Articles

Evidence summary

An RCT found that MVM supplementation for one academic year didn’t improve academic achievement more than placebo in 640 children, 8 to 12 years of age, from low-income urban families.¹ Scores on the Terra Nova academic achievement test of reading, math, language, science, and social sciences didn’t differ between students taking MVM supplements or placebo.

Another RCT that compared MVM supplementation with placebo among 245 children between 6 and 12 years of age found no clinically significant improvements in IQ scores overall. However, within a small subset, more children who took MVM showed a clinical increase in IQ than children who took placebo.²

Investigators randomized children to daily MVM supplementation (50% of the US recommended daily allowance) or placebo for 3 months, then measured their Wechsler IQ scores. Overall, the MVM group scored 2.5 points higher (95% confidence interval [CI], 1.85-3.15) than the placebo group (a 15-point change is clinically significant).

More children taking MVM supplements (44) than placebo (25) showed increases in nonverbal IQ scores of 15 or more points (35% compared with 21%; P<.01). The authors speculate that this result may be attributable to the fact that one in 7 schoolchildren was undernourished. A major weakness of the study was its 16% attrition rate.

Fortified milk reduced disease in young children in India

A community-based, double-blind RCT found that milk fortified with vitamins A, C, and E plus minerals reduced common illnesses over the course of a year more than unfortified milk among 633 children 1 to 3 years of age living in a peri-urban area of India.³

Children who drank fortified milk had fewer days of fever (9.1 compared with 9.7 days for placebo; P=.005), a lower incidence of diarrhea (odds ratio [OR]=0.82; 95% CI, 0.73-0.93), and a decreased rate of lower respiratory illness (OR=0.74; 95% CI, 0.57-0.97). Children 2 years and younger showed the greatest effect.

Asthma and food allergies: The data are mixed

An inception cohort study found an association between early MV use and a higher risk of asthma and food allergies.⁴ Investigators evaluated more than 8000 American women and their newborns over the first 3 years of life. The study population included more families with low socioeconomic status (50%), blacks (51%), and infants born before 37 weeks’ gestation (23%) than the general US population.

Exclusively formula-fed infants who took MV in the first 6 months were more likely to develop asthma (OR=1.27; 95% CI, 1.04-1.56) and food allergies (OR=1.6; 95% CI, 1.2-2.2) than formula-fed infants who didn’t take MV.

However, a birth cohort study of 2470 Swedish children that analyzed health data from parental questionnaires and compared serum immunoglobulin E (IgE) concentrations at 8 years of age found no association between MV use within the past 12 months and clinical allergic disease or specific IgE concentrations.⁵ Children who took MV at age 4 years or earlier had lower rates of IgE sensitization to food allergens at 8 years (OR=0.61; 95% CI, 0.39-0.97).

Recommendations

According to the American Academy of Pediatrics Committee on Nutrition,⁶ healthy children who are growing normally and consume a varied diet don’t need routine supplementation with vitamins and minerals. The Committee states that if parents wish to give their children supplements, a standard pediatric multivitamin generally poses no risk.

Evidence summary

An RCT found that MVM supplementation for one academic year didn’t improve academic achievement more than placebo in 640 children, 8 to 12 years of age, from low-income urban families.¹ Scores on the Terra Nova academic achievement test of reading, math, language, science, and social sciences didn’t differ between students taking MVM supplements or placebo.

Another RCT that compared MVM supplementation with placebo among 245 children between 6 and 12 years of age found no clinically significant improvements in IQ scores overall. However, within a small subset, more children who took MVM showed a clinical increase in IQ than children who took placebo.²

Investigators randomized children to daily MVM supplementation (50% of the US recommended daily allowance) or placebo for 3 months, then measured their Wechsler IQ scores. Overall, the MVM group scored 2.5 points higher (95% confidence interval [CI], 1.85-3.15) than the placebo group (a 15-point change is clinically significant).

More children taking MVM supplements (44) than placebo (25) showed increases in nonverbal IQ scores of 15 or more points (35% compared with 21%; P<.01). The authors speculate that this result may be attributable to the fact that one in 7 schoolchildren was undernourished. A major weakness of the study was its 16% attrition rate.

Fortified milk reduced disease in young children in India

A community-based, double-blind RCT found that milk fortified with vitamins A, C, and E plus minerals reduced common illnesses over the course of a year more than unfortified milk among 633 children 1 to 3 years of age living in a peri-urban area of India.³

Children who drank fortified milk had fewer days of fever (9.1 compared with 9.7 days for placebo; P=.005), a lower incidence of diarrhea (odds ratio [OR]=0.82; 95% CI, 0.73-0.93), and a decreased rate of lower respiratory illness (OR=0.74; 95% CI, 0.57-0.97). Children 2 years and younger showed the greatest effect.

Asthma and food allergies: The data are mixed

An inception cohort study found an association between early MV use and a higher risk of asthma and food allergies.⁴ Investigators evaluated more than 8000 American women and their newborns over the first 3 years of life. The study population included more families with low socioeconomic status (50%), blacks (51%), and infants born before 37 weeks’ gestation (23%) than the general US population.

Exclusively formula-fed infants who took MV in the first 6 months were more likely to develop asthma (OR=1.27; 95% CI, 1.04-1.56) and food allergies (OR=1.6; 95% CI, 1.2-2.2) than formula-fed infants who didn’t take MV.

However, a birth cohort study of 2470 Swedish children that analyzed health data from parental questionnaires and compared serum immunoglobulin E (IgE) concentrations at 8 years of age found no association between MV use within the past 12 months and clinical allergic disease or specific IgE concentrations.⁵ Children who took MV at age 4 years or earlier had lower rates of IgE sensitization to food allergens at 8 years (OR=0.61; 95% CI, 0.39-0.97).

Recommendations

According to the American Academy of Pediatrics Committee on Nutrition,⁶ healthy children who are growing normally and consume a varied diet don’t need routine supplementation with vitamins and minerals. The Committee states that if parents wish to give their children supplements, a standard pediatric multivitamin generally poses no risk.

Evidence summary

An RCT found that MVM supplementation for one academic year didn’t improve academic achievement more than placebo in 640 children, 8 to 12 years of age, from low-income urban families.¹ Scores on the Terra Nova academic achievement test of reading, math, language, science, and social sciences didn’t differ between students taking MVM supplements or placebo.

Another RCT that compared MVM supplementation with placebo among 245 children between 6 and 12 years of age found no clinically significant improvements in IQ scores overall. However, within a small subset, more children who took MVM showed a clinical increase in IQ than children who took placebo.²

Investigators randomized children to daily MVM supplementation (50% of the US recommended daily allowance) or placebo for 3 months, then measured their Wechsler IQ scores. Overall, the MVM group scored 2.5 points higher (95% confidence interval [CI], 1.85-3.15) than the placebo group (a 15-point change is clinically significant).

More children taking MVM supplements (44) than placebo (25) showed increases in nonverbal IQ scores of 15 or more points (35% compared with 21%; P<.01). The authors speculate that this result may be attributable to the fact that one in 7 schoolchildren was undernourished. A major weakness of the study was its 16% attrition rate.

Fortified milk reduced disease in young children in India

A community-based, double-blind RCT found that milk fortified with vitamins A, C, and E plus minerals reduced common illnesses over the course of a year more than unfortified milk among 633 children 1 to 3 years of age living in a peri-urban area of India.³

Children who drank fortified milk had fewer days of fever (9.1 compared with 9.7 days for placebo; P=.005), a lower incidence of diarrhea (odds ratio [OR]=0.82; 95% CI, 0.73-0.93), and a decreased rate of lower respiratory illness (OR=0.74; 95% CI, 0.57-0.97). Children 2 years and younger showed the greatest effect.

Asthma and food allergies: The data are mixed

An inception cohort study found an association between early MV use and a higher risk of asthma and food allergies.⁴ Investigators evaluated more than 8000 American women and their newborns over the first 3 years of life. The study population included more families with low socioeconomic status (50%), blacks (51%), and infants born before 37 weeks’ gestation (23%) than the general US population.

Exclusively formula-fed infants who took MV in the first 6 months were more likely to develop asthma (OR=1.27; 95% CI, 1.04-1.56) and food allergies (OR=1.6; 95% CI, 1.2-2.2) than formula-fed infants who didn’t take MV.

However, a birth cohort study of 2470 Swedish children that analyzed health data from parental questionnaires and compared serum immunoglobulin E (IgE) concentrations at 8 years of age found no association between MV use within the past 12 months and clinical allergic disease or specific IgE concentrations.⁵ Children who took MV at age 4 years or earlier had lower rates of IgE sensitization to food allergens at 8 years (OR=0.61; 95% CI, 0.39-0.97).

Recommendations

According to the American Academy of Pediatrics Committee on Nutrition,⁶ healthy children who are growing normally and consume a varied diet don’t need routine supplementation with vitamins and minerals. The Committee states that if parents wish to give their children supplements, a standard pediatric multivitamin generally poses no risk.

