable impact on the outcome of treatment.⁸ Clearly, the placebo effect has gone from being simply a nuisance that interferes with the evaluation of new treatments to a variable worthy of study and application in its own right. Although, for the most part, clinical practice has not kept up with these advances.

Placebos seem to have their greatest impact on the subjective symptoms of disease—pain, distress, and discouragement. It should come as no surprise, then, that placebos are particularly effective in certain psychiatric conditions. In some forms of anxiety and depressive disorders, for example, distress is the illness, and placebos reliably bring relief. Patients with panic disorder, mild to moderate depression, or generalized anxiety disorder get almost as much relief with placebo as they do with conventional treatment (about one-half improve with placebo).^9-11 But <20% of those with obsessive-compulsive disorder improve with placebo, and placebo response rates are also low in patients with schizophrenia or dementia. Mania, attention-deficit/hyperactivity disorder (ADHD), and severe depression fall somewhere in the middle.³

Harnessing the placebo response

There may be a few circumstances in psychiatric practice when it makes sense to intentionally prescribe a placebo as treatment, and we discuss those below. But far more frequently, what we know about the elements that contribute to the placebo effect can be applied to enhance the benefits of any treatment. Patients might be best served if deliberate mobilization of the placebo effect was a standard adjunct to conventional clinical care.

Various components of the treatment situation, collectively referred to as placebo, are a powerful antidote for illness, and some of these healing components exert their influence without special activity on the clinician’s part:

• Simply seeking psychiatric care can bring relief by providing some sense of control over distressing symptoms. The standard trappings of the office or clinic and customary office procedures—from the presentation of one’s insurance card to taking a history—offer reassurance and evoke the expectation that improvement or recovery is around the corner.
• The comfort provided by the psychiatrist’s presence is enhanced when patients feel that they are in the hands of a recognized healer. Psychiatrists inspire confidence when they look like a psychiatrist, or more precisely, like the patient’s idea of what a psychiatrist should look like. In our culture, that means a white coat or business attire.

A thorough evaluation is one of the common treatment elements that does the most to reduce distress and inspire confidence. The quality of an evaluation bears a strong relationship to patients’ satisfaction with the medical encounter, and can influence the amount of disability they suffer.^3,12-15

Although guidelines for conducting effective psychiatric interviews have been around for almost 100 years, psychiatrists vary considerably in the extent to which they elicit complete and accurate information, build rapport, give patients the sense that they are listened to, and provide a thorough assessment. The degree to which patients feel that the clinician is responsive to their concerns depends as much on the style of the interview as on the amount of time devoted to it. Nonverbal behavior can carry the message that the clinician is paying full attention. Something as simple as not answering the phone during an interview (this seems obvious, but a surprising and troubling number of mental health professionals take phone calls during interviews and treatment sessions) conveys an important message about the importance that the clinician places on the patient’s problems.³

The idea that the treatment situation itself provides reassurance and reduces distress, and in doing so, powers a good bit of the placebo effect, is enshrined in such concepts as the importance of good bedside manner. Many feel that the doctor’s thoughtful attention, positive regard, and optimism—so valued by patients—are justified on humanitarian grounds alone; actual evidence that this caring behavior contributes to healing isn’t required. To many, the healing properties of the treatment situation are self-evident. But as the costs of health care snowball and the demands for efficiency and cost-effectiveness rise, the time that psychiatrists can devote to patients has dwindled. Third-party payors demand evidence, beyond intuition and common sense, that diagnostic procedures and treatments have some usefulness, and rightly so.

Is there any evidence that the common components of the treatment situation provide benefit?³ More specifically, does the quality of the doctor–patient relationship and the patient’s feelings about a therapeutic encounter promote healing? Several studies suggest that the doctor–patient relationship has a demonstrable impact on symptom relief.¹⁶ In 1 study, oncologists were randomly assigned to receive a Communication Skills Training (CST) program or not. CST included a 1.5-day face-to-face workshop and 6 hours of monthly videoconferencing that focused on improving communication skills with patients.¹⁷ Lessons included building rapport, engaging in appropriate eye contact, and normalizing difficult experiences. One week after initially consulting with their physician, patients who saw an oncologist in the CST group experienced less anxiety and depression than those who saw an oncologist who did not receive CST. The benefit of CST for patient anxiety mostly persisted at a 3-month follow-up.

A recent meta-analysis pooled the results of 47 studies to examine the relationship between how much trust patients have for their doctors and health outcomes. There was a small to medium association: More trust was associated with greater improvement.¹⁸ It is possible that a good doctor–patient relationship enhances expectancies. However, it is also likely that a positive therapeutic relationship is inherently soothing and reduces distress or dysfunction independent of expectation. Regardless of the precise mechanism, these studies warrant attention. We all understand that it is important on ethical grounds to treat patients with respect and kindness. Research shows that this type of behavior also promotes recovery.

Patient expectations. The idea that expectation of improvement has a major impact on treatment outcome is firmly grounded in research on the placebo effect. Studies have shown that what people expect to experience as an outcome of treatment has a substantial impact on what they actually experience. In a classic study, a doctor told some patients with symptoms of minor illness that they would feel better soon and another group with the same symptoms that he didn’t know what ailed them.¹⁹ Two weeks later, 64% of patients in the “positive expectation” group were improved, compared with only 39% of patients in the “negative” group. In another study, adults were exposed to an allergen that caused a skin reaction.²⁰ Hand lotion (ie, a therapeutically inert substance) was then spread on the skin. Patients were led to believe that the cream would either alleviate or exacerbate the itching. The experimentally-induced wheal-and-flare was measured in both groups a few minutes after the allergen and cream were applied. The wheal-and-flare were worse for participants in the group that expected exacerbation.

Not uncommonly, expectation can have more impact on clinical outcome than a drug’s pharmacologic activity. In a double-blind placebo-controlled study, patients with depression were treated with St. John’s wort, sertraline, or placebo.²¹ They improved to the same extent with all 3 treatments. But when patients were asked to guess the treatment to which they had been assigned, those who thought they had received placebo showed little improvement, irrespective of which intervention they actually received, and those who guessed they had been given St. John’s wort or sertraline showed uniformly large improvement, irrespective of which intervention they actually received (including placebo). The researchers concluded that “Patient beliefs regarding treatment may have a stronger association with clinical outcome than the actual medication received.”

Psychiatrists who wish to use all the therapeutic tools at their disposal must attend to and manage patient expectations. One part of channeling a patient’s expectation is to thoroughly assess the patient’s beliefs regarding the efficacy of various treatments. If a patient’s uncle said that a certain drug is a miracle cure for anxiety, and the patient believes it to be true, then that expectation must be taken into consideration. Many patients prefer alternative treatments to conventional therapies. As long as there is no reason to think an alternative treatment will cause harm, a compromise might be reasonable. For example, if a patient with schizophrenia wants to treat her symptoms with herbal tea, the psychiatrist could say, “In addition to the tea, I recommend that you also take clozapine. The combination is likely to improve your symptoms.”³ More than anything else, the words a psychiatrist uses when recommending treatment shape the patient’s expectations. “You should be feeling a lot less anxious soon after you start taking this” has a different effect than “Try this. It may help.”

Prescribing ‘open-label’ placebo

There may be some limited circumstances where an actual placebo (eg, a sugar pill) might be suitable as a treatment. These include when placebo and conventional treatment provide similar results and a patient is reluctant to take conventional medicine, or when there is no effective conventional treatment. The deceptive prescription of placebo (providing placebo and calling it a drug) has a long history and was considered ethical—and recommended by medical authorities—until the latter half of the 20th century. This practice was deemed unethical in the 1980s, because it was dishonest and violated patient autonomy. Because it was widely believed that placebos given openly would be ineffective, the end of placebo treatment seemed at hand. An intriguing body of evidence, however, suggests that placebos can be effective even when patients know they are taking a placebo. Patients given an “open-label” placebo are told something along the lines of “the pill being prescribed contains no medicine, but some people improve with it, perhaps because the pill stimulates the body’s self-healing.” Open-label placebo has been evaluated for depression,²² low back pain,²³ irritable bowel syndrome,²⁴ neurosis,²⁵ allergic rhinitis,²⁶ and anxiety.²⁷ Most of these studies are small, and some were uncontrolled. Yet they consistently have shown that symptoms improve with a nondeceptive placebo, and improve to a greater extent than with no treatment.

The most recent trial is a promising example of the potential of open-label placebos. In this study, 96 patients with chronic low back pain were randomly assigned to 3 weeks of treatment as usual (TAU) or 3 weeks of TAU plus open-label placebo.²³ Patients who received open-label placebo were educated about the placebo effect and shown a film clip describing promising results of a prior open-label placebo study. They were then given placebo pills to be take once daily, and clearly told the pills contained no active medication. After 3 weeks, patients in the TAU plus placebo group reported less pain and less disability than patients who received TAU without a placebo. Some patients even requested a placebo prescription at the end of the study.

The placebo response provides a rational basis for prescribing innocuous alternative therapies with no intrinsic therapeutic value. Patients who prefer and believe in the effectiveness of alternative remedies—herbal compounds, massage, magnets, homeopathic solutions, etc.—can be recommended these treatments to mobilize a placebo response.

Using a conditioning model. Prescribing a placebo to obtain a conditioned drug response has enormous but untapped clinical potential. Both animal and human research indicates that a wide range of drug responses, from immune suppression to motor stimulation, can be conditioned (a neutral stimulus, such as a pill or injection, associated with drug administration can in itself evoke the drug effect). In many conditioning or dose-extending models, a particular response to real medication (such as pain relief after analgesics) first becomes conditioned due to repeated exposure to the drug given in a particular vehicle. Then, the treatment shifts to some doses comprising of real medicine and some doses comprising of placebo. Because the drug response has been conditioned, it is thought that the response to an identically appearing placebo will mirror the drug response. The active drug often is only replaced by placebo for certain doses under a schedule of partial reinforcement, given the ubiquity of extinction (the conditioned response lessens when the conditioned stimulus is presented alone on repeated trials).

In 1 version of a conditioning study, children with ADHD were randomized to 1 of 3 groups.²⁸ One group (full dose) took the standard dose of medication for 2 months, a second group (reduced dose) took a standard dose during 1 month followed by a half dose during the second month, and children in the third group (reduced dose with placebo) took the standard dose plus a visually distinctive placebo during the first month, followed by a half dose plus the visually distinctive placebo during the second month. Not surprisingly, ADHD symptoms were worse among children in the reduced-dose group. However, there was no difference between those in the reduced-dose with placebo group and those in the full-dose group. It appears as though the symptom reduction associated with a 100% dose was an unconditioned response that could be mimicked with the addition of a placebo pill.

In another study, patients with psoriasis were randomly assigned to receive a full dose of active medication (0.1% triamcinolone cream) twice a day, or a full dose of active medication for 25% to 50% of the doses, with a placebo (moisturizing cream) given for the other 50% to 75% of the doses.²⁹ Relapse rates were not statistically different between groups.

“It is a mystery how a ubiquitous treatment used since antiquity was unknown, unnamed, and unidentified until recently. It is even more remarkable because this is the only treatment common to all societies and cultures.”¹

The treatment discussed above is not a specific pill, surgery, plant, or herb. Rather, the authors are referring to placebo. Indeed, the history of medical treatment is largely a chronicle of placebos. When subjected to scientific scrutiny, the overwhelming majority of treatments have turned out to be devoid of intrinsic therapeutic value; they derived their benefits from the placebo effect. Despite these benefits, the term “placebo” comes with unfortunate baggage. Latin for “I shall please,” it is the first word of the Christian vespers for the dead. In the 12th century these vespers were commonly referred to as placebos. By the 1300s, the term had become secular and pejorative, suggesting a flatterer or sycophant. When the word entered medical terminology in the late 18th century, the negative connotation stuck. A placebo was defined as a medicine given to please patients rather than to benefit them. In the modern era, the lack of pharmacologic activity became part of the definition as well.

The word placebo brings with it connotations of deception, fakery, and ineffectiveness. But one of the things about placebos that contribute mightily to the health care community’s aversion toward them is, in fact, their effectiveness. They bring relief across a wide range of medical conditions.² In doing so, placebos impugn the value of our most cherished remedies, hamper the development of new therapeutics, and threaten our livelihoods as health professionals.³

Placebos often are conceptualized as any treatment that lacks intrinsic therapeutic value, such as sugar pills. But looking at what placebo treatment actually entails, both in placebo-controlled treatment trials and in clinical settings, suggests a more comprehensive definition. Placebos encompass all the elements common to any treatment or healing situation. These include a recognized healer, evaluation, diagnosis, prognosis, plausible treatment, and most importantly, the expectation that one will recover. Along these lines, the placebo response can be thought of as the response to the common elements of the treatment or healing situation.³

Research regarding the placebo effect has mushroomed in the past 2 decades. Over this time, we have learned a good deal about both the mechanisms underlying the placebo effect and how the placebo effect can be applied to enhance the benefit of conventional treatment. Brain imaging technology has revealed that when placebo treatment alleviates pain, Parkinson’s disease, and depression, brain changes occur that are similar to those observed with active pharma­cologic treatment.^4,5 Recent studies also show that deliberate, open (nondeceptive) use of placebo can improve the symptoms of several conditions, including depression, pain, and irritable bowel syndrome.⁶ Furthermore, intermittent substitution of placebo pills for pharmacologically active treatment in a conditioning paradigm can be as effective as the “real” treatment.⁷ Also, research over the past decade has verified that certain common features of the treatment situation, particularly the quality of the doctor–patient encounter, contribute to the placebo response and have a demonstrable impact on the outcome of treatment.⁸ Clearly, the placebo effect has gone from being simply a nuisance that interferes with the evaluation of new treatments to a variable worthy of study and application in its own right. Although, for the most part, clinical practice has not kept up with these advances.

Placebos seem to have their greatest impact on the subjective symptoms of disease—pain, distress, and discouragement. It should come as no surprise, then, that placebos are particularly effective in certain psychiatric conditions. In some forms of anxiety and depressive disorders, for example, distress is the illness, and placebos reliably bring relief. Patients with panic disorder, mild to moderate depression, or generalized anxiety disorder get almost as much relief with placebo as they do with conventional treatment (about one-half improve with placebo).^9-11 But <20% of those with obsessive-compulsive disorder improve with placebo, and placebo response rates are also low in patients with schizophrenia or dementia. Mania, attention-deficit/hyperactivity disorder (ADHD), and severe depression fall somewhere in the middle.³

Harnessing the placebo response

There may be a few circumstances in psychiatric practice when it makes sense to intentionally prescribe a placebo as treatment, and we discuss those below. But far more frequently, what we know about the elements that contribute to the placebo effect can be applied to enhance the benefits of any treatment. Patients might be best served if deliberate mobilization of the placebo effect was a standard adjunct to conventional clinical care.

Various components of the treatment situation, collectively referred to as placebo, are a powerful antidote for illness, and some of these healing components exert their influence without special activity on the clinician’s part:

• Simply seeking psychiatric care can bring relief by providing some sense of control over distressing symptoms. The standard trappings of the office or clinic and customary office procedures—from the presentation of one’s insurance card to taking a history—offer reassurance and evoke the expectation that improvement or recovery is around the corner.
• The comfort provided by the psychiatrist’s presence is enhanced when patients feel that they are in the hands of a recognized healer. Psychiatrists inspire confidence when they look like a psychiatrist, or more precisely, like the patient’s idea of what a psychiatrist should look like. In our culture, that means a white coat or business attire.

A thorough evaluation is one of the common treatment elements that does the most to reduce distress and inspire confidence. The quality of an evaluation bears a strong relationship to patients’ satisfaction with the medical encounter, and can influence the amount of disability they suffer.^3,12-15

Although guidelines for conducting effective psychiatric interviews have been around for almost 100 years, psychiatrists vary considerably in the extent to which they elicit complete and accurate information, build rapport, give patients the sense that they are listened to, and provide a thorough assessment. The degree to which patients feel that the clinician is responsive to their concerns depends as much on the style of the interview as on the amount of time devoted to it. Nonverbal behavior can carry the message that the clinician is paying full attention. Something as simple as not answering the phone during an interview (this seems obvious, but a surprising and troubling number of mental health professionals take phone calls during interviews and treatment sessions) conveys an important message about the importance that the clinician places on the patient’s problems.³

The idea that the treatment situation itself provides reassurance and reduces distress, and in doing so, powers a good bit of the placebo effect, is enshrined in such concepts as the importance of good bedside manner. Many feel that the doctor’s thoughtful attention, positive regard, and optimism—so valued by patients—are justified on humanitarian grounds alone; actual evidence that this caring behavior contributes to healing isn’t required. To many, the healing properties of the treatment situation are self-evident. But as the costs of health care snowball and the demands for efficiency and cost-effectiveness rise, the time that psychiatrists can devote to patients has dwindled. Third-party payors demand evidence, beyond intuition and common sense, that diagnostic procedures and treatments have some usefulness, and rightly so.

Is there any evidence that the common components of the treatment situation provide benefit?³ More specifically, does the quality of the doctor–patient relationship and the patient’s feelings about a therapeutic encounter promote healing? Several studies suggest that the doctor–patient relationship has a demonstrable impact on symptom relief.¹⁶ In 1 study, oncologists were randomly assigned to receive a Communication Skills Training (CST) program or not. CST included a 1.5-day face-to-face workshop and 6 hours of monthly videoconferencing that focused on improving communication skills with patients.¹⁷ Lessons included building rapport, engaging in appropriate eye contact, and normalizing difficult experiences. One week after initially consulting with their physician, patients who saw an oncologist in the CST group experienced less anxiety and depression than those who saw an oncologist who did not receive CST. The benefit of CST for patient anxiety mostly persisted at a 3-month follow-up.

A recent meta-analysis pooled the results of 47 studies to examine the relationship between how much trust patients have for their doctors and health outcomes. There was a small to medium association: More trust was associated with greater improvement.¹⁸ It is possible that a good doctor–patient relationship enhances expectancies. However, it is also likely that a positive therapeutic relationship is inherently soothing and reduces distress or dysfunction independent of expectation. Regardless of the precise mechanism, these studies warrant attention. We all understand that it is important on ethical grounds to treat patients with respect and kindness. Research shows that this type of behavior also promotes recovery.

Patient expectations. The idea that expectation of improvement has a major impact on treatment outcome is firmly grounded in research on the placebo effect. Studies have shown that what people expect to experience as an outcome of treatment has a substantial impact on what they actually experience. In a classic study, a doctor told some patients with symptoms of minor illness that they would feel better soon and another group with the same symptoms that he didn’t know what ailed them.¹⁹ Two weeks later, 64% of patients in the “positive expectation” group were improved, compared with only 39% of patients in the “negative” group. In another study, adults were exposed to an allergen that caused a skin reaction.²⁰ Hand lotion (ie, a therapeutically inert substance) was then spread on the skin. Patients were led to believe that the cream would either alleviate or exacerbate the itching. The experimentally-induced wheal-and-flare was measured in both groups a few minutes after the allergen and cream were applied. The wheal-and-flare were worse for participants in the group that expected exacerbation.

Not uncommonly, expectation can have more impact on clinical outcome than a drug’s pharmacologic activity. In a double-blind placebo-controlled study, patients with depression were treated with St. John’s wort, sertraline, or placebo.²¹ They improved to the same extent with all 3 treatments. But when patients were asked to guess the treatment to which they had been assigned, those who thought they had received placebo showed little improvement, irrespective of which intervention they actually received, and those who guessed they had been given St. John’s wort or sertraline showed uniformly large improvement, irrespective of which intervention they actually received (including placebo). The researchers concluded that “Patient beliefs regarding treatment may have a stronger association with clinical outcome than the actual medication received.”

Psychiatrists who wish to use all the therapeutic tools at their disposal must attend to and manage patient expectations. One part of channeling a patient’s expectation is to thoroughly assess the patient’s beliefs regarding the efficacy of various treatments. If a patient’s uncle said that a certain drug is a miracle cure for anxiety, and the patient believes it to be true, then that expectation must be taken into consideration. Many patients prefer alternative treatments to conventional therapies. As long as there is no reason to think an alternative treatment will cause harm, a compromise might be reasonable. For example, if a patient with schizophrenia wants to treat her symptoms with herbal tea, the psychiatrist could say, “In addition to the tea, I recommend that you also take clozapine. The combination is likely to improve your symptoms.”³ More than anything else, the words a psychiatrist uses when recommending treatment shape the patient’s expectations. “You should be feeling a lot less anxious soon after you start taking this” has a different effect than “Try this. It may help.”

Prescribing ‘open-label’ placebo

There may be some limited circumstances where an actual placebo (eg, a sugar pill) might be suitable as a treatment. These include when placebo and conventional treatment provide similar results and a patient is reluctant to take conventional medicine, or when there is no effective conventional treatment. The deceptive prescription of placebo (providing placebo and calling it a drug) has a long history and was considered ethical—and recommended by medical authorities—until the latter half of the 20th century. This practice was deemed unethical in the 1980s, because it was dishonest and violated patient autonomy. Because it was widely believed that placebos given openly would be ineffective, the end of placebo treatment seemed at hand. An intriguing body of evidence, however, suggests that placebos can be effective even when patients know they are taking a placebo. Patients given an “open-label” placebo are told something along the lines of “the pill being prescribed contains no medicine, but some people improve with it, perhaps because the pill stimulates the body’s self-healing.” Open-label placebo has been evaluated for depression,²² low back pain,²³ irritable bowel syndrome,²⁴ neurosis,²⁵ allergic rhinitis,²⁶ and anxiety.²⁷ Most of these studies are small, and some were uncontrolled. Yet they consistently have shown that symptoms improve with a nondeceptive placebo, and improve to a greater extent than with no treatment.

The most recent trial is a promising example of the potential of open-label placebos. In this study, 96 patients with chronic low back pain were randomly assigned to 3 weeks of treatment as usual (TAU) or 3 weeks of TAU plus open-label placebo.²³ Patients who received open-label placebo were educated about the placebo effect and shown a film clip describing promising results of a prior open-label placebo study. They were then given placebo pills to be take once daily, and clearly told the pills contained no active medication. After 3 weeks, patients in the TAU plus placebo group reported less pain and less disability than patients who received TAU without a placebo. Some patients even requested a placebo prescription at the end of the study.

The placebo response provides a rational basis for prescribing innocuous alternative therapies with no intrinsic therapeutic value. Patients who prefer and believe in the effectiveness of alternative remedies—herbal compounds, massage, magnets, homeopathic solutions, etc.—can be recommended these treatments to mobilize a placebo response.

Using a conditioning model. Prescribing a placebo to obtain a conditioned drug response has enormous but untapped clinical potential. Both animal and human research indicates that a wide range of drug responses, from immune suppression to motor stimulation, can be conditioned (a neutral stimulus, such as a pill or injection, associated with drug administration can in itself evoke the drug effect). In many conditioning or dose-extending models, a particular response to real medication (such as pain relief after analgesics) first becomes conditioned due to repeated exposure to the drug given in a particular vehicle. Then, the treatment shifts to some doses comprising of real medicine and some doses comprising of placebo. Because the drug response has been conditioned, it is thought that the response to an identically appearing placebo will mirror the drug response. The active drug often is only replaced by placebo for certain doses under a schedule of partial reinforcement, given the ubiquity of extinction (the conditioned response lessens when the conditioned stimulus is presented alone on repeated trials).

In 1 version of a conditioning study, children with ADHD were randomized to 1 of 3 groups.²⁸ One group (full dose) took the standard dose of medication for 2 months, a second group (reduced dose) took a standard dose during 1 month followed by a half dose during the second month, and children in the third group (reduced dose with placebo) took the standard dose plus a visually distinctive placebo during the first month, followed by a half dose plus the visually distinctive placebo during the second month. Not surprisingly, ADHD symptoms were worse among children in the reduced-dose group. However, there was no difference between those in the reduced-dose with placebo group and those in the full-dose group. It appears as though the symptom reduction associated with a 100% dose was an unconditioned response that could be mimicked with the addition of a placebo pill.

In another study, patients with psoriasis were randomly assigned to receive a full dose of active medication (0.1% triamcinolone cream) twice a day, or a full dose of active medication for 25% to 50% of the doses, with a placebo (moisturizing cream) given for the other 50% to 75% of the doses.²⁹ Relapse rates were not statistically different between groups.

Although the biological aspects of psychiatry are crucial, psychotherapy is an integral part of psychiatry. Unfortunately, the emphasis on psychotherapy training in psychiatry residency programs has declined compared with a decade or more ago. In an era of dwindling psychotherapy training and resources, the quality and type of psychotherapy training has become more variable. In addition to helping maintain the therapeutic alliance, nuanced psychotherapy by a trained professional can be transformational by helping patients to:

• process complex life events and emotions
• feel understood
• overcome psychological barriers to recovery
• enhance self-esteem.

For the therapist, effective psychotherapy usually involves warmth and empathy, active listening, an authentic and genuine positive regard for the patient and his (her) experiences. The therapist should also pay attention to her (his) own emotions and internal responses, as well as the patient’s, while having the perseverance to address and help the patient work through challenging aspects.

When providing psychotherapy for adult patients, consider these basic, but salient points that are often overlooked.

Refrain from making life decisions for patients, except in exceptional circumstances, such as in situations of abuse and other crises.¹ Telling an adult patient what to do about life decisions that he finds challenging fits more under life coaching than psychotherapy. Through therapy, patients should be helped in processing the pros and cons of certain decisions and in navigating the decision-making process to arrive at a decision that makes the most sense to them. Also, it’s not uncommon for thera­peutic relationships to rupture when therapists give advice such as suggesting that a patient divorce his spouse, date a certain individual, or have children.

