Clinical Inquiries

Evidence summary

Analgesic rebound headaches are seen in 1% of the population, mostly middle-aged women with underlying migraines.^1,2 Also termed analgesic-overuse headaches, they are defined by the International Headache Society guidelines as headaches occurring more than 15 days per month, mild to moderate in intensity, developing or worsening with analgesic overuse, and resolving or reverting to the prior underlying headache pattern within 2 months of discontinuing the analgesic(s).³

A case series studied 50 patients with rebound headaches for 5 or more days a week at baseline.⁴ Patients were educated regarding analgesic overuse headaches, after which their analgesics were abruptly discontinued, and they were followed up to a year. Subcutaneous DHE was used as needed for symptomatic relief of excruciating headaches. At study completion, 78% of patients had adequately stopped analgesics. The goal of greater than 6 consecutive headache-free days was achieved in 74% patients in an average of 84 days.

A 9-week double-blind, placebo-controlled trial randomized 20 nondepressed patients with analgesic overuse headache to receive amitriptyline or active placebo (trihexyphenidyl).⁵ Patients were admitted to the hospital for 1 week and withdrawn from all analgesics. The 2 groups had similar baseline characteristics. During the hospitalization, the amitriptyline treatment group received intravenous amitriptyline escalating from 25 to 75 mg. During the following month, oral study medications were continued, and patients took low doses of aspirin or acetaminophen, as needed. There was no significant difference between the 2 groups with regard to analgesic use. At completion of this low-powered study, no difference was found between the 2 groups in headache frequency or analgesic use, although certain components of a quality-of-life scale were better in the amitriptyline group.

An open-label trial of patients with chronic migraine and analgesic overuse in a headache sub-specialty center abruptly withdrew 150 participants from analgesics and quasi-randomized them to 3 groups: prednisone (tapering from 60 to 20 mg over 6 days), naratriptan (Amerge) (2.5 mg twice daily for 6 days), or no prophylactic treatment.⁶ Patients given the active substances were told it would reduce withdrawal symptoms; patients given placebo were not given this advice. All patients received education about the pathophysiology of rebound headaches, kept a headache diary, and were phoned weekly to ensure compliance. In addition, they all received capsules containing gradually increasing doses of atenolol, nortriptyline, and flunarazine (a calcium channel blocker not FDA-approved.) Indo-methacin and chlorpromazine were used as needed. Results from the first 6 days showed no difference in headaches between the 3 groups; however, significantly more patients used chlorpromazine in the “no pharmacologic treatment” group

By the end of 5 weeks, headache frequency was significantly reduced in all groups from baseline; however, there were no differences between groups in headache frequency or intensity in this small study. Of note, there were statistically fewer withdrawal symptoms and less use of rescue medications among patients who received the initial prophylactic treatments. The indomethacin rescue use was 24%, 18%, and 14% of patients for the no prophylactic treatment, prednisone, and naratriptan groups respectively, while chlorpromazine rescue use was 14%, 0%, and 0%, respectively. The number of patients needed to treat to prevent any withdrawal symptoms (nausea, vomiting, nervousness, dizziness, etc.) was 1 for every 3.5 for naratriptan, and 6.4 for prednisone.

Recommendations from others

The American Council for Headache Education recommends discontinuing all analgesics.⁷ It notes some patients may need prophylactic medication (although no specific agent is recommended), and hospitalization may be indicated for withdrawal for patients who have abused narcotics. A headache textbook recommends 1 of 2 approaches for patients undergoing outpatient treatment: (1) gradual tapering of the offending medication with substitution of a long-acting nonsteroidal anti-inflammatory drug (NSAID) and initiation of preventive therapy, or (2) abrupt discontinuation of the offending medication and initiation followed by gradual tapering of a “transitional” medication such as NSAIDs, DHE, corticosteroids, or triptans. The authors recommend an intravenous DHE protocol for treatment failures and patients requiring inpatient treatment.⁸

Evidence summary

Analgesic rebound headaches are seen in 1% of the population, mostly middle-aged women with underlying migraines.^1,2 Also termed analgesic-overuse headaches, they are defined by the International Headache Society guidelines as headaches occurring more than 15 days per month, mild to moderate in intensity, developing or worsening with analgesic overuse, and resolving or reverting to the prior underlying headache pattern within 2 months of discontinuing the analgesic(s).³

A case series studied 50 patients with rebound headaches for 5 or more days a week at baseline.⁴ Patients were educated regarding analgesic overuse headaches, after which their analgesics were abruptly discontinued, and they were followed up to a year. Subcutaneous DHE was used as needed for symptomatic relief of excruciating headaches. At study completion, 78% of patients had adequately stopped analgesics. The goal of greater than 6 consecutive headache-free days was achieved in 74% patients in an average of 84 days.

A 9-week double-blind, placebo-controlled trial randomized 20 nondepressed patients with analgesic overuse headache to receive amitriptyline or active placebo (trihexyphenidyl).⁵ Patients were admitted to the hospital for 1 week and withdrawn from all analgesics. The 2 groups had similar baseline characteristics. During the hospitalization, the amitriptyline treatment group received intravenous amitriptyline escalating from 25 to 75 mg. During the following month, oral study medications were continued, and patients took low doses of aspirin or acetaminophen, as needed. There was no significant difference between the 2 groups with regard to analgesic use. At completion of this low-powered study, no difference was found between the 2 groups in headache frequency or analgesic use, although certain components of a quality-of-life scale were better in the amitriptyline group.