Evidence summary

A 2008 Cochrane review examined 7 RCTs comparing ICS with placebo in 1989 patients 30 to 52 years of age with mild asthma or COPD. The reviewers found no evidence of increased bone turnover, decreased bone mineral density, or increased vertebral fracture in the ICS group compared with the placebo group at 2 to 3 years’ follow-up (odds ratio [OR] for fracture=1.87; 95% confidence interval [CI], 0.5-7.0).

Steroid doses ranged from 200 to 4000 mcg beclomethasone equivalent ICS per day.¹ A 100-mcg beclomethasone equivalent ICS dose is 50 mcg fluticasone, 80 mcg budesonide, or 200 mcg triamcinolone.²

A 2008 meta-analysis of 11 RCTs that examined a number of adverse effects of ICS in adult patients with COPD found 3 studies (8131 patients) that reported no significant increase in fracture risk at 36 months in the ICS group compared with the placebo group (OR=1.09; 95% CI, 0.89-1.33). Steroid doses ranged from 1000 to 2000 mcg beclomethasone equivalent ICS per day.³

Some studies suggest an association between dose and risk
A 2008 meta-analysis that included patients with COPD or asthma, average age 43 to 81 years, showed no difference in fracture risk overall at 1 to 4 years’ follow-up (OR=1.02; 95% CI, 0.96-1.08). This analysis examined 4 RCTs, 6 case-control studies, and 3 cohort studies.⁴

A subgroup analysis of patients taking higher-dose ICS (>1500 mcg beclomethasone equivalent ICS per day) that pooled data from case-control and cohort studies suggested an increased risk of fracture (OR=1.30; 95% CI, 1.07-1.58).⁴

Investigators identified a possible dose-dependent relationship in another meta-analysis of 5 case-control studies (43,783 cases, 259,936 controls).⁵ The meta-analysis included 4 of the studies examined in the previously discussed meta-analysis.⁴

The investigators found a relative risk of 1.12 (95% CI, 1.0-1.26) for nonvertebral fracture for each 1000-mcg increase in beclomethasone equivalent ICS dose per day.⁵ Longer follow-up time wasn’t associated with greater fracture risk.

But the relationship isn’t clear
Although some non-RCT studies discussed here show that higher doses of steroids may lead to increased fracture risk, the strength of this association isn’t clear. The authors of the Cochrane review and the meta-analyses point out that a significant number of confounding factors can put asthma and COPD patients at increased risk for fracture. They include age, smoking status, inactivity, and severity of underlying lung disease. The fact that different authors controlled differently for these factors introduced heterogeneity into the meta-analyses described here.^1,3-5

Recommendations

Guidelines for the Diagnosis and Management of Asthma from the National Heart, Lung, and Blood Institute state that “most benefit is achieved with relatively low doses of ICS, whereas the risk of adverse effects increases with dose. … ICS use may be associated with a dose-dependent reduction in bone mineral content, although low or medium doses appear to have no major adverse effect. Elderly patients may be more at risk due to preexisting osteoporosis, changes in estrogen levels that affect calcium utilization, and a sedentary lifestyle.”⁶

Evidence summary

A 2008 Cochrane review examined 7 RCTs comparing ICS with placebo in 1989 patients 30 to 52 years of age with mild asthma or COPD. The reviewers found no evidence of increased bone turnover, decreased bone mineral density, or increased vertebral fracture in the ICS group compared with the placebo group at 2 to 3 years’ follow-up (odds ratio [OR] for fracture=1.87; 95% confidence interval [CI], 0.5-7.0).

Steroid doses ranged from 200 to 4000 mcg beclomethasone equivalent ICS per day.¹ A 100-mcg beclomethasone equivalent ICS dose is 50 mcg fluticasone, 80 mcg budesonide, or 200 mcg triamcinolone.²

A 2008 meta-analysis of 11 RCTs that examined a number of adverse effects of ICS in adult patients with COPD found 3 studies (8131 patients) that reported no significant increase in fracture risk at 36 months in the ICS group compared with the placebo group (OR=1.09; 95% CI, 0.89-1.33). Steroid doses ranged from 1000 to 2000 mcg beclomethasone equivalent ICS per day.³

Some studies suggest an association between dose and risk
A 2008 meta-analysis that included patients with COPD or asthma, average age 43 to 81 years, showed no difference in fracture risk overall at 1 to 4 years’ follow-up (OR=1.02; 95% CI, 0.96-1.08). This analysis examined 4 RCTs, 6 case-control studies, and 3 cohort studies.⁴

A subgroup analysis of patients taking higher-dose ICS (>1500 mcg beclomethasone equivalent ICS per day) that pooled data from case-control and cohort studies suggested an increased risk of fracture (OR=1.30; 95% CI, 1.07-1.58).⁴

Investigators identified a possible dose-dependent relationship in another meta-analysis of 5 case-control studies (43,783 cases, 259,936 controls).⁵ The meta-analysis included 4 of the studies examined in the previously discussed meta-analysis.⁴

The investigators found a relative risk of 1.12 (95% CI, 1.0-1.26) for nonvertebral fracture for each 1000-mcg increase in beclomethasone equivalent ICS dose per day.⁵ Longer follow-up time wasn’t associated with greater fracture risk.

But the relationship isn’t clear
Although some non-RCT studies discussed here show that higher doses of steroids may lead to increased fracture risk, the strength of this association isn’t clear. The authors of the Cochrane review and the meta-analyses point out that a significant number of confounding factors can put asthma and COPD patients at increased risk for fracture. They include age, smoking status, inactivity, and severity of underlying lung disease. The fact that different authors controlled differently for these factors introduced heterogeneity into the meta-analyses described here.^1,3-5

Recommendations

Guidelines for the Diagnosis and Management of Asthma from the National Heart, Lung, and Blood Institute state that “most benefit is achieved with relatively low doses of ICS, whereas the risk of adverse effects increases with dose. … ICS use may be associated with a dose-dependent reduction in bone mineral content, although low or medium doses appear to have no major adverse effect. Elderly patients may be more at risk due to preexisting osteoporosis, changes in estrogen levels that affect calcium utilization, and a sedentary lifestyle.”⁶

Evidence summary

A 2008 Cochrane review examined 7 RCTs comparing ICS with placebo in 1989 patients 30 to 52 years of age with mild asthma or COPD. The reviewers found no evidence of increased bone turnover, decreased bone mineral density, or increased vertebral fracture in the ICS group compared with the placebo group at 2 to 3 years’ follow-up (odds ratio [OR] for fracture=1.87; 95% confidence interval [CI], 0.5-7.0).

Steroid doses ranged from 200 to 4000 mcg beclomethasone equivalent ICS per day.¹ A 100-mcg beclomethasone equivalent ICS dose is 50 mcg fluticasone, 80 mcg budesonide, or 200 mcg triamcinolone.²

A 2008 meta-analysis of 11 RCTs that examined a number of adverse effects of ICS in adult patients with COPD found 3 studies (8131 patients) that reported no significant increase in fracture risk at 36 months in the ICS group compared with the placebo group (OR=1.09; 95% CI, 0.89-1.33). Steroid doses ranged from 1000 to 2000 mcg beclomethasone equivalent ICS per day.³

Some studies suggest an association between dose and risk
A 2008 meta-analysis that included patients with COPD or asthma, average age 43 to 81 years, showed no difference in fracture risk overall at 1 to 4 years’ follow-up (OR=1.02; 95% CI, 0.96-1.08). This analysis examined 4 RCTs, 6 case-control studies, and 3 cohort studies.⁴

A subgroup analysis of patients taking higher-dose ICS (>1500 mcg beclomethasone equivalent ICS per day) that pooled data from case-control and cohort studies suggested an increased risk of fracture (OR=1.30; 95% CI, 1.07-1.58).⁴

Investigators identified a possible dose-dependent relationship in another meta-analysis of 5 case-control studies (43,783 cases, 259,936 controls).⁵ The meta-analysis included 4 of the studies examined in the previously discussed meta-analysis.⁴

The investigators found a relative risk of 1.12 (95% CI, 1.0-1.26) for nonvertebral fracture for each 1000-mcg increase in beclomethasone equivalent ICS dose per day.⁵ Longer follow-up time wasn’t associated with greater fracture risk.