There are many reasons why giving advice in psychotherapy is not recommended. Giving advice can be an impediment to the therapeutic process.² What is good advice for one patient may not be good for another. Therapists who give advice often do so from their own lens and perspective. This perspective may not only be different from the patient’s priorities and life circumstances, but the therapist also may have inadequate information about the patient’s situation,^1,2 which could lead to providing advice that could even harm the patient. In addition, providing advice might prevent a patient from gaining adequate agency or self-directedness while promoting an unhealthy dependence on the therapist and reinforcing the patient’s self-doubt or lack of confidence. In these cases, the patient may later resent the therapist for the advice.

Address the ‘here and now.’¹ Pay attention to immediate issues or themes that emerge, and address them with the patient gently and thoughtfully, as appropriate. Ignoring these may create risks of missing vital, underlying material that could reveal more of the patient’s inner world, as these themes can sometimes reflect other themes of the patient’s life outside of treatment.

Acknowledging and empathizing, when appropriate, are key initial steps that help decrease resistance and facilitate the therapeutic process.

Explore the affect. Paying attention to the patient’s emotional state is critical.³ This holds true for all types of psychotherapy. For example, if a patient suddenly becomes tearful when telling his story or describing recent events, this is usually a sign that the subject matter affects or holds value to the patient in a significant or meaningful way and should be further explored.

‘Meet the patient where they are.’ This doesn’t mean you should yield to the patient or give in to his demands. It implies that you should assess the patient’s readiness for a particular intervention and devise interventions from that standpoint, exploring the patient’s ambivalence, noticing resistance, and continuing to acknowledge and empathize with where the patient is in life or treatment. When utilized judiciously, this technique can help the therapist align with the patient, and help the patient move forward through resistance and ambivalence.

Be nonjudgmental and empathetic. Patients place trust in their therapists when they disclose thoughts or emotions that are sensitive, meaningful, or close to the heart. A nonjudgmental response helps the patient accept his experiences and emotions. Being empathetic requires putting oneself in another’s shoes; it does not mean agreeing with the patient. Of course, if you learn that your patient abused a child or an older adult, you are required to report it to the appropriate state agency. In addition, follow the duty to warn and protect in case of any other safety issues, as appropriate.

Do not assume. Open-ended questions and exploration are key. For example, a patient told her resident therapist that her father recently passed away. The therapist expressed to the patient how hard this must be for her. However, the patient said she was relieved by her father’s death, because he had been abusive to her for years. Because of the therapist’s comment, the patient doubted her own reaction and felt guilty for not being more upset about her father’s death.

Avoid over-identifying with your patient. If you find yourself over-identifying with a patient because you have a common background or life events, seek supervision. Over-identification not only can pose barriers to objectively identifying patterns and trends in the patient’s behavior or presentation but also can increase the risk of crossing boundaries or even minimizing the patient’s experience. Exercise caution if you find yourself wanting to be liked by your patient; this is a common mistake among beginning therapists.⁴

Seek supervision. If you are feeling angry, frustrated, indifferent, or overly attached toward a patient, recognize this countertransference and seek consultation or supervision from an experienced colleague or supervisor. These emotions can be valuable tools that shed light not only on the patient’s life and the session itself, but also help you identify any other factors, such as your own feelings or experiences, that might be contributing to these reactions.

Although the biological aspects of psychiatry are crucial, psychotherapy is an integral part of psychiatry. Unfortunately, the emphasis on psychotherapy training in psychiatry residency programs has declined compared with a decade or more ago. In an era of dwindling psychotherapy training and resources, the quality and type of psychotherapy training has become more variable. In addition to helping maintain the therapeutic alliance, nuanced psychotherapy by a trained professional can be transformational by helping patients to:

• process complex life events and emotions
• feel understood
• overcome psychological barriers to recovery
• enhance self-esteem.

For the therapist, effective psychotherapy usually involves warmth and empathy, active listening, an authentic and genuine positive regard for the patient and his (her) experiences. The therapist should also pay attention to her (his) own emotions and internal responses, as well as the patient’s, while having the perseverance to address and help the patient work through challenging aspects.

When providing psychotherapy for adult patients, consider these basic, but salient points that are often overlooked.

Refrain from making life decisions for patients, except in exceptional circumstances, such as in situations of abuse and other crises.¹ Telling an adult patient what to do about life decisions that he finds challenging fits more under life coaching than psychotherapy. Through therapy, patients should be helped in processing the pros and cons of certain decisions and in navigating the decision-making process to arrive at a decision that makes the most sense to them. Also, it’s not uncommon for thera­peutic relationships to rupture when therapists give advice such as suggesting that a patient divorce his spouse, date a certain individual, or have children.

There are many reasons why giving advice in psychotherapy is not recommended. Giving advice can be an impediment to the therapeutic process.² What is good advice for one patient may not be good for another. Therapists who give advice often do so from their own lens and perspective. This perspective may not only be different from the patient’s priorities and life circumstances, but the therapist also may have inadequate information about the patient’s situation,^1,2 which could lead to providing advice that could even harm the patient. In addition, providing advice might prevent a patient from gaining adequate agency or self-directedness while promoting an unhealthy dependence on the therapist and reinforcing the patient’s self-doubt or lack of confidence. In these cases, the patient may later resent the therapist for the advice.

Address the ‘here and now.’¹ Pay attention to immediate issues or themes that emerge, and address them with the patient gently and thoughtfully, as appropriate. Ignoring these may create risks of missing vital, underlying material that could reveal more of the patient’s inner world, as these themes can sometimes reflect other themes of the patient’s life outside of treatment.

Acknowledging and empathizing, when appropriate, are key initial steps that help decrease resistance and facilitate the therapeutic process.

Explore the affect. Paying attention to the patient’s emotional state is critical.³ This holds true for all types of psychotherapy. For example, if a patient suddenly becomes tearful when telling his story or describing recent events, this is usually a sign that the subject matter affects or holds value to the patient in a significant or meaningful way and should be further explored.

‘Meet the patient where they are.’ This doesn’t mean you should yield to the patient or give in to his demands. It implies that you should assess the patient’s readiness for a particular intervention and devise interventions from that standpoint, exploring the patient’s ambivalence, noticing resistance, and continuing to acknowledge and empathize with where the patient is in life or treatment. When utilized judiciously, this technique can help the therapist align with the patient, and help the patient move forward through resistance and ambivalence.

Be nonjudgmental and empathetic. Patients place trust in their therapists when they disclose thoughts or emotions that are sensitive, meaningful, or close to the heart. A nonjudgmental response helps the patient accept his experiences and emotions. Being empathetic requires putting oneself in another’s shoes; it does not mean agreeing with the patient. Of course, if you learn that your patient abused a child or an older adult, you are required to report it to the appropriate state agency. In addition, follow the duty to warn and protect in case of any other safety issues, as appropriate.

Do not assume. Open-ended questions and exploration are key. For example, a patient told her resident therapist that her father recently passed away. The therapist expressed to the patient how hard this must be for her. However, the patient said she was relieved by her father’s death, because he had been abusive to her for years. Because of the therapist’s comment, the patient doubted her own reaction and felt guilty for not being more upset about her father’s death.

Avoid over-identifying with your patient. If you find yourself over-identifying with a patient because you have a common background or life events, seek supervision. Over-identification not only can pose barriers to objectively identifying patterns and trends in the patient’s behavior or presentation but also can increase the risk of crossing boundaries or even minimizing the patient’s experience. Exercise caution if you find yourself wanting to be liked by your patient; this is a common mistake among beginning therapists.⁴

Seek supervision. If you are feeling angry, frustrated, indifferent, or overly attached toward a patient, recognize this countertransference and seek consultation or supervision from an experienced colleague or supervisor. These emotions can be valuable tools that shed light not only on the patient’s life and the session itself, but also help you identify any other factors, such as your own feelings or experiences, that might be contributing to these reactions.

Although the biological aspects of psychiatry are crucial, psychotherapy is an integral part of psychiatry. Unfortunately, the emphasis on psychotherapy training in psychiatry residency programs has declined compared with a decade or more ago. In an era of dwindling psychotherapy training and resources, the quality and type of psychotherapy training has become more variable. In addition to helping maintain the therapeutic alliance, nuanced psychotherapy by a trained professional can be transformational by helping patients to:

• process complex life events and emotions
• feel understood
• overcome psychological barriers to recovery
• enhance self-esteem.

For the therapist, effective psychotherapy usually involves warmth and empathy, active listening, an authentic and genuine positive regard for the patient and his (her) experiences. The therapist should also pay attention to her (his) own emotions and internal responses, as well as the patient’s, while having the perseverance to address and help the patient work through challenging aspects.

When providing psychotherapy for adult patients, consider these basic, but salient points that are often overlooked.

Refrain from making life decisions for patients, except in exceptional circumstances, such as in situations of abuse and other crises.¹ Telling an adult patient what to do about life decisions that he finds challenging fits more under life coaching than psychotherapy. Through therapy, patients should be helped in processing the pros and cons of certain decisions and in navigating the decision-making process to arrive at a decision that makes the most sense to them. Also, it’s not uncommon for thera­peutic relationships to rupture when therapists give advice such as suggesting that a patient divorce his spouse, date a certain individual, or have children.

There are many reasons why giving advice in psychotherapy is not recommended. Giving advice can be an impediment to the therapeutic process.² What is good advice for one patient may not be good for another. Therapists who give advice often do so from their own lens and perspective. This perspective may not only be different from the patient’s priorities and life circumstances, but the therapist also may have inadequate information about the patient’s situation,^1,2 which could lead to providing advice that could even harm the patient. In addition, providing advice might prevent a patient from gaining adequate agency or self-directedness while promoting an unhealthy dependence on the therapist and reinforcing the patient’s self-doubt or lack of confidence. In these cases, the patient may later resent the therapist for the advice.

Address the ‘here and now.’¹ Pay attention to immediate issues or themes that emerge, and address them with the patient gently and thoughtfully, as appropriate. Ignoring these may create risks of missing vital, underlying material that could reveal more of the patient’s inner world, as these themes can sometimes reflect other themes of the patient’s life outside of treatment.

Acknowledging and empathizing, when appropriate, are key initial steps that help decrease resistance and facilitate the therapeutic process.

Explore the affect. Paying attention to the patient’s emotional state is critical.³ This holds true for all types of psychotherapy. For example, if a patient suddenly becomes tearful when telling his story or describing recent events, this is usually a sign that the subject matter affects or holds value to the patient in a significant or meaningful way and should be further explored.

‘Meet the patient where they are.’ This doesn’t mean you should yield to the patient or give in to his demands. It implies that you should assess the patient’s readiness for a particular intervention and devise interventions from that standpoint, exploring the patient’s ambivalence, noticing resistance, and continuing to acknowledge and empathize with where the patient is in life or treatment. When utilized judiciously, this technique can help the therapist align with the patient, and help the patient move forward through resistance and ambivalence.

Be nonjudgmental and empathetic. Patients place trust in their therapists when they disclose thoughts or emotions that are sensitive, meaningful, or close to the heart. A nonjudgmental response helps the patient accept his experiences and emotions. Being empathetic requires putting oneself in another’s shoes; it does not mean agreeing with the patient. Of course, if you learn that your patient abused a child or an older adult, you are required to report it to the appropriate state agency. In addition, follow the duty to warn and protect in case of any other safety issues, as appropriate.

Do not assume. Open-ended questions and exploration are key. For example, a patient told her resident therapist that her father recently passed away. The therapist expressed to the patient how hard this must be for her. However, the patient said she was relieved by her father’s death, because he had been abusive to her for years. Because of the therapist’s comment, the patient doubted her own reaction and felt guilty for not being more upset about her father’s death.

Avoid over-identifying with your patient. If you find yourself over-identifying with a patient because you have a common background or life events, seek supervision. Over-identification not only can pose barriers to objectively identifying patterns and trends in the patient’s behavior or presentation but also can increase the risk of crossing boundaries or even minimizing the patient’s experience. Exercise caution if you find yourself wanting to be liked by your patient; this is a common mistake among beginning therapists.⁴

Seek supervision. If you are feeling angry, frustrated, indifferent, or overly attached toward a patient, recognize this countertransference and seek consultation or supervision from an experienced colleague or supervisor. These emotions can be valuable tools that shed light not only on the patient’s life and the session itself, but also help you identify any other factors, such as your own feelings or experiences, that might be contributing to these reactions.

Despite a lack of evidence, polypharmacy often is used to treat autism spectrum disorder (ASD),¹ while educational techniques are underutilized. Compared with the general population, children with ASD may be more prone to the adverse effects of the medications used to treat symptoms, such as antipsychotics and antidepressants.² Therefore, when addressing problem behaviors, such as tantrums, aggressiveness, or self-injury, in a patient with ASD, before prescribing a medication, consider the ABCs of these behaviors.³

Antecedents. What happened before the behavior occurred? Where and when did the behavior occur? Was the individual unable to get a desired tangible item, such as a preferred food, toy, or another object? Was the individual told complete a task that he (she) did not want to do? Did the individual see someone else getting attention?

Behaviors. What behavior(s) occurred after each antecedent?

Consequences. What happened after the behavior occurred? Did the caregiver give the individual the item he wanted? Was the individual able to get out of doing work that he did not want to do or become the center of attention?

Having parents document the ABCs is useful not only for finding out why a behavior occurred, but also for objectively determining if and how a medication is affecting the frequency of a behavior. Charts that parents can use to document ABC data are available online (eg, http://www.positively­autism.com/downloads/datasheet_abc.pdf). Once this data is collected, it can be used to implement appropriate interventions, which I describe as DEFG.

Differential reinforcement of other behaviors is a procedure that provides positive reinforcement for not engaging in a problem behavior or for staying on task. For example, use a token board to reward positive behaviors, with physical tokens or written marks. However, some patients require immediate reinforcement. I suggest that parents or caregivers carry small pieces of preferred food to give to the patient to reinforce positive behavior.

Exercise. A review of 18 studies reported that physical exercise, such as jogging, weight training, and bike riding, can help reduce problem behaviors in individuals with ASD.⁴ Among 64 participants with ASD, there was a decrease in aggression, stereotypy, off-task behavior, and elopement, and improvements in on-task and motor behavior such as playing catch.

Function. Refer to the ABCs to determine why a specific problem behavior is occurring. Each behavior can have 1 or multiple functions; therefore, develop a plan specific to the reason the patient engages in the behavior. For example, if the individual engages in a behavior to avoid a task, the parent or caregiver can give individual tokens that the individual can later exchange for a break, instead of engaging in the problem behavior to avoid the task. If a behavior appears to be done for attention, instruct the caregivers to provide frequent periods of attention when the individual engages in positive behaviors.

Go to the appropriate placement. By law, persons age ≤21 have the right to an education and to make meaningful progress. If a patient with ASD exhibits behaviors that interfere with learning, he is entitled to a placement that can provide intensive applied behavior analysis. If you feel that the child needs a different school, write an evaluation for the parent or guardian to submit to the school district and clearly outline the patient’s needs and requirements.

Despite a lack of evidence, polypharmacy often is used to treat autism spectrum disorder (ASD),¹ while educational techniques are underutilized. Compared with the general population, children with ASD may be more prone to the adverse effects of the medications used to treat symptoms, such as antipsychotics and antidepressants.² Therefore, when addressing problem behaviors, such as tantrums, aggressiveness, or self-injury, in a patient with ASD, before prescribing a medication, consider the ABCs of these behaviors.³

Antecedents. What happened before the behavior occurred? Where and when did the behavior occur? Was the individual unable to get a desired tangible item, such as a preferred food, toy, or another object? Was the individual told complete a task that he (she) did not want to do? Did the individual see someone else getting attention?

Behaviors. What behavior(s) occurred after each antecedent?

Consequences. What happened after the behavior occurred? Did the caregiver give the individual the item he wanted? Was the individual able to get out of doing work that he did not want to do or become the center of attention?

Having parents document the ABCs is useful not only for finding out why a behavior occurred, but also for objectively determining if and how a medication is affecting the frequency of a behavior. Charts that parents can use to document ABC data are available online (eg, http://www.positively­autism.com/downloads/datasheet_abc.pdf). Once this data is collected, it can be used to implement appropriate interventions, which I describe as DEFG.

Differential reinforcement of other behaviors is a procedure that provides positive reinforcement for not engaging in a problem behavior or for staying on task. For example, use a token board to reward positive behaviors, with physical tokens or written marks. However, some patients require immediate reinforcement. I suggest that parents or caregivers carry small pieces of preferred food to give to the patient to reinforce positive behavior.

Exercise. A review of 18 studies reported that physical exercise, such as jogging, weight training, and bike riding, can help reduce problem behaviors in individuals with ASD.⁴ Among 64 participants with ASD, there was a decrease in aggression, stereotypy, off-task behavior, and elopement, and improvements in on-task and motor behavior such as playing catch.

Function. Refer to the ABCs to determine why a specific problem behavior is occurring. Each behavior can have 1 or multiple functions; therefore, develop a plan specific to the reason the patient engages in the behavior. For example, if the individual engages in a behavior to avoid a task, the parent or caregiver can give individual tokens that the individual can later exchange for a break, instead of engaging in the problem behavior to avoid the task. If a behavior appears to be done for attention, instruct the caregivers to provide frequent periods of attention when the individual engages in positive behaviors.

Go to the appropriate placement. By law, persons age ≤21 have the right to an education and to make meaningful progress. If a patient with ASD exhibits behaviors that interfere with learning, he is entitled to a placement that can provide intensive applied behavior analysis. If you feel that the child needs a different school, write an evaluation for the parent or guardian to submit to the school district and clearly outline the patient’s needs and requirements.

Despite a lack of evidence, polypharmacy often is used to treat autism spectrum disorder (ASD),¹ while educational techniques are underutilized. Compared with the general population, children with ASD may be more prone to the adverse effects of the medications used to treat symptoms, such as antipsychotics and antidepressants.² Therefore, when addressing problem behaviors, such as tantrums, aggressiveness, or self-injury, in a patient with ASD, before prescribing a medication, consider the ABCs of these behaviors.³

Antecedents. What happened before the behavior occurred? Where and when did the behavior occur? Was the individual unable to get a desired tangible item, such as a preferred food, toy, or another object? Was the individual told complete a task that he (she) did not want to do? Did the individual see someone else getting attention?

Behaviors. What behavior(s) occurred after each antecedent?

Consequences. What happened after the behavior occurred? Did the caregiver give the individual the item he wanted? Was the individual able to get out of doing work that he did not want to do or become the center of attention?

Having parents document the ABCs is useful not only for finding out why a behavior occurred, but also for objectively determining if and how a medication is affecting the frequency of a behavior. Charts that parents can use to document ABC data are available online (eg, http://www.positively­autism.com/downloads/datasheet_abc.pdf). Once this data is collected, it can be used to implement appropriate interventions, which I describe as DEFG.

Differential reinforcement of other behaviors is a procedure that provides positive reinforcement for not engaging in a problem behavior or for staying on task. For example, use a token board to reward positive behaviors, with physical tokens or written marks. However, some patients require immediate reinforcement. I suggest that parents or caregivers carry small pieces of preferred food to give to the patient to reinforce positive behavior.

Exercise. A review of 18 studies reported that physical exercise, such as jogging, weight training, and bike riding, can help reduce problem behaviors in individuals with ASD.⁴ Among 64 participants with ASD, there was a decrease in aggression, stereotypy, off-task behavior, and elopement, and improvements in on-task and motor behavior such as playing catch.

Function. Refer to the ABCs to determine why a specific problem behavior is occurring. Each behavior can have 1 or multiple functions; therefore, develop a plan specific to the reason the patient engages in the behavior. For example, if the individual engages in a behavior to avoid a task, the parent or caregiver can give individual tokens that the individual can later exchange for a break, instead of engaging in the problem behavior to avoid the task. If a behavior appears to be done for attention, instruct the caregivers to provide frequent periods of attention when the individual engages in positive behaviors.

Go to the appropriate placement. By law, persons age ≤21 have the right to an education and to make meaningful progress. If a patient with ASD exhibits behaviors that interfere with learning, he is entitled to a placement that can provide intensive applied behavior analysis. If you feel that the child needs a different school, write an evaluation for the parent or guardian to submit to the school district and clearly outline the patient’s needs and requirements.

Most psychiatrists are required to sign an employment contract before taking a job, but few of us have received any training on reviewing such contracts. We often rely on coworkers and attorneys to navigate this process for us. However, the contract is crucial, because it outlines your employer’s clinical and administrative expectations for the position, and it gives you the opportunity to lay out what you want.¹ Because an employment contract is legally binding, you should thoroughly read it and look for clauses that may not work in your best interest. Although not a complete list, the following items should be reviewed before signing a contract.^1,2
 

Benefits. Make sure you are offered a reasonable salary, but balance the dollar amount with benefits such as:

• continuing medical education allowances
• educational loan forgiveness
• health/malpractice/disability insurance
• retirement benefits
• compensation for call schedule.

In some cases, there may be a delay before you are eligible to obtain certain benefits.

Work expectations. Many contracts state that the position is “full-time” or have other nonspecific parameters for work expectations. You should inquire about objective work parameters, such as duty hours, the average frequency of the current call schedule, timeframe for completing medical documentation, and penalties for not meeting clinical or administrative requirements, so you are not surprised by:

• working longer-than-planned shifts
• performing on-call duties
• working on days that you were not expecting
• having your credentialing status placed in jeopardy.

Some group practices allow for a half-day of no scheduled appointments with patients, so you can complete paperwork and return phone calls.

Noncompete clause. This restricts you from working within a certain geographic area or for a competing employer for a finite time period after the contract terminates or expires. A noncompete clause could restrict you from practicing within a large geographical area, especially if the job is located in a densely populated area. Some noncompete clauses do not include a temporal or geographic restriction, but can limit your ability to bring patients with you to a new practice or facility when the contract expires.

Malpractice insurance. Two types of malpractice insurance are occurrence and claims-made:

• Occurrence insurance protects you whenever an action is brought against you, even if the action is brought after the contract terminates or expires.
• Claims-made insurance provides coverage if the policy with the same insurer was in effect when the malpractice was committed and when the actual action was commenced.

Although claims-made insurance is less expensive, it can leave you without coverage should you leave your employer and no longer maintain the same insurance policy. Claims-made can be converted into occurrence through the purchase of a tail endorsement. If the employer does not offer you tail coverage, then it is your responsibility to pay for this insurance, which can be expensive.

Termination language. Every contract features a termination section that lists potential causes for terminating your employment. This list is usually not exhaustive, but it sets the framework for a realistic view of reasonable causes. Contracts also commonly contain provisions that permit termination “without cause” after notice of termination is provided. Although you could negotiate for more notice time, “without cause” clauses are unlikely to be removed from the contract.

Most psychiatrists are required to sign an employment contract before taking a job, but few of us have received any training on reviewing such contracts. We often rely on coworkers and attorneys to navigate this process for us. However, the contract is crucial, because it outlines your employer’s clinical and administrative expectations for the position, and it gives you the opportunity to lay out what you want.¹ Because an employment contract is legally binding, you should thoroughly read it and look for clauses that may not work in your best interest. Although not a complete list, the following items should be reviewed before signing a contract.^1,2
 

Benefits. Make sure you are offered a reasonable salary, but balance the dollar amount with benefits such as:

• continuing medical education allowances
• educational loan forgiveness
• health/malpractice/disability insurance
• retirement benefits
• compensation for call schedule.

In some cases, there may be a delay before you are eligible to obtain certain benefits.

Work expectations. Many contracts state that the position is “full-time” or have other nonspecific parameters for work expectations. You should inquire about objective work parameters, such as duty hours, the average frequency of the current call schedule, timeframe for completing medical documentation, and penalties for not meeting clinical or administrative requirements, so you are not surprised by:

• working longer-than-planned shifts
• performing on-call duties
• working on days that you were not expecting
• having your credentialing status placed in jeopardy.

Some group practices allow for a half-day of no scheduled appointments with patients, so you can complete paperwork and return phone calls.

Noncompete clause. This restricts you from working within a certain geographic area or for a competing employer for a finite time period after the contract terminates or expires. A noncompete clause could restrict you from practicing within a large geographical area, especially if the job is located in a densely populated area. Some noncompete clauses do not include a temporal or geographic restriction, but can limit your ability to bring patients with you to a new practice or facility when the contract expires.

Malpractice insurance. Two types of malpractice insurance are occurrence and claims-made:

• Occurrence insurance protects you whenever an action is brought against you, even if the action is brought after the contract terminates or expires.
• Claims-made insurance provides coverage if the policy with the same insurer was in effect when the malpractice was committed and when the actual action was commenced.

Although claims-made insurance is less expensive, it can leave you without coverage should you leave your employer and no longer maintain the same insurance policy. Claims-made can be converted into occurrence through the purchase of a tail endorsement. If the employer does not offer you tail coverage, then it is your responsibility to pay for this insurance, which can be expensive.

Termination language. Every contract features a termination section that lists potential causes for terminating your employment. This list is usually not exhaustive, but it sets the framework for a realistic view of reasonable causes. Contracts also commonly contain provisions that permit termination “without cause” after notice of termination is provided. Although you could negotiate for more notice time, “without cause” clauses are unlikely to be removed from the contract.

Most psychiatrists are required to sign an employment contract before taking a job, but few of us have received any training on reviewing such contracts. We often rely on coworkers and attorneys to navigate this process for us. However, the contract is crucial, because it outlines your employer’s clinical and administrative expectations for the position, and it gives you the opportunity to lay out what you want.¹ Because an employment contract is legally binding, you should thoroughly read it and look for clauses that may not work in your best interest. Although not a complete list, the following items should be reviewed before signing a contract.^1,2
 

Benefits. Make sure you are offered a reasonable salary, but balance the dollar amount with benefits such as:

• continuing medical education allowances
• educational loan forgiveness
• health/malpractice/disability insurance
• retirement benefits
• compensation for call schedule.