An open-label trial of patients with chronic migraine and analgesic overuse in a headache sub-specialty center abruptly withdrew 150 participants from analgesics and quasi-randomized them to 3 groups: prednisone (tapering from 60 to 20 mg over 6 days), naratriptan (Amerge) (2.5 mg twice daily for 6 days), or no prophylactic treatment.⁶ Patients given the active substances were told it would reduce withdrawal symptoms; patients given placebo were not given this advice. All patients received education about the pathophysiology of rebound headaches, kept a headache diary, and were phoned weekly to ensure compliance. In addition, they all received capsules containing gradually increasing doses of atenolol, nortriptyline, and flunarazine (a calcium channel blocker not FDA-approved.) Indo-methacin and chlorpromazine were used as needed. Results from the first 6 days showed no difference in headaches between the 3 groups; however, significantly more patients used chlorpromazine in the “no pharmacologic treatment” group

By the end of 5 weeks, headache frequency was significantly reduced in all groups from baseline; however, there were no differences between groups in headache frequency or intensity in this small study. Of note, there were statistically fewer withdrawal symptoms and less use of rescue medications among patients who received the initial prophylactic treatments. The indomethacin rescue use was 24%, 18%, and 14% of patients for the no prophylactic treatment, prednisone, and naratriptan groups respectively, while chlorpromazine rescue use was 14%, 0%, and 0%, respectively. The number of patients needed to treat to prevent any withdrawal symptoms (nausea, vomiting, nervousness, dizziness, etc.) was 1 for every 3.5 for naratriptan, and 6.4 for prednisone.

Recommendations from others

The American Council for Headache Education recommends discontinuing all analgesics.⁷ It notes some patients may need prophylactic medication (although no specific agent is recommended), and hospitalization may be indicated for withdrawal for patients who have abused narcotics. A headache textbook recommends 1 of 2 approaches for patients undergoing outpatient treatment: (1) gradual tapering of the offending medication with substitution of a long-acting nonsteroidal anti-inflammatory drug (NSAID) and initiation of preventive therapy, or (2) abrupt discontinuation of the offending medication and initiation followed by gradual tapering of a “transitional” medication such as NSAIDs, DHE, corticosteroids, or triptans. The authors recommend an intravenous DHE protocol for treatment failures and patients requiring inpatient treatment.⁸

Evidence summary

Analgesic rebound headaches are seen in 1% of the population, mostly middle-aged women with underlying migraines.^1,2 Also termed analgesic-overuse headaches, they are defined by the International Headache Society guidelines as headaches occurring more than 15 days per month, mild to moderate in intensity, developing or worsening with analgesic overuse, and resolving or reverting to the prior underlying headache pattern within 2 months of discontinuing the analgesic(s).³

A case series studied 50 patients with rebound headaches for 5 or more days a week at baseline.⁴ Patients were educated regarding analgesic overuse headaches, after which their analgesics were abruptly discontinued, and they were followed up to a year. Subcutaneous DHE was used as needed for symptomatic relief of excruciating headaches. At study completion, 78% of patients had adequately stopped analgesics. The goal of greater than 6 consecutive headache-free days was achieved in 74% patients in an average of 84 days.

A 9-week double-blind, placebo-controlled trial randomized 20 nondepressed patients with analgesic overuse headache to receive amitriptyline or active placebo (trihexyphenidyl).⁵ Patients were admitted to the hospital for 1 week and withdrawn from all analgesics. The 2 groups had similar baseline characteristics. During the hospitalization, the amitriptyline treatment group received intravenous amitriptyline escalating from 25 to 75 mg. During the following month, oral study medications were continued, and patients took low doses of aspirin or acetaminophen, as needed. There was no significant difference between the 2 groups with regard to analgesic use. At completion of this low-powered study, no difference was found between the 2 groups in headache frequency or analgesic use, although certain components of a quality-of-life scale were better in the amitriptyline group.

An open-label trial of patients with chronic migraine and analgesic overuse in a headache sub-specialty center abruptly withdrew 150 participants from analgesics and quasi-randomized them to 3 groups: prednisone (tapering from 60 to 20 mg over 6 days), naratriptan (Amerge) (2.5 mg twice daily for 6 days), or no prophylactic treatment.⁶ Patients given the active substances were told it would reduce withdrawal symptoms; patients given placebo were not given this advice. All patients received education about the pathophysiology of rebound headaches, kept a headache diary, and were phoned weekly to ensure compliance. In addition, they all received capsules containing gradually increasing doses of atenolol, nortriptyline, and flunarazine (a calcium channel blocker not FDA-approved.) Indo-methacin and chlorpromazine were used as needed. Results from the first 6 days showed no difference in headaches between the 3 groups; however, significantly more patients used chlorpromazine in the “no pharmacologic treatment” group

By the end of 5 weeks, headache frequency was significantly reduced in all groups from baseline; however, there were no differences between groups in headache frequency or intensity in this small study. Of note, there were statistically fewer withdrawal symptoms and less use of rescue medications among patients who received the initial prophylactic treatments. The indomethacin rescue use was 24%, 18%, and 14% of patients for the no prophylactic treatment, prednisone, and naratriptan groups respectively, while chlorpromazine rescue use was 14%, 0%, and 0%, respectively. The number of patients needed to treat to prevent any withdrawal symptoms (nausea, vomiting, nervousness, dizziness, etc.) was 1 for every 3.5 for naratriptan, and 6.4 for prednisone.