But the relationship isn’t clear
Although some non-RCT studies discussed here show that higher doses of steroids may lead to increased fracture risk, the strength of this association isn’t clear. The authors of the Cochrane review and the meta-analyses point out that a significant number of confounding factors can put asthma and COPD patients at increased risk for fracture. They include age, smoking status, inactivity, and severity of underlying lung disease. The fact that different authors controlled differently for these factors introduced heterogeneity into the meta-analyses described here.^1,3-5

Recommendations

Guidelines for the Diagnosis and Management of Asthma from the National Heart, Lung, and Blood Institute state that “most benefit is achieved with relatively low doses of ICS, whereas the risk of adverse effects increases with dose. … ICS use may be associated with a dose-dependent reduction in bone mineral content, although low or medium doses appear to have no major adverse effect. Elderly patients may be more at risk due to preexisting osteoporosis, changes in estrogen levels that affect calcium utilization, and a sedentary lifestyle.”⁶

Most acute respiratory tract infections (ARTIs) are caused by viruses, do not require antibiotics, and resolve spontaneously.^1,2 And yet, unnecessary prescribing of antibiotics for ARTIs continues—accounting for approximately half of all such prescriptions²—despite its well-known contribution to antimicrobial resistance, a public health threat as declared by the Institute of Medicine, the Centers for Disease Control and Prevention (CDC), and the World Health Organization (WHO).^3-5

Even though the CDC has widely disseminated clinical guidelines for ARTI^6-10 and annually publicizes recommendations for ARTI management during “Get Smart About Antibiotics Week,”¹¹ it appears that providers have difficulty implementing the guidelines.^12-14 Granted, antibiotic prescription rates in general have declined somewhat, but the use of broad-spectrum antibiotics (macrolides and fluoroquinolones) and antibiotics for older Americans has increased.¹²

There are several plausible reasons for overprescribing. Patients have expectations for treatment based on prior experience or on a false assumption that their illness is bacterial in origin.¹⁴ Providers may be concerned that certain individuals are at risk of complications if not treated. Patient race, health maintenance organization membership, and insurance status have all been implicated as factors related to antimicrobial overutilization.^12-16 It can be perceived as time consuming to educate patients about the likely viral nature of their illness and the lack of utility and increased risks in taking unneeded antibiotics.¹⁷ Furthermore, attempts at patient and physician education (eg, physician performance feedback) do not always reduce antibiotic overuse.^18-20

We wanted to know the state of ARTI antibiotic use in our practice and whether we could identify goals for improvement through quality interventions. We sought to determine the distribution of ARTI final diagnoses in our practice, the frequency and types of antibiotics prescribed, and factors associated with antibiotic prescribing.

Methods

Setting and subjects
Subjects were adult patients seen at Mayo Clinic Family Medicine offices in Arizona between December 14, 2009, and March 4, 2010. We created a convenience sample from visits scheduled for patients with ARTI symptoms. We encouraged, but did not require, clinic staff to use a standardized data collection form to document symptoms, physical examination findings, diagnostic impressions, and prescription decisions that were then entered into an Excel spreadsheet. At one of our 2 sites, clinicians (attending physicians, nurse practitioners, and resident physicians) used the form at the point of care to enroll a portion of the sample population. A retrospective chart audit (with or without use of the form) was the means of selecting the remainder of the sample at this site and the entire sample at our second site. We obtained informed consent from all patients enrolled with the data collection form. The Mayo Foundation Institutional Review Board approved the project.

We defined an ARTI as a new illness occurring within the previous 3 weeks, associated with cough, sinus pain, nasal congestion or rhinorrhea, sore throat, or fever. We excluded patients who had a longer duration of symptoms, a previous evaluation, or a noninfectious diagnosis. We included ARTI patients with concomitant asthma or chronic obstructive pulmonary disease (COPD).

We enrolled 438 patients. Two hundred thirty-one (53%) consented prospectively to data collection with our standardized form; 207 (47%) were reviewed by retrospective chart audit. The mean age of subjects was 54 years (range 18-94, intraquartile range 45-69). Cough was the most frequent chief complaint (58%).

Statistical analysis
We calculated the frequency of each ARTI final diagnosis and its associated antibiotic prescription rate. We also tested for associations between clinical features and the provision of antibiotics. We hypothesized that our providers would be more likely to prescribe antibiotics for patients of advanced age and in the presence of other risk factors for complications.

Results

We determined patient risks for ARTI complication in the prospective data collection group only. Of the 231 patients, 147 (64%) had at least one risk for complication, the most common being age ≥65 (37%). Other risks were employment as a health care worker (12%), asthma (11%), atherosclerotic heart disease (8%), COPD (7%), and tobacco use (5%).

Final diagnoses for all patients appear in TABLE 1. We allowed clinicians to report more than one diagnosis, resulting in 501 final diagnoses reported for 438 patients (63 received 2 final diagnoses). Sinusitis was diagnosed most frequently (32%). Other common diagnoses were viral upper respiratory infection (URI) and acute bronchitis (29% and 24%, respectively).

Antibiotics most often prescribed. Three hundred four ARTI patients (69%) received antibiotic prescriptions. Macrolides were most commonly prescribed (167/304 [55%]). Two hundred eight ARTI patients (68%) received broad-spectrum antibiotics (macrolides or fluoroquinolones); 96 (32%) received narrow-spectrum agents (penicillin, cephalosporin, sulfa, or tetracycline derivatives). TABLE 2 lists the frequency of antibiotic prescription and the antibiotic class most frequently prescribed for each ARTI diagnosis.

Factors associated with increased prescribing included specific history and physical exam findings (TABLE 3). A major determinant of treatment was duration of illness. Those who received antibiotics had a mean duration of illness of 8.3 days, compared with 7.0 days for those not receiving antibiotic therapy (P = .03).

The rate of antibiotic prescribing varied by provider type (TABLE 4). Four resident physicians (all of whom were investigators) prescribed least often, followed by attending physicians, then nurse practitioners. Investigators were significantly less likely to prescribe antimicrobials than noninvestigators (P<.001). We assessed whether use of our standardized data collection form affected prescribing rates. When we excluded patients whose data were entered with this form, no difference in rates was seen.

We also noted wide ranges of prescribing rates between individual providers. While all providers enrolled patients, numbers ranged from one to 51, with a mean of 18. For those who enrolled ≥10 subjects, prescribing rates ranged from a low of 29% (8/28) for a resident physician investigator to 93% (63/68) for 4 noninvestigator attending physicians.

Factors not associated with increased prescribing. We had hypothesized that specific patient characteristics (age and medical complication) would be associated with provision of antimicrobials. However, there was no correlation between patient age and rate of prescribing. The 304 patients who received an antibiotic had a mean age of 54 years (standard deviation [SD]=18), as did the 134 who did not receive one (mean age, 54; SD=20; P=.95). There was a nonsignificant trend for a reduced rate of prescribing for patients younger than age 30. For patients 18 to 29 years old, the rate was 60% (31/52); for those ≥30 years, it was 71% (273/386; odds ratio [OR]=1.64; 95% confidence interval, 0.90-2.97).

Similarly, presence of medical complication did not significantly affect antibiotic prescribing rates. Patients with any risk factor for complication (age >65, diabetes, atherosclerotic heart disease, heart failure, COPD, asthma, tobacco smoking, or active cancer treatment) had a 62% prescription rate (91/147), which was the same as that of patients without such risks (52/84 [62%]; P=1.0).

TABLE 1
Final diagnoses for 438 patients with ARTI

TABLE 2
Antibiotic use and type prescribed for ARTI varied by diagnosis

TABLE 3
Historical features, exam findings associated with antibiotic prescribing

TABLE 4
Antibiotic prescription rates for ARTI varied by provider type, investigator status

Discussion

Providers in our practice had surprisingly high rates of antibiotic prescribing for ARTIs (69% overall). By comparison, the overall antibiotic use rate for ARTIs in the most recent National Ambulatory Medical Care Survey (NAMCS) analysis (1995-2006) was 58%.¹² The prescribing rate for office settings alone was just 52%. Steinman’s analysis of NAMCS data from 1997-1999 revealed an overall rate of 63%.¹³

Data analyzed from >4200 Medicare enrollees seen for ARTI visits revealed great variation in prescribing rates by office site: 21% to 88%, with a median rate of 54%.²⁰ The rate varied by final diagnoses: sinusitis, 69%; bronchitis, 59%; pharyngitis, 50%; and URI, 26%. A rate of 77% was recently reported in a Veterans Administration office setting.²¹ Those with sinusitis and bronchitis similarly received more prescriptions than those with acute pharyngitis and URI.

In addition to our high overall rate, we also diagnosed patients with sinusitis and bronchitis frequently (32% and 24% of all patients, respectively), perhaps as false justification for prescribing antibiotics (provided for 99% and 91%, respectively). Also noteworthy is that more than one-third of URI patients in our practice received antibiotics.

We had expected, but did not see, differences in prescribing rates between older and younger patients, as well as those with and without risk factors for complications. Our expectations were based on NAMCS data, which have demonstrated increasing use of antibiotics in older patients.²

Treatment for those with bronchitis was surprisingly frequent; 91% received antibiotics. A Cochrane systematic review attributes slight symptom benefit to antibiotic use (improvement in cough by about one day).²² This benefit, however, is rarely seen in patients who have been ill for <1 week. The magnitude of this benefit must be weighed against the cost and adverse effects of antibiotics and the potential for promoting antimicrobial resistance. Most patients’ symptoms are mild and self-limited, and risks may exceed benefits.