In some cases, there may be a delay before you are eligible to obtain certain benefits.

Work expectations. Many contracts state that the position is “full-time” or have other nonspecific parameters for work expectations. You should inquire about objective work parameters, such as duty hours, the average frequency of the current call schedule, timeframe for completing medical documentation, and penalties for not meeting clinical or administrative requirements, so you are not surprised by:

• working longer-than-planned shifts
• performing on-call duties
• working on days that you were not expecting
• having your credentialing status placed in jeopardy.

Some group practices allow for a half-day of no scheduled appointments with patients, so you can complete paperwork and return phone calls.

Noncompete clause. This restricts you from working within a certain geographic area or for a competing employer for a finite time period after the contract terminates or expires. A noncompete clause could restrict you from practicing within a large geographical area, especially if the job is located in a densely populated area. Some noncompete clauses do not include a temporal or geographic restriction, but can limit your ability to bring patients with you to a new practice or facility when the contract expires.

Malpractice insurance. Two types of malpractice insurance are occurrence and claims-made:

• Occurrence insurance protects you whenever an action is brought against you, even if the action is brought after the contract terminates or expires.
• Claims-made insurance provides coverage if the policy with the same insurer was in effect when the malpractice was committed and when the actual action was commenced.

Although claims-made insurance is less expensive, it can leave you without coverage should you leave your employer and no longer maintain the same insurance policy. Claims-made can be converted into occurrence through the purchase of a tail endorsement. If the employer does not offer you tail coverage, then it is your responsibility to pay for this insurance, which can be expensive.

Termination language. Every contract features a termination section that lists potential causes for terminating your employment. This list is usually not exhaustive, but it sets the framework for a realistic view of reasonable causes. Contracts also commonly contain provisions that permit termination “without cause” after notice of termination is provided. Although you could negotiate for more notice time, “without cause” clauses are unlikely to be removed from the contract.

Patient satisfaction with medical care during hospitalization is a common quality metric.^1,2 Studies showing higher patient satisfaction have reported lower 30-day hospital readmissions³ and improved overall health.^4,5 Conversely, communication failures are associated with dissatisfaction among hospitalized patients and adverse outcomes.^6,7 A lack of familiarity with hospital providers weakens collaborative decision making and prevents high-quality patient care.^8,9

Bedside visual tools, such as whiteboards and pictures of medical staff, have been widely used to enhance communication between patients, families, and providers.^10,11 Results of studies evaluating these tools are varied. For example, 1 study found that 98% of patients were better able to identify physicians when their names were written on whiteboards.¹² Yet in another, only 21.1% of patients were more likely to correctly identify ≥1 physicians using pictures.¹³ Thus, despite widespread use,¹¹ whether visual tools improve patient satisfaction and patient care more broadly remains unclear.^14,15

We performed a systematic review to answer the following 3 questions: first, what is the effect of visual tools on outcomes (ie, provider identification, understanding of providers’ roles, patient–provider communication, and satisfaction); second, does impact vary by type of visual tool (eg, whiteboards vs pictures of providers); and third, what factors (eg, study design, patient population) are associated with provider identification, communication, and patient satisfaction?

Search Strategy

We used the Preferred Reporting Items for Systematic Reviews and Meta-Analysis when performing this review.¹⁶ A research librarian (WT) conducted serial searches for studies reporting the use of bedside visual tools for hospitalized patients in Medline (via OVID), Embase, SCOPUS, Web of Science, CINAHL, and Cochrane DSR and CENTRAL. Controlled vocabularies (ie, Medical Subject Headings terms) were used to identify synonyms for visual tools of interest. Additional studies were identified manually through bibliographies and meeting abstracts. No study design, publication date, or language restrictions were placed on the search, which was conducted between April 2016 and February 2017 (see supplementary Appendix A).

Study Selection

Two reviewers (AG and KT) independently assessed study eligibility; discrepancies were resolved by a third reviewer (VC). We included all adult or pediatric English language studies in which the effect of visual tool(s) on patient outcomes was reported. Visual tools were defined as the bedside display of information or an instrument given to patients to convey information regarding providers or medical care. Patient-reported outcomes included the following: (a) physician identification, (b) understanding of provider roles, (c) patient–provider communication, and (d) patient satisfaction with care. Providers were defined as physicians, residents, interns, medical students, nurse practitioners, or nurses. We excluded studies that were not original research (eg, conference abstracts, not peer reviewed), reported qualitative data without quantitative outcomes, or did not include a bedside visual tool. Given our interest in hospitalized general medicine patients, studies conducted in emergency departments, surgical units, obstetrics and gynecology wards, and intensive care units were excluded.

Data Extraction and Analysis

Data were extracted independently and in duplicate from all studies by using a template adapted from the Cochrane Collaboration.¹⁷ For all studies, we abstracted study design, type of visual tool (eg, whiteboards), unit setting (eg, medical), population studied (eg, adult vs pediatric), and outcomes reported (ie, physician identification, understanding of provider roles, communication, and satisfaction with care). Reviewers independently assessed and categorized the impact of tools on reported outcomes.

To standardize and compare outcomes across studies, the following were used to denote a positive association between visual tools and relevant outcomes: a greater number of physicians correctly identified by name/picture or title/role; the use of terms such as “high,” “agreed,” or “significant” on surveys; or ≥4 Likert scores for domains of identification, understanding of roles, communication, and satisfaction with care. Conversely, the inability to identify providers compared to the control/baseline; poor recall of titles/roles; lower Likert-scale scores (ie, ≤2); or survey terms such as “poor,” “disagreed,” or “insignificant” were considered to connote negative impact. Studies in which Likert scores were rated neither high nor low (ie, 3), or in which patients neither agreed nor disagreed on value were considered neutral.

Owing to clinical heterogeneity within studies, meta-analyses were not performed. Descriptive statistics were used to describe study outcomes. A priori¹⁸ studies were evaluated according to the following categories: design (eg, randomized vs observational), outcomes (eg, patient satisfaction), intervention (type of visual tool), and patient population (adult or pediatric). Because pediatric patients have underdeveloped communication skills and include parents and/or guardians, data from pediatric studies were tabulated and reported separately to those from adult studies.

Quality Assessment

As recommended by the Cochrane Collaboration, 2 reviewers (AG, KT) assessed the risk of study bias by using the Downs and Black Scale.^17,19 Discrepancies in assessment were resolved by a third reviewer (VC). This instrument uses a point-based system to estimate the quality of a study by rating domains such as internal and external validity, bias, and confounding. In keeping with prior systematic reviews,^18,20,21 studies with a score of ≥18 were considered high quality. Interrater agreement for the adjudication of study quality was calculated using the Cohen κ statistic.

Eleven studies were conducted on adult hospitalized pa­tients ^{12-14,22-24,26,27,29,30,33} and included 3 randomized controlled studies.^14,27,33

Results by Outcomes Provider Identification Nine studies measured patients’ ability to identify providers with the use of visual aids, and all 9 reported improvements in this outcome. Visual tools used to measure provider identification included pictures (n = 5),^{13,14,23,27,33} whiteboards (n = 3),^12,22,30 and patient portals (n = 1).²⁶ Within studies that used pictures, individual pictures (n = 2)^13,23 and handouts with pictures of multiple providers (n = 3) were used.^14,27,33 In 2 studies, care team members such as a dietitian, physiotherapist or pharmacist, were included when measuring identification.^14,33

Understanding Providers’ RolesSix studies assessed the effect of visual tools on patients’ understanding of provider roles.^{13,14,22,26,27,33} Four studies reported a positive effect with the use of pictures,^27,33 whiteboards,²² and patient portals.²⁶ However, 2 studies reported either no difference or negative impressions. Appel et al.¹⁴ reported no difference in the understanding of physician roles using a handout of providers’ pictures and titles. Arora et al.¹³ used individual pictures of physicians with descriptions of roles and found a negative association, as demonstrated by fewer patients rating their understanding of physicians’ roles as excellent or very good in the intervention period (45.6%) compared with the baseline (55.3%).

 

Patient–Provider Communication

Three studies evaluated the influence of visual tools on communication.^14,24,29 Using pictures, Appel et al.¹⁴ found no difference in the perceived quality of communication. Singh et al.²⁹ used whiteboards and reported improved communication scores for physicians and nurses. With notepads, patients surveyed by Farberg et al.²⁴ stated that the tool improved provider communication.

Patient Satisfaction

Five studies assessed patient satisfaction related to the use of visual tools. ^{22,23,27,30,33} One study reported satisfaction as positive with the use of individual pictures.²³ Two studies that used handouts with pictures of all team members reported either a positive³³ or neutral²⁷ impact on satisfaction. Studies that used whiteboards reported a positive association with satisfaction^22,30 despite differences in content, such as the inclusion of prewritten prompts for writing goals of care and scheduled tests³⁰ versus the name of the nurse and their education level.²²

Results by Type of Visual Tool Pictures

Five studies that used pictures reported a positive effect on provider identification.^{13,14,23,27,33} Two^27,33 of 4 studies^13,14,27,33 that assessed patients’ understanding of team member roles reported a positive influence, while 1 reported no difference.¹⁴ A fourth study demonstrated a negative association, perhaps due to differences in the description of providers’ roles listed on the tool.¹³ Only 1 study examined the influence of pictures on patient–provider communication, and this study found no difference.¹⁴ Satisfaction with care via the use of pictures varied between positive (2 studies)^23,33 and neutral (1 study).²⁷

Whiteboards

Four studies tested the use of whiteboards; of these, 3 reported a positive influence on provider identification.^12,22,30 One study reported a positive impact on patient–provider communication.²⁹ Two studies noted a positive effect on patient satisfaction.^22,30 Notably, the responsibility for updating whiteboards differed between the studies (ie, nurses only²² vs residents, medical students, and nurses).³⁰

Patient Portal

In 1 study, an electronic portal that included names with pictures of providers, descriptions of their roles, lists of medications, and scheduled tests and/or procedures was used as a visual tool. The portal improved patients’ identification of physicians and patients’ understanding of roles. However, improvements in the knowledge of medication changes and planned tests and/or procedures during hospitalization were not observed.²⁶ This finding would suggest limitations in the hospitalized patient’s knowledge of the plan of care, which could potentially weaken patient–provider communication.

Notepads

Only 1 study assessed the use of formatted notepads on patient–provider communication and noted a positive association. Notepads used prompts for different categories (eg, diagnosis/treatment, medications, etc) to encourage patient questions for providers.²⁴

Five studies were conducted on hospitalized pediatric units.^{15,25,28,31,32} All studies surveyed the parents, guardians, or caregivers of pediatric patients. One study excluded patients ≥12 years of age because of legal differences in access to adolescent health information,³² while another interviewed parents and/or guardians of teenagers.¹⁵

Results by Outcomes Provider Identification and Understanding of Physicians’ Roles

Four studies that assessed the influence of visual tools on provider identification and understanding of roles reported a positive association.^15,25,28,31 Visual tools varied between pictures (n = 2),^15,31 patient portal (n = 1),²⁸ and whiteboards and pictures combined (n = 1).²⁵ The measurement of outcomes varied between surveys with free text responses,²⁸ multiple choice questions,²⁵ and 1-5 Likert scales.^15,31

Patient–Provider Communication

Two studies assessed the impact of patient portal use on communication and reported a positive association.^28,32 The 2 portals autopopulated names, pictures, and roles of providers from electronic medical records. Singh et al.²⁸ used a portal that was also available in Spanish and accommodated for non-English speakers. Kelly et al.³² reported that 90% of parents perceived that portal use was associated with reduced errors in care, with 8% finding errors in their child’s medication list.

Patient Satisfaction

Three studies assessed patient satisfaction via the use of visual tools.^15,28,31 Singh et al.²⁸ noted a positive influence on satisfaction via a patient portal. Dudas et al.¹⁵ used a single-page handout with names and pictures of each provider, along with information regarding the training and roles of each provider. Distribution of these handouts to patients by investigators led to a positive influence on satisfaction. While Unaka et al.³¹ used a similar handout, they asked residents to distribute them and found no significant difference in satisfaction scores between the intervention (66%) and control group (62%).

Results by Type of Visual Tool Pictures

Two studies reported a positive impact on provider identification and understanding of roles with the use of pictures.^15,31 Dudas et al.¹⁵ demonstrated a 4.8-fold increase in the odds of parents identifying a medical student, as compared with the control. Similarly, after adjusting for length of stay and prior hospitalization, Unaka et al.³¹ reported that a higher percentage of patients correctly identified providers using this approach.

Whiteboard and Picture

One study evaluated the simultaneous use of whiteboards and pictures to improve the identification of providers. The study noted improved identification of supervising doctors and increased recognition of roles for supervising doctors, residents, and medical students.²⁵

Patient Portal

Two studies used patient portals as visual tools. Singh et al.²⁸ assessed the use of a patient portal with names, roles, and pictures of treatment team members. Use of this tool was positively associated with provider identification, understanding of roles, communication, and satisfaction. Kelly et al.³² noted that 60% of parents felt that portal use improved healthcare team communication.

The risk of bias was assessed for both adult and pediatric studies in aggregate. The average risk of bias using the Downs and Black Scale was 17.81 (range 14-22, standard deviation [SD] 2.20). Of the 16 included studies, 9 were rated at a low risk of bias (score

• >

18).^13-15,26-31 Risk of bias was greatest for measures of external validity (mean 2.88, range 2-3, SD 0.34), internal validity (mean 4.06, range 3-6, SD 1.00), and confounding (mean 2.69, range 1-6, SD 1.35). Two of 3 randomized controlled trials had a low risk of bias.^14,27 Interrater reliability for study quality adjudication was 0.90, suggesting excellent agreement (see supplementary Appendix B).

In this systematic review, the effects of visual tools on outcomes, such as provider identification, understanding of roles, patient–provider communication, and satisfaction with care, were variable. The majority of included studies were conducted on adult patients (n = 11).^{12-14,22-24,26,27,29,30,33} Pictures were the most frequently used tool (n = 7)^{13-15,23,27,31,33} and consequently had the greatest sample size across the review (n = 1297). While pictures had a positive influence on provider identification in all studies, comprehension of provider roles and satisfaction were variable. Although the content of whiteboards varied between studies, they showed favorable effects on provider identification (3 of 4 studies)^12,22,30 and satisfaction (2 of 2 studies).^22,30 While electronic medical record-based tools had a positive influence on outcomes,^26,28 only 1 accounted for language preferences.²⁸ Formatted notepads positively influenced patient–provider communication, but their use was limited by literacy.²⁴ Collectively, these data suggest that visual tools have varying effects on patient-reported outcomes, likely owing to differences in study design, interventions, and evaluation methods.

Theoretically, visual tools should facilitate easier identification of providers and engender collaborative relationships. However, such tools do not replace face-to-face patient–provider and family discussions. Rather, these enhancements best serve as a medium to asynchronously display information to patients and family members. Indeed, within the included studies, we found that the use of visual tools was effective in improving satisfaction (6/8 studies), identification (13/13 studies), and understanding of provider roles (8/10 studies). Thus, it is reasonable to say that, in conjunction with excellent clinical care, these tools have an important role in improving care delivery in the hospital.

Despite this promise, we noted that the effectiveness of individual tools varied, a fact that may relate to differences across studies. First, inconsistencies in the format and/or content of the tools were noted. For example, within studies using pictures, tools varied from individual photographs of each team member^13,23 to 1-page handouts with pictures of all team members.^14,15,31 Such differences in presentation could affect spatial recognition in identifying providers, as single photos are known to be easier to process than multiple images at the same time.³⁴ Second, no study evaluated patient preference of a visual tool. Thus, personal preferences for pictures versus whiteboards versus electronic modalities or a combination of tools might affect outcomes. Additionally, the utility of visual tools in visually impaired, confused, or non-English-speaking patients may limit effectiveness. Future studies that address these aspects and account for patient preferences may better elucidate the role of visual tools in hospitals.

Our results should be considered in the context of several limitations. First, only 3 studies used randomized trial designs; thus, confounding from unmeasured variables inherent to observational designs is possible. Second, none of the interventions tested were blinded to providers, raising the possibility of a Hawthorne effect (ie, alteration of provider behavior in response to awareness of being observed).³⁵ Third, all studies were conducted at single centers, and only 9 of 16 studies were rated at a low risk of bias; thus, caution in broad extrapolations of this literature is necessary.

However, our study has several strengths, including a thorough search of heterogeneous literature, inclusion of both adult and pediatric populations, and a focus on myriad patient-reported outcomes. Second, by contrasting outcomes and measurement strategies across studies, our review helps explicate differences in results related to variation in outcome measurement or presentation of visual data. Third, because we frame results by outcome and type of visual tool used, we are able to identify strengths and weaknesses of individual tools in novel ways. Finally, our data suggest that the use of picture-based techniques and whiteboards are among the most promising visual interventions. Future studies that pair graphic designers with patients to improve the layout of these tools might prove valuable. Additionally, because the measurement of outcomes is confounded by aspects such as lack of controls, severity of illness, and language barriers, a randomized design would help provide greater clarity regarding effectiveness.

In conclusion, we found that visual tools appear to foster recognition of providers and understanding of their roles. However, variability of format, content, and measurement of outcomes hinders the identification of a single optimal approach. Future work using randomized controlled trial designs and standardized tools and measurements would be welcomed.

Acknowledgments

The authors thank Laura Appel, Kevin O’Leary, and Siddharth Singh for providing unpublished data and clarifications to help these analyses.

Disclosure

 Anupama Goyal is the guarantor. Anupama Goyal and Komalpreet Tur performed primary data abstraction and analysis. Anupama Goyal, Scott Flanders, Jason Mann, and Vineet Chopra drafted the manuscript. All authors contributed to the development of the selection criteria, the risk of bias assessment strategy, and the data extraction criteria. Anupama Goyal, Jason Mann, Whitney Townsend, and Vineet Chopra developed the search strategy. Vineet Chopra provided systematic review expertise. All authors read, provided feedback, and approved the final manuscript. The authors declare that they have no conflicts of interest.

Patient satisfaction with medical care during hospitalization is a common quality metric.^1,2 Studies showing higher patient satisfaction have reported lower 30-day hospital readmissions³ and improved overall health.^4,5 Conversely, communication failures are associated with dissatisfaction among hospitalized patients and adverse outcomes.^6,7 A lack of familiarity with hospital providers weakens collaborative decision making and prevents high-quality patient care.^8,9

Bedside visual tools, such as whiteboards and pictures of medical staff, have been widely used to enhance communication between patients, families, and providers.^10,11 Results of studies evaluating these tools are varied. For example, 1 study found that 98% of patients were better able to identify physicians when their names were written on whiteboards.¹² Yet in another, only 21.1% of patients were more likely to correctly identify ≥1 physicians using pictures.¹³ Thus, despite widespread use,¹¹ whether visual tools improve patient satisfaction and patient care more broadly remains unclear.^14,15

We performed a systematic review to answer the following 3 questions: first, what is the effect of visual tools on outcomes (ie, provider identification, understanding of providers’ roles, patient–provider communication, and satisfaction); second, does impact vary by type of visual tool (eg, whiteboards vs pictures of providers); and third, what factors (eg, study design, patient population) are associated with provider identification, communication, and patient satisfaction?

Search Strategy

We used the Preferred Reporting Items for Systematic Reviews and Meta-Analysis when performing this review.¹⁶ A research librarian (WT) conducted serial searches for studies reporting the use of bedside visual tools for hospitalized patients in Medline (via OVID), Embase, SCOPUS, Web of Science, CINAHL, and Cochrane DSR and CENTRAL. Controlled vocabularies (ie, Medical Subject Headings terms) were used to identify synonyms for visual tools of interest. Additional studies were identified manually through bibliographies and meeting abstracts. No study design, publication date, or language restrictions were placed on the search, which was conducted between April 2016 and February 2017 (see supplementary Appendix A).

Study Selection

Two reviewers (AG and KT) independently assessed study eligibility; discrepancies were resolved by a third reviewer (VC). We included all adult or pediatric English language studies in which the effect of visual tool(s) on patient outcomes was reported. Visual tools were defined as the bedside display of information or an instrument given to patients to convey information regarding providers or medical care. Patient-reported outcomes included the following: (a) physician identification, (b) understanding of provider roles, (c) patient–provider communication, and (d) patient satisfaction with care. Providers were defined as physicians, residents, interns, medical students, nurse practitioners, or nurses. We excluded studies that were not original research (eg, conference abstracts, not peer reviewed), reported qualitative data without quantitative outcomes, or did not include a bedside visual tool. Given our interest in hospitalized general medicine patients, studies conducted in emergency departments, surgical units, obstetrics and gynecology wards, and intensive care units were excluded.

Data Extraction and Analysis

Data were extracted independently and in duplicate from all studies by using a template adapted from the Cochrane Collaboration.¹⁷ For all studies, we abstracted study design, type of visual tool (eg, whiteboards), unit setting (eg, medical), population studied (eg, adult vs pediatric), and outcomes reported (ie, physician identification, understanding of provider roles, communication, and satisfaction with care). Reviewers independently assessed and categorized the impact of tools on reported outcomes.

To standardize and compare outcomes across studies, the following were used to denote a positive association between visual tools and relevant outcomes: a greater number of physicians correctly identified by name/picture or title/role; the use of terms such as “high,” “agreed,” or “significant” on surveys; or ≥4 Likert scores for domains of identification, understanding of roles, communication, and satisfaction with care. Conversely, the inability to identify providers compared to the control/baseline; poor recall of titles/roles; lower Likert-scale scores (ie, ≤2); or survey terms such as “poor,” “disagreed,” or “insignificant” were considered to connote negative impact. Studies in which Likert scores were rated neither high nor low (ie, 3), or in which patients neither agreed nor disagreed on value were considered neutral.

Owing to clinical heterogeneity within studies, meta-analyses were not performed. Descriptive statistics were used to describe study outcomes. A priori¹⁸ studies were evaluated according to the following categories: design (eg, randomized vs observational), outcomes (eg, patient satisfaction), intervention (type of visual tool), and patient population (adult or pediatric). Because pediatric patients have underdeveloped communication skills and include parents and/or guardians, data from pediatric studies were tabulated and reported separately to those from adult studies.

Quality Assessment

As recommended by the Cochrane Collaboration, 2 reviewers (AG, KT) assessed the risk of study bias by using the Downs and Black Scale.^17,19 Discrepancies in assessment were resolved by a third reviewer (VC). This instrument uses a point-based system to estimate the quality of a study by rating domains such as internal and external validity, bias, and confounding. In keeping with prior systematic reviews,^18,20,21 studies with a score of ≥18 were considered high quality. Interrater agreement for the adjudication of study quality was calculated using the Cohen κ statistic.

Eleven studies were conducted on adult hospitalized pa­tients ^{12-14,22-24,26,27,29,30,33} and included 3 randomized controlled studies.^14,27,33

Results by Outcomes Provider Identification Nine studies measured patients’ ability to identify providers with the use of visual aids, and all 9 reported improvements in this outcome. Visual tools used to measure provider identification included pictures (n = 5),^{13,14,23,27,33} whiteboards (n = 3),^12,22,30 and patient portals (n = 1).²⁶ Within studies that used pictures, individual pictures (n = 2)^13,23 and handouts with pictures of multiple providers (n = 3) were used.^14,27,33 In 2 studies, care team members such as a dietitian, physiotherapist or pharmacist, were included when measuring identification.^14,33

Understanding Providers’ RolesSix studies assessed the effect of visual tools on patients’ understanding of provider roles.^{13,14,22,26,27,33} Four studies reported a positive effect with the use of pictures,^27,33 whiteboards,²² and patient portals.²⁶ However, 2 studies reported either no difference or negative impressions. Appel et al.¹⁴ reported no difference in the understanding of physician roles using a handout of providers’ pictures and titles. Arora et al.¹³ used individual pictures of physicians with descriptions of roles and found a negative association, as demonstrated by fewer patients rating their understanding of physicians’ roles as excellent or very good in the intervention period (45.6%) compared with the baseline (55.3%).

 

Patient–Provider Communication

Three studies evaluated the influence of visual tools on communication.^14,24,29 Using pictures, Appel et al.¹⁴ found no difference in the perceived quality of communication. Singh et al.²⁹ used whiteboards and reported improved communication scores for physicians and nurses. With notepads, patients surveyed by Farberg et al.²⁴ stated that the tool improved provider communication.

Patient Satisfaction

Five studies assessed patient satisfaction related to the use of visual tools. ^{22,23,27,30,33} One study reported satisfaction as positive with the use of individual pictures.²³ Two studies that used handouts with pictures of all team members reported either a positive³³ or neutral²⁷ impact on satisfaction. Studies that used whiteboards reported a positive association with satisfaction^22,30 despite differences in content, such as the inclusion of prewritten prompts for writing goals of care and scheduled tests³⁰ versus the name of the nurse and their education level.²²

Results by Type of Visual Tool Pictures

Five studies that used pictures reported a positive effect on provider identification.^{13,14,23,27,33} Two^27,33 of 4 studies^13,14,27,33 that assessed patients’ understanding of team member roles reported a positive influence, while 1 reported no difference.¹⁴ A fourth study demonstrated a negative association, perhaps due to differences in the description of providers’ roles listed on the tool.¹³ Only 1 study examined the influence of pictures on patient–provider communication, and this study found no difference.¹⁴ Satisfaction with care via the use of pictures varied between positive (2 studies)^23,33 and neutral (1 study).²⁷

Whiteboards

Four studies tested the use of whiteboards; of these, 3 reported a positive influence on provider identification.^12,22,30 One study reported a positive impact on patient–provider communication.²⁹ Two studies noted a positive effect on patient satisfaction.^22,30 Notably, the responsibility for updating whiteboards differed between the studies (ie, nurses only²² vs residents, medical students, and nurses).³⁰

Patient Portal

In 1 study, an electronic portal that included names with pictures of providers, descriptions of their roles, lists of medications, and scheduled tests and/or procedures was used as a visual tool. The portal improved patients’ identification of physicians and patients’ understanding of roles. However, improvements in the knowledge of medication changes and planned tests and/or procedures during hospitalization were not observed.²⁶ This finding would suggest limitations in the hospitalized patient’s knowledge of the plan of care, which could potentially weaken patient–provider communication.