Recommendations from others

The American Council for Headache Education recommends discontinuing all analgesics.⁷ It notes some patients may need prophylactic medication (although no specific agent is recommended), and hospitalization may be indicated for withdrawal for patients who have abused narcotics. A headache textbook recommends 1 of 2 approaches for patients undergoing outpatient treatment: (1) gradual tapering of the offending medication with substitution of a long-acting nonsteroidal anti-inflammatory drug (NSAID) and initiation of preventive therapy, or (2) abrupt discontinuation of the offending medication and initiation followed by gradual tapering of a “transitional” medication such as NSAIDs, DHE, corticosteroids, or triptans. The authors recommend an intravenous DHE protocol for treatment failures and patients requiring inpatient treatment.⁸

Evidence summary

A prospective study of 1128 children identified as anemic with a screening hemoglobin level showed that subsequent testing—which included mean corpuscular volume, protoporphyrin, transferrin, and ferritin measurements—did not reliably distinguish potential responders from nonresponders to a 3-month trial of empiric iron therapy.¹ In fact, more than half of the responders would have been missed if treatment had been restricted to infants with abnormal mean corpuscular volume or iron studies.

Because of the simplicity, low cost, and relative safety of iron therapy for infants, this trial suggests that a therapeutic trial of iron be given first, reserving further work-up for the small number of infants that still have unexplained hemoglobin concentrations of <11.0 g/dL after a therapeutic trial. Similar results were found in a prospective controlled treatment trial among Alaskan Native children² as well as a trial of empiric iron therapy among infants with anemia.³

Another prospective study of 970 healthy infants identified 62 infants with a heel-stick capillary hematocrit of <33%. Of these, 31 had repeat hematocrit values of <33% as confirmed by subsequent heel-stick complete blood count measurement. Twenty of these anemic infants (65%) completed the study protocol, which included a 1-month trial of iron, a follow-up complete blood count, and hemoglobin electrophoresis for those infants with persistent microcytosis or positive sickle preparation (performed at initial screening for all African American infants). Six infants (30%) had an increase in hemoglobin concentration of 1.0 g/dL or more and were presumed to be iron-deficient; they went on to receive an additional 2 months of iron therapy. Two of these were found to have co-existing alpha-thalassemia. Of the remainder, 11 (55%) were determined to have a low-normal hematocrit (mean=31.5 ± 0.9), 1 had alpha thalassemia alone, 1 had coexisting alpha-thalassemia and hemoglobin AS, and 1 had hemoglobin SC. Review of data showed that abnormal diagnoses (iron deficiency, thalassemia, and sickle cell trait or disease) were found in 9 of 11 infants with high RDW and in none of the 9 with normal RDW. The authors concluded that RDW alone appears to be predictive of identifiable causes of anemia when used to screen healthy 12-month-old babies.⁴

A recent Cochrane review suggests there is a clinically significant benefit for the treatment of iron-deficiency anemia; however, there is a need for further randomized controlled trials with long-term follow-up.⁵ A randomized controlled trial of iron supplementation vs placebo in 278 infants testing positive for iron-deficiency anemia demonstrated that once daily, moderate-dose ferrous sulfate (FeSO₄) therapy (3 mg/kg/d of elemental iron) given to fasting 1-year-old infants results in no more gastrointestinal side effects than placebo therapy.⁶ Another study demonstrated that iron sulfate drops (40 mg elemental iron divided 3 times a day) or a single daily dose of microencapsulated ferrous fumarate sprinkles (80 mg elemental iron) plus ascorbic acid resulted in a similar rate of successful treatment of anemia without side effects.⁷

In a retrospective cohort study⁸ of 1358 innercity children aged 9 to 36 months who underwent screening, 343 (25%) had anemia (Hgb <11 g/dL); of these, 239 (72%) were prescribed iron and 95 (28%) were not. Responders were defined as those with a hemoglobin value of greater than 11 g/dL or an increase of 1 g/dL documented within 6 months of the initial screening visit. Follow-up rates for both groups were low (~50%), but of those prescribed iron, 107 of 150 (71%) responded to treatment compared with 27 of 48 (68%) of those who did not receive iron. Since similar response rates were seen among infants who did and infants who did not receive iron therapy, proving the benefit of routine screening followed by a trial of iron may be problematic in populations with higher rates of anemia, low follow-up rates, and high spontaneous resolution rates.

Recommendations from others

The United States Preventive Services Task Force,⁹ American Academy of Family Physicians,¹⁰ and American Academy of Pediatrics¹¹ recommend screening infants for iron-deficiency anemia but do not address appropriate follow-up for positive screens.

The Centers for Disease Control and Prevention (CDC) guidelines recommend performing a confirmatory hemoglobin and hematocrit after a positive anemia screening. If anemia is confirmed and the child is not ill, then treat with iron replacement (3 mg elemental iron/kg/daily) for 4 weeks followed by a repeat test. An increase in hemoglobin concentration ≥1 g/dL or in hematocrit ≥3% confirms the diagnosis of iron-deficiency anemia. If iron-deficiency anemia is confirmed, they recommend continuing iron therapy for 2 more months (3 months total treatment), and rechecking hemoglobin or hematocrit 6 months after successful treatment is completed. Nonresponders, despite compliance with the iron supplementation regimen and the absence of acute illness, should undergo further evaluation including mean corpuscular volume, RDW, and serum ferritin concentration.¹²

Evidence summary

A prospective study of 1128 children identified as anemic with a screening hemoglobin level showed that subsequent testing—which included mean corpuscular volume, protoporphyrin, transferrin, and ferritin measurements—did not reliably distinguish potential responders from nonresponders to a 3-month trial of empiric iron therapy.¹ In fact, more than half of the responders would have been missed if treatment had been restricted to infants with abnormal mean corpuscular volume or iron studies.