Guidelines state, “The widespread use of antibiotics for the treatment of acute bronchitis is not justified and vigorous efforts to curtail their use should be encouraged.”²³ The CDC agrees, noting that “routine antibiotic treatment of uncomplicated acute bronchitis is not recommended, regardless of duration of cough.”¹⁰

As observed in another study,¹⁴ a clinical factor associated with prescribing decisions at our practice was the duration of illness. Patients in our practice had been ill, on average, 8 days before presenting to the office. Over time, our encounters with regular patients may have taught them to wait until their symptoms are prolonged or progressive before seeking evaluation.

We saw large differences in prescribing rates between providers, and hope this means there is room for improvement by addressing reasons for variability. Education about individual prescribing behaviors may motivate those with the highest rates of use to improve.

We noted high rates of broad-spectrum antibiotic use. This is consistent with other research findings of a shift away from narrow-spectrum agents.¹² We did not determine the frequency of allergies to narrow-spectrum agents. Anecdotally, the opinion of some patients was that narrow-spectrum medicines “just don’t work,” given their experience of persistent cold symptoms when using such agents.

Quality-improvement processes such as DMAIC (Define, Measure, Analyze, Improve, Control) or PDSA (Plan, Do, Study, Act) require collection of baseline data so that interventions can be tailored to meet the root causes identified.²⁴ This project determined preintervention practice behaviors and allowed us to create quality metrics that could define our future success.

Study limitations. One obvious reason for the prescribing variability noted above is that those who helped plan and implement the project knew their practice behaviors were being reviewed and had studied the relevant practice guidelines. Whether non-investigator providers were up to date with recommendations and could carefully select appropriate treatment candidates is unclear.

This study was of our practice alone, and findings may not be generalizable to other practices. We encourage physicians to similarly examine their own prescribing habits in order to set practice-improvement goals.

CORRESPONDENCE Michael L. Grover, DO, Department of Family Medicine, Mayo Clinic, 13737 N 92nd Street, Scottsdale, AZ 85260; [email protected]

Most acute respiratory tract infections (ARTIs) are caused by viruses, do not require antibiotics, and resolve spontaneously.^1,2 And yet, unnecessary prescribing of antibiotics for ARTIs continues—accounting for approximately half of all such prescriptions²—despite its well-known contribution to antimicrobial resistance, a public health threat as declared by the Institute of Medicine, the Centers for Disease Control and Prevention (CDC), and the World Health Organization (WHO).^3-5

Even though the CDC has widely disseminated clinical guidelines for ARTI^6-10 and annually publicizes recommendations for ARTI management during “Get Smart About Antibiotics Week,”¹¹ it appears that providers have difficulty implementing the guidelines.^12-14 Granted, antibiotic prescription rates in general have declined somewhat, but the use of broad-spectrum antibiotics (macrolides and fluoroquinolones) and antibiotics for older Americans has increased.¹²

There are several plausible reasons for overprescribing. Patients have expectations for treatment based on prior experience or on a false assumption that their illness is bacterial in origin.¹⁴ Providers may be concerned that certain individuals are at risk of complications if not treated. Patient race, health maintenance organization membership, and insurance status have all been implicated as factors related to antimicrobial overutilization.^12-16 It can be perceived as time consuming to educate patients about the likely viral nature of their illness and the lack of utility and increased risks in taking unneeded antibiotics.¹⁷ Furthermore, attempts at patient and physician education (eg, physician performance feedback) do not always reduce antibiotic overuse.^18-20

We wanted to know the state of ARTI antibiotic use in our practice and whether we could identify goals for improvement through quality interventions. We sought to determine the distribution of ARTI final diagnoses in our practice, the frequency and types of antibiotics prescribed, and factors associated with antibiotic prescribing.

Methods

Setting and subjects
Subjects were adult patients seen at Mayo Clinic Family Medicine offices in Arizona between December 14, 2009, and March 4, 2010. We created a convenience sample from visits scheduled for patients with ARTI symptoms. We encouraged, but did not require, clinic staff to use a standardized data collection form to document symptoms, physical examination findings, diagnostic impressions, and prescription decisions that were then entered into an Excel spreadsheet. At one of our 2 sites, clinicians (attending physicians, nurse practitioners, and resident physicians) used the form at the point of care to enroll a portion of the sample population. A retrospective chart audit (with or without use of the form) was the means of selecting the remainder of the sample at this site and the entire sample at our second site. We obtained informed consent from all patients enrolled with the data collection form. The Mayo Foundation Institutional Review Board approved the project.

We defined an ARTI as a new illness occurring within the previous 3 weeks, associated with cough, sinus pain, nasal congestion or rhinorrhea, sore throat, or fever. We excluded patients who had a longer duration of symptoms, a previous evaluation, or a noninfectious diagnosis. We included ARTI patients with concomitant asthma or chronic obstructive pulmonary disease (COPD).

We enrolled 438 patients. Two hundred thirty-one (53%) consented prospectively to data collection with our standardized form; 207 (47%) were reviewed by retrospective chart audit. The mean age of subjects was 54 years (range 18-94, intraquartile range 45-69). Cough was the most frequent chief complaint (58%).

Statistical analysis
We calculated the frequency of each ARTI final diagnosis and its associated antibiotic prescription rate. We also tested for associations between clinical features and the provision of antibiotics. We hypothesized that our providers would be more likely to prescribe antibiotics for patients of advanced age and in the presence of other risk factors for complications.

Results

We determined patient risks for ARTI complication in the prospective data collection group only. Of the 231 patients, 147 (64%) had at least one risk for complication, the most common being age ≥65 (37%). Other risks were employment as a health care worker (12%), asthma (11%), atherosclerotic heart disease (8%), COPD (7%), and tobacco use (5%).

Final diagnoses for all patients appear in TABLE 1. We allowed clinicians to report more than one diagnosis, resulting in 501 final diagnoses reported for 438 patients (63 received 2 final diagnoses). Sinusitis was diagnosed most frequently (32%). Other common diagnoses were viral upper respiratory infection (URI) and acute bronchitis (29% and 24%, respectively).

Antibiotics most often prescribed. Three hundred four ARTI patients (69%) received antibiotic prescriptions. Macrolides were most commonly prescribed (167/304 [55%]). Two hundred eight ARTI patients (68%) received broad-spectrum antibiotics (macrolides or fluoroquinolones); 96 (32%) received narrow-spectrum agents (penicillin, cephalosporin, sulfa, or tetracycline derivatives). TABLE 2 lists the frequency of antibiotic prescription and the antibiotic class most frequently prescribed for each ARTI diagnosis.

Factors associated with increased prescribing included specific history and physical exam findings (TABLE 3). A major determinant of treatment was duration of illness. Those who received antibiotics had a mean duration of illness of 8.3 days, compared with 7.0 days for those not receiving antibiotic therapy (P = .03).

The rate of antibiotic prescribing varied by provider type (TABLE 4). Four resident physicians (all of whom were investigators) prescribed least often, followed by attending physicians, then nurse practitioners. Investigators were significantly less likely to prescribe antimicrobials than noninvestigators (P<.001). We assessed whether use of our standardized data collection form affected prescribing rates. When we excluded patients whose data were entered with this form, no difference in rates was seen.

We also noted wide ranges of prescribing rates between individual providers. While all providers enrolled patients, numbers ranged from one to 51, with a mean of 18. For those who enrolled ≥10 subjects, prescribing rates ranged from a low of 29% (8/28) for a resident physician investigator to 93% (63/68) for 4 noninvestigator attending physicians.

Factors not associated with increased prescribing. We had hypothesized that specific patient characteristics (age and medical complication) would be associated with provision of antimicrobials. However, there was no correlation between patient age and rate of prescribing. The 304 patients who received an antibiotic had a mean age of 54 years (standard deviation [SD]=18), as did the 134 who did not receive one (mean age, 54; SD=20; P=.95). There was a nonsignificant trend for a reduced rate of prescribing for patients younger than age 30. For patients 18 to 29 years old, the rate was 60% (31/52); for those ≥30 years, it was 71% (273/386; odds ratio [OR]=1.64; 95% confidence interval, 0.90-2.97).

Similarly, presence of medical complication did not significantly affect antibiotic prescribing rates. Patients with any risk factor for complication (age >65, diabetes, atherosclerotic heart disease, heart failure, COPD, asthma, tobacco smoking, or active cancer treatment) had a 62% prescription rate (91/147), which was the same as that of patients without such risks (52/84 [62%]; P=1.0).

TABLE 1
Final diagnoses for 438 patients with ARTI

TABLE 2
Antibiotic use and type prescribed for ARTI varied by diagnosis

TABLE 3
Historical features, exam findings associated with antibiotic prescribing

TABLE 4
Antibiotic prescription rates for ARTI varied by provider type, investigator status

Discussion

Providers in our practice had surprisingly high rates of antibiotic prescribing for ARTIs (69% overall). By comparison, the overall antibiotic use rate for ARTIs in the most recent National Ambulatory Medical Care Survey (NAMCS) analysis (1995-2006) was 58%.¹² The prescribing rate for office settings alone was just 52%. Steinman’s analysis of NAMCS data from 1997-1999 revealed an overall rate of 63%.¹³

Data analyzed from >4200 Medicare enrollees seen for ARTI visits revealed great variation in prescribing rates by office site: 21% to 88%, with a median rate of 54%.²⁰ The rate varied by final diagnoses: sinusitis, 69%; bronchitis, 59%; pharyngitis, 50%; and URI, 26%. A rate of 77% was recently reported in a Veterans Administration office setting.²¹ Those with sinusitis and bronchitis similarly received more prescriptions than those with acute pharyngitis and URI.