Notepads

Only 1 study assessed the use of formatted notepads on patient–provider communication and noted a positive association. Notepads used prompts for different categories (eg, diagnosis/treatment, medications, etc) to encourage patient questions for providers.²⁴

Five studies were conducted on hospitalized pediatric units.^{15,25,28,31,32} All studies surveyed the parents, guardians, or caregivers of pediatric patients. One study excluded patients ≥12 years of age because of legal differences in access to adolescent health information,³² while another interviewed parents and/or guardians of teenagers.¹⁵

Results by Outcomes Provider Identification and Understanding of Physicians’ Roles

Four studies that assessed the influence of visual tools on provider identification and understanding of roles reported a positive association.^15,25,28,31 Visual tools varied between pictures (n = 2),^15,31 patient portal (n = 1),²⁸ and whiteboards and pictures combined (n = 1).²⁵ The measurement of outcomes varied between surveys with free text responses,²⁸ multiple choice questions,²⁵ and 1-5 Likert scales.^15,31

Patient–Provider Communication

Two studies assessed the impact of patient portal use on communication and reported a positive association.^28,32 The 2 portals autopopulated names, pictures, and roles of providers from electronic medical records. Singh et al.²⁸ used a portal that was also available in Spanish and accommodated for non-English speakers. Kelly et al.³² reported that 90% of parents perceived that portal use was associated with reduced errors in care, with 8% finding errors in their child’s medication list.

Patient Satisfaction

Three studies assessed patient satisfaction via the use of visual tools.^15,28,31 Singh et al.²⁸ noted a positive influence on satisfaction via a patient portal. Dudas et al.¹⁵ used a single-page handout with names and pictures of each provider, along with information regarding the training and roles of each provider. Distribution of these handouts to patients by investigators led to a positive influence on satisfaction. While Unaka et al.³¹ used a similar handout, they asked residents to distribute them and found no significant difference in satisfaction scores between the intervention (66%) and control group (62%).

Results by Type of Visual Tool Pictures

Two studies reported a positive impact on provider identification and understanding of roles with the use of pictures.^15,31 Dudas et al.¹⁵ demonstrated a 4.8-fold increase in the odds of parents identifying a medical student, as compared with the control. Similarly, after adjusting for length of stay and prior hospitalization, Unaka et al.³¹ reported that a higher percentage of patients correctly identified providers using this approach.

Whiteboard and Picture

One study evaluated the simultaneous use of whiteboards and pictures to improve the identification of providers. The study noted improved identification of supervising doctors and increased recognition of roles for supervising doctors, residents, and medical students.²⁵

Patient Portal

Two studies used patient portals as visual tools. Singh et al.²⁸ assessed the use of a patient portal with names, roles, and pictures of treatment team members. Use of this tool was positively associated with provider identification, understanding of roles, communication, and satisfaction. Kelly et al.³² noted that 60% of parents felt that portal use improved healthcare team communication.

The risk of bias was assessed for both adult and pediatric studies in aggregate. The average risk of bias using the Downs and Black Scale was 17.81 (range 14-22, standard deviation [SD] 2.20). Of the 16 included studies, 9 were rated at a low risk of bias (score

• >

18).^13-15,26-31 Risk of bias was greatest for measures of external validity (mean 2.88, range 2-3, SD 0.34), internal validity (mean 4.06, range 3-6, SD 1.00), and confounding (mean 2.69, range 1-6, SD 1.35). Two of 3 randomized controlled trials had a low risk of bias.^14,27 Interrater reliability for study quality adjudication was 0.90, suggesting excellent agreement (see supplementary Appendix B).

In this systematic review, the effects of visual tools on outcomes, such as provider identification, understanding of roles, patient–provider communication, and satisfaction with care, were variable. The majority of included studies were conducted on adult patients (n = 11).^{12-14,22-24,26,27,29,30,33} Pictures were the most frequently used tool (n = 7)^{13-15,23,27,31,33} and consequently had the greatest sample size across the review (n = 1297). While pictures had a positive influence on provider identification in all studies, comprehension of provider roles and satisfaction were variable. Although the content of whiteboards varied between studies, they showed favorable effects on provider identification (3 of 4 studies)^12,22,30 and satisfaction (2 of 2 studies).^22,30 While electronic medical record-based tools had a positive influence on outcomes,^26,28 only 1 accounted for language preferences.²⁸ Formatted notepads positively influenced patient–provider communication, but their use was limited by literacy.²⁴ Collectively, these data suggest that visual tools have varying effects on patient-reported outcomes, likely owing to differences in study design, interventions, and evaluation methods.

Theoretically, visual tools should facilitate easier identification of providers and engender collaborative relationships. However, such tools do not replace face-to-face patient–provider and family discussions. Rather, these enhancements best serve as a medium to asynchronously display information to patients and family members. Indeed, within the included studies, we found that the use of visual tools was effective in improving satisfaction (6/8 studies), identification (13/13 studies), and understanding of provider roles (8/10 studies). Thus, it is reasonable to say that, in conjunction with excellent clinical care, these tools have an important role in improving care delivery in the hospital.

Despite this promise, we noted that the effectiveness of individual tools varied, a fact that may relate to differences across studies. First, inconsistencies in the format and/or content of the tools were noted. For example, within studies using pictures, tools varied from individual photographs of each team member^13,23 to 1-page handouts with pictures of all team members.^14,15,31 Such differences in presentation could affect spatial recognition in identifying providers, as single photos are known to be easier to process than multiple images at the same time.³⁴ Second, no study evaluated patient preference of a visual tool. Thus, personal preferences for pictures versus whiteboards versus electronic modalities or a combination of tools might affect outcomes. Additionally, the utility of visual tools in visually impaired, confused, or non-English-speaking patients may limit effectiveness. Future studies that address these aspects and account for patient preferences may better elucidate the role of visual tools in hospitals.

Our results should be considered in the context of several limitations. First, only 3 studies used randomized trial designs; thus, confounding from unmeasured variables inherent to observational designs is possible. Second, none of the interventions tested were blinded to providers, raising the possibility of a Hawthorne effect (ie, alteration of provider behavior in response to awareness of being observed).³⁵ Third, all studies were conducted at single centers, and only 9 of 16 studies were rated at a low risk of bias; thus, caution in broad extrapolations of this literature is necessary.

However, our study has several strengths, including a thorough search of heterogeneous literature, inclusion of both adult and pediatric populations, and a focus on myriad patient-reported outcomes. Second, by contrasting outcomes and measurement strategies across studies, our review helps explicate differences in results related to variation in outcome measurement or presentation of visual data. Third, because we frame results by outcome and type of visual tool used, we are able to identify strengths and weaknesses of individual tools in novel ways. Finally, our data suggest that the use of picture-based techniques and whiteboards are among the most promising visual interventions. Future studies that pair graphic designers with patients to improve the layout of these tools might prove valuable. Additionally, because the measurement of outcomes is confounded by aspects such as lack of controls, severity of illness, and language barriers, a randomized design would help provide greater clarity regarding effectiveness.

In conclusion, we found that visual tools appear to foster recognition of providers and understanding of their roles. However, variability of format, content, and measurement of outcomes hinders the identification of a single optimal approach. Future work using randomized controlled trial designs and standardized tools and measurements would be welcomed.

Acknowledgments

The authors thank Laura Appel, Kevin O’Leary, and Siddharth Singh for providing unpublished data and clarifications to help these analyses.

Disclosure

 Anupama Goyal is the guarantor. Anupama Goyal and Komalpreet Tur performed primary data abstraction and analysis. Anupama Goyal, Scott Flanders, Jason Mann, and Vineet Chopra drafted the manuscript. All authors contributed to the development of the selection criteria, the risk of bias assessment strategy, and the data extraction criteria. Anupama Goyal, Jason Mann, Whitney Townsend, and Vineet Chopra developed the search strategy. Vineet Chopra provided systematic review expertise. All authors read, provided feedback, and approved the final manuscript. The authors declare that they have no conflicts of interest.

Patient satisfaction with medical care during hospitalization is a common quality metric.^1,2 Studies showing higher patient satisfaction have reported lower 30-day hospital readmissions³ and improved overall health.^4,5 Conversely, communication failures are associated with dissatisfaction among hospitalized patients and adverse outcomes.^6,7 A lack of familiarity with hospital providers weakens collaborative decision making and prevents high-quality patient care.^8,9

Bedside visual tools, such as whiteboards and pictures of medical staff, have been widely used to enhance communication between patients, families, and providers.^10,11 Results of studies evaluating these tools are varied. For example, 1 study found that 98% of patients were better able to identify physicians when their names were written on whiteboards.¹² Yet in another, only 21.1% of patients were more likely to correctly identify ≥1 physicians using pictures.¹³ Thus, despite widespread use,¹¹ whether visual tools improve patient satisfaction and patient care more broadly remains unclear.^14,15

We performed a systematic review to answer the following 3 questions: first, what is the effect of visual tools on outcomes (ie, provider identification, understanding of providers’ roles, patient–provider communication, and satisfaction); second, does impact vary by type of visual tool (eg, whiteboards vs pictures of providers); and third, what factors (eg, study design, patient population) are associated with provider identification, communication, and patient satisfaction?

Search Strategy

We used the Preferred Reporting Items for Systematic Reviews and Meta-Analysis when performing this review.¹⁶ A research librarian (WT) conducted serial searches for studies reporting the use of bedside visual tools for hospitalized patients in Medline (via OVID), Embase, SCOPUS, Web of Science, CINAHL, and Cochrane DSR and CENTRAL. Controlled vocabularies (ie, Medical Subject Headings terms) were used to identify synonyms for visual tools of interest. Additional studies were identified manually through bibliographies and meeting abstracts. No study design, publication date, or language restrictions were placed on the search, which was conducted between April 2016 and February 2017 (see supplementary Appendix A).

Study Selection

Two reviewers (AG and KT) independently assessed study eligibility; discrepancies were resolved by a third reviewer (VC). We included all adult or pediatric English language studies in which the effect of visual tool(s) on patient outcomes was reported. Visual tools were defined as the bedside display of information or an instrument given to patients to convey information regarding providers or medical care. Patient-reported outcomes included the following: (a) physician identification, (b) understanding of provider roles, (c) patient–provider communication, and (d) patient satisfaction with care. Providers were defined as physicians, residents, interns, medical students, nurse practitioners, or nurses. We excluded studies that were not original research (eg, conference abstracts, not peer reviewed), reported qualitative data without quantitative outcomes, or did not include a bedside visual tool. Given our interest in hospitalized general medicine patients, studies conducted in emergency departments, surgical units, obstetrics and gynecology wards, and intensive care units were excluded.

Data Extraction and Analysis

Data were extracted independently and in duplicate from all studies by using a template adapted from the Cochrane Collaboration.¹⁷ For all studies, we abstracted study design, type of visual tool (eg, whiteboards), unit setting (eg, medical), population studied (eg, adult vs pediatric), and outcomes reported (ie, physician identification, understanding of provider roles, communication, and satisfaction with care). Reviewers independently assessed and categorized the impact of tools on reported outcomes.

To standardize and compare outcomes across studies, the following were used to denote a positive association between visual tools and relevant outcomes: a greater number of physicians correctly identified by name/picture or title/role; the use of terms such as “high,” “agreed,” or “significant” on surveys; or ≥4 Likert scores for domains of identification, understanding of roles, communication, and satisfaction with care. Conversely, the inability to identify providers compared to the control/baseline; poor recall of titles/roles; lower Likert-scale scores (ie, ≤2); or survey terms such as “poor,” “disagreed,” or “insignificant” were considered to connote negative impact. Studies in which Likert scores were rated neither high nor low (ie, 3), or in which patients neither agreed nor disagreed on value were considered neutral.

Owing to clinical heterogeneity within studies, meta-analyses were not performed. Descriptive statistics were used to describe study outcomes. A priori¹⁸ studies were evaluated according to the following categories: design (eg, randomized vs observational), outcomes (eg, patient satisfaction), intervention (type of visual tool), and patient population (adult or pediatric). Because pediatric patients have underdeveloped communication skills and include parents and/or guardians, data from pediatric studies were tabulated and reported separately to those from adult studies.

Quality Assessment

As recommended by the Cochrane Collaboration, 2 reviewers (AG, KT) assessed the risk of study bias by using the Downs and Black Scale.^17,19 Discrepancies in assessment were resolved by a third reviewer (VC). This instrument uses a point-based system to estimate the quality of a study by rating domains such as internal and external validity, bias, and confounding. In keeping with prior systematic reviews,^18,20,21 studies with a score of ≥18 were considered high quality. Interrater agreement for the adjudication of study quality was calculated using the Cohen κ statistic.

Eleven studies were conducted on adult hospitalized pa­tients ^{12-14,22-24,26,27,29,30,33} and included 3 randomized controlled studies.^14,27,33

Results by Outcomes Provider Identification Nine studies measured patients’ ability to identify providers with the use of visual aids, and all 9 reported improvements in this outcome. Visual tools used to measure provider identification included pictures (n = 5),^{13,14,23,27,33} whiteboards (n = 3),^12,22,30 and patient portals (n = 1).²⁶ Within studies that used pictures, individual pictures (n = 2)^13,23 and handouts with pictures of multiple providers (n = 3) were used.^14,27,33 In 2 studies, care team members such as a dietitian, physiotherapist or pharmacist, were included when measuring identification.^14,33

Understanding Providers’ RolesSix studies assessed the effect of visual tools on patients’ understanding of provider roles.^{13,14,22,26,27,33} Four studies reported a positive effect with the use of pictures,^27,33 whiteboards,²² and patient portals.²⁶ However, 2 studies reported either no difference or negative impressions. Appel et al.¹⁴ reported no difference in the understanding of physician roles using a handout of providers’ pictures and titles. Arora et al.¹³ used individual pictures of physicians with descriptions of roles and found a negative association, as demonstrated by fewer patients rating their understanding of physicians’ roles as excellent or very good in the intervention period (45.6%) compared with the baseline (55.3%).

 

Patient–Provider Communication

Three studies evaluated the influence of visual tools on communication.^14,24,29 Using pictures, Appel et al.¹⁴ found no difference in the perceived quality of communication. Singh et al.²⁹ used whiteboards and reported improved communication scores for physicians and nurses. With notepads, patients surveyed by Farberg et al.²⁴ stated that the tool improved provider communication.

Patient Satisfaction

Five studies assessed patient satisfaction related to the use of visual tools. ^{22,23,27,30,33} One study reported satisfaction as positive with the use of individual pictures.²³ Two studies that used handouts with pictures of all team members reported either a positive³³ or neutral²⁷ impact on satisfaction. Studies that used whiteboards reported a positive association with satisfaction^22,30 despite differences in content, such as the inclusion of prewritten prompts for writing goals of care and scheduled tests³⁰ versus the name of the nurse and their education level.²²

Results by Type of Visual Tool Pictures

Five studies that used pictures reported a positive effect on provider identification.^{13,14,23,27,33} Two^27,33 of 4 studies^13,14,27,33 that assessed patients’ understanding of team member roles reported a positive influence, while 1 reported no difference.¹⁴ A fourth study demonstrated a negative association, perhaps due to differences in the description of providers’ roles listed on the tool.¹³ Only 1 study examined the influence of pictures on patient–provider communication, and this study found no difference.¹⁴ Satisfaction with care via the use of pictures varied between positive (2 studies)^23,33 and neutral (1 study).²⁷

Whiteboards

Four studies tested the use of whiteboards; of these, 3 reported a positive influence on provider identification.^12,22,30 One study reported a positive impact on patient–provider communication.²⁹ Two studies noted a positive effect on patient satisfaction.^22,30 Notably, the responsibility for updating whiteboards differed between the studies (ie, nurses only²² vs residents, medical students, and nurses).³⁰

Patient Portal

In 1 study, an electronic portal that included names with pictures of providers, descriptions of their roles, lists of medications, and scheduled tests and/or procedures was used as a visual tool. The portal improved patients’ identification of physicians and patients’ understanding of roles. However, improvements in the knowledge of medication changes and planned tests and/or procedures during hospitalization were not observed.²⁶ This finding would suggest limitations in the hospitalized patient’s knowledge of the plan of care, which could potentially weaken patient–provider communication.

Notepads

Only 1 study assessed the use of formatted notepads on patient–provider communication and noted a positive association. Notepads used prompts for different categories (eg, diagnosis/treatment, medications, etc) to encourage patient questions for providers.²⁴

Five studies were conducted on hospitalized pediatric units.^{15,25,28,31,32} All studies surveyed the parents, guardians, or caregivers of pediatric patients. One study excluded patients ≥12 years of age because of legal differences in access to adolescent health information,³² while another interviewed parents and/or guardians of teenagers.¹⁵

Results by Outcomes Provider Identification and Understanding of Physicians’ Roles

Four studies that assessed the influence of visual tools on provider identification and understanding of roles reported a positive association.^15,25,28,31 Visual tools varied between pictures (n = 2),^15,31 patient portal (n = 1),²⁸ and whiteboards and pictures combined (n = 1).²⁵ The measurement of outcomes varied between surveys with free text responses,²⁸ multiple choice questions,²⁵ and 1-5 Likert scales.^15,31

Patient–Provider Communication

Two studies assessed the impact of patient portal use on communication and reported a positive association.^28,32 The 2 portals autopopulated names, pictures, and roles of providers from electronic medical records. Singh et al.²⁸ used a portal that was also available in Spanish and accommodated for non-English speakers. Kelly et al.³² reported that 90% of parents perceived that portal use was associated with reduced errors in care, with 8% finding errors in their child’s medication list.

Patient Satisfaction

Three studies assessed patient satisfaction via the use of visual tools.^15,28,31 Singh et al.²⁸ noted a positive influence on satisfaction via a patient portal. Dudas et al.¹⁵ used a single-page handout with names and pictures of each provider, along with information regarding the training and roles of each provider. Distribution of these handouts to patients by investigators led to a positive influence on satisfaction. While Unaka et al.³¹ used a similar handout, they asked residents to distribute them and found no significant difference in satisfaction scores between the intervention (66%) and control group (62%).

Results by Type of Visual Tool Pictures

Two studies reported a positive impact on provider identification and understanding of roles with the use of pictures.^15,31 Dudas et al.¹⁵ demonstrated a 4.8-fold increase in the odds of parents identifying a medical student, as compared with the control. Similarly, after adjusting for length of stay and prior hospitalization, Unaka et al.³¹ reported that a higher percentage of patients correctly identified providers using this approach.

Whiteboard and Picture

One study evaluated the simultaneous use of whiteboards and pictures to improve the identification of providers. The study noted improved identification of supervising doctors and increased recognition of roles for supervising doctors, residents, and medical students.²⁵

Patient Portal

Two studies used patient portals as visual tools. Singh et al.²⁸ assessed the use of a patient portal with names, roles, and pictures of treatment team members. Use of this tool was positively associated with provider identification, understanding of roles, communication, and satisfaction. Kelly et al.³² noted that 60% of parents felt that portal use improved healthcare team communication.

The risk of bias was assessed for both adult and pediatric studies in aggregate. The average risk of bias using the Downs and Black Scale was 17.81 (range 14-22, standard deviation [SD] 2.20). Of the 16 included studies, 9 were rated at a low risk of bias (score

• >

18).^13-15,26-31 Risk of bias was greatest for measures of external validity (mean 2.88, range 2-3, SD 0.34), internal validity (mean 4.06, range 3-6, SD 1.00), and confounding (mean 2.69, range 1-6, SD 1.35). Two of 3 randomized controlled trials had a low risk of bias.^14,27 Interrater reliability for study quality adjudication was 0.90, suggesting excellent agreement (see supplementary Appendix B).

In this systematic review, the effects of visual tools on outcomes, such as provider identification, understanding of roles, patient–provider communication, and satisfaction with care, were variable. The majority of included studies were conducted on adult patients (n = 11).^{12-14,22-24,26,27,29,30,33} Pictures were the most frequently used tool (n = 7)^{13-15,23,27,31,33} and consequently had the greatest sample size across the review (n = 1297). While pictures had a positive influence on provider identification in all studies, comprehension of provider roles and satisfaction were variable. Although the content of whiteboards varied between studies, they showed favorable effects on provider identification (3 of 4 studies)^12,22,30 and satisfaction (2 of 2 studies).^22,30 While electronic medical record-based tools had a positive influence on outcomes,^26,28 only 1 accounted for language preferences.²⁸ Formatted notepads positively influenced patient–provider communication, but their use was limited by literacy.²⁴ Collectively, these data suggest that visual tools have varying effects on patient-reported outcomes, likely owing to differences in study design, interventions, and evaluation methods.

Theoretically, visual tools should facilitate easier identification of providers and engender collaborative relationships. However, such tools do not replace face-to-face patient–provider and family discussions. Rather, these enhancements best serve as a medium to asynchronously display information to patients and family members. Indeed, within the included studies, we found that the use of visual tools was effective in improving satisfaction (6/8 studies), identification (13/13 studies), and understanding of provider roles (8/10 studies). Thus, it is reasonable to say that, in conjunction with excellent clinical care, these tools have an important role in improving care delivery in the hospital.

Despite this promise, we noted that the effectiveness of individual tools varied, a fact that may relate to differences across studies. First, inconsistencies in the format and/or content of the tools were noted. For example, within studies using pictures, tools varied from individual photographs of each team member^13,23 to 1-page handouts with pictures of all team members.^14,15,31 Such differences in presentation could affect spatial recognition in identifying providers, as single photos are known to be easier to process than multiple images at the same time.³⁴ Second, no study evaluated patient preference of a visual tool. Thus, personal preferences for pictures versus whiteboards versus electronic modalities or a combination of tools might affect outcomes. Additionally, the utility of visual tools in visually impaired, confused, or non-English-speaking patients may limit effectiveness. Future studies that address these aspects and account for patient preferences may better elucidate the role of visual tools in hospitals.

Our results should be considered in the context of several limitations. First, only 3 studies used randomized trial designs; thus, confounding from unmeasured variables inherent to observational designs is possible. Second, none of the interventions tested were blinded to providers, raising the possibility of a Hawthorne effect (ie, alteration of provider behavior in response to awareness of being observed).³⁵ Third, all studies were conducted at single centers, and only 9 of 16 studies were rated at a low risk of bias; thus, caution in broad extrapolations of this literature is necessary.

However, our study has several strengths, including a thorough search of heterogeneous literature, inclusion of both adult and pediatric populations, and a focus on myriad patient-reported outcomes. Second, by contrasting outcomes and measurement strategies across studies, our review helps explicate differences in results related to variation in outcome measurement or presentation of visual data. Third, because we frame results by outcome and type of visual tool used, we are able to identify strengths and weaknesses of individual tools in novel ways. Finally, our data suggest that the use of picture-based techniques and whiteboards are among the most promising visual interventions. Future studies that pair graphic designers with patients to improve the layout of these tools might prove valuable. Additionally, because the measurement of outcomes is confounded by aspects such as lack of controls, severity of illness, and language barriers, a randomized design would help provide greater clarity regarding effectiveness.

In conclusion, we found that visual tools appear to foster recognition of providers and understanding of their roles. However, variability of format, content, and measurement of outcomes hinders the identification of a single optimal approach. Future work using randomized controlled trial designs and standardized tools and measurements would be welcomed.

Acknowledgments

The authors thank Laura Appel, Kevin O’Leary, and Siddharth Singh for providing unpublished data and clarifications to help these analyses.

Disclosure

 Anupama Goyal is the guarantor. Anupama Goyal and Komalpreet Tur performed primary data abstraction and analysis. Anupama Goyal, Scott Flanders, Jason Mann, and Vineet Chopra drafted the manuscript. All authors contributed to the development of the selection criteria, the risk of bias assessment strategy, and the data extraction criteria. Anupama Goyal, Jason Mann, Whitney Townsend, and Vineet Chopra developed the search strategy. Vineet Chopra provided systematic review expertise. All authors read, provided feedback, and approved the final manuscript. The authors declare that they have no conflicts of interest.

Raynaud phenomenon is an overactive vascular response to cold and emotional stress that results in cutaneous color changes and sensory symptoms of the digits (Figure 1). It can occur in isolation as primary Raynaud phenomenon or secondary to another disease process. It is thought to be triggered by a heightened sympathetic vasoconstrictive response of small arteriovenous anastomoses in the fingers, toes, ears, and tip of the nose. These structures play a key role in maintaining a stable core body temperature by cutaneous thermoregulation.¹

Secondary Raynaud phenomenon can be seen with a wide array of systemic conditions as well as environmental and drug exposures. It is a frequent feature of autoimmune rheumatic conditions such as systemic sclerosis, mixed connective tissue disease, systemic lupus erythematosus, and dermatomyositis. Less commonly, cryoproteinemias, paraneoplastic syndromes, hypothyroidism, and carpal tunnel syndrome can be associated with or cause Raynaud phenomenon. Vibratory trauma (eg, from using a jackhammer) and drugs (eg, vasopressors, stimulants, ergots, chemotherapeutic agents) can also cause Raynaud phenomenon.¹

A variety of disorders that cause vasospasm or vascular occlusion of the peripheral circulation can mimic typical Raynaud phenomenon, including peripheral nerve injury,² complex regional pain syndrome,³ occlusive vascular disease, vasculitis, acrocyanosis,⁴ and thoracic outlet syndrome.