Because of the simplicity, low cost, and relative safety of iron therapy for infants, this trial suggests that a therapeutic trial of iron be given first, reserving further work-up for the small number of infants that still have unexplained hemoglobin concentrations of <11.0 g/dL after a therapeutic trial. Similar results were found in a prospective controlled treatment trial among Alaskan Native children² as well as a trial of empiric iron therapy among infants with anemia.³

Another prospective study of 970 healthy infants identified 62 infants with a heel-stick capillary hematocrit of <33%. Of these, 31 had repeat hematocrit values of <33% as confirmed by subsequent heel-stick complete blood count measurement. Twenty of these anemic infants (65%) completed the study protocol, which included a 1-month trial of iron, a follow-up complete blood count, and hemoglobin electrophoresis for those infants with persistent microcytosis or positive sickle preparation (performed at initial screening for all African American infants). Six infants (30%) had an increase in hemoglobin concentration of 1.0 g/dL or more and were presumed to be iron-deficient; they went on to receive an additional 2 months of iron therapy. Two of these were found to have co-existing alpha-thalassemia. Of the remainder, 11 (55%) were determined to have a low-normal hematocrit (mean=31.5 ± 0.9), 1 had alpha thalassemia alone, 1 had coexisting alpha-thalassemia and hemoglobin AS, and 1 had hemoglobin SC. Review of data showed that abnormal diagnoses (iron deficiency, thalassemia, and sickle cell trait or disease) were found in 9 of 11 infants with high RDW and in none of the 9 with normal RDW. The authors concluded that RDW alone appears to be predictive of identifiable causes of anemia when used to screen healthy 12-month-old babies.⁴

A recent Cochrane review suggests there is a clinically significant benefit for the treatment of iron-deficiency anemia; however, there is a need for further randomized controlled trials with long-term follow-up.⁵ A randomized controlled trial of iron supplementation vs placebo in 278 infants testing positive for iron-deficiency anemia demonstrated that once daily, moderate-dose ferrous sulfate (FeSO₄) therapy (3 mg/kg/d of elemental iron) given to fasting 1-year-old infants results in no more gastrointestinal side effects than placebo therapy.⁶ Another study demonstrated that iron sulfate drops (40 mg elemental iron divided 3 times a day) or a single daily dose of microencapsulated ferrous fumarate sprinkles (80 mg elemental iron) plus ascorbic acid resulted in a similar rate of successful treatment of anemia without side effects.⁷

In a retrospective cohort study⁸ of 1358 innercity children aged 9 to 36 months who underwent screening, 343 (25%) had anemia (Hgb <11 g/dL); of these, 239 (72%) were prescribed iron and 95 (28%) were not. Responders were defined as those with a hemoglobin value of greater than 11 g/dL or an increase of 1 g/dL documented within 6 months of the initial screening visit. Follow-up rates for both groups were low (~50%), but of those prescribed iron, 107 of 150 (71%) responded to treatment compared with 27 of 48 (68%) of those who did not receive iron. Since similar response rates were seen among infants who did and infants who did not receive iron therapy, proving the benefit of routine screening followed by a trial of iron may be problematic in populations with higher rates of anemia, low follow-up rates, and high spontaneous resolution rates.

Recommendations from others

The United States Preventive Services Task Force,⁹ American Academy of Family Physicians,¹⁰ and American Academy of Pediatrics¹¹ recommend screening infants for iron-deficiency anemia but do not address appropriate follow-up for positive screens.

The Centers for Disease Control and Prevention (CDC) guidelines recommend performing a confirmatory hemoglobin and hematocrit after a positive anemia screening. If anemia is confirmed and the child is not ill, then treat with iron replacement (3 mg elemental iron/kg/daily) for 4 weeks followed by a repeat test. An increase in hemoglobin concentration ≥1 g/dL or in hematocrit ≥3% confirms the diagnosis of iron-deficiency anemia. If iron-deficiency anemia is confirmed, they recommend continuing iron therapy for 2 more months (3 months total treatment), and rechecking hemoglobin or hematocrit 6 months after successful treatment is completed. Nonresponders, despite compliance with the iron supplementation regimen and the absence of acute illness, should undergo further evaluation including mean corpuscular volume, RDW, and serum ferritin concentration.¹²

Evidence summary

A prospective study of 1128 children identified as anemic with a screening hemoglobin level showed that subsequent testing—which included mean corpuscular volume, protoporphyrin, transferrin, and ferritin measurements—did not reliably distinguish potential responders from nonresponders to a 3-month trial of empiric iron therapy.¹ In fact, more than half of the responders would have been missed if treatment had been restricted to infants with abnormal mean corpuscular volume or iron studies.