In addition to our high overall rate, we also diagnosed patients with sinusitis and bronchitis frequently (32% and 24% of all patients, respectively), perhaps as false justification for prescribing antibiotics (provided for 99% and 91%, respectively). Also noteworthy is that more than one-third of URI patients in our practice received antibiotics.

We had expected, but did not see, differences in prescribing rates between older and younger patients, as well as those with and without risk factors for complications. Our expectations were based on NAMCS data, which have demonstrated increasing use of antibiotics in older patients.²

Treatment for those with bronchitis was surprisingly frequent; 91% received antibiotics. A Cochrane systematic review attributes slight symptom benefit to antibiotic use (improvement in cough by about one day).²² This benefit, however, is rarely seen in patients who have been ill for <1 week. The magnitude of this benefit must be weighed against the cost and adverse effects of antibiotics and the potential for promoting antimicrobial resistance. Most patients’ symptoms are mild and self-limited, and risks may exceed benefits.

Guidelines state, “The widespread use of antibiotics for the treatment of acute bronchitis is not justified and vigorous efforts to curtail their use should be encouraged.”²³ The CDC agrees, noting that “routine antibiotic treatment of uncomplicated acute bronchitis is not recommended, regardless of duration of cough.”¹⁰

As observed in another study,¹⁴ a clinical factor associated with prescribing decisions at our practice was the duration of illness. Patients in our practice had been ill, on average, 8 days before presenting to the office. Over time, our encounters with regular patients may have taught them to wait until their symptoms are prolonged or progressive before seeking evaluation.

We saw large differences in prescribing rates between providers, and hope this means there is room for improvement by addressing reasons for variability. Education about individual prescribing behaviors may motivate those with the highest rates of use to improve.

We noted high rates of broad-spectrum antibiotic use. This is consistent with other research findings of a shift away from narrow-spectrum agents.¹² We did not determine the frequency of allergies to narrow-spectrum agents. Anecdotally, the opinion of some patients was that narrow-spectrum medicines “just don’t work,” given their experience of persistent cold symptoms when using such agents.

Quality-improvement processes such as DMAIC (Define, Measure, Analyze, Improve, Control) or PDSA (Plan, Do, Study, Act) require collection of baseline data so that interventions can be tailored to meet the root causes identified.²⁴ This project determined preintervention practice behaviors and allowed us to create quality metrics that could define our future success.

Study limitations. One obvious reason for the prescribing variability noted above is that those who helped plan and implement the project knew their practice behaviors were being reviewed and had studied the relevant practice guidelines. Whether non-investigator providers were up to date with recommendations and could carefully select appropriate treatment candidates is unclear.

This study was of our practice alone, and findings may not be generalizable to other practices. We encourage physicians to similarly examine their own prescribing habits in order to set practice-improvement goals.

CORRESPONDENCE Michael L. Grover, DO, Department of Family Medicine, Mayo Clinic, 13737 N 92nd Street, Scottsdale, AZ 85260; [email protected]

Most acute respiratory tract infections (ARTIs) are caused by viruses, do not require antibiotics, and resolve spontaneously.^1,2 And yet, unnecessary prescribing of antibiotics for ARTIs continues—accounting for approximately half of all such prescriptions²—despite its well-known contribution to antimicrobial resistance, a public health threat as declared by the Institute of Medicine, the Centers for Disease Control and Prevention (CDC), and the World Health Organization (WHO).^3-5

Even though the CDC has widely disseminated clinical guidelines for ARTI^6-10 and annually publicizes recommendations for ARTI management during “Get Smart About Antibiotics Week,”¹¹ it appears that providers have difficulty implementing the guidelines.^12-14 Granted, antibiotic prescription rates in general have declined somewhat, but the use of broad-spectrum antibiotics (macrolides and fluoroquinolones) and antibiotics for older Americans has increased.¹²

There are several plausible reasons for overprescribing. Patients have expectations for treatment based on prior experience or on a false assumption that their illness is bacterial in origin.¹⁴ Providers may be concerned that certain individuals are at risk of complications if not treated. Patient race, health maintenance organization membership, and insurance status have all been implicated as factors related to antimicrobial overutilization.^12-16 It can be perceived as time consuming to educate patients about the likely viral nature of their illness and the lack of utility and increased risks in taking unneeded antibiotics.¹⁷ Furthermore, attempts at patient and physician education (eg, physician performance feedback) do not always reduce antibiotic overuse.^18-20

We wanted to know the state of ARTI antibiotic use in our practice and whether we could identify goals for improvement through quality interventions. We sought to determine the distribution of ARTI final diagnoses in our practice, the frequency and types of antibiotics prescribed, and factors associated with antibiotic prescribing.

Methods

Setting and subjects
Subjects were adult patients seen at Mayo Clinic Family Medicine offices in Arizona between December 14, 2009, and March 4, 2010. We created a convenience sample from visits scheduled for patients with ARTI symptoms. We encouraged, but did not require, clinic staff to use a standardized data collection form to document symptoms, physical examination findings, diagnostic impressions, and prescription decisions that were then entered into an Excel spreadsheet. At one of our 2 sites, clinicians (attending physicians, nurse practitioners, and resident physicians) used the form at the point of care to enroll a portion of the sample population. A retrospective chart audit (with or without use of the form) was the means of selecting the remainder of the sample at this site and the entire sample at our second site. We obtained informed consent from all patients enrolled with the data collection form. The Mayo Foundation Institutional Review Board approved the project.

We defined an ARTI as a new illness occurring within the previous 3 weeks, associated with cough, sinus pain, nasal congestion or rhinorrhea, sore throat, or fever. We excluded patients who had a longer duration of symptoms, a previous evaluation, or a noninfectious diagnosis. We included ARTI patients with concomitant asthma or chronic obstructive pulmonary disease (COPD).

We enrolled 438 patients. Two hundred thirty-one (53%) consented prospectively to data collection with our standardized form; 207 (47%) were reviewed by retrospective chart audit. The mean age of subjects was 54 years (range 18-94, intraquartile range 45-69). Cough was the most frequent chief complaint (58%).

Statistical analysis
We calculated the frequency of each ARTI final diagnosis and its associated antibiotic prescription rate. We also tested for associations between clinical features and the provision of antibiotics. We hypothesized that our providers would be more likely to prescribe antibiotics for patients of advanced age and in the presence of other risk factors for complications.

Results

We determined patient risks for ARTI complication in the prospective data collection group only. Of the 231 patients, 147 (64%) had at least one risk for complication, the most common being age ≥65 (37%). Other risks were employment as a health care worker (12%), asthma (11%), atherosclerotic heart disease (8%), COPD (7%), and tobacco use (5%).

Final diagnoses for all patients appear in TABLE 1. We allowed clinicians to report more than one diagnosis, resulting in 501 final diagnoses reported for 438 patients (63 received 2 final diagnoses). Sinusitis was diagnosed most frequently (32%). Other common diagnoses were viral upper respiratory infection (URI) and acute bronchitis (29% and 24%, respectively).

Antibiotics most often prescribed. Three hundred four ARTI patients (69%) received antibiotic prescriptions. Macrolides were most commonly prescribed (167/304 [55%]). Two hundred eight ARTI patients (68%) received broad-spectrum antibiotics (macrolides or fluoroquinolones); 96 (32%) received narrow-spectrum agents (penicillin, cephalosporin, sulfa, or tetracycline derivatives). TABLE 2 lists the frequency of antibiotic prescription and the antibiotic class most frequently prescribed for each ARTI diagnosis.

Factors associated with increased prescribing included specific history and physical exam findings (TABLE 3). A major determinant of treatment was duration of illness. Those who received antibiotics had a mean duration of illness of 8.3 days, compared with 7.0 days for those not receiving antibiotic therapy (P = .03).

The rate of antibiotic prescribing varied by provider type (TABLE 4). Four resident physicians (all of whom were investigators) prescribed least often, followed by attending physicians, then nurse practitioners. Investigators were significantly less likely to prescribe antimicrobials than noninvestigators (P<.001). We assessed whether use of our standardized data collection form affected prescribing rates. When we excluded patients whose data were entered with this form, no difference in rates was seen.

We also noted wide ranges of prescribing rates between individual providers. While all providers enrolled patients, numbers ranged from one to 51, with a mean of 18. For those who enrolled ≥10 subjects, prescribing rates ranged from a low of 29% (8/28) for a resident physician investigator to 93% (63/68) for 4 noninvestigator attending physicians.