The prevalence of Raynaud phenomenon is not exactly known, in part due to geographic differences in climate and variation in methods of assessment. However, a 2015 systematic review and meta-analysis of primary Raynaud phenomenon determined a pooled prevalence of 4.85% (95% confidence interval [CI] 2.08%–8.71%) in the general population.⁵ Accordingly, accurate identification and management of this condition is a useful skill for the internist.

COLD SENSITIVITY AND COLOR CHANGES

Because there are no confirmatory diagnostic tests for this condition, there are no formal diagnostic criteria. However, many experts agree that Raynaud phenomenon can be diagnosed clinically when patients report:

• Unusual sensitivity of the fingers to cold, manifesting as pain or paresthesia (eg, tingling, pricking, numbness), and
• Color changes of the fingers when exposed to cold, specifically pale white or blue-black, or both.⁶

Provocative testing such as submerging patients’ hands in cold water is not recommended, as it is distressing to the patient and inconsistent in triggering an event.

Pain is a symptom of critical digital ischemia.

The skin color changes are due to rapid alterations in blood flow in digital skin. The pale white is due to markedly reduced or absent flow secondary to intense vasoconstriction, the blue-black is due to hypoxemic venous stasis, and the red blush is due to hyperemic reperfusion (Figure 1). However, not all patients have all 3 phases of the classic triphasic color changes, and color changes may not follow a set sequence.

Raynaud phenomenon can also occur in other areas of the body that have thermoregulatory vessels, such as the toes, ears, nipples, tongue, and nose. While some patients with Raynaud phenomenon have a finger that is more sensitive than the others, repeated isolated single-digit or asymmetric events without typical progression to all fingers suggest a secondary local structural disease requiring further investigation (see below).

Symptoms related to Raynaud often mimic sensory changes including paresthesias, numbness, aching, and clumsiness of the hand. Abnormal vascular reactivity has been implicated as a causative factor in several disorders, such as migraine headache, preeclampsia, and variant angina. While case reports, case series, and some controlled studies have linked Raynaud phenomenon and these conditions, there is no solid evidence of a systemic vaso­spastic disorder in patients with primary Raynaud phenomenon.

Raynaud phenomenon is triggered by more than just a cold ambient temperature. Provocation can occur during movement from warmer to relatively cooler temperatures, as well as during episodes of elevated sympathetic activity (eg, emotional distress or fear). In fact, maintaining full body warmth as well as emotional equilibrium are the most important strategies to reduce the frequency of attacks.

PRIMARY VS SECONDARY RAYNAUD PHENOMENON

To distinguish between primary and secondary Raynaud phenomenon, a careful history and physical examination are paramount.

Primary Raynaud phenomenon

In uncomplicated primary Raynaud phenomenon, the episodes typically last 15 to 20 minutes after rewarming and usually start in a single finger and spread to other digits symmetrically and bilaterally.⁷ The thumb is often spared, and ischemic digital ulcers do not occur. Vasoconstrictive episodes are mild.

Females under age 20 are most commonly affected. In our experience, a young woman with the above clinical picture, no signs or symptoms suggestive of connective tissue disease (see below), and normal nailfold capillaries can be diagnosed as having primary Raynaud phenomenon without any further workup.

Careful clinical follow-up is recommended, because if an occult secondary process is indeed present, most patients will begin to show additional symptoms or signs of it within 2 years of the onset of Raynaud phenomenon.

Should a clinician be unfamiliar with nailfold capillary examination, or if symptoms (eg, fatigue or arthralgia) or signs (eg, rash, arthritis) suggestive of connective tissue disease are present, referral to a rheumatologist for further evaluation is appropriate. Results of further diagnostic testing dictated by the history and physical such as a screening antinuclear antibody test can be sent before referral.

Secondary Raynaud phenomenon

Several clinical features suggest secondary Raynaud phenomenon and warrant referral to a rheumatologist:

• Age 20 or older at onset
• Frequent severe vasoconstrictive episodes
• Male sex
• Thumb involvement

• 

  Signs of an autoimmune rheumatic disease, eg, sclerodactyly, cutaneous or mucosal matted telangiectasia, inflammatory arthritis, an abnormal lung examination, severe digital ischemia with ulceration or gangrene, or nailfold capillary dilation or dropout (Figure 2)⁸

  Isolated single-limb or 1-finger ischemic events, seen in macrovascular occlusive disease or inflammatory disease mimicking Raynaud phenomenon (eg, atherosclerosis, vasculitis); when isolated acute ischemic events occur in the upper or lower extremity, a further workup is necessary.

Figure 3 shows our approach to evaluation.

NONPHARMACOLOGIC THERAPY

Cold avoidance and stress management are first-line therapies for preventing Raynaud attacks and must be part of any treatment strategy. Digital arteries and thermoregulatory vessels of the skin are predominantly under sympathetic adrenergic control, so temperature changes and emotional stressors trigger vasoconstriction. Patients should be counseled to:

Keep the whole body warm. Patients should wear multiple layers of clothing, a hat, warm gloves, and warm socks. Commercially available hand-warmers can help, especially for patients who live in cold climates.

Learn to avoid or manage stress. Good communication, attention to the patient’s needs, and regular follow-up for reassurance are paramount. For some patients, psychotropic medications to manage mood may help. Behavioral approaches have been suggested for acute stress management. One approach, autogenic training, is a form of relaxation with temperature biofeedback in which finger temperature data are provided to patients to help them learn to relax by monitoring their internal states and changes in temperature. However, there are no strong data to support the routine use of this technique or the use of one behavioral approach over another. Trials have generally been of low quality and limited by small sample size.⁹

Stop smoking!¹⁰

Stop a Raynaud attack should one occur, eg, place the hands under warm water or in a warm part of the body, such as under legs when sitting. This can help speed recovery.

In addition, the physician should:

Eliminate vasoconstricting agents such as nonselective beta-blockers, ergots, triptans, and amphetamines.

PHARMACOLOGIC THERAPY

For many patients, nonpharmacologic interventions are enough to decrease the severity and frequency of attacks. However, if Raynaud phenomenon continues to negatively affect quality of life, drug therapy can be added (Table 1).

Calcium channel blockers

Calcium channel blockers are first-line agents for both primary and secondary Raynaud phenomenon that does not adequately respond to nonpharmacologic interventions. These agents are effective, available, and reasonably inexpensive. 

Dihydropyridine calcium channel blockers such as nifedipine and amlodipine are commonly used. Both drugs are acceptable options, though some patients may respond better to one than the other in terms of symptoms and side effects. Nondihydropyridines such as diltiazem can also be used, but they have less potent vasodilatory effects because they are less selective for vascular smooth muscle.

These medications should be started at the lowest dose and titrated up over several weeks as tolerated to achieve their maximal effect. Intermittent therapy (eg, during the winter months only) is reasonable for primary Raynaud without risk of digital ulceration, as relief of symptoms and improvement in quality of life are the main indications for therapy in this circumstance.

A 2016 Cochrane review and meta-analysis of the use of calcium channel blockers to treat primary Raynaud phenomenon included 7 randomized controlled trials with 296 patients treated with either nifedipine or nicardipine.¹¹ There was moderate-quality evidence that these drugs minimally decreased the frequency of attacks (standardized mean difference of 0.23; 95% CI 0.08–0.38, P = .003). This translated to 1.72 fewer attacks per week with treatment than with no pharmacologic therapy (95% CI 0.60–2.84). When analyzed individually, only nifedipine was effective; nicardipine did not decrease the frequency of attacks.

Unfortunately, calcium channel blockers failed to decrease the severity of attacks (according to unvalidated severity scoring systems) or make any differences in physiologic measurement outcomes. Attacks were not completely eliminated, just less frequent than before treatment.¹¹

Most commonly reported side effects included headache, flushing, hypotension, edema, and, rarely, gastrointestinal reflux. Use of these medications may be limited by hypotension.

The review was limited by the small sample size, short duration of treatment, and relatively low doses of calcium channel blockers used in the available studies.¹¹

A 2005 meta-analysis also indicated a statistically significant decrease of 2.8 to 5 attacks per week with nifedipine treatment, though this study also included some patients with secondary Raynaud phenomenon.¹²

Phosphodiesterase type 5 inhibitors

When calcium channel blockers do not adequately control symptoms, phosphodiesterase type 5 (PDE5) inhibitors can be added or substituted. These medications work by preventing breakdown of cyclic guanosine monophosphate, which induces relaxation in vascular smooth muscle and vasodilation.

Sildenafil can be started at a low dose (20 mg daily) and up-titrated to the maximum dose (20 mg 3 times daily) as tolerated.

A 2014 meta-analysis of 6 randomized controlled trials included 244 patients with secondary Raynaud phenomenon treated with sildenafil, tadalafil, or vardenafil.¹³ These drugs decreased the daily frequency of attacks by about 0.5 per day vs placebo (–0.49,  95% CI –0.71 to –0.28, P <  .0001). PDE5 inhibitors also decreased the severity of attacks (based on the Raynaud’s Condition Score, a popular scoring system) and the duration of attacks by a statistically significant amount.

Almost all patients in these 6 trials were on PDE5 monotherapy. Data on the cumulative benefit of calcium channel blocker and PDE5 inhibitor combination therapy are not yet available. Not all patients tolerate combination therapy, as it can cause symptomatic hypotension, but it can be a successful option in some.

There are also no data showing that either calcium channel blockers or PDE5 inhibitors are superior, though the former are less expensive. A small double-blind, randomized, crossover study of udenafil vs amlodipine in the treatment of secondary Raynaud phenomenon showed that both medications significantly decreased the frequency of attacks and had comparable efficacy.¹⁴

Cost and insurance coverage. We have generally been successful in obtaining coverage for this off-label use of PDE5 inhibitors, though additional effort may be required. No drug (not even a calcium channel blocker) is approved by the US Food and Drug Administration for use in Raynaud phenomenon. In our experience, a letter of appeal outlining the rationale for use and citing supporting publications can lead to successful coverage of a medication. If the drug is still not approved, the patient either pays for it out of pocket or another agent is selected. In certain circumstances, pharmaceutical companies may provide prescription assistance for compassionate use of these drugs in Raynaud phenomenon, although this also takes letter-writing, phone calls, or both on the part of the physician.

Topical nitrates

Patients who have an unsatisfactory response to calcium channel blockers with or without PDE5 inhibitors can try topical nitrates, available as sustained-release transdermal patches, tapes, creams, gels, and ointments.

Small trials have noted slight improvement in the Raynaud Condition Score¹⁵ and finger temperature¹⁶ with these therapies. Another trial noted decreased frequency of attacks and symptoms with the use of sustained-release glyceryl trinitrate patches, but use was limited by intolerable headache.¹⁷

In our experience, topical nitrates are most helpful for patients who have 1 or a few digits that are more severely affected than the others, and we reserve these drugs for this indication. Localized vasodilation can provide targeted rapid relief of more ischemic areas.

Topical nitroglycerin can be applied to the base of the ischemic digit for 6 to 12 hours. Preparations vary, and patients should be closely monitored for dose response and tolerance.

Combining a topical nitrate with a calcium channel blocker is safe, but the use of a nitrate with a PDE5 inhibitor is contraindicated due to the risk of hypotension. The use of topical nitrates may be limited by systemic side effects such as headache and flushing and a lack of benefit over time.

Other therapies

If the aforementioned agents are not tolerated or not effective, there is limited evidence that other therapies reduce the frequency and sometimes the severity of attacks. These are not first-line agents but may be tried when other options have been exhausted and symptoms persist. There are no data to support combining these therapies, but in our experience doing so may help some patients in whom drug-drug interactions are not prohibitive.

Prazosin, an alpha-1-adrenergic receptor antagonist, was reported to improve Raynaud phenomenon in 2 small studies in the 1980s, but we do not use it since better options are available. In addition, the vasoactive blood vessels involved do not have alpha-1 receptors, so there is no theoretical basis for using prazosin.^18,19

Fluoxetine, a selective serotonin reuptake inhibitor, reduced the frequency and severity of attacks in a 6-week crossover study with nifedipine.²⁰

Losartan, an angiotensin II receptor blocker, also reduced the severity and frequency of attacks when compared with nifedipine.²¹

Pentoxifylline, a nonselective phosphodiesterase inhibitor, showed some benefit in a trial in 11 patients with primary Raynaud.²²

Atorvastatin, a lipid-lowering drug, reduced the number of digital ulcers in patients with secondary Raynaud already on first-line vasodilatory therapy, and might be added in this situation.²³

Botulinum toxin A injections have some data to support their use, but evidence is based on uncontrolled case series.²⁴ A controlled trial in scleroderma patients with severe Raynaud phenomenon found botulinum toxin to be no better than placebo.²⁵

Prostacyclin preparations are available. Intermittent intravenous doses of prostacyclin analogues over several days can be used in resistant cases. Oral prostacyclin agents have not shown consistent benefit. New prostacyclin receptor agonists are under investigation.

Overall, we move to other options only in patients with persistent symptoms that impair quality of life, or in patients with recurrent digital ischemic lesions that have not responded to calcium channel blockers and PDE5 inhibitors or nitrates, either alone or in combination.

DIGITAL ULCERATION AND ACUTE DIGITAL ISCHEMIC CRISIS

Patients with secondary Raynaud phenomenon may be at risk of recurrent digital ulceration and acute digital ischemia with gangrene. These patients should be comanaged with a rheumatologist so that the underlying disease process is fully addressed. Digital ulcers should be inspected closely for signs of infection, which may require treatment with antibiotics.

Acute digital ischemia is a medical emergency and should prompt inpatient admission with warming, emotional regulation, and pain control (often with narcotics) to decrease sympathetic vasoconstriction. These patients require aggressive vasodilatory therapy to reverse the ischemic event.

A short-acting calcium channel blocker or combination therapy with a calcium channel blocker and a PDE5 inhibitor or topical nitrate should be started. If there is no benefit, then transient intravenous vasodilatory therapy with a prostacyclin (epoprostenol) or localized digital sympathectomy is used to prevent digital loss.

The endothelin receptor inhibitor bosentan has been shown to decrease recurrent digital ulcers in patients with scleroderma, and while bosentan does not decrease the frequency of Raynaud attacks, it can be used in this select group to prevent new digital ulcers.

Treatment options may be limited by insurance coverage or access to intravenous infusions.

TAKE-HOME RECOMMENDATIONS

For many patients with primary or secondary Raynaud phenomenon, nonpharmacologic interventions are all that are required to decrease the frequency of attacks and improve quality of life. The goal should not be to eliminate attacks completely, as aggressive drug treatment may cause more harm than benefit. From our perspective, the goals of treatment should be to improve quality of life and prevent ischemic complications.

Pharmacologic therapies should be added only if attacks remain poorly controlled with incapacitating symptoms, or if the patient has digital ischemic ulcers. Calcium channel blockers are first-line therapy, given proven efficacy and low cost, and should be titrated to the maximum tolerated dose before adding or substituting other agents.

Raynaud phenomenon is an overactive vascular response to cold and emotional stress that results in cutaneous color changes and sensory symptoms of the digits (Figure 1). It can occur in isolation as primary Raynaud phenomenon or secondary to another disease process. It is thought to be triggered by a heightened sympathetic vasoconstrictive response of small arteriovenous anastomoses in the fingers, toes, ears, and tip of the nose. These structures play a key role in maintaining a stable core body temperature by cutaneous thermoregulation.¹

Secondary Raynaud phenomenon can be seen with a wide array of systemic conditions as well as environmental and drug exposures. It is a frequent feature of autoimmune rheumatic conditions such as systemic sclerosis, mixed connective tissue disease, systemic lupus erythematosus, and dermatomyositis. Less commonly, cryoproteinemias, paraneoplastic syndromes, hypothyroidism, and carpal tunnel syndrome can be associated with or cause Raynaud phenomenon. Vibratory trauma (eg, from using a jackhammer) and drugs (eg, vasopressors, stimulants, ergots, chemotherapeutic agents) can also cause Raynaud phenomenon.¹

A variety of disorders that cause vasospasm or vascular occlusion of the peripheral circulation can mimic typical Raynaud phenomenon, including peripheral nerve injury,² complex regional pain syndrome,³ occlusive vascular disease, vasculitis, acrocyanosis,⁴ and thoracic outlet syndrome.

The prevalence of Raynaud phenomenon is not exactly known, in part due to geographic differences in climate and variation in methods of assessment. However, a 2015 systematic review and meta-analysis of primary Raynaud phenomenon determined a pooled prevalence of 4.85% (95% confidence interval [CI] 2.08%–8.71%) in the general population.⁵ Accordingly, accurate identification and management of this condition is a useful skill for the internist.

COLD SENSITIVITY AND COLOR CHANGES

Because there are no confirmatory diagnostic tests for this condition, there are no formal diagnostic criteria. However, many experts agree that Raynaud phenomenon can be diagnosed clinically when patients report:

• Unusual sensitivity of the fingers to cold, manifesting as pain or paresthesia (eg, tingling, pricking, numbness), and
• Color changes of the fingers when exposed to cold, specifically pale white or blue-black, or both.⁶

Provocative testing such as submerging patients’ hands in cold water is not recommended, as it is distressing to the patient and inconsistent in triggering an event.

Pain is a symptom of critical digital ischemia.

The skin color changes are due to rapid alterations in blood flow in digital skin. The pale white is due to markedly reduced or absent flow secondary to intense vasoconstriction, the blue-black is due to hypoxemic venous stasis, and the red blush is due to hyperemic reperfusion (Figure 1). However, not all patients have all 3 phases of the classic triphasic color changes, and color changes may not follow a set sequence.

Raynaud phenomenon can also occur in other areas of the body that have thermoregulatory vessels, such as the toes, ears, nipples, tongue, and nose. While some patients with Raynaud phenomenon have a finger that is more sensitive than the others, repeated isolated single-digit or asymmetric events without typical progression to all fingers suggest a secondary local structural disease requiring further investigation (see below).

Symptoms related to Raynaud often mimic sensory changes including paresthesias, numbness, aching, and clumsiness of the hand. Abnormal vascular reactivity has been implicated as a causative factor in several disorders, such as migraine headache, preeclampsia, and variant angina. While case reports, case series, and some controlled studies have linked Raynaud phenomenon and these conditions, there is no solid evidence of a systemic vaso­spastic disorder in patients with primary Raynaud phenomenon.

Raynaud phenomenon is triggered by more than just a cold ambient temperature. Provocation can occur during movement from warmer to relatively cooler temperatures, as well as during episodes of elevated sympathetic activity (eg, emotional distress or fear). In fact, maintaining full body warmth as well as emotional equilibrium are the most important strategies to reduce the frequency of attacks.

PRIMARY VS SECONDARY RAYNAUD PHENOMENON

To distinguish between primary and secondary Raynaud phenomenon, a careful history and physical examination are paramount.

Primary Raynaud phenomenon

In uncomplicated primary Raynaud phenomenon, the episodes typically last 15 to 20 minutes after rewarming and usually start in a single finger and spread to other digits symmetrically and bilaterally.⁷ The thumb is often spared, and ischemic digital ulcers do not occur. Vasoconstrictive episodes are mild.

Females under age 20 are most commonly affected. In our experience, a young woman with the above clinical picture, no signs or symptoms suggestive of connective tissue disease (see below), and normal nailfold capillaries can be diagnosed as having primary Raynaud phenomenon without any further workup.

Careful clinical follow-up is recommended, because if an occult secondary process is indeed present, most patients will begin to show additional symptoms or signs of it within 2 years of the onset of Raynaud phenomenon.

Should a clinician be unfamiliar with nailfold capillary examination, or if symptoms (eg, fatigue or arthralgia) or signs (eg, rash, arthritis) suggestive of connective tissue disease are present, referral to a rheumatologist for further evaluation is appropriate. Results of further diagnostic testing dictated by the history and physical such as a screening antinuclear antibody test can be sent before referral.

Secondary Raynaud phenomenon

Several clinical features suggest secondary Raynaud phenomenon and warrant referral to a rheumatologist:

• Age 20 or older at onset
• Frequent severe vasoconstrictive episodes
• Male sex
• Thumb involvement

• 

  Signs of an autoimmune rheumatic disease, eg, sclerodactyly, cutaneous or mucosal matted telangiectasia, inflammatory arthritis, an abnormal lung examination, severe digital ischemia with ulceration or gangrene, or nailfold capillary dilation or dropout (Figure 2)⁸

  Isolated single-limb or 1-finger ischemic events, seen in macrovascular occlusive disease or inflammatory disease mimicking Raynaud phenomenon (eg, atherosclerosis, vasculitis); when isolated acute ischemic events occur in the upper or lower extremity, a further workup is necessary.

Figure 3 shows our approach to evaluation.

NONPHARMACOLOGIC THERAPY

Cold avoidance and stress management are first-line therapies for preventing Raynaud attacks and must be part of any treatment strategy. Digital arteries and thermoregulatory vessels of the skin are predominantly under sympathetic adrenergic control, so temperature changes and emotional stressors trigger vasoconstriction. Patients should be counseled to:

Keep the whole body warm. Patients should wear multiple layers of clothing, a hat, warm gloves, and warm socks. Commercially available hand-warmers can help, especially for patients who live in cold climates.

Learn to avoid or manage stress. Good communication, attention to the patient’s needs, and regular follow-up for reassurance are paramount. For some patients, psychotropic medications to manage mood may help. Behavioral approaches have been suggested for acute stress management. One approach, autogenic training, is a form of relaxation with temperature biofeedback in which finger temperature data are provided to patients to help them learn to relax by monitoring their internal states and changes in temperature. However, there are no strong data to support the routine use of this technique or the use of one behavioral approach over another. Trials have generally been of low quality and limited by small sample size.⁹

Stop smoking!¹⁰

Stop a Raynaud attack should one occur, eg, place the hands under warm water or in a warm part of the body, such as under legs when sitting. This can help speed recovery.

In addition, the physician should:

Eliminate vasoconstricting agents such as nonselective beta-blockers, ergots, triptans, and amphetamines.

PHARMACOLOGIC THERAPY

For many patients, nonpharmacologic interventions are enough to decrease the severity and frequency of attacks. However, if Raynaud phenomenon continues to negatively affect quality of life, drug therapy can be added (Table 1).

Calcium channel blockers

Calcium channel blockers are first-line agents for both primary and secondary Raynaud phenomenon that does not adequately respond to nonpharmacologic interventions. These agents are effective, available, and reasonably inexpensive. 

Dihydropyridine calcium channel blockers such as nifedipine and amlodipine are commonly used. Both drugs are acceptable options, though some patients may respond better to one than the other in terms of symptoms and side effects. Nondihydropyridines such as diltiazem can also be used, but they have less potent vasodilatory effects because they are less selective for vascular smooth muscle.

These medications should be started at the lowest dose and titrated up over several weeks as tolerated to achieve their maximal effect. Intermittent therapy (eg, during the winter months only) is reasonable for primary Raynaud without risk of digital ulceration, as relief of symptoms and improvement in quality of life are the main indications for therapy in this circumstance.

A 2016 Cochrane review and meta-analysis of the use of calcium channel blockers to treat primary Raynaud phenomenon included 7 randomized controlled trials with 296 patients treated with either nifedipine or nicardipine.¹¹ There was moderate-quality evidence that these drugs minimally decreased the frequency of attacks (standardized mean difference of 0.23; 95% CI 0.08–0.38, P = .003). This translated to 1.72 fewer attacks per week with treatment than with no pharmacologic therapy (95% CI 0.60–2.84). When analyzed individually, only nifedipine was effective; nicardipine did not decrease the frequency of attacks.

Unfortunately, calcium channel blockers failed to decrease the severity of attacks (according to unvalidated severity scoring systems) or make any differences in physiologic measurement outcomes. Attacks were not completely eliminated, just less frequent than before treatment.¹¹

Most commonly reported side effects included headache, flushing, hypotension, edema, and, rarely, gastrointestinal reflux. Use of these medications may be limited by hypotension.

The review was limited by the small sample size, short duration of treatment, and relatively low doses of calcium channel blockers used in the available studies.¹¹

A 2005 meta-analysis also indicated a statistically significant decrease of 2.8 to 5 attacks per week with nifedipine treatment, though this study also included some patients with secondary Raynaud phenomenon.¹²

Phosphodiesterase type 5 inhibitors

When calcium channel blockers do not adequately control symptoms, phosphodiesterase type 5 (PDE5) inhibitors can be added or substituted. These medications work by preventing breakdown of cyclic guanosine monophosphate, which induces relaxation in vascular smooth muscle and vasodilation.

Sildenafil can be started at a low dose (20 mg daily) and up-titrated to the maximum dose (20 mg 3 times daily) as tolerated.

A 2014 meta-analysis of 6 randomized controlled trials included 244 patients with secondary Raynaud phenomenon treated with sildenafil, tadalafil, or vardenafil.¹³ These drugs decreased the daily frequency of attacks by about 0.5 per day vs placebo (–0.49,  95% CI –0.71 to –0.28, P <  .0001). PDE5 inhibitors also decreased the severity of attacks (based on the Raynaud’s Condition Score, a popular scoring system) and the duration of attacks by a statistically significant amount.

Almost all patients in these 6 trials were on PDE5 monotherapy. Data on the cumulative benefit of calcium channel blocker and PDE5 inhibitor combination therapy are not yet available. Not all patients tolerate combination therapy, as it can cause symptomatic hypotension, but it can be a successful option in some.

There are also no data showing that either calcium channel blockers or PDE5 inhibitors are superior, though the former are less expensive. A small double-blind, randomized, crossover study of udenafil vs amlodipine in the treatment of secondary Raynaud phenomenon showed that both medications significantly decreased the frequency of attacks and had comparable efficacy.¹⁴

Cost and insurance coverage. We have generally been successful in obtaining coverage for this off-label use of PDE5 inhibitors, though additional effort may be required. No drug (not even a calcium channel blocker) is approved by the US Food and Drug Administration for use in Raynaud phenomenon. In our experience, a letter of appeal outlining the rationale for use and citing supporting publications can lead to successful coverage of a medication. If the drug is still not approved, the patient either pays for it out of pocket or another agent is selected. In certain circumstances, pharmaceutical companies may provide prescription assistance for compassionate use of these drugs in Raynaud phenomenon, although this also takes letter-writing, phone calls, or both on the part of the physician.