Because of the simplicity, low cost, and relative safety of iron therapy for infants, this trial suggests that a therapeutic trial of iron be given first, reserving further work-up for the small number of infants that still have unexplained hemoglobin concentrations of <11.0 g/dL after a therapeutic trial. Similar results were found in a prospective controlled treatment trial among Alaskan Native children² as well as a trial of empiric iron therapy among infants with anemia.³

Another prospective study of 970 healthy infants identified 62 infants with a heel-stick capillary hematocrit of <33%. Of these, 31 had repeat hematocrit values of <33% as confirmed by subsequent heel-stick complete blood count measurement. Twenty of these anemic infants (65%) completed the study protocol, which included a 1-month trial of iron, a follow-up complete blood count, and hemoglobin electrophoresis for those infants with persistent microcytosis or positive sickle preparation (performed at initial screening for all African American infants). Six infants (30%) had an increase in hemoglobin concentration of 1.0 g/dL or more and were presumed to be iron-deficient; they went on to receive an additional 2 months of iron therapy. Two of these were found to have co-existing alpha-thalassemia. Of the remainder, 11 (55%) were determined to have a low-normal hematocrit (mean=31.5 ± 0.9), 1 had alpha thalassemia alone, 1 had coexisting alpha-thalassemia and hemoglobin AS, and 1 had hemoglobin SC. Review of data showed that abnormal diagnoses (iron deficiency, thalassemia, and sickle cell trait or disease) were found in 9 of 11 infants with high RDW and in none of the 9 with normal RDW. The authors concluded that RDW alone appears to be predictive of identifiable causes of anemia when used to screen healthy 12-month-old babies.⁴

A recent Cochrane review suggests there is a clinically significant benefit for the treatment of iron-deficiency anemia; however, there is a need for further randomized controlled trials with long-term follow-up.⁵ A randomized controlled trial of iron supplementation vs placebo in 278 infants testing positive for iron-deficiency anemia demonstrated that once daily, moderate-dose ferrous sulfate (FeSO₄) therapy (3 mg/kg/d of elemental iron) given to fasting 1-year-old infants results in no more gastrointestinal side effects than placebo therapy.⁶ Another study demonstrated that iron sulfate drops (40 mg elemental iron divided 3 times a day) or a single daily dose of microencapsulated ferrous fumarate sprinkles (80 mg elemental iron) plus ascorbic acid resulted in a similar rate of successful treatment of anemia without side effects.⁷

In a retrospective cohort study⁸ of 1358 innercity children aged 9 to 36 months who underwent screening, 343 (25%) had anemia (Hgb <11 g/dL); of these, 239 (72%) were prescribed iron and 95 (28%) were not. Responders were defined as those with a hemoglobin value of greater than 11 g/dL or an increase of 1 g/dL documented within 6 months of the initial screening visit. Follow-up rates for both groups were low (~50%), but of those prescribed iron, 107 of 150 (71%) responded to treatment compared with 27 of 48 (68%) of those who did not receive iron. Since similar response rates were seen among infants who did and infants who did not receive iron therapy, proving the benefit of routine screening followed by a trial of iron may be problematic in populations with higher rates of anemia, low follow-up rates, and high spontaneous resolution rates.

Recommendations from others

The United States Preventive Services Task Force,⁹ American Academy of Family Physicians,¹⁰ and American Academy of Pediatrics¹¹ recommend screening infants for iron-deficiency anemia but do not address appropriate follow-up for positive screens.

The Centers for Disease Control and Prevention (CDC) guidelines recommend performing a confirmatory hemoglobin and hematocrit after a positive anemia screening. If anemia is confirmed and the child is not ill, then treat with iron replacement (3 mg elemental iron/kg/daily) for 4 weeks followed by a repeat test. An increase in hemoglobin concentration ≥1 g/dL or in hematocrit ≥3% confirms the diagnosis of iron-deficiency anemia. If iron-deficiency anemia is confirmed, they recommend continuing iron therapy for 2 more months (3 months total treatment), and rechecking hemoglobin or hematocrit 6 months after successful treatment is completed. Nonresponders, despite compliance with the iron supplementation regimen and the absence of acute illness, should undergo further evaluation including mean corpuscular volume, RDW, and serum ferritin concentration.¹²

TABLE
Evidence-based use of C-reactive protein in cardiovascular disease

Evidence summary

C-reactive protein is a nonspecific serum marker of inflammatory response. While it is elevated in a variety of conditions, a link has been suggested between CRP and pathogenesis of clinical cardiovascular disease.¹

Several retrospective studies have reported risk ratios for developing cardiovascular disease, ranging from 2.3 to 4.4 when comparing subjects with the highest levels of hs-CRP with those who have the lowest levels.^2-9 Though systematic bias in retrospective study design limits the interpretation of these findings, the findings are of some benefit to answering this question when large, prospective, randomized studies are not available.

One of the largest and most recent of these studies reports adjusted odds for development of coronary artery disease of 1.45 (95% confidence interval [CI], 1.25–1.68) for subjects in the top third of hs-CRP levels compared with those in the bottom third.⁹ Odds ratios (OR) for other predictors of coronary artery disease are higher than this, in particular total cholesterol (OR=2.35; 95% CI, 2.03–2.74), cigarette smoking (OR=1.87; 95% CI, 1.62–2.22), and elevated systolic blood pressure (OR=1.50; 95% CI, 1.30–1.73). This shows that hs-CRP does not contribute as much as these factors to the established risk profile for coronary heart disease.

These same authors go on to provide a systematic review of 22 prospective studies of hs-CRP involving 7068 patients, which showed that an elevated hs-CRP was associated with higher odds of developing coronary artery disease (OR=1.58; 95% CI, 1.48–1.68). They also examined the largest 4 studies in their review (which included 4107 cases) and found a slightly lower OR of 1.49 (95% CI, 1.37–1.62). This meta-analysis included only studies published since 2000 because earlier studies, which had yielded higher odds for hs-CRP, suggested a pattern consistent with publication bias.