Factors not associated with increased prescribing. We had hypothesized that specific patient characteristics (age and medical complication) would be associated with provision of antimicrobials. However, there was no correlation between patient age and rate of prescribing. The 304 patients who received an antibiotic had a mean age of 54 years (standard deviation [SD]=18), as did the 134 who did not receive one (mean age, 54; SD=20; P=.95). There was a nonsignificant trend for a reduced rate of prescribing for patients younger than age 30. For patients 18 to 29 years old, the rate was 60% (31/52); for those ≥30 years, it was 71% (273/386; odds ratio [OR]=1.64; 95% confidence interval, 0.90-2.97).

Similarly, presence of medical complication did not significantly affect antibiotic prescribing rates. Patients with any risk factor for complication (age >65, diabetes, atherosclerotic heart disease, heart failure, COPD, asthma, tobacco smoking, or active cancer treatment) had a 62% prescription rate (91/147), which was the same as that of patients without such risks (52/84 [62%]; P=1.0).

TABLE 1
Final diagnoses for 438 patients with ARTI

TABLE 2
Antibiotic use and type prescribed for ARTI varied by diagnosis

TABLE 3
Historical features, exam findings associated with antibiotic prescribing

TABLE 4
Antibiotic prescription rates for ARTI varied by provider type, investigator status

Discussion

Providers in our practice had surprisingly high rates of antibiotic prescribing for ARTIs (69% overall). By comparison, the overall antibiotic use rate for ARTIs in the most recent National Ambulatory Medical Care Survey (NAMCS) analysis (1995-2006) was 58%.¹² The prescribing rate for office settings alone was just 52%. Steinman’s analysis of NAMCS data from 1997-1999 revealed an overall rate of 63%.¹³

Data analyzed from >4200 Medicare enrollees seen for ARTI visits revealed great variation in prescribing rates by office site: 21% to 88%, with a median rate of 54%.²⁰ The rate varied by final diagnoses: sinusitis, 69%; bronchitis, 59%; pharyngitis, 50%; and URI, 26%. A rate of 77% was recently reported in a Veterans Administration office setting.²¹ Those with sinusitis and bronchitis similarly received more prescriptions than those with acute pharyngitis and URI.

In addition to our high overall rate, we also diagnosed patients with sinusitis and bronchitis frequently (32% and 24% of all patients, respectively), perhaps as false justification for prescribing antibiotics (provided for 99% and 91%, respectively). Also noteworthy is that more than one-third of URI patients in our practice received antibiotics.

We had expected, but did not see, differences in prescribing rates between older and younger patients, as well as those with and without risk factors for complications. Our expectations were based on NAMCS data, which have demonstrated increasing use of antibiotics in older patients.²

Treatment for those with bronchitis was surprisingly frequent; 91% received antibiotics. A Cochrane systematic review attributes slight symptom benefit to antibiotic use (improvement in cough by about one day).²² This benefit, however, is rarely seen in patients who have been ill for <1 week. The magnitude of this benefit must be weighed against the cost and adverse effects of antibiotics and the potential for promoting antimicrobial resistance. Most patients’ symptoms are mild and self-limited, and risks may exceed benefits.

Guidelines state, “The widespread use of antibiotics for the treatment of acute bronchitis is not justified and vigorous efforts to curtail their use should be encouraged.”²³ The CDC agrees, noting that “routine antibiotic treatment of uncomplicated acute bronchitis is not recommended, regardless of duration of cough.”¹⁰

As observed in another study,¹⁴ a clinical factor associated with prescribing decisions at our practice was the duration of illness. Patients in our practice had been ill, on average, 8 days before presenting to the office. Over time, our encounters with regular patients may have taught them to wait until their symptoms are prolonged or progressive before seeking evaluation.

We saw large differences in prescribing rates between providers, and hope this means there is room for improvement by addressing reasons for variability. Education about individual prescribing behaviors may motivate those with the highest rates of use to improve.

We noted high rates of broad-spectrum antibiotic use. This is consistent with other research findings of a shift away from narrow-spectrum agents.¹² We did not determine the frequency of allergies to narrow-spectrum agents. Anecdotally, the opinion of some patients was that narrow-spectrum medicines “just don’t work,” given their experience of persistent cold symptoms when using such agents.

Quality-improvement processes such as DMAIC (Define, Measure, Analyze, Improve, Control) or PDSA (Plan, Do, Study, Act) require collection of baseline data so that interventions can be tailored to meet the root causes identified.²⁴ This project determined preintervention practice behaviors and allowed us to create quality metrics that could define our future success.

Study limitations. One obvious reason for the prescribing variability noted above is that those who helped plan and implement the project knew their practice behaviors were being reviewed and had studied the relevant practice guidelines. Whether non-investigator providers were up to date with recommendations and could carefully select appropriate treatment candidates is unclear.

This study was of our practice alone, and findings may not be generalizable to other practices. We encourage physicians to similarly examine their own prescribing habits in order to set practice-improvement goals.

CORRESPONDENCE Michael L. Grover, DO, Department of Family Medicine, Mayo Clinic, 13737 N 92nd Street, Scottsdale, AZ 85260; [email protected]

Evidence summary

A systematic review from the US Surgeon General’s office of studies addressing the relationship between secondhand smoke exposure and asthma severity in children from 0 to 18 years of age found that children with asthma who were exposed to secondhand smoke had “greater disease severity” than unexposed children.¹ The studies—including 8 prospective and retrospective cohort studies (N=6095), one case-control study (N=149), and 11 uncontrolled case series (N=2932)—were performed in the United States, Canada, the United Kingdom, Sweden, Singapore, South Africa, Kenya, and Nigeria.

Investigators found a significant worsening of asthma caused by secondhand smoke in 6 of 11 clinic-based studies and 2 of 9 population-based studies. Children with asthma who were exposed to secondhand smoke had more doctor visits, more frequent flares, and higher disease severity scores than children who weren’t exposed. Heterogeneity among the studies prevented a meta-analysis of data on severity of asthma.

Where there’s smoke, there are worse health outcomes
Three of 4 subsequent cohort studies found poorer health outcomes among children with asthma who were exposed to smoking than children who weren’t. The first study, of 523 children 4 to 16 years of age with physician-diagnosed asthma, correlated smoke exposure, as indicated by serum cotinine levels, with pulmonary function tests and clinical outcomes.² Children with high serum cotinine levels (>0.63 mg/mL) were more likely to have asthma symptoms monthly or more often, as reported by the family (adjusted odds ratio [OR]=2.7; 95% confidence interval [CI], 1.1-6.5), than children with low cotinine levels (<0.116 ng/mL). High cotinine levels weren’t associated with significant changes in forced expiratory volume in one second, decreased school attendance, or increased physician visits.

Another study of 438 children ages 2 to 12 years with physician-diagnosed asthma and at least one parent who smoked, correlated salivary cotinine levels with the likelihood of contacting a physician for asthma symptoms.³ Children with high salivary cotinine levels (>4.5 ng/mL) had higher asthma-related physician contact rates than children with low cotinine levels (≤2 ng/mL) (incidence rate ratio=1.2; 95% CI, 1.1-1.4).

A third study evaluated asthma treatment response in 167 children from families throughout France who were 6 to 12 years of age and recently diagnosed with mild or moderate persistent asthma.⁴ Investigators performed pulmonary function tests and collected data on symptoms every 4 months for 3 years. Children who lived with someone who smoked were less likely to have controlled asthma symptoms (OR=0.34; 95% CI, 0.13–0.91).

The fourth study, of 126 urban children ages 6 to 12 years with physician-diagnosed asthma and in-home smoke exposure, correlated urinary cotinine levels and rates of clinical illness. It found no significant differences in parent-reported illness between children with higher urinary cotinine levels and children with lower levels.⁵

Recommendations

The National Asthma Education and Prevention Program Expert Panel recommends that physicians ask patients about their smoking status and refer adults who have children with asthma to smoking cessation programs.⁶ The panel further recommends that clinicians advise people with asthma to avoid smoking and limit exposure to environmental tobacco smoke.

Evidence summary

A systematic review from the US Surgeon General’s office of studies addressing the relationship between secondhand smoke exposure and asthma severity in children from 0 to 18 years of age found that children with asthma who were exposed to secondhand smoke had “greater disease severity” than unexposed children.¹ The studies—including 8 prospective and retrospective cohort studies (N=6095), one case-control study (N=149), and 11 uncontrolled case series (N=2932)—were performed in the United States, Canada, the United Kingdom, Sweden, Singapore, South Africa, Kenya, and Nigeria.

Investigators found a significant worsening of asthma caused by secondhand smoke in 6 of 11 clinic-based studies and 2 of 9 population-based studies. Children with asthma who were exposed to secondhand smoke had more doctor visits, more frequent flares, and higher disease severity scores than children who weren’t exposed. Heterogeneity among the studies prevented a meta-analysis of data on severity of asthma.