Topical nitrates

Patients who have an unsatisfactory response to calcium channel blockers with or without PDE5 inhibitors can try topical nitrates, available as sustained-release transdermal patches, tapes, creams, gels, and ointments.

Small trials have noted slight improvement in the Raynaud Condition Score¹⁵ and finger temperature¹⁶ with these therapies. Another trial noted decreased frequency of attacks and symptoms with the use of sustained-release glyceryl trinitrate patches, but use was limited by intolerable headache.¹⁷

In our experience, topical nitrates are most helpful for patients who have 1 or a few digits that are more severely affected than the others, and we reserve these drugs for this indication. Localized vasodilation can provide targeted rapid relief of more ischemic areas.

Topical nitroglycerin can be applied to the base of the ischemic digit for 6 to 12 hours. Preparations vary, and patients should be closely monitored for dose response and tolerance.

Combining a topical nitrate with a calcium channel blocker is safe, but the use of a nitrate with a PDE5 inhibitor is contraindicated due to the risk of hypotension. The use of topical nitrates may be limited by systemic side effects such as headache and flushing and a lack of benefit over time.

Other therapies

If the aforementioned agents are not tolerated or not effective, there is limited evidence that other therapies reduce the frequency and sometimes the severity of attacks. These are not first-line agents but may be tried when other options have been exhausted and symptoms persist. There are no data to support combining these therapies, but in our experience doing so may help some patients in whom drug-drug interactions are not prohibitive.

Prazosin, an alpha-1-adrenergic receptor antagonist, was reported to improve Raynaud phenomenon in 2 small studies in the 1980s, but we do not use it since better options are available. In addition, the vasoactive blood vessels involved do not have alpha-1 receptors, so there is no theoretical basis for using prazosin.^18,19

Fluoxetine, a selective serotonin reuptake inhibitor, reduced the frequency and severity of attacks in a 6-week crossover study with nifedipine.²⁰

Losartan, an angiotensin II receptor blocker, also reduced the severity and frequency of attacks when compared with nifedipine.²¹

Pentoxifylline, a nonselective phosphodiesterase inhibitor, showed some benefit in a trial in 11 patients with primary Raynaud.²²

Atorvastatin, a lipid-lowering drug, reduced the number of digital ulcers in patients with secondary Raynaud already on first-line vasodilatory therapy, and might be added in this situation.²³

Botulinum toxin A injections have some data to support their use, but evidence is based on uncontrolled case series.²⁴ A controlled trial in scleroderma patients with severe Raynaud phenomenon found botulinum toxin to be no better than placebo.²⁵

Prostacyclin preparations are available. Intermittent intravenous doses of prostacyclin analogues over several days can be used in resistant cases. Oral prostacyclin agents have not shown consistent benefit. New prostacyclin receptor agonists are under investigation.

Overall, we move to other options only in patients with persistent symptoms that impair quality of life, or in patients with recurrent digital ischemic lesions that have not responded to calcium channel blockers and PDE5 inhibitors or nitrates, either alone or in combination.

DIGITAL ULCERATION AND ACUTE DIGITAL ISCHEMIC CRISIS

Patients with secondary Raynaud phenomenon may be at risk of recurrent digital ulceration and acute digital ischemia with gangrene. These patients should be comanaged with a rheumatologist so that the underlying disease process is fully addressed. Digital ulcers should be inspected closely for signs of infection, which may require treatment with antibiotics.

Acute digital ischemia is a medical emergency and should prompt inpatient admission with warming, emotional regulation, and pain control (often with narcotics) to decrease sympathetic vasoconstriction. These patients require aggressive vasodilatory therapy to reverse the ischemic event.

A short-acting calcium channel blocker or combination therapy with a calcium channel blocker and a PDE5 inhibitor or topical nitrate should be started. If there is no benefit, then transient intravenous vasodilatory therapy with a prostacyclin (epoprostenol) or localized digital sympathectomy is used to prevent digital loss.

The endothelin receptor inhibitor bosentan has been shown to decrease recurrent digital ulcers in patients with scleroderma, and while bosentan does not decrease the frequency of Raynaud attacks, it can be used in this select group to prevent new digital ulcers.

Treatment options may be limited by insurance coverage or access to intravenous infusions.

TAKE-HOME RECOMMENDATIONS

For many patients with primary or secondary Raynaud phenomenon, nonpharmacologic interventions are all that are required to decrease the frequency of attacks and improve quality of life. The goal should not be to eliminate attacks completely, as aggressive drug treatment may cause more harm than benefit. From our perspective, the goals of treatment should be to improve quality of life and prevent ischemic complications.

Pharmacologic therapies should be added only if attacks remain poorly controlled with incapacitating symptoms, or if the patient has digital ischemic ulcers. Calcium channel blockers are first-line therapy, given proven efficacy and low cost, and should be titrated to the maximum tolerated dose before adding or substituting other agents.

Raynaud phenomenon is an overactive vascular response to cold and emotional stress that results in cutaneous color changes and sensory symptoms of the digits (Figure 1). It can occur in isolation as primary Raynaud phenomenon or secondary to another disease process. It is thought to be triggered by a heightened sympathetic vasoconstrictive response of small arteriovenous anastomoses in the fingers, toes, ears, and tip of the nose. These structures play a key role in maintaining a stable core body temperature by cutaneous thermoregulation.¹

Secondary Raynaud phenomenon can be seen with a wide array of systemic conditions as well as environmental and drug exposures. It is a frequent feature of autoimmune rheumatic conditions such as systemic sclerosis, mixed connective tissue disease, systemic lupus erythematosus, and dermatomyositis. Less commonly, cryoproteinemias, paraneoplastic syndromes, hypothyroidism, and carpal tunnel syndrome can be associated with or cause Raynaud phenomenon. Vibratory trauma (eg, from using a jackhammer) and drugs (eg, vasopressors, stimulants, ergots, chemotherapeutic agents) can also cause Raynaud phenomenon.¹

A variety of disorders that cause vasospasm or vascular occlusion of the peripheral circulation can mimic typical Raynaud phenomenon, including peripheral nerve injury,² complex regional pain syndrome,³ occlusive vascular disease, vasculitis, acrocyanosis,⁴ and thoracic outlet syndrome.

The prevalence of Raynaud phenomenon is not exactly known, in part due to geographic differences in climate and variation in methods of assessment. However, a 2015 systematic review and meta-analysis of primary Raynaud phenomenon determined a pooled prevalence of 4.85% (95% confidence interval [CI] 2.08%–8.71%) in the general population.⁵ Accordingly, accurate identification and management of this condition is a useful skill for the internist.

COLD SENSITIVITY AND COLOR CHANGES

Because there are no confirmatory diagnostic tests for this condition, there are no formal diagnostic criteria. However, many experts agree that Raynaud phenomenon can be diagnosed clinically when patients report:

• Unusual sensitivity of the fingers to cold, manifesting as pain or paresthesia (eg, tingling, pricking, numbness), and
• Color changes of the fingers when exposed to cold, specifically pale white or blue-black, or both.⁶

Provocative testing such as submerging patients’ hands in cold water is not recommended, as it is distressing to the patient and inconsistent in triggering an event.

Pain is a symptom of critical digital ischemia.

The skin color changes are due to rapid alterations in blood flow in digital skin. The pale white is due to markedly reduced or absent flow secondary to intense vasoconstriction, the blue-black is due to hypoxemic venous stasis, and the red blush is due to hyperemic reperfusion (Figure 1). However, not all patients have all 3 phases of the classic triphasic color changes, and color changes may not follow a set sequence.

Raynaud phenomenon can also occur in other areas of the body that have thermoregulatory vessels, such as the toes, ears, nipples, tongue, and nose. While some patients with Raynaud phenomenon have a finger that is more sensitive than the others, repeated isolated single-digit or asymmetric events without typical progression to all fingers suggest a secondary local structural disease requiring further investigation (see below).

Symptoms related to Raynaud often mimic sensory changes including paresthesias, numbness, aching, and clumsiness of the hand. Abnormal vascular reactivity has been implicated as a causative factor in several disorders, such as migraine headache, preeclampsia, and variant angina. While case reports, case series, and some controlled studies have linked Raynaud phenomenon and these conditions, there is no solid evidence of a systemic vaso­spastic disorder in patients with primary Raynaud phenomenon.

Raynaud phenomenon is triggered by more than just a cold ambient temperature. Provocation can occur during movement from warmer to relatively cooler temperatures, as well as during episodes of elevated sympathetic activity (eg, emotional distress or fear). In fact, maintaining full body warmth as well as emotional equilibrium are the most important strategies to reduce the frequency of attacks.

PRIMARY VS SECONDARY RAYNAUD PHENOMENON

To distinguish between primary and secondary Raynaud phenomenon, a careful history and physical examination are paramount.

Primary Raynaud phenomenon

In uncomplicated primary Raynaud phenomenon, the episodes typically last 15 to 20 minutes after rewarming and usually start in a single finger and spread to other digits symmetrically and bilaterally.⁷ The thumb is often spared, and ischemic digital ulcers do not occur. Vasoconstrictive episodes are mild.

Females under age 20 are most commonly affected. In our experience, a young woman with the above clinical picture, no signs or symptoms suggestive of connective tissue disease (see below), and normal nailfold capillaries can be diagnosed as having primary Raynaud phenomenon without any further workup.

Careful clinical follow-up is recommended, because if an occult secondary process is indeed present, most patients will begin to show additional symptoms or signs of it within 2 years of the onset of Raynaud phenomenon.

Should a clinician be unfamiliar with nailfold capillary examination, or if symptoms (eg, fatigue or arthralgia) or signs (eg, rash, arthritis) suggestive of connective tissue disease are present, referral to a rheumatologist for further evaluation is appropriate. Results of further diagnostic testing dictated by the history and physical such as a screening antinuclear antibody test can be sent before referral.

Secondary Raynaud phenomenon

Several clinical features suggest secondary Raynaud phenomenon and warrant referral to a rheumatologist:

• Age 20 or older at onset
• Frequent severe vasoconstrictive episodes
• Male sex
• Thumb involvement

• 

  Signs of an autoimmune rheumatic disease, eg, sclerodactyly, cutaneous or mucosal matted telangiectasia, inflammatory arthritis, an abnormal lung examination, severe digital ischemia with ulceration or gangrene, or nailfold capillary dilation or dropout (Figure 2)⁸

  Isolated single-limb or 1-finger ischemic events, seen in macrovascular occlusive disease or inflammatory disease mimicking Raynaud phenomenon (eg, atherosclerosis, vasculitis); when isolated acute ischemic events occur in the upper or lower extremity, a further workup is necessary.

Figure 3 shows our approach to evaluation.

NONPHARMACOLOGIC THERAPY

Cold avoidance and stress management are first-line therapies for preventing Raynaud attacks and must be part of any treatment strategy. Digital arteries and thermoregulatory vessels of the skin are predominantly under sympathetic adrenergic control, so temperature changes and emotional stressors trigger vasoconstriction. Patients should be counseled to:

Keep the whole body warm. Patients should wear multiple layers of clothing, a hat, warm gloves, and warm socks. Commercially available hand-warmers can help, especially for patients who live in cold climates.

Learn to avoid or manage stress. Good communication, attention to the patient’s needs, and regular follow-up for reassurance are paramount. For some patients, psychotropic medications to manage mood may help. Behavioral approaches have been suggested for acute stress management. One approach, autogenic training, is a form of relaxation with temperature biofeedback in which finger temperature data are provided to patients to help them learn to relax by monitoring their internal states and changes in temperature. However, there are no strong data to support the routine use of this technique or the use of one behavioral approach over another. Trials have generally been of low quality and limited by small sample size.⁹

Stop smoking!¹⁰

Stop a Raynaud attack should one occur, eg, place the hands under warm water or in a warm part of the body, such as under legs when sitting. This can help speed recovery.

In addition, the physician should:

Eliminate vasoconstricting agents such as nonselective beta-blockers, ergots, triptans, and amphetamines.

PHARMACOLOGIC THERAPY

For many patients, nonpharmacologic interventions are enough to decrease the severity and frequency of attacks. However, if Raynaud phenomenon continues to negatively affect quality of life, drug therapy can be added (Table 1).

Calcium channel blockers

Calcium channel blockers are first-line agents for both primary and secondary Raynaud phenomenon that does not adequately respond to nonpharmacologic interventions. These agents are effective, available, and reasonably inexpensive. 

Dihydropyridine calcium channel blockers such as nifedipine and amlodipine are commonly used. Both drugs are acceptable options, though some patients may respond better to one than the other in terms of symptoms and side effects. Nondihydropyridines such as diltiazem can also be used, but they have less potent vasodilatory effects because they are less selective for vascular smooth muscle.

These medications should be started at the lowest dose and titrated up over several weeks as tolerated to achieve their maximal effect. Intermittent therapy (eg, during the winter months only) is reasonable for primary Raynaud without risk of digital ulceration, as relief of symptoms and improvement in quality of life are the main indications for therapy in this circumstance.

A 2016 Cochrane review and meta-analysis of the use of calcium channel blockers to treat primary Raynaud phenomenon included 7 randomized controlled trials with 296 patients treated with either nifedipine or nicardipine.¹¹ There was moderate-quality evidence that these drugs minimally decreased the frequency of attacks (standardized mean difference of 0.23; 95% CI 0.08–0.38, P = .003). This translated to 1.72 fewer attacks per week with treatment than with no pharmacologic therapy (95% CI 0.60–2.84). When analyzed individually, only nifedipine was effective; nicardipine did not decrease the frequency of attacks.

Unfortunately, calcium channel blockers failed to decrease the severity of attacks (according to unvalidated severity scoring systems) or make any differences in physiologic measurement outcomes. Attacks were not completely eliminated, just less frequent than before treatment.¹¹

Most commonly reported side effects included headache, flushing, hypotension, edema, and, rarely, gastrointestinal reflux. Use of these medications may be limited by hypotension.

The review was limited by the small sample size, short duration of treatment, and relatively low doses of calcium channel blockers used in the available studies.¹¹

A 2005 meta-analysis also indicated a statistically significant decrease of 2.8 to 5 attacks per week with nifedipine treatment, though this study also included some patients with secondary Raynaud phenomenon.¹²

Phosphodiesterase type 5 inhibitors

When calcium channel blockers do not adequately control symptoms, phosphodiesterase type 5 (PDE5) inhibitors can be added or substituted. These medications work by preventing breakdown of cyclic guanosine monophosphate, which induces relaxation in vascular smooth muscle and vasodilation.

Sildenafil can be started at a low dose (20 mg daily) and up-titrated to the maximum dose (20 mg 3 times daily) as tolerated.

A 2014 meta-analysis of 6 randomized controlled trials included 244 patients with secondary Raynaud phenomenon treated with sildenafil, tadalafil, or vardenafil.¹³ These drugs decreased the daily frequency of attacks by about 0.5 per day vs placebo (–0.49,  95% CI –0.71 to –0.28, P <  .0001). PDE5 inhibitors also decreased the severity of attacks (based on the Raynaud’s Condition Score, a popular scoring system) and the duration of attacks by a statistically significant amount.

Almost all patients in these 6 trials were on PDE5 monotherapy. Data on the cumulative benefit of calcium channel blocker and PDE5 inhibitor combination therapy are not yet available. Not all patients tolerate combination therapy, as it can cause symptomatic hypotension, but it can be a successful option in some.

There are also no data showing that either calcium channel blockers or PDE5 inhibitors are superior, though the former are less expensive. A small double-blind, randomized, crossover study of udenafil vs amlodipine in the treatment of secondary Raynaud phenomenon showed that both medications significantly decreased the frequency of attacks and had comparable efficacy.¹⁴

Cost and insurance coverage. We have generally been successful in obtaining coverage for this off-label use of PDE5 inhibitors, though additional effort may be required. No drug (not even a calcium channel blocker) is approved by the US Food and Drug Administration for use in Raynaud phenomenon. In our experience, a letter of appeal outlining the rationale for use and citing supporting publications can lead to successful coverage of a medication. If the drug is still not approved, the patient either pays for it out of pocket or another agent is selected. In certain circumstances, pharmaceutical companies may provide prescription assistance for compassionate use of these drugs in Raynaud phenomenon, although this also takes letter-writing, phone calls, or both on the part of the physician.

Topical nitrates

Patients who have an unsatisfactory response to calcium channel blockers with or without PDE5 inhibitors can try topical nitrates, available as sustained-release transdermal patches, tapes, creams, gels, and ointments.

Small trials have noted slight improvement in the Raynaud Condition Score¹⁵ and finger temperature¹⁶ with these therapies. Another trial noted decreased frequency of attacks and symptoms with the use of sustained-release glyceryl trinitrate patches, but use was limited by intolerable headache.¹⁷

In our experience, topical nitrates are most helpful for patients who have 1 or a few digits that are more severely affected than the others, and we reserve these drugs for this indication. Localized vasodilation can provide targeted rapid relief of more ischemic areas.

Topical nitroglycerin can be applied to the base of the ischemic digit for 6 to 12 hours. Preparations vary, and patients should be closely monitored for dose response and tolerance.

Combining a topical nitrate with a calcium channel blocker is safe, but the use of a nitrate with a PDE5 inhibitor is contraindicated due to the risk of hypotension. The use of topical nitrates may be limited by systemic side effects such as headache and flushing and a lack of benefit over time.

Other therapies

If the aforementioned agents are not tolerated or not effective, there is limited evidence that other therapies reduce the frequency and sometimes the severity of attacks. These are not first-line agents but may be tried when other options have been exhausted and symptoms persist. There are no data to support combining these therapies, but in our experience doing so may help some patients in whom drug-drug interactions are not prohibitive.

Prazosin, an alpha-1-adrenergic receptor antagonist, was reported to improve Raynaud phenomenon in 2 small studies in the 1980s, but we do not use it since better options are available. In addition, the vasoactive blood vessels involved do not have alpha-1 receptors, so there is no theoretical basis for using prazosin.^18,19

Fluoxetine, a selective serotonin reuptake inhibitor, reduced the frequency and severity of attacks in a 6-week crossover study with nifedipine.²⁰

Losartan, an angiotensin II receptor blocker, also reduced the severity and frequency of attacks when compared with nifedipine.²¹

Pentoxifylline, a nonselective phosphodiesterase inhibitor, showed some benefit in a trial in 11 patients with primary Raynaud.²²

Atorvastatin, a lipid-lowering drug, reduced the number of digital ulcers in patients with secondary Raynaud already on first-line vasodilatory therapy, and might be added in this situation.²³

Botulinum toxin A injections have some data to support their use, but evidence is based on uncontrolled case series.²⁴ A controlled trial in scleroderma patients with severe Raynaud phenomenon found botulinum toxin to be no better than placebo.²⁵

Prostacyclin preparations are available. Intermittent intravenous doses of prostacyclin analogues over several days can be used in resistant cases. Oral prostacyclin agents have not shown consistent benefit. New prostacyclin receptor agonists are under investigation.

Overall, we move to other options only in patients with persistent symptoms that impair quality of life, or in patients with recurrent digital ischemic lesions that have not responded to calcium channel blockers and PDE5 inhibitors or nitrates, either alone or in combination.

DIGITAL ULCERATION AND ACUTE DIGITAL ISCHEMIC CRISIS

Patients with secondary Raynaud phenomenon may be at risk of recurrent digital ulceration and acute digital ischemia with gangrene. These patients should be comanaged with a rheumatologist so that the underlying disease process is fully addressed. Digital ulcers should be inspected closely for signs of infection, which may require treatment with antibiotics.

Acute digital ischemia is a medical emergency and should prompt inpatient admission with warming, emotional regulation, and pain control (often with narcotics) to decrease sympathetic vasoconstriction. These patients require aggressive vasodilatory therapy to reverse the ischemic event.

A short-acting calcium channel blocker or combination therapy with a calcium channel blocker and a PDE5 inhibitor or topical nitrate should be started. If there is no benefit, then transient intravenous vasodilatory therapy with a prostacyclin (epoprostenol) or localized digital sympathectomy is used to prevent digital loss.

The endothelin receptor inhibitor bosentan has been shown to decrease recurrent digital ulcers in patients with scleroderma, and while bosentan does not decrease the frequency of Raynaud attacks, it can be used in this select group to prevent new digital ulcers.

Treatment options may be limited by insurance coverage or access to intravenous infusions.

TAKE-HOME RECOMMENDATIONS

For many patients with primary or secondary Raynaud phenomenon, nonpharmacologic interventions are all that are required to decrease the frequency of attacks and improve quality of life. The goal should not be to eliminate attacks completely, as aggressive drug treatment may cause more harm than benefit. From our perspective, the goals of treatment should be to improve quality of life and prevent ischemic complications.

Pharmacologic therapies should be added only if attacks remain poorly controlled with incapacitating symptoms, or if the patient has digital ischemic ulcers. Calcium channel blockers are first-line therapy, given proven efficacy and low cost, and should be titrated to the maximum tolerated dose before adding or substituting other agents.

Antipsychotics are FDA-approved as a primary treatment for schizophrenia and bipolar disorder and as adjunctive therapy for major depressive disorder. In the United States, approximately 26% of antipsychotic prescriptions written for these indications are for individuals age >65.¹ Additionally, antipsychotics are widely used to treat behavioral symptoms associated with dementia.¹ The rapid expansion of the use of second-generation antipsychotics (SGAs), in particular, has been driven in part by their lower risk for extrapyramidal symptoms (EPS) compared with first-generation antipsychotics (FGAs).¹ However, a growing body of data indicates that all antipsychotics have a range of adverse effects in older patients. This focus is critical in light of demographic trends—in the next 10 to 15 years, the population age >60 will grow 3.5 times more rapidly than the general population.²

In this context, psychiatrists need information on the relative risks of antipsychotics for older patients. This 3-part series summarizes findings and recommendations on safety and tolerability when prescribing antipsychotics in older individuals with chronic psychotic disorders, such as schizophrenia, bipolar disorder, depression, and dementia. This review aims to:

• briefly summarize the major studies and analyses relevant to older patients with these diagnoses
• provide a summative opinion on safety and tolerability issues in these older adults
• highlight the gaps in the evidence base and areas that need additional research.

Part 1 focuses on older adults with schizophrenia or bipolar disorder. Subsequent articles will focus on prescribing antipsychotics to older adults with depression and those with dementia.

Schizophrenia

Summary of benefits, place in treatment armamentarium. Individuals with schizophrenia have a shorter life expectancy than that of the general population mostly as a result of suicide and comorbid physical illnesses,³ but the number of patients with schizophrenia age >55 will double over the next 2 decades.⁴ With aging, both positive and negative symptoms may be a focus of treatment (Table 1).^5,6 Antipsychotics are a first-line treatment for older patients with schizophrenia with few medication alternatives.⁷ Safety risks associated with antipsychotics in older people span a broad spectrum (Table 2).⁸

A 6-week prospective RCT evaluated paliperidone extended-release vs placebo in 114 older adults (age ≥65 years; mean age, 70 years) with schizophrenia.¹⁴ There was an optional 24-week extension of open-label treatment with paliperidone. Mean daily dose of paliperidone was 8.4 mg. Efficacy measures did not show consistent statistically significant differences between treatment groups. Discontinuation rates were similar between paliperidone (7%) vs placebo (8%). Serious adverse events occurred in 3% of paliperidone-treated vs 8% of placebo-treated patients. Elevated prolactin levels occurred in one-half of paliperidone-treated patients. There were no prolactin or glucose treatment-related adverse events or significant mean changes in body weight for either paliperidone-treated or placebo-treated patients. Safety findings in the 24-week, open-label extension group were consistent with the RCT results.

Howanitz et al¹⁵ conducted a 12-week, prospective RCT that compared clozapine (mean dose, 300 mg/d) with chlorpromazine (mean dose, 600 mg/d) in 42 older adults (mean age, 67 years) with schizophrenia. Drop-out rate prior to 5 weeks was 19% and similar between groups. Common adverse effects included sialorrhea, hematologic abnormalities, sedation, tachycardia, EPS, and weight gain. Although both drugs were effective, more patients taking clozapine had tachycardia and weight gain, while more chlorpromazine patients reported sedation.

There have been other, less rigorous studies.^7,8 Most of these studies evaluated risperidone and olanzapine, and most were conducted in “younger” geriatric patients (age <75 years). Although patients who participate in clinical trials may be healthier than “typical” patients, adverse effects such as EPS, sedation, and weight gain were still relatively common in these studies.

Other clinical data. A major consideration in treating older adults with schizophrenia is balancing the need to administer an antipsychotic dose high enough to alleviate psychotic symptoms while minimizing dose-dependent adverse effects. There is a U-shaped relationship between age and vulnerability to antipsychotic adverse effects,^16,17 wherein adverse effects are highest at younger and older ages. Evidence supports using the lowest effective antipsychotic dose for geriatric patients with schizophrenia. Positive emission tomography (PET) studies suggest that older patients develop EPS with lower doses despite lower receptor occupancy.^17,18 A recent study of 35 older patients (mean age, 60.1 years) with schizophrenia obtained PET, clinical measures, and blood pharmacokinetic measures before and after reduction of risperidone or olanzapine doses.¹⁸ A ≥40% reduction in dose was associated with reduced adverse effects, particularly EPS and elevation of prolactin levels. Moreover, the therapeutic window of striatal D2/D3 receptor occupancy appeared to be 50% to 60% in these older patients, compared with 65% to 80% in younger patients.