Two very recent studies evaluating statin therapy for CVD suggest that CRP may be monitored as an independent factor for predicting CVD outcomes for patients undergoing aggressive lipid therapy.^10,11 These randomized, masked trials suggest that CRP is directly predictive of recurrent events among patients with known CVD. Its usefulness may be greatest when trying to decide whether to pursue aggressive (high-dose) statin therapy for these patients.

It is not clear whether hs-CRP is a direct, causative marker for atherosclerosis or whether it is simply a proxy marker elevated in conjunction with other known risk factors. This issue, combined with the fact that its elevation does not contribute as significantly as other risk factors, makes hs-CRP an inappropriate screening test for cardiovascular disease in the healthy adult population. If results continue to accrue supporting the relationship between statin therapy and reduction of CVD outcomes attributable to CRP, we may find that monitoring CRP levels becomes appropriate in the management of patients with known moderate or severe risk or known disease.

Recommendations from others

A consensus statement from the American Heart Association and the Centers for Disease Control and Prevention discourages use of hs-CRP for screening in the healthy adult population. It offers support for using hs-CRP for assessment of patients at medium risk levels for whom the test will alter treatment decisions.¹ Guidelines from the Institute for Clinical Systems Improvement for lipid management in adults state that, “non-traditional risk factors (C-reactive protein [CRP] and total homocysteine) have been shown to have some predictive values in screening vascular disease. The value of screening for these risk factors is not yet known.”¹²

TABLE
Evidence-based use of C-reactive protein in cardiovascular disease

Evidence summary

C-reactive protein is a nonspecific serum marker of inflammatory response. While it is elevated in a variety of conditions, a link has been suggested between CRP and pathogenesis of clinical cardiovascular disease.¹

Several retrospective studies have reported risk ratios for developing cardiovascular disease, ranging from 2.3 to 4.4 when comparing subjects with the highest levels of hs-CRP with those who have the lowest levels.^2-9 Though systematic bias in retrospective study design limits the interpretation of these findings, the findings are of some benefit to answering this question when large, prospective, randomized studies are not available.

One of the largest and most recent of these studies reports adjusted odds for development of coronary artery disease of 1.45 (95% confidence interval [CI], 1.25–1.68) for subjects in the top third of hs-CRP levels compared with those in the bottom third.⁹ Odds ratios (OR) for other predictors of coronary artery disease are higher than this, in particular total cholesterol (OR=2.35; 95% CI, 2.03–2.74), cigarette smoking (OR=1.87; 95% CI, 1.62–2.22), and elevated systolic blood pressure (OR=1.50; 95% CI, 1.30–1.73). This shows that hs-CRP does not contribute as much as these factors to the established risk profile for coronary heart disease.

These same authors go on to provide a systematic review of 22 prospective studies of hs-CRP involving 7068 patients, which showed that an elevated hs-CRP was associated with higher odds of developing coronary artery disease (OR=1.58; 95% CI, 1.48–1.68). They also examined the largest 4 studies in their review (which included 4107 cases) and found a slightly lower OR of 1.49 (95% CI, 1.37–1.62). This meta-analysis included only studies published since 2000 because earlier studies, which had yielded higher odds for hs-CRP, suggested a pattern consistent with publication bias.

Two very recent studies evaluating statin therapy for CVD suggest that CRP may be monitored as an independent factor for predicting CVD outcomes for patients undergoing aggressive lipid therapy.^10,11 These randomized, masked trials suggest that CRP is directly predictive of recurrent events among patients with known CVD. Its usefulness may be greatest when trying to decide whether to pursue aggressive (high-dose) statin therapy for these patients.

It is not clear whether hs-CRP is a direct, causative marker for atherosclerosis or whether it is simply a proxy marker elevated in conjunction with other known risk factors. This issue, combined with the fact that its elevation does not contribute as significantly as other risk factors, makes hs-CRP an inappropriate screening test for cardiovascular disease in the healthy adult population. If results continue to accrue supporting the relationship between statin therapy and reduction of CVD outcomes attributable to CRP, we may find that monitoring CRP levels becomes appropriate in the management of patients with known moderate or severe risk or known disease.

Recommendations from others

A consensus statement from the American Heart Association and the Centers for Disease Control and Prevention discourages use of hs-CRP for screening in the healthy adult population. It offers support for using hs-CRP for assessment of patients at medium risk levels for whom the test will alter treatment decisions.¹ Guidelines from the Institute for Clinical Systems Improvement for lipid management in adults state that, “non-traditional risk factors (C-reactive protein [CRP] and total homocysteine) have been shown to have some predictive values in screening vascular disease. The value of screening for these risk factors is not yet known.”¹²

TABLE
Evidence-based use of C-reactive protein in cardiovascular disease

Evidence summary

C-reactive protein is a nonspecific serum marker of inflammatory response. While it is elevated in a variety of conditions, a link has been suggested between CRP and pathogenesis of clinical cardiovascular disease.¹

Several retrospective studies have reported risk ratios for developing cardiovascular disease, ranging from 2.3 to 4.4 when comparing subjects with the highest levels of hs-CRP with those who have the lowest levels.^2-9 Though systematic bias in retrospective study design limits the interpretation of these findings, the findings are of some benefit to answering this question when large, prospective, randomized studies are not available.