Where there’s smoke, there are worse health outcomes
Three of 4 subsequent cohort studies found poorer health outcomes among children with asthma who were exposed to smoking than children who weren’t. The first study, of 523 children 4 to 16 years of age with physician-diagnosed asthma, correlated smoke exposure, as indicated by serum cotinine levels, with pulmonary function tests and clinical outcomes.² Children with high serum cotinine levels (>0.63 mg/mL) were more likely to have asthma symptoms monthly or more often, as reported by the family (adjusted odds ratio [OR]=2.7; 95% confidence interval [CI], 1.1-6.5), than children with low cotinine levels (<0.116 ng/mL). High cotinine levels weren’t associated with significant changes in forced expiratory volume in one second, decreased school attendance, or increased physician visits.

Another study of 438 children ages 2 to 12 years with physician-diagnosed asthma and at least one parent who smoked, correlated salivary cotinine levels with the likelihood of contacting a physician for asthma symptoms.³ Children with high salivary cotinine levels (>4.5 ng/mL) had higher asthma-related physician contact rates than children with low cotinine levels (≤2 ng/mL) (incidence rate ratio=1.2; 95% CI, 1.1-1.4).

A third study evaluated asthma treatment response in 167 children from families throughout France who were 6 to 12 years of age and recently diagnosed with mild or moderate persistent asthma.⁴ Investigators performed pulmonary function tests and collected data on symptoms every 4 months for 3 years. Children who lived with someone who smoked were less likely to have controlled asthma symptoms (OR=0.34; 95% CI, 0.13–0.91).

The fourth study, of 126 urban children ages 6 to 12 years with physician-diagnosed asthma and in-home smoke exposure, correlated urinary cotinine levels and rates of clinical illness. It found no significant differences in parent-reported illness between children with higher urinary cotinine levels and children with lower levels.⁵

Recommendations

The National Asthma Education and Prevention Program Expert Panel recommends that physicians ask patients about their smoking status and refer adults who have children with asthma to smoking cessation programs.⁶ The panel further recommends that clinicians advise people with asthma to avoid smoking and limit exposure to environmental tobacco smoke.

Evidence summary

A systematic review from the US Surgeon General’s office of studies addressing the relationship between secondhand smoke exposure and asthma severity in children from 0 to 18 years of age found that children with asthma who were exposed to secondhand smoke had “greater disease severity” than unexposed children.¹ The studies—including 8 prospective and retrospective cohort studies (N=6095), one case-control study (N=149), and 11 uncontrolled case series (N=2932)—were performed in the United States, Canada, the United Kingdom, Sweden, Singapore, South Africa, Kenya, and Nigeria.

Investigators found a significant worsening of asthma caused by secondhand smoke in 6 of 11 clinic-based studies and 2 of 9 population-based studies. Children with asthma who were exposed to secondhand smoke had more doctor visits, more frequent flares, and higher disease severity scores than children who weren’t exposed. Heterogeneity among the studies prevented a meta-analysis of data on severity of asthma.

Where there’s smoke, there are worse health outcomes
Three of 4 subsequent cohort studies found poorer health outcomes among children with asthma who were exposed to smoking than children who weren’t. The first study, of 523 children 4 to 16 years of age with physician-diagnosed asthma, correlated smoke exposure, as indicated by serum cotinine levels, with pulmonary function tests and clinical outcomes.² Children with high serum cotinine levels (>0.63 mg/mL) were more likely to have asthma symptoms monthly or more often, as reported by the family (adjusted odds ratio [OR]=2.7; 95% confidence interval [CI], 1.1-6.5), than children with low cotinine levels (<0.116 ng/mL). High cotinine levels weren’t associated with significant changes in forced expiratory volume in one second, decreased school attendance, or increased physician visits.

Another study of 438 children ages 2 to 12 years with physician-diagnosed asthma and at least one parent who smoked, correlated salivary cotinine levels with the likelihood of contacting a physician for asthma symptoms.³ Children with high salivary cotinine levels (>4.5 ng/mL) had higher asthma-related physician contact rates than children with low cotinine levels (≤2 ng/mL) (incidence rate ratio=1.2; 95% CI, 1.1-1.4).

A third study evaluated asthma treatment response in 167 children from families throughout France who were 6 to 12 years of age and recently diagnosed with mild or moderate persistent asthma.⁴ Investigators performed pulmonary function tests and collected data on symptoms every 4 months for 3 years. Children who lived with someone who smoked were less likely to have controlled asthma symptoms (OR=0.34; 95% CI, 0.13–0.91).

The fourth study, of 126 urban children ages 6 to 12 years with physician-diagnosed asthma and in-home smoke exposure, correlated urinary cotinine levels and rates of clinical illness. It found no significant differences in parent-reported illness between children with higher urinary cotinine levels and children with lower levels.⁵

Recommendations

The National Asthma Education and Prevention Program Expert Panel recommends that physicians ask patients about their smoking status and refer adults who have children with asthma to smoking cessation programs.⁶ The panel further recommends that clinicians advise people with asthma to avoid smoking and limit exposure to environmental tobacco smoke.

ILLUSTRATIVE CASE

A 53-year-old man asks you to prescribe Chantix to help him stop smoking. He has made several attempts to quit in the past, but never managed to stop for more than 6 months— and has smoked a pack a day for 30 years. The patient does not have a history of heart disease, but he is on statin therapy for hyperlipidemia. What should you tell him about varenicline’s potential benefits and risks?

Tobacco use remains the largest preventable contributor to death and disease in the United States.² In smokers with coronary heart disease, smoking cessation is associated with a 36% reduction in all-cause mortality (relative risk [RR], 0.64; 95% confidence interval [CI], 0.58-0.71)—a risk reduction greater than that of statins (29%), aspirin (15%), beta-blockers (23%), or ACE inhibitors (23%).³

Varenicline now has 2 black box warnings
In its 2009 update on recommendations for smoking cessation, the United States Preventive Services Task Force cited NRT and controlled-release bupropion, as well as varenicline, as effective smoking cessation aids.⁴ Varenicline received US Food and Drug Administration (FDA) approval in 2006. In 2009, the FDA added a black box warning based on evidence of its adverse neuropsychiatric effects, including suicidality.⁵

In July 2011, the FDA required another label change,⁶ based on a double-blind RCT published in 2010 showing that for patients with CVD, varenicline is associated with an increased risk.⁷ As a partial nicotine agonist, varenicline could confer some of the CV risk associated with nicotine abuse.⁸ The FDA has asked its manufacturer, Pfizer Inc, to conduct further studies.⁶ The meta-analysis reviewed below—which was not associated with Pfizer or the FDA—was published in September 2011, just a couple of months after the label change.¹

STUDY SUMMARY: Risk of ischemic or arrhythmic event is small but significant

Singh et al searched for double-blind RCTs that tested varenicline against a control in tobacco users.¹ All included studies had to have reported adverse CV events. The primary outcome was any ischemic or arrhythmic CV event.

The researchers found 15 such studies (n=8216), which ranged in duration from 7 to 52 weeks. Most used a placebo control, but some included bupropion or NRT. The researchers used a Peto odds ratio (OR) for the meta-analysis, useful when combining uncommon events and including studies with no events.⁹

Compared with placebo, varenicline significantly increased the risk of CV events (odds ratio [OR], 1.72; 95% CI, 1.09-2.71). The incidence of CV events was 1.06% (52 of 4908) among varenicline users vs 0.82% (27 of 3308) in the controls (number needed to harm [NNH]=417).

The limited number of deaths (1.4% among patients taking varenicline vs 2.1% in the placebo groups) prevented analysis of mortality risk. The study with the most statistical power, which accounted for 57% of the overall effect, was the only one that included patients with known stable CV disease. (None included patients with unstable CV disease, whose risk may be greater.) Even when this study was removed, however, the outcome (OR, 2.54; 95% CI, 1.26-5.12) was consistent with the primary result for CV events. A sensitivity analysis comparing the risk associated with varenicline with that of either NRT or bupropion yielded similar results (OR, 1.67; 95% CI, 1.07-26.2). For a higher risk population with stable CVD (5.6% annual risk at baseline), the authors estimated an overall NNH of 28 per year (95% CI, 13-213).

WHAT’S NEW: Evidence of CV risk is cause for concern

This meta-analysis provides evidence that varenicline is associated with a small but significant harmful effect on CV outcomes. The methods Singh et al used for review and article selection appear to be sound, and analysis of the included studies reveals little likelihood of publication bias.

CAVEATS: For many, benefits of quitting outweigh the risks

The absolute risk of a CV event found in this meta-analysis was small—just 0.24%. What’s more, the primary outcome was a composite of a diverse group of outcomes, some more serious than others. And, when compared with the highly positive effects of smoking cessation, the benefit-harm analysis still appears to favor varenicline for most patients. The estimated number needed to treat to get one person to stop smoking for ≥24 weeks is about 10 (95% CI, 8-13).⁸

CHALLENGES TO IMPLEMENTATION: Finding time to educate patients

The additional time needed to discuss the CV and neuropsychiatric risks of varenicline will be a challenge to physicians working in busy outpatient settings. Proper documentation of this discussion is prudent, however, given the increase in risk with this medication.