Long-term risks of antipsychotic treatment across the lifespan are less clear, with evidence suggesting both lower and higher mortality risk.^19,20 It is difficult to fully disentangle the long-term risks of antipsychotics from the cumulative effects of lifestyle and comorbidity among individuals who have lived with schizophrenia for decades. Large naturalistic studies that include substantial numbers of older people with schizophrenia might be a way to elicit more information on long-term safety. The Schizophrenia Outpatient Health Outcome (SOHO) study was a large naturalistic trial that recruited >10,000 individuals with schizophrenia in 10 European countries.²¹ Although the SOHO study found differences between antipsychotics and adverse effects, such as EPS, weight gain, and sexual dysfunction, because the mean age of these patients was approximately 40 years and the follow-up period was only 3 years, it is difficult to draw conclusions that could be relevant to older individuals who have had schizophrenia for decades.

Bipolar Disorder

A secondary analysis of mixed-age, RCTs examined response in older adults (age ≥55 years) with bipolar I depression who received lurasidone as monotherapy or adjunctive therapy.³² In the monotherapy study, these patients were randomized to 6 weeks of lurasidone 20 to 60 mg/d, lurasidone 80 to 120 mg/d, or placebo. In the adjunctive therapy study, they were randomized to lurasidone 20 to 120 mg/d or placebo with either lithium or valproate. There were 83 older adults (17.1% of the sample) in the monotherapy study and 53 (15.6%) in the adjunctive therapy study. Mean improvement in depression was significantly higher for both doses of lurasidone monotherapy than placebo. Adjunctive lurasidone was not associated with statistically significant improvement vs placebo. The most frequent adverse events in older patients on lurasidone monotherapy 20 to 60 mg/d or 80 to 120 mg/d were nausea (18.5% and 9.7%, respectively) and somnolence (11.1% and 0%, respectively). Akathisia (9.7%) and insomnia (9.7%) were the most common adverse events in the group receiving 80 to 120 mg/d, with the rate of akathisia exhibiting a dose-related increase. Weight change with lurasidone was similar to placebo, and there were no clinically meaningful group changes in vital signs, electrocardiography, or laboratory parameters.

A small (N = 20) open study found improvement in older adults with bipolar depression with aripiprazole (mean dose, 10.3 mg/d).³³ Adverse effects included restlessness and weight gain (n = 3, 9% each), sedation (n = 2, 10%), and drooling and diarrhea/loose stools (n = 1, 5% each). In another small study (N = 15) using asenapine (mean dose, 11.2 mg/d) in mainly older bipolar patients with depression, the most common adverse effects were gastrointestinal (GI) discomfort (n = 5, 33%) and restlessness, tremors, cognitive difficulties, and sluggishness (n = 2, 13% each).³⁴

Clinical trials: Bipolar mania. Researchers conducted a pooled analysis of two 12-week randomized trials comparing quetiapine with placebo in a mixed-age sample with bipolar mania.³⁵ In a subgroup of 59 older patients (mean age, 62.9 years), manic symptoms improved significantly more with quetiapine (modal dose, 550 mg/d) than with placebo. Adverse effects reported by >10% of older patients were dry mouth, somnolence, postural hypotension, insomnia, weight gain, and dizziness. Insomnia was reported by >10% of patients receiving placebo.

In a case series of 11 elderly patients with mania receiving asenapine, Baruch et al³⁶ reported a 63% remission rate. One patient discontinued the study because of a new rash, 1 discontinued after developing peripheral edema, and 3 patients reported mild sedation.

Beyer et al³⁷ reported on a post hoc analysis of 94 older adults (mean age, 57.1 years; range, 50.1 to 74.8 years) with acute bipolar mania receiving olanzapine (n = 47), divalproex (n = 31), or placebo (n = 16) in a pooled olanzapine clinical trials database. Patients receiving olanzapine or divalproex had improvement in mania; those receiving placebo did not improve. Safety findings were comparable with reports in younger patients with mania.

Other clinical data. Adverse effects found in mixed-age samples using secondary analyses of clinical trials need to be interpreted with caution because these types of studies usually exclude individuals with significant medical comorbidity. Medical burden, cognitive impairment, or concomitant medications generally necessitate slower drug titration and lower total daily dosing. For example, a secondary analysis of the U.S. National Institute of Health-funded Systematic Treatment Enhancement Program for Bipolar Disorder study, which had broader inclusion criteria than most clinical trials, reported that, although recovery rates in older adults with bipolar disorder were fairly good (78.5%), lower doses of risperidone were used in older vs younger patients.³⁸

Clinical considerations

Interpretation of the relative risks of antipsychotics in older people must be tempered by the caveat that there is limited high-quality data (Table 4). Antipsychotics are the first-line therapy for older patients with schizophrenia, although their use is supported by a small number of prospective RCTs. SGAs are preferred because of their lower propensity to cause EPS and other motor adverse effects. Older persons with schizophrenia have an EPS threshold lower than younger patients and determining the lowest effective dosage may minimize EPS and cognitive adverse effects. As individuals with long-standing schizophrenia get older, their antipsychotic dosages may need to be reduced, and clinicians need to monitor for adverse effects that are more common among older people, such as tardive dyskinesia and metabolic abnormalities. In healthy, “younger” geriatric patients, monitoring for adverse effects may be similar to monitoring of younger patients. Patients who are older or frail may need more frequent assessment.

Like older adults with schizophrenia, geriatric patients with bipolar disorder have reduced drug tolerability and experience more adverse effects than younger patients. There are no prospective controlled studies that evaluated using antipsychotics in older patients with bipolar disorder. In older bipolar patients, the most problematic adverse effects of antipsychotics are akathisia, parkinsonism, other EPS, sedation and dizziness (which may increase fall risk), and GI discomfort. A key tolerability and safety consideration when treating older adults with bipolar disorder is the role of antipsychotics in relation to the use of lithium and mood stabilizers. Some studies have suggested that lithium has neuroprotective effects when used long-term; however, at least 1 report suggested that long-term antipsychotic treatment may be associated with neurodegeneration.³⁹

The literature does not provide strong evidence on the many clinical variations that we see in routine practice settings, such as combinations of drug treatments or drugs prescribed to patients with specific comorbid conditions. There is a need for large cohort studies that monitor treatment course, medical comorbidity, and prognosis. Additionally, well-designed clinical trials such as the DART-AD, which investigated longer-term trajectories of people with dementia taking antipsychotics, should serve as a model for the type of research that is needed to better understand outcome variability among older people with chronic psychotic or bipolar disorders.⁴⁰

Antipsychotics are FDA-approved as a primary treatment for schizophrenia and bipolar disorder and as adjunctive therapy for major depressive disorder. In the United States, approximately 26% of antipsychotic prescriptions written for these indications are for individuals age >65.¹ Additionally, antipsychotics are widely used to treat behavioral symptoms associated with dementia.¹ The rapid expansion of the use of second-generation antipsychotics (SGAs), in particular, has been driven in part by their lower risk for extrapyramidal symptoms (EPS) compared with first-generation antipsychotics (FGAs).¹ However, a growing body of data indicates that all antipsychotics have a range of adverse effects in older patients. This focus is critical in light of demographic trends—in the next 10 to 15 years, the population age >60 will grow 3.5 times more rapidly than the general population.²

In this context, psychiatrists need information on the relative risks of antipsychotics for older patients. This 3-part series summarizes findings and recommendations on safety and tolerability when prescribing antipsychotics in older individuals with chronic psychotic disorders, such as schizophrenia, bipolar disorder, depression, and dementia. This review aims to:

• briefly summarize the major studies and analyses relevant to older patients with these diagnoses
• provide a summative opinion on safety and tolerability issues in these older adults
• highlight the gaps in the evidence base and areas that need additional research.

Part 1 focuses on older adults with schizophrenia or bipolar disorder. Subsequent articles will focus on prescribing antipsychotics to older adults with depression and those with dementia.

Schizophrenia

Summary of benefits, place in treatment armamentarium. Individuals with schizophrenia have a shorter life expectancy than that of the general population mostly as a result of suicide and comorbid physical illnesses,³ but the number of patients with schizophrenia age >55 will double over the next 2 decades.⁴ With aging, both positive and negative symptoms may be a focus of treatment (Table 1).^5,6 Antipsychotics are a first-line treatment for older patients with schizophrenia with few medication alternatives.⁷ Safety risks associated with antipsychotics in older people span a broad spectrum (Table 2).⁸

A 6-week prospective RCT evaluated paliperidone extended-release vs placebo in 114 older adults (age ≥65 years; mean age, 70 years) with schizophrenia.¹⁴ There was an optional 24-week extension of open-label treatment with paliperidone. Mean daily dose of paliperidone was 8.4 mg. Efficacy measures did not show consistent statistically significant differences between treatment groups. Discontinuation rates were similar between paliperidone (7%) vs placebo (8%). Serious adverse events occurred in 3% of paliperidone-treated vs 8% of placebo-treated patients. Elevated prolactin levels occurred in one-half of paliperidone-treated patients. There were no prolactin or glucose treatment-related adverse events or significant mean changes in body weight for either paliperidone-treated or placebo-treated patients. Safety findings in the 24-week, open-label extension group were consistent with the RCT results.

Howanitz et al¹⁵ conducted a 12-week, prospective RCT that compared clozapine (mean dose, 300 mg/d) with chlorpromazine (mean dose, 600 mg/d) in 42 older adults (mean age, 67 years) with schizophrenia. Drop-out rate prior to 5 weeks was 19% and similar between groups. Common adverse effects included sialorrhea, hematologic abnormalities, sedation, tachycardia, EPS, and weight gain. Although both drugs were effective, more patients taking clozapine had tachycardia and weight gain, while more chlorpromazine patients reported sedation.

There have been other, less rigorous studies.^7,8 Most of these studies evaluated risperidone and olanzapine, and most were conducted in “younger” geriatric patients (age <75 years). Although patients who participate in clinical trials may be healthier than “typical” patients, adverse effects such as EPS, sedation, and weight gain were still relatively common in these studies.

Other clinical data. A major consideration in treating older adults with schizophrenia is balancing the need to administer an antipsychotic dose high enough to alleviate psychotic symptoms while minimizing dose-dependent adverse effects. There is a U-shaped relationship between age and vulnerability to antipsychotic adverse effects,^16,17 wherein adverse effects are highest at younger and older ages. Evidence supports using the lowest effective antipsychotic dose for geriatric patients with schizophrenia. Positive emission tomography (PET) studies suggest that older patients develop EPS with lower doses despite lower receptor occupancy.^17,18 A recent study of 35 older patients (mean age, 60.1 years) with schizophrenia obtained PET, clinical measures, and blood pharmacokinetic measures before and after reduction of risperidone or olanzapine doses.¹⁸ A ≥40% reduction in dose was associated with reduced adverse effects, particularly EPS and elevation of prolactin levels. Moreover, the therapeutic window of striatal D2/D3 receptor occupancy appeared to be 50% to 60% in these older patients, compared with 65% to 80% in younger patients.

Long-term risks of antipsychotic treatment across the lifespan are less clear, with evidence suggesting both lower and higher mortality risk.^19,20 It is difficult to fully disentangle the long-term risks of antipsychotics from the cumulative effects of lifestyle and comorbidity among individuals who have lived with schizophrenia for decades. Large naturalistic studies that include substantial numbers of older people with schizophrenia might be a way to elicit more information on long-term safety. The Schizophrenia Outpatient Health Outcome (SOHO) study was a large naturalistic trial that recruited >10,000 individuals with schizophrenia in 10 European countries.²¹ Although the SOHO study found differences between antipsychotics and adverse effects, such as EPS, weight gain, and sexual dysfunction, because the mean age of these patients was approximately 40 years and the follow-up period was only 3 years, it is difficult to draw conclusions that could be relevant to older individuals who have had schizophrenia for decades.

Bipolar Disorder

A secondary analysis of mixed-age, RCTs examined response in older adults (age ≥55 years) with bipolar I depression who received lurasidone as monotherapy or adjunctive therapy.³² In the monotherapy study, these patients were randomized to 6 weeks of lurasidone 20 to 60 mg/d, lurasidone 80 to 120 mg/d, or placebo. In the adjunctive therapy study, they were randomized to lurasidone 20 to 120 mg/d or placebo with either lithium or valproate. There were 83 older adults (17.1% of the sample) in the monotherapy study and 53 (15.6%) in the adjunctive therapy study. Mean improvement in depression was significantly higher for both doses of lurasidone monotherapy than placebo. Adjunctive lurasidone was not associated with statistically significant improvement vs placebo. The most frequent adverse events in older patients on lurasidone monotherapy 20 to 60 mg/d or 80 to 120 mg/d were nausea (18.5% and 9.7%, respectively) and somnolence (11.1% and 0%, respectively). Akathisia (9.7%) and insomnia (9.7%) were the most common adverse events in the group receiving 80 to 120 mg/d, with the rate of akathisia exhibiting a dose-related increase. Weight change with lurasidone was similar to placebo, and there were no clinically meaningful group changes in vital signs, electrocardiography, or laboratory parameters.

A small (N = 20) open study found improvement in older adults with bipolar depression with aripiprazole (mean dose, 10.3 mg/d).³³ Adverse effects included restlessness and weight gain (n = 3, 9% each), sedation (n = 2, 10%), and drooling and diarrhea/loose stools (n = 1, 5% each). In another small study (N = 15) using asenapine (mean dose, 11.2 mg/d) in mainly older bipolar patients with depression, the most common adverse effects were gastrointestinal (GI) discomfort (n = 5, 33%) and restlessness, tremors, cognitive difficulties, and sluggishness (n = 2, 13% each).³⁴

Clinical trials: Bipolar mania. Researchers conducted a pooled analysis of two 12-week randomized trials comparing quetiapine with placebo in a mixed-age sample with bipolar mania.³⁵ In a subgroup of 59 older patients (mean age, 62.9 years), manic symptoms improved significantly more with quetiapine (modal dose, 550 mg/d) than with placebo. Adverse effects reported by >10% of older patients were dry mouth, somnolence, postural hypotension, insomnia, weight gain, and dizziness. Insomnia was reported by >10% of patients receiving placebo.

In a case series of 11 elderly patients with mania receiving asenapine, Baruch et al³⁶ reported a 63% remission rate. One patient discontinued the study because of a new rash, 1 discontinued after developing peripheral edema, and 3 patients reported mild sedation.

Beyer et al³⁷ reported on a post hoc analysis of 94 older adults (mean age, 57.1 years; range, 50.1 to 74.8 years) with acute bipolar mania receiving olanzapine (n = 47), divalproex (n = 31), or placebo (n = 16) in a pooled olanzapine clinical trials database. Patients receiving olanzapine or divalproex had improvement in mania; those receiving placebo did not improve. Safety findings were comparable with reports in younger patients with mania.

Other clinical data. Adverse effects found in mixed-age samples using secondary analyses of clinical trials need to be interpreted with caution because these types of studies usually exclude individuals with significant medical comorbidity. Medical burden, cognitive impairment, or concomitant medications generally necessitate slower drug titration and lower total daily dosing. For example, a secondary analysis of the U.S. National Institute of Health-funded Systematic Treatment Enhancement Program for Bipolar Disorder study, which had broader inclusion criteria than most clinical trials, reported that, although recovery rates in older adults with bipolar disorder were fairly good (78.5%), lower doses of risperidone were used in older vs younger patients.³⁸

Clinical considerations

Interpretation of the relative risks of antipsychotics in older people must be tempered by the caveat that there is limited high-quality data (Table 4). Antipsychotics are the first-line therapy for older patients with schizophrenia, although their use is supported by a small number of prospective RCTs. SGAs are preferred because of their lower propensity to cause EPS and other motor adverse effects. Older persons with schizophrenia have an EPS threshold lower than younger patients and determining the lowest effective dosage may minimize EPS and cognitive adverse effects. As individuals with long-standing schizophrenia get older, their antipsychotic dosages may need to be reduced, and clinicians need to monitor for adverse effects that are more common among older people, such as tardive dyskinesia and metabolic abnormalities. In healthy, “younger” geriatric patients, monitoring for adverse effects may be similar to monitoring of younger patients. Patients who are older or frail may need more frequent assessment.

Like older adults with schizophrenia, geriatric patients with bipolar disorder have reduced drug tolerability and experience more adverse effects than younger patients. There are no prospective controlled studies that evaluated using antipsychotics in older patients with bipolar disorder. In older bipolar patients, the most problematic adverse effects of antipsychotics are akathisia, parkinsonism, other EPS, sedation and dizziness (which may increase fall risk), and GI discomfort. A key tolerability and safety consideration when treating older adults with bipolar disorder is the role of antipsychotics in relation to the use of lithium and mood stabilizers. Some studies have suggested that lithium has neuroprotective effects when used long-term; however, at least 1 report suggested that long-term antipsychotic treatment may be associated with neurodegeneration.³⁹

The literature does not provide strong evidence on the many clinical variations that we see in routine practice settings, such as combinations of drug treatments or drugs prescribed to patients with specific comorbid conditions. There is a need for large cohort studies that monitor treatment course, medical comorbidity, and prognosis. Additionally, well-designed clinical trials such as the DART-AD, which investigated longer-term trajectories of people with dementia taking antipsychotics, should serve as a model for the type of research that is needed to better understand outcome variability among older people with chronic psychotic or bipolar disorders.⁴⁰

Antipsychotics are FDA-approved as a primary treatment for schizophrenia and bipolar disorder and as adjunctive therapy for major depressive disorder. In the United States, approximately 26% of antipsychotic prescriptions written for these indications are for individuals age >65.¹ Additionally, antipsychotics are widely used to treat behavioral symptoms associated with dementia.¹ The rapid expansion of the use of second-generation antipsychotics (SGAs), in particular, has been driven in part by their lower risk for extrapyramidal symptoms (EPS) compared with first-generation antipsychotics (FGAs).¹ However, a growing body of data indicates that all antipsychotics have a range of adverse effects in older patients. This focus is critical in light of demographic trends—in the next 10 to 15 years, the population age >60 will grow 3.5 times more rapidly than the general population.²

In this context, psychiatrists need information on the relative risks of antipsychotics for older patients. This 3-part series summarizes findings and recommendations on safety and tolerability when prescribing antipsychotics in older individuals with chronic psychotic disorders, such as schizophrenia, bipolar disorder, depression, and dementia. This review aims to:

• briefly summarize the major studies and analyses relevant to older patients with these diagnoses
• provide a summative opinion on safety and tolerability issues in these older adults
• highlight the gaps in the evidence base and areas that need additional research.

Part 1 focuses on older adults with schizophrenia or bipolar disorder. Subsequent articles will focus on prescribing antipsychotics to older adults with depression and those with dementia.

Schizophrenia

Summary of benefits, place in treatment armamentarium. Individuals with schizophrenia have a shorter life expectancy than that of the general population mostly as a result of suicide and comorbid physical illnesses,³ but the number of patients with schizophrenia age >55 will double over the next 2 decades.⁴ With aging, both positive and negative symptoms may be a focus of treatment (Table 1).^5,6 Antipsychotics are a first-line treatment for older patients with schizophrenia with few medication alternatives.⁷ Safety risks associated with antipsychotics in older people span a broad spectrum (Table 2).⁸

A 6-week prospective RCT evaluated paliperidone extended-release vs placebo in 114 older adults (age ≥65 years; mean age, 70 years) with schizophrenia.¹⁴ There was an optional 24-week extension of open-label treatment with paliperidone. Mean daily dose of paliperidone was 8.4 mg. Efficacy measures did not show consistent statistically significant differences between treatment groups. Discontinuation rates were similar between paliperidone (7%) vs placebo (8%). Serious adverse events occurred in 3% of paliperidone-treated vs 8% of placebo-treated patients. Elevated prolactin levels occurred in one-half of paliperidone-treated patients. There were no prolactin or glucose treatment-related adverse events or significant mean changes in body weight for either paliperidone-treated or placebo-treated patients. Safety findings in the 24-week, open-label extension group were consistent with the RCT results.

Howanitz et al¹⁵ conducted a 12-week, prospective RCT that compared clozapine (mean dose, 300 mg/d) with chlorpromazine (mean dose, 600 mg/d) in 42 older adults (mean age, 67 years) with schizophrenia. Drop-out rate prior to 5 weeks was 19% and similar between groups. Common adverse effects included sialorrhea, hematologic abnormalities, sedation, tachycardia, EPS, and weight gain. Although both drugs were effective, more patients taking clozapine had tachycardia and weight gain, while more chlorpromazine patients reported sedation.

There have been other, less rigorous studies.^7,8 Most of these studies evaluated risperidone and olanzapine, and most were conducted in “younger” geriatric patients (age <75 years). Although patients who participate in clinical trials may be healthier than “typical” patients, adverse effects such as EPS, sedation, and weight gain were still relatively common in these studies.

Other clinical data. A major consideration in treating older adults with schizophrenia is balancing the need to administer an antipsychotic dose high enough to alleviate psychotic symptoms while minimizing dose-dependent adverse effects. There is a U-shaped relationship between age and vulnerability to antipsychotic adverse effects,^16,17 wherein adverse effects are highest at younger and older ages. Evidence supports using the lowest effective antipsychotic dose for geriatric patients with schizophrenia. Positive emission tomography (PET) studies suggest that older patients develop EPS with lower doses despite lower receptor occupancy.^17,18 A recent study of 35 older patients (mean age, 60.1 years) with schizophrenia obtained PET, clinical measures, and blood pharmacokinetic measures before and after reduction of risperidone or olanzapine doses.¹⁸ A ≥40% reduction in dose was associated with reduced adverse effects, particularly EPS and elevation of prolactin levels. Moreover, the therapeutic window of striatal D2/D3 receptor occupancy appeared to be 50% to 60% in these older patients, compared with 65% to 80% in younger patients.

Long-term risks of antipsychotic treatment across the lifespan are less clear, with evidence suggesting both lower and higher mortality risk.^19,20 It is difficult to fully disentangle the long-term risks of antipsychotics from the cumulative effects of lifestyle and comorbidity among individuals who have lived with schizophrenia for decades. Large naturalistic studies that include substantial numbers of older people with schizophrenia might be a way to elicit more information on long-term safety. The Schizophrenia Outpatient Health Outcome (SOHO) study was a large naturalistic trial that recruited >10,000 individuals with schizophrenia in 10 European countries.²¹ Although the SOHO study found differences between antipsychotics and adverse effects, such as EPS, weight gain, and sexual dysfunction, because the mean age of these patients was approximately 40 years and the follow-up period was only 3 years, it is difficult to draw conclusions that could be relevant to older individuals who have had schizophrenia for decades.

Bipolar Disorder

A secondary analysis of mixed-age, RCTs examined response in older adults (age ≥55 years) with bipolar I depression who received lurasidone as monotherapy or adjunctive therapy.³² In the monotherapy study, these patients were randomized to 6 weeks of lurasidone 20 to 60 mg/d, lurasidone 80 to 120 mg/d, or placebo. In the adjunctive therapy study, they were randomized to lurasidone 20 to 120 mg/d or placebo with either lithium or valproate. There were 83 older adults (17.1% of the sample) in the monotherapy study and 53 (15.6%) in the adjunctive therapy study. Mean improvement in depression was significantly higher for both doses of lurasidone monotherapy than placebo. Adjunctive lurasidone was not associated with statistically significant improvement vs placebo. The most frequent adverse events in older patients on lurasidone monotherapy 20 to 60 mg/d or 80 to 120 mg/d were nausea (18.5% and 9.7%, respectively) and somnolence (11.1% and 0%, respectively). Akathisia (9.7%) and insomnia (9.7%) were the most common adverse events in the group receiving 80 to 120 mg/d, with the rate of akathisia exhibiting a dose-related increase. Weight change with lurasidone was similar to placebo, and there were no clinically meaningful group changes in vital signs, electrocardiography, or laboratory parameters.

A small (N = 20) open study found improvement in older adults with bipolar depression with aripiprazole (mean dose, 10.3 mg/d).³³ Adverse effects included restlessness and weight gain (n = 3, 9% each), sedation (n = 2, 10%), and drooling and diarrhea/loose stools (n = 1, 5% each). In another small study (N = 15) using asenapine (mean dose, 11.2 mg/d) in mainly older bipolar patients with depression, the most common adverse effects were gastrointestinal (GI) discomfort (n = 5, 33%) and restlessness, tremors, cognitive difficulties, and sluggishness (n = 2, 13% each).³⁴

Clinical trials: Bipolar mania. Researchers conducted a pooled analysis of two 12-week randomized trials comparing quetiapine with placebo in a mixed-age sample with bipolar mania.³⁵ In a subgroup of 59 older patients (mean age, 62.9 years), manic symptoms improved significantly more with quetiapine (modal dose, 550 mg/d) than with placebo. Adverse effects reported by >10% of older patients were dry mouth, somnolence, postural hypotension, insomnia, weight gain, and dizziness. Insomnia was reported by >10% of patients receiving placebo.

In a case series of 11 elderly patients with mania receiving asenapine, Baruch et al³⁶ reported a 63% remission rate. One patient discontinued the study because of a new rash, 1 discontinued after developing peripheral edema, and 3 patients reported mild sedation.

Beyer et al³⁷ reported on a post hoc analysis of 94 older adults (mean age, 57.1 years; range, 50.1 to 74.8 years) with acute bipolar mania receiving olanzapine (n = 47), divalproex (n = 31), or placebo (n = 16) in a pooled olanzapine clinical trials database. Patients receiving olanzapine or divalproex had improvement in mania; those receiving placebo did not improve. Safety findings were comparable with reports in younger patients with mania.