One of the largest and most recent of these studies reports adjusted odds for development of coronary artery disease of 1.45 (95% confidence interval [CI], 1.25–1.68) for subjects in the top third of hs-CRP levels compared with those in the bottom third.⁹ Odds ratios (OR) for other predictors of coronary artery disease are higher than this, in particular total cholesterol (OR=2.35; 95% CI, 2.03–2.74), cigarette smoking (OR=1.87; 95% CI, 1.62–2.22), and elevated systolic blood pressure (OR=1.50; 95% CI, 1.30–1.73). This shows that hs-CRP does not contribute as much as these factors to the established risk profile for coronary heart disease.

These same authors go on to provide a systematic review of 22 prospective studies of hs-CRP involving 7068 patients, which showed that an elevated hs-CRP was associated with higher odds of developing coronary artery disease (OR=1.58; 95% CI, 1.48–1.68). They also examined the largest 4 studies in their review (which included 4107 cases) and found a slightly lower OR of 1.49 (95% CI, 1.37–1.62). This meta-analysis included only studies published since 2000 because earlier studies, which had yielded higher odds for hs-CRP, suggested a pattern consistent with publication bias.

Two very recent studies evaluating statin therapy for CVD suggest that CRP may be monitored as an independent factor for predicting CVD outcomes for patients undergoing aggressive lipid therapy.^10,11 These randomized, masked trials suggest that CRP is directly predictive of recurrent events among patients with known CVD. Its usefulness may be greatest when trying to decide whether to pursue aggressive (high-dose) statin therapy for these patients.

It is not clear whether hs-CRP is a direct, causative marker for atherosclerosis or whether it is simply a proxy marker elevated in conjunction with other known risk factors. This issue, combined with the fact that its elevation does not contribute as significantly as other risk factors, makes hs-CRP an inappropriate screening test for cardiovascular disease in the healthy adult population. If results continue to accrue supporting the relationship between statin therapy and reduction of CVD outcomes attributable to CRP, we may find that monitoring CRP levels becomes appropriate in the management of patients with known moderate or severe risk or known disease.

Recommendations from others

A consensus statement from the American Heart Association and the Centers for Disease Control and Prevention discourages use of hs-CRP for screening in the healthy adult population. It offers support for using hs-CRP for assessment of patients at medium risk levels for whom the test will alter treatment decisions.¹ Guidelines from the Institute for Clinical Systems Improvement for lipid management in adults state that, “non-traditional risk factors (C-reactive protein [CRP] and total homocysteine) have been shown to have some predictive values in screening vascular disease. The value of screening for these risk factors is not yet known.”¹²

Evidence summary

Three forms of niacin exist: immediate-release (IR), sustained-release/long-acting (SR/LA), and extended-release (ER), which is currently available only as Niaspan.¹ Published incidence of niacin-induced hepatotoxicity varies according to the definition of hepatotoxicity, with a 0% to 46% rate of elevated hepatic enzymes. Hepatotoxicity includes mild liver enzyme elevations, steatosis, hepatitis, abnormal liver biopsies, or fulminant hepatic failure.^2,3 Between 1982 and 1992, 11 case reports have linked IR nicotinic acid to a wide range of hepatotoxicities. For patients taking LA/SR niacin doses ≥3 g/d or switching from the IR to the LA product, 21 case reports have linked LA/SR niacin with adverse outcomes.^3,4 In several of the LA/SR cases, patients were rechal-lenged with IR formulations with no recurrent hepatocellular damage.^3,4 In these case reports, onset of hepatotoxicity ranged from 2 days to 18 months. In a retrospective cohort of 969 veterans taking LA/SR niacin, those who developed hepatotoxicity had onset between 1 and 28 months of initiating treatment.² Studies evaluating the risk of hepatotoxicity with niacin alone and in combination with statins are summarized in the Table .

Because LA/SR niacin has an active metabolite (nicotinamide), hepatotoxicity is more likely to occur with the LA/SR formulation than with IR niacin.³ In a small prospective comparative study of IR and LA/SR niacin (n=46), 0/23 patients taking IR niacin exhibited hepatic toxicity, compared with 12/23 (52%) of patients taking the LA/SR formulation.⁵ In this study, patients receiving 1 g/d of LA/SR niacin had increases in transaminases similar to those of patients on 3 g/d of IR niacin. It is therefore recommended that if a patient cannot tolerate IR niacin and is switched to the LA/SR form, the dosage be reduced by 50% to 70%.⁵ At doses >2 g/d of LA/SR niacin, mean transaminases approached 3 times the upper limit of normal (ULN), supporting recommendations not to exceed this dose for LA/SR niacin.⁵

Several LA/SR products exist, and their differing pharmacologic and clinical properties necessitate monitoring as though starting anew when changing from one LA/SR formulation to another.¹ Because of the unfavorable risk-benefit ratio of LA/SR formulations compared with other niacin formulations, production and marketing of many LA/SR niacin brands has ceased. The ER formulation (Niaspan), only available by prescription, has a balanced metabolism resulting in less hepatotoxicity (<1%).^1,6 Expert opinion mandates continued annual monitoring of liver function tests (LFT) for all patients, including those on a stable ER niacin dose, no new risk factors for hepatotoxicity, and a series of normal LFTs.⁷