Acknowledgement

The Purls Surveillance System is supported in part by Grant Number UL1RR024999 from the National Center for Research Resources, a Clinical Translational Science Award to the University of Chicago. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources or the National Institutes of Health.

Click here to view PURL METHODOLOGY

ILLUSTRATIVE CASE

A 53-year-old man asks you to prescribe Chantix to help him stop smoking. He has made several attempts to quit in the past, but never managed to stop for more than 6 months— and has smoked a pack a day for 30 years. The patient does not have a history of heart disease, but he is on statin therapy for hyperlipidemia. What should you tell him about varenicline’s potential benefits and risks?

Tobacco use remains the largest preventable contributor to death and disease in the United States.² In smokers with coronary heart disease, smoking cessation is associated with a 36% reduction in all-cause mortality (relative risk [RR], 0.64; 95% confidence interval [CI], 0.58-0.71)—a risk reduction greater than that of statins (29%), aspirin (15%), beta-blockers (23%), or ACE inhibitors (23%).³

Varenicline now has 2 black box warnings
In its 2009 update on recommendations for smoking cessation, the United States Preventive Services Task Force cited NRT and controlled-release bupropion, as well as varenicline, as effective smoking cessation aids.⁴ Varenicline received US Food and Drug Administration (FDA) approval in 2006. In 2009, the FDA added a black box warning based on evidence of its adverse neuropsychiatric effects, including suicidality.⁵

In July 2011, the FDA required another label change,⁶ based on a double-blind RCT published in 2010 showing that for patients with CVD, varenicline is associated with an increased risk.⁷ As a partial nicotine agonist, varenicline could confer some of the CV risk associated with nicotine abuse.⁸ The FDA has asked its manufacturer, Pfizer Inc, to conduct further studies.⁶ The meta-analysis reviewed below—which was not associated with Pfizer or the FDA—was published in September 2011, just a couple of months after the label change.¹

STUDY SUMMARY: Risk of ischemic or arrhythmic event is small but significant

Singh et al searched for double-blind RCTs that tested varenicline against a control in tobacco users.¹ All included studies had to have reported adverse CV events. The primary outcome was any ischemic or arrhythmic CV event.

The researchers found 15 such studies (n=8216), which ranged in duration from 7 to 52 weeks. Most used a placebo control, but some included bupropion or NRT. The researchers used a Peto odds ratio (OR) for the meta-analysis, useful when combining uncommon events and including studies with no events.⁹

Compared with placebo, varenicline significantly increased the risk of CV events (odds ratio [OR], 1.72; 95% CI, 1.09-2.71). The incidence of CV events was 1.06% (52 of 4908) among varenicline users vs 0.82% (27 of 3308) in the controls (number needed to harm [NNH]=417).

The limited number of deaths (1.4% among patients taking varenicline vs 2.1% in the placebo groups) prevented analysis of mortality risk. The study with the most statistical power, which accounted for 57% of the overall effect, was the only one that included patients with known stable CV disease. (None included patients with unstable CV disease, whose risk may be greater.) Even when this study was removed, however, the outcome (OR, 2.54; 95% CI, 1.26-5.12) was consistent with the primary result for CV events. A sensitivity analysis comparing the risk associated with varenicline with that of either NRT or bupropion yielded similar results (OR, 1.67; 95% CI, 1.07-26.2). For a higher risk population with stable CVD (5.6% annual risk at baseline), the authors estimated an overall NNH of 28 per year (95% CI, 13-213).

WHAT’S NEW: Evidence of CV risk is cause for concern

This meta-analysis provides evidence that varenicline is associated with a small but significant harmful effect on CV outcomes. The methods Singh et al used for review and article selection appear to be sound, and analysis of the included studies reveals little likelihood of publication bias.

CAVEATS: For many, benefits of quitting outweigh the risks

The absolute risk of a CV event found in this meta-analysis was small—just 0.24%. What’s more, the primary outcome was a composite of a diverse group of outcomes, some more serious than others. And, when compared with the highly positive effects of smoking cessation, the benefit-harm analysis still appears to favor varenicline for most patients. The estimated number needed to treat to get one person to stop smoking for ≥24 weeks is about 10 (95% CI, 8-13).⁸

CHALLENGES TO IMPLEMENTATION: Finding time to educate patients

The additional time needed to discuss the CV and neuropsychiatric risks of varenicline will be a challenge to physicians working in busy outpatient settings. Proper documentation of this discussion is prudent, however, given the increase in risk with this medication.

Acknowledgement

The Purls Surveillance System is supported in part by Grant Number UL1RR024999 from the National Center for Research Resources, a Clinical Translational Science Award to the University of Chicago. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources or the National Institutes of Health.

Click here to view PURL METHODOLOGY

ILLUSTRATIVE CASE

A 53-year-old man asks you to prescribe Chantix to help him stop smoking. He has made several attempts to quit in the past, but never managed to stop for more than 6 months— and has smoked a pack a day for 30 years. The patient does not have a history of heart disease, but he is on statin therapy for hyperlipidemia. What should you tell him about varenicline’s potential benefits and risks?

Tobacco use remains the largest preventable contributor to death and disease in the United States.² In smokers with coronary heart disease, smoking cessation is associated with a 36% reduction in all-cause mortality (relative risk [RR], 0.64; 95% confidence interval [CI], 0.58-0.71)—a risk reduction greater than that of statins (29%), aspirin (15%), beta-blockers (23%), or ACE inhibitors (23%).³

Varenicline now has 2 black box warnings
In its 2009 update on recommendations for smoking cessation, the United States Preventive Services Task Force cited NRT and controlled-release bupropion, as well as varenicline, as effective smoking cessation aids.⁴ Varenicline received US Food and Drug Administration (FDA) approval in 2006. In 2009, the FDA added a black box warning based on evidence of its adverse neuropsychiatric effects, including suicidality.⁵

In July 2011, the FDA required another label change,⁶ based on a double-blind RCT published in 2010 showing that for patients with CVD, varenicline is associated with an increased risk.⁷ As a partial nicotine agonist, varenicline could confer some of the CV risk associated with nicotine abuse.⁸ The FDA has asked its manufacturer, Pfizer Inc, to conduct further studies.⁶ The meta-analysis reviewed below—which was not associated with Pfizer or the FDA—was published in September 2011, just a couple of months after the label change.¹

STUDY SUMMARY: Risk of ischemic or arrhythmic event is small but significant

Singh et al searched for double-blind RCTs that tested varenicline against a control in tobacco users.¹ All included studies had to have reported adverse CV events. The primary outcome was any ischemic or arrhythmic CV event.

The researchers found 15 such studies (n=8216), which ranged in duration from 7 to 52 weeks. Most used a placebo control, but some included bupropion or NRT. The researchers used a Peto odds ratio (OR) for the meta-analysis, useful when combining uncommon events and including studies with no events.⁹

Compared with placebo, varenicline significantly increased the risk of CV events (odds ratio [OR], 1.72; 95% CI, 1.09-2.71). The incidence of CV events was 1.06% (52 of 4908) among varenicline users vs 0.82% (27 of 3308) in the controls (number needed to harm [NNH]=417).

The limited number of deaths (1.4% among patients taking varenicline vs 2.1% in the placebo groups) prevented analysis of mortality risk. The study with the most statistical power, which accounted for 57% of the overall effect, was the only one that included patients with known stable CV disease. (None included patients with unstable CV disease, whose risk may be greater.) Even when this study was removed, however, the outcome (OR, 2.54; 95% CI, 1.26-5.12) was consistent with the primary result for CV events. A sensitivity analysis comparing the risk associated with varenicline with that of either NRT or bupropion yielded similar results (OR, 1.67; 95% CI, 1.07-26.2). For a higher risk population with stable CVD (5.6% annual risk at baseline), the authors estimated an overall NNH of 28 per year (95% CI, 13-213).

WHAT’S NEW: Evidence of CV risk is cause for concern

This meta-analysis provides evidence that varenicline is associated with a small but significant harmful effect on CV outcomes. The methods Singh et al used for review and article selection appear to be sound, and analysis of the included studies reveals little likelihood of publication bias.

CAVEATS: For many, benefits of quitting outweigh the risks

The absolute risk of a CV event found in this meta-analysis was small—just 0.24%. What’s more, the primary outcome was a composite of a diverse group of outcomes, some more serious than others. And, when compared with the highly positive effects of smoking cessation, the benefit-harm analysis still appears to favor varenicline for most patients. The estimated number needed to treat to get one person to stop smoking for ≥24 weeks is about 10 (95% CI, 8-13).⁸

CHALLENGES TO IMPLEMENTATION: Finding time to educate patients

The additional time needed to discuss the CV and neuropsychiatric risks of varenicline will be a challenge to physicians working in busy outpatient settings. Proper documentation of this discussion is prudent, however, given the increase in risk with this medication.

Acknowledgement

The Purls Surveillance System is supported in part by Grant Number UL1RR024999 from the National Center for Research Resources, a Clinical Translational Science Award to the University of Chicago. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Center for Research Resources or the National Institutes of Health.

Click here to view PURL METHODOLOGY