Other clinical data. Adverse effects found in mixed-age samples using secondary analyses of clinical trials need to be interpreted with caution because these types of studies usually exclude individuals with significant medical comorbidity. Medical burden, cognitive impairment, or concomitant medications generally necessitate slower drug titration and lower total daily dosing. For example, a secondary analysis of the U.S. National Institute of Health-funded Systematic Treatment Enhancement Program for Bipolar Disorder study, which had broader inclusion criteria than most clinical trials, reported that, although recovery rates in older adults with bipolar disorder were fairly good (78.5%), lower doses of risperidone were used in older vs younger patients.³⁸

Clinical considerations

Interpretation of the relative risks of antipsychotics in older people must be tempered by the caveat that there is limited high-quality data (Table 4). Antipsychotics are the first-line therapy for older patients with schizophrenia, although their use is supported by a small number of prospective RCTs. SGAs are preferred because of their lower propensity to cause EPS and other motor adverse effects. Older persons with schizophrenia have an EPS threshold lower than younger patients and determining the lowest effective dosage may minimize EPS and cognitive adverse effects. As individuals with long-standing schizophrenia get older, their antipsychotic dosages may need to be reduced, and clinicians need to monitor for adverse effects that are more common among older people, such as tardive dyskinesia and metabolic abnormalities. In healthy, “younger” geriatric patients, monitoring for adverse effects may be similar to monitoring of younger patients. Patients who are older or frail may need more frequent assessment.

Like older adults with schizophrenia, geriatric patients with bipolar disorder have reduced drug tolerability and experience more adverse effects than younger patients. There are no prospective controlled studies that evaluated using antipsychotics in older patients with bipolar disorder. In older bipolar patients, the most problematic adverse effects of antipsychotics are akathisia, parkinsonism, other EPS, sedation and dizziness (which may increase fall risk), and GI discomfort. A key tolerability and safety consideration when treating older adults with bipolar disorder is the role of antipsychotics in relation to the use of lithium and mood stabilizers. Some studies have suggested that lithium has neuroprotective effects when used long-term; however, at least 1 report suggested that long-term antipsychotic treatment may be associated with neurodegeneration.³⁹

The literature does not provide strong evidence on the many clinical variations that we see in routine practice settings, such as combinations of drug treatments or drugs prescribed to patients with specific comorbid conditions. There is a need for large cohort studies that monitor treatment course, medical comorbidity, and prognosis. Additionally, well-designed clinical trials such as the DART-AD, which investigated longer-term trajectories of people with dementia taking antipsychotics, should serve as a model for the type of research that is needed to better understand outcome variability among older people with chronic psychotic or bipolar disorders.⁴⁰

Borderline personality disorder (BPD) is associated with impaired psychosocial functioning,^1-4 reduced health-related quality of life,⁵ high utilization of services,^6,7 and excess mortality.^8-10 Although BPD occurs in up to 40% of psychiatric inpatients¹¹ and 10% of outpatients,¹² it is underrecognized.^13-15 Often, patients with BPD do not receive an accurate diagnosis until ≥10 years after initially seeking treatment.¹⁶ The treatment and clinical implications of failing to recognize BPD include overprescribing medication and underutilizing empirically effective psychotherapies.¹⁴

This review summarizes studies of the underdiagnosis of BPD in routine clinical practice, describes which patients should be screened, and reviews alternative approaches to screening.

Underrecognition of BPD

The Rhode Island Methods to Improve Diagnostic Assessment and Services (MIDAS) project is an ongoing clinical research study involving the integration of research assessment methods into routine clinical practice.¹⁷ In an early report from the MIDAS project, BPD diagnoses derived from structured and unstructured clinical interviews were compared between 2 groups of psychiatric outpatients in the same practice.¹⁵ Individuals in the structured interview cohort were 35 times more often diagnosed with BPD than individuals evaluated with an unstructured clinical interview. Importantly, when the information from the structured interview was presented to the clinicians, BPD was more likely to be diagnosed clinically.

Other studies^13,16 also found that the rate of diagnosing BPD was higher when the diagnosis was based on a semi-structured diagnostic interview compared with an unstructured clinical interview, and that clinicians were reluctant to diagnose BPD during their routine intake diagnostic evaluation.

Clinicians, however, do not use semi-structured interviews in their practice, and they also do not tend to diagnose personality disorders (PDs) based on direct questioning, as they typically would when assessing a symptom-based disorder such as depression or anxiety. Rather, clinicians report that they rely on longitudinal observations to diagnose PDs.¹⁸ However, the results from the MIDAS project were inconsistent with clinicians’ reports. When clinicians were presented with the results of the semi-structured interview, they usually would diagnose BPD, even though it was the initial evaluation. If clinicians actually relied on longitudinal observations and considered information based on the direct question approach of research interviews to be irrelevant or invalid, then the results from the semi-structured interview should not have influenced the rate at which they diagnosed BPD. This suggests that the primary issue in diagnosing PDs is not the need for longi­tudinal observation but rather the need for more information, and that there is a role for screening questionnaires.

One potential criticism of studies demonstrating underrecognition of BPD in clinical practice is that patients typically were interviewed when they presented for treatment, when most were depressed or anxious. The possible pathologizing effects of psychiatric state on personality have been known for years.¹⁹ However, a large body of literature examining the treatment, prognostic, familial, and biological correlates of PDs supports the validity of diagnosing PDs in this manner. Moreover, from a clinical perspective, the sooner a clinician is aware of the presence of BPD, the more likely this information can be used for treatment planning.

Who should be screened for BPD?

BPD is underrecognized and underdiagnosed because patients with BPD often also have comorbid mood, anxiety, or substance use disorders.^20,21 The symptoms associated with these disorders are typically the chief concern of patients with undiagnosed BPD who present for treatment. Patients with BPD rarely present for an intake evaluation and state that they are struggling with abandonment fears, chronic feelings of emptiness, or an identity disturbance. If patients identified these problems as their chief concerns, BPD would be easier to recognize.

Although several studies have documented the frequency of BPD in patients with a specific psychiatric diagnosis such as major depressive disorder (MDD) or attention-deficit/hyperactivity disorder,^22-26 the MIDAS project examined the frequency of BPD in patients with various diagnoses and evaluated which disorders were associated with a significantly increased rate of BPD.²⁷ The highest rate of BPD was found in patients with bipolar disorder. Approximately 25% of patients with bipolar II disorder and one-third of those with bipolar I disorder were diagnosed with BPD; these rates were significantly higher than the rate of BPD in patients without these disorders (Table 1²⁷). The rate of BPD was second highest in patients with a principal diagnosis of posttraumatic stress disorder (PTSD) and MDD; however, the rate of BPD in these patients was not significantly elevated compared with patients who did not have these principal diagnoses. Three disorders were associated with a significantly lower rate of BPD: adjustment disorder, dysthymic disorder, and generalized anxiety disorder.

A brief review of screening statistics

Screening tests for most psychiatric disorders are based on multi-item scales in which a total score is computed from a sum of item scores, and a cutoff point is established to determine who does and does not screen positive on the test. However, sensitivity, specificity, and positive and negative predictive values are not invariant properties of a screening test with a continuous score distribution. Rather, the performance statistics of a scale can be altered by changing the threshold score to distinguish cases from non-cases. When the screening threshold is lowered, sensitivity increases and specificity decreases.

For screening, a broad net needs to be cast so that all (or almost all) cases are included. Therefore, the cutoff score should be set low to prioritize the sensitivity of the instrument. A screening scale also should have high negative predictive value so that the clinician can be confident that patients who screen negative on the test do not have the disorder.

Screening questionnaires for BPD

Several questionnaires have been developed to screen for PDs (Table 2^28-35). Some screen for each of the DSM PDs,^28,36-42 and some screen more broadly for the presence or absence of any PD.^29,43,44 The most commonly studied self-report scale for BPD is the McLean Screening Instrument for Borderline Personality Disorder (MSI-BPD),³⁰ a 10-item self-report scale derived from a subset of questions from the BPD module of a semi-structured diagnostic interview.

The initial validation study³⁰ found that the optimal cutoff score was 7, which resulted in a sensitivity of 81% and specificity of 89%. Three studies have evaluated the scale in adolescents and young adults,^45-47 and 3 studies examined the scale in adult outpatients.^48-50 Across all 6 studies, at the optimal cutoff scores determined in each study, the sensitivity of the MSI-BPD ranged from 68% to 94% (mean, 80%) and the specificity ranged from 66% to 80% (mean, 72%).

Problems with screening questionnaires. Although screening scales have been developed for many psychiatric disorders, they have not been widely used in mental health settings. In a previous commentary, I argued that the conceptual justification for using self-report screening scales for single disorders in psychiatric settings is questionable.⁵¹ Another problem with screening scales is their potential misuse as case-finding instruments. In the literature on bipolar disorder screening, several researchers misconstrued a positive screen to indicate caseness.⁵¹ Although this is not a problem with the screening measures or the selection of a cutoff score, caution must be taken to not confuse screening with diagnosis.⁵²

Screening for BPD as part of your diagnostic interview

An alternative approach to using self-administered questionnaires for screening is for clinicians to include questions in their evaluation as part of a psychiatric review of systems. When conducting a diagnostic interview, clinicians typically screen for disorders that are comorbid with the principal diagnosis by asking about the comorbid disorders’ necessary features or “gate criteria.” For example, in a patient with a principal diagnosis of MDD, the clinician would inquire about the presence of panic attacks, excessive worry, or substance use to screen for the presence of panic disorder, generalized anxiety disorder, or a substance use disorder. In contrast, for polythetically defined disorders such as BPD, there is no single gate criterion, because the disorder is diagnosed based on the presence of at least 5 of 9 criteria and no single one of these criteria is required to be present to establish the diagnosis.

As part of the MIDAS project, the psycho­metric properties of the BPD criteria were examined to determine if it was possible to identify 1 or 2 criteria that could serve as gate criteria to screen for the disorder. If the sensitivity of 1 criterion or a combination of 2 BPD criteria was sufficiently high (ie, >90%), then the assessment of this criterion (or these criteria) could be included in a psychiatric review of systems, thus potentially improving the detection of BPD. Researchers hypothesized that affective instability, considered first by Linehan⁵³ and later by other theorists⁵⁴ to be of central importance to the clinical manifestations of BPD, could function as a gate criterion. In the sample of 3,674 psychiatric outpatients who were evaluated with a semi-structured interview, the sensitivity of the affective instability criterion was 92.8%, and the negative predictive value of the criterion was 99%.

Borderline personality disorder (BPD) is associated with impaired psychosocial functioning,^1-4 reduced health-related quality of life,⁵ high utilization of services,^6,7 and excess mortality.^8-10 Although BPD occurs in up to 40% of psychiatric inpatients¹¹ and 10% of outpatients,¹² it is underrecognized.^13-15 Often, patients with BPD do not receive an accurate diagnosis until ≥10 years after initially seeking treatment.¹⁶ The treatment and clinical implications of failing to recognize BPD include overprescribing medication and underutilizing empirically effective psychotherapies.¹⁴

This review summarizes studies of the underdiagnosis of BPD in routine clinical practice, describes which patients should be screened, and reviews alternative approaches to screening.

Underrecognition of BPD

The Rhode Island Methods to Improve Diagnostic Assessment and Services (MIDAS) project is an ongoing clinical research study involving the integration of research assessment methods into routine clinical practice.¹⁷ In an early report from the MIDAS project, BPD diagnoses derived from structured and unstructured clinical interviews were compared between 2 groups of psychiatric outpatients in the same practice.¹⁵ Individuals in the structured interview cohort were 35 times more often diagnosed with BPD than individuals evaluated with an unstructured clinical interview. Importantly, when the information from the structured interview was presented to the clinicians, BPD was more likely to be diagnosed clinically.

Other studies^13,16 also found that the rate of diagnosing BPD was higher when the diagnosis was based on a semi-structured diagnostic interview compared with an unstructured clinical interview, and that clinicians were reluctant to diagnose BPD during their routine intake diagnostic evaluation.

Clinicians, however, do not use semi-structured interviews in their practice, and they also do not tend to diagnose personality disorders (PDs) based on direct questioning, as they typically would when assessing a symptom-based disorder such as depression or anxiety. Rather, clinicians report that they rely on longitudinal observations to diagnose PDs.¹⁸ However, the results from the MIDAS project were inconsistent with clinicians’ reports. When clinicians were presented with the results of the semi-structured interview, they usually would diagnose BPD, even though it was the initial evaluation. If clinicians actually relied on longitudinal observations and considered information based on the direct question approach of research interviews to be irrelevant or invalid, then the results from the semi-structured interview should not have influenced the rate at which they diagnosed BPD. This suggests that the primary issue in diagnosing PDs is not the need for longi­tudinal observation but rather the need for more information, and that there is a role for screening questionnaires.

One potential criticism of studies demonstrating underrecognition of BPD in clinical practice is that patients typically were interviewed when they presented for treatment, when most were depressed or anxious. The possible pathologizing effects of psychiatric state on personality have been known for years.¹⁹ However, a large body of literature examining the treatment, prognostic, familial, and biological correlates of PDs supports the validity of diagnosing PDs in this manner. Moreover, from a clinical perspective, the sooner a clinician is aware of the presence of BPD, the more likely this information can be used for treatment planning.

Who should be screened for BPD?

BPD is underrecognized and underdiagnosed because patients with BPD often also have comorbid mood, anxiety, or substance use disorders.^20,21 The symptoms associated with these disorders are typically the chief concern of patients with undiagnosed BPD who present for treatment. Patients with BPD rarely present for an intake evaluation and state that they are struggling with abandonment fears, chronic feelings of emptiness, or an identity disturbance. If patients identified these problems as their chief concerns, BPD would be easier to recognize.

Although several studies have documented the frequency of BPD in patients with a specific psychiatric diagnosis such as major depressive disorder (MDD) or attention-deficit/hyperactivity disorder,^22-26 the MIDAS project examined the frequency of BPD in patients with various diagnoses and evaluated which disorders were associated with a significantly increased rate of BPD.²⁷ The highest rate of BPD was found in patients with bipolar disorder. Approximately 25% of patients with bipolar II disorder and one-third of those with bipolar I disorder were diagnosed with BPD; these rates were significantly higher than the rate of BPD in patients without these disorders (Table 1²⁷). The rate of BPD was second highest in patients with a principal diagnosis of posttraumatic stress disorder (PTSD) and MDD; however, the rate of BPD in these patients was not significantly elevated compared with patients who did not have these principal diagnoses. Three disorders were associated with a significantly lower rate of BPD: adjustment disorder, dysthymic disorder, and generalized anxiety disorder.

A brief review of screening statistics

Screening tests for most psychiatric disorders are based on multi-item scales in which a total score is computed from a sum of item scores, and a cutoff point is established to determine who does and does not screen positive on the test. However, sensitivity, specificity, and positive and negative predictive values are not invariant properties of a screening test with a continuous score distribution. Rather, the performance statistics of a scale can be altered by changing the threshold score to distinguish cases from non-cases. When the screening threshold is lowered, sensitivity increases and specificity decreases.

For screening, a broad net needs to be cast so that all (or almost all) cases are included. Therefore, the cutoff score should be set low to prioritize the sensitivity of the instrument. A screening scale also should have high negative predictive value so that the clinician can be confident that patients who screen negative on the test do not have the disorder.

Screening questionnaires for BPD

Several questionnaires have been developed to screen for PDs (Table 2^28-35). Some screen for each of the DSM PDs,^28,36-42 and some screen more broadly for the presence or absence of any PD.^29,43,44 The most commonly studied self-report scale for BPD is the McLean Screening Instrument for Borderline Personality Disorder (MSI-BPD),³⁰ a 10-item self-report scale derived from a subset of questions from the BPD module of a semi-structured diagnostic interview.

The initial validation study³⁰ found that the optimal cutoff score was 7, which resulted in a sensitivity of 81% and specificity of 89%. Three studies have evaluated the scale in adolescents and young adults,^45-47 and 3 studies examined the scale in adult outpatients.^48-50 Across all 6 studies, at the optimal cutoff scores determined in each study, the sensitivity of the MSI-BPD ranged from 68% to 94% (mean, 80%) and the specificity ranged from 66% to 80% (mean, 72%).

Problems with screening questionnaires. Although screening scales have been developed for many psychiatric disorders, they have not been widely used in mental health settings. In a previous commentary, I argued that the conceptual justification for using self-report screening scales for single disorders in psychiatric settings is questionable.⁵¹ Another problem with screening scales is their potential misuse as case-finding instruments. In the literature on bipolar disorder screening, several researchers misconstrued a positive screen to indicate caseness.⁵¹ Although this is not a problem with the screening measures or the selection of a cutoff score, caution must be taken to not confuse screening with diagnosis.⁵²

Screening for BPD as part of your diagnostic interview

An alternative approach to using self-administered questionnaires for screening is for clinicians to include questions in their evaluation as part of a psychiatric review of systems. When conducting a diagnostic interview, clinicians typically screen for disorders that are comorbid with the principal diagnosis by asking about the comorbid disorders’ necessary features or “gate criteria.” For example, in a patient with a principal diagnosis of MDD, the clinician would inquire about the presence of panic attacks, excessive worry, or substance use to screen for the presence of panic disorder, generalized anxiety disorder, or a substance use disorder. In contrast, for polythetically defined disorders such as BPD, there is no single gate criterion, because the disorder is diagnosed based on the presence of at least 5 of 9 criteria and no single one of these criteria is required to be present to establish the diagnosis.

As part of the MIDAS project, the psycho­metric properties of the BPD criteria were examined to determine if it was possible to identify 1 or 2 criteria that could serve as gate criteria to screen for the disorder. If the sensitivity of 1 criterion or a combination of 2 BPD criteria was sufficiently high (ie, >90%), then the assessment of this criterion (or these criteria) could be included in a psychiatric review of systems, thus potentially improving the detection of BPD. Researchers hypothesized that affective instability, considered first by Linehan⁵³ and later by other theorists⁵⁴ to be of central importance to the clinical manifestations of BPD, could function as a gate criterion. In the sample of 3,674 psychiatric outpatients who were evaluated with a semi-structured interview, the sensitivity of the affective instability criterion was 92.8%, and the negative predictive value of the criterion was 99%.

Borderline personality disorder (BPD) is associated with impaired psychosocial functioning,^1-4 reduced health-related quality of life,⁵ high utilization of services,^6,7 and excess mortality.^8-10 Although BPD occurs in up to 40% of psychiatric inpatients¹¹ and 10% of outpatients,¹² it is underrecognized.^13-15 Often, patients with BPD do not receive an accurate diagnosis until ≥10 years after initially seeking treatment.¹⁶ The treatment and clinical implications of failing to recognize BPD include overprescribing medication and underutilizing empirically effective psychotherapies.¹⁴

This review summarizes studies of the underdiagnosis of BPD in routine clinical practice, describes which patients should be screened, and reviews alternative approaches to screening.

Underrecognition of BPD

The Rhode Island Methods to Improve Diagnostic Assessment and Services (MIDAS) project is an ongoing clinical research study involving the integration of research assessment methods into routine clinical practice.¹⁷ In an early report from the MIDAS project, BPD diagnoses derived from structured and unstructured clinical interviews were compared between 2 groups of psychiatric outpatients in the same practice.¹⁵ Individuals in the structured interview cohort were 35 times more often diagnosed with BPD than individuals evaluated with an unstructured clinical interview. Importantly, when the information from the structured interview was presented to the clinicians, BPD was more likely to be diagnosed clinically.

Other studies^13,16 also found that the rate of diagnosing BPD was higher when the diagnosis was based on a semi-structured diagnostic interview compared with an unstructured clinical interview, and that clinicians were reluctant to diagnose BPD during their routine intake diagnostic evaluation.

Clinicians, however, do not use semi-structured interviews in their practice, and they also do not tend to diagnose personality disorders (PDs) based on direct questioning, as they typically would when assessing a symptom-based disorder such as depression or anxiety. Rather, clinicians report that they rely on longitudinal observations to diagnose PDs.¹⁸ However, the results from the MIDAS project were inconsistent with clinicians’ reports. When clinicians were presented with the results of the semi-structured interview, they usually would diagnose BPD, even though it was the initial evaluation. If clinicians actually relied on longitudinal observations and considered information based on the direct question approach of research interviews to be irrelevant or invalid, then the results from the semi-structured interview should not have influenced the rate at which they diagnosed BPD. This suggests that the primary issue in diagnosing PDs is not the need for longi­tudinal observation but rather the need for more information, and that there is a role for screening questionnaires.

One potential criticism of studies demonstrating underrecognition of BPD in clinical practice is that patients typically were interviewed when they presented for treatment, when most were depressed or anxious. The possible pathologizing effects of psychiatric state on personality have been known for years.¹⁹ However, a large body of literature examining the treatment, prognostic, familial, and biological correlates of PDs supports the validity of diagnosing PDs in this manner. Moreover, from a clinical perspective, the sooner a clinician is aware of the presence of BPD, the more likely this information can be used for treatment planning.

Who should be screened for BPD?

BPD is underrecognized and underdiagnosed because patients with BPD often also have comorbid mood, anxiety, or substance use disorders.^20,21 The symptoms associated with these disorders are typically the chief concern of patients with undiagnosed BPD who present for treatment. Patients with BPD rarely present for an intake evaluation and state that they are struggling with abandonment fears, chronic feelings of emptiness, or an identity disturbance. If patients identified these problems as their chief concerns, BPD would be easier to recognize.

Although several studies have documented the frequency of BPD in patients with a specific psychiatric diagnosis such as major depressive disorder (MDD) or attention-deficit/hyperactivity disorder,^22-26 the MIDAS project examined the frequency of BPD in patients with various diagnoses and evaluated which disorders were associated with a significantly increased rate of BPD.²⁷ The highest rate of BPD was found in patients with bipolar disorder. Approximately 25% of patients with bipolar II disorder and one-third of those with bipolar I disorder were diagnosed with BPD; these rates were significantly higher than the rate of BPD in patients without these disorders (Table 1²⁷). The rate of BPD was second highest in patients with a principal diagnosis of posttraumatic stress disorder (PTSD) and MDD; however, the rate of BPD in these patients was not significantly elevated compared with patients who did not have these principal diagnoses. Three disorders were associated with a significantly lower rate of BPD: adjustment disorder, dysthymic disorder, and generalized anxiety disorder.

A brief review of screening statistics

Screening tests for most psychiatric disorders are based on multi-item scales in which a total score is computed from a sum of item scores, and a cutoff point is established to determine who does and does not screen positive on the test. However, sensitivity, specificity, and positive and negative predictive values are not invariant properties of a screening test with a continuous score distribution. Rather, the performance statistics of a scale can be altered by changing the threshold score to distinguish cases from non-cases. When the screening threshold is lowered, sensitivity increases and specificity decreases.

For screening, a broad net needs to be cast so that all (or almost all) cases are included. Therefore, the cutoff score should be set low to prioritize the sensitivity of the instrument. A screening scale also should have high negative predictive value so that the clinician can be confident that patients who screen negative on the test do not have the disorder.

Screening questionnaires for BPD

Several questionnaires have been developed to screen for PDs (Table 2^28-35). Some screen for each of the DSM PDs,^28,36-42 and some screen more broadly for the presence or absence of any PD.^29,43,44 The most commonly studied self-report scale for BPD is the McLean Screening Instrument for Borderline Personality Disorder (MSI-BPD),³⁰ a 10-item self-report scale derived from a subset of questions from the BPD module of a semi-structured diagnostic interview.

The initial validation study³⁰ found that the optimal cutoff score was 7, which resulted in a sensitivity of 81% and specificity of 89%. Three studies have evaluated the scale in adolescents and young adults,^45-47 and 3 studies examined the scale in adult outpatients.^48-50 Across all 6 studies, at the optimal cutoff scores determined in each study, the sensitivity of the MSI-BPD ranged from 68% to 94% (mean, 80%) and the specificity ranged from 66% to 80% (mean, 72%).

Problems with screening questionnaires. Although screening scales have been developed for many psychiatric disorders, they have not been widely used in mental health settings. In a previous commentary, I argued that the conceptual justification for using self-report screening scales for single disorders in psychiatric settings is questionable.⁵¹ Another problem with screening scales is their potential misuse as case-finding instruments. In the literature on bipolar disorder screening, several researchers misconstrued a positive screen to indicate caseness.⁵¹ Although this is not a problem with the screening measures or the selection of a cutoff score, caution must be taken to not confuse screening with diagnosis.⁵²

Screening for BPD as part of your diagnostic interview

An alternative approach to using self-administered questionnaires for screening is for clinicians to include questions in their evaluation as part of a psychiatric review of systems. When conducting a diagnostic interview, clinicians typically screen for disorders that are comorbid with the principal diagnosis by asking about the comorbid disorders’ necessary features or “gate criteria.” For example, in a patient with a principal diagnosis of MDD, the clinician would inquire about the presence of panic attacks, excessive worry, or substance use to screen for the presence of panic disorder, generalized anxiety disorder, or a substance use disorder. In contrast, for polythetically defined disorders such as BPD, there is no single gate criterion, because the disorder is diagnosed based on the presence of at least 5 of 9 criteria and no single one of these criteria is required to be present to establish the diagnosis.

As part of the MIDAS project, the psycho­metric properties of the BPD criteria were examined to determine if it was possible to identify 1 or 2 criteria that could serve as gate criteria to screen for the disorder. If the sensitivity of 1 criterion or a combination of 2 BPD criteria was sufficiently high (ie, >90%), then the assessment of this criterion (or these criteria) could be included in a psychiatric review of systems, thus potentially improving the detection of BPD. Researchers hypothesized that affective instability, considered first by Linehan⁵³ and later by other theorists⁵⁴ to be of central importance to the clinical manifestations of BPD, could function as a gate criterion. In the sample of 3,674 psychiatric outpatients who were evaluated with a semi-structured interview, the sensitivity of the affective instability criterion was 92.8%, and the negative predictive value of the criterion was 99%.