TABLE
Studies of niacin toxicity

Recommendations from others

Elevated hepatic enzymes <3 times the ULN may occur but usually resolve with continued therapy or reduced doses. Enzymes >3 times the ULN require discontinuation of therapy.⁸ The American Society of Health-System Pharmacists (ASHP) recommends screening at baseline, every 2 to 3 months for the first year and every 6 to 12 months there-after.⁸ The ASHP also recommends that patients be started on IR niacin products, with consideration of ER products only when IR products are not tolerated or alternative products are ineffective. ASHP makes no mention of LA/SR products in their recommendations.⁸ They recommend more frequent monitoring for high-risk patients—risks include doses >2 g/d for LA/SR and >3 g/d for IR; LA/SR formulations; switching between formulations; taking concomitant drugs that interact (ie, sulfonylureas); excessive alcohol use (undefined); and preexisting liver disease (based on a bivariate analysis of factors associated with increased risk of hepatic toxicity from a single retrospective cohort study)⁵ — and for patients who demonstrate signs/symptoms of toxicity (nausea, vomiting, malaise, loss of appetite, right upper quadrant pain, jaundice, and dark urine).⁸ The National Cholesterol Education Program Expert Panel update in 2004 recommended obtaining ALT/AST initially, 6 to 8 weeks after reaching a daily dose of 1500 mg, 6 to 8 weeks after reaching the maximum daily dose, then annually or more frequently if indicated.⁷

Evidence summary

Three forms of niacin exist: immediate-release (IR), sustained-release/long-acting (SR/LA), and extended-release (ER), which is currently available only as Niaspan.¹ Published incidence of niacin-induced hepatotoxicity varies according to the definition of hepatotoxicity, with a 0% to 46% rate of elevated hepatic enzymes. Hepatotoxicity includes mild liver enzyme elevations, steatosis, hepatitis, abnormal liver biopsies, or fulminant hepatic failure.^2,3 Between 1982 and 1992, 11 case reports have linked IR nicotinic acid to a wide range of hepatotoxicities. For patients taking LA/SR niacin doses ≥3 g/d or switching from the IR to the LA product, 21 case reports have linked LA/SR niacin with adverse outcomes.^3,4 In several of the LA/SR cases, patients were rechal-lenged with IR formulations with no recurrent hepatocellular damage.^3,4 In these case reports, onset of hepatotoxicity ranged from 2 days to 18 months. In a retrospective cohort of 969 veterans taking LA/SR niacin, those who developed hepatotoxicity had onset between 1 and 28 months of initiating treatment.² Studies evaluating the risk of hepatotoxicity with niacin alone and in combination with statins are summarized in the Table .

Because LA/SR niacin has an active metabolite (nicotinamide), hepatotoxicity is more likely to occur with the LA/SR formulation than with IR niacin.³ In a small prospective comparative study of IR and LA/SR niacin (n=46), 0/23 patients taking IR niacin exhibited hepatic toxicity, compared with 12/23 (52%) of patients taking the LA/SR formulation.⁵ In this study, patients receiving 1 g/d of LA/SR niacin had increases in transaminases similar to those of patients on 3 g/d of IR niacin. It is therefore recommended that if a patient cannot tolerate IR niacin and is switched to the LA/SR form, the dosage be reduced by 50% to 70%.⁵ At doses >2 g/d of LA/SR niacin, mean transaminases approached 3 times the upper limit of normal (ULN), supporting recommendations not to exceed this dose for LA/SR niacin.⁵

Several LA/SR products exist, and their differing pharmacologic and clinical properties necessitate monitoring as though starting anew when changing from one LA/SR formulation to another.¹ Because of the unfavorable risk-benefit ratio of LA/SR formulations compared with other niacin formulations, production and marketing of many LA/SR niacin brands has ceased. The ER formulation (Niaspan), only available by prescription, has a balanced metabolism resulting in less hepatotoxicity (<1%).^1,6 Expert opinion mandates continued annual monitoring of liver function tests (LFT) for all patients, including those on a stable ER niacin dose, no new risk factors for hepatotoxicity, and a series of normal LFTs.⁷

TABLE
Studies of niacin toxicity

Recommendations from others

Elevated hepatic enzymes <3 times the ULN may occur but usually resolve with continued therapy or reduced doses. Enzymes >3 times the ULN require discontinuation of therapy.⁸ The American Society of Health-System Pharmacists (ASHP) recommends screening at baseline, every 2 to 3 months for the first year and every 6 to 12 months there-after.⁸ The ASHP also recommends that patients be started on IR niacin products, with consideration of ER products only when IR products are not tolerated or alternative products are ineffective. ASHP makes no mention of LA/SR products in their recommendations.⁸ They recommend more frequent monitoring for high-risk patients—risks include doses >2 g/d for LA/SR and >3 g/d for IR; LA/SR formulations; switching between formulations; taking concomitant drugs that interact (ie, sulfonylureas); excessive alcohol use (undefined); and preexisting liver disease (based on a bivariate analysis of factors associated with increased risk of hepatic toxicity from a single retrospective cohort study)⁵ — and for patients who demonstrate signs/symptoms of toxicity (nausea, vomiting, malaise, loss of appetite, right upper quadrant pain, jaundice, and dark urine).⁸ The National Cholesterol Education Program Expert Panel update in 2004 recommended obtaining ALT/AST initially, 6 to 8 weeks after reaching a daily dose of 1500 mg, 6 to 8 weeks after reaching the maximum daily dose, then annually or more frequently if indicated.⁷

Evidence summary

Three forms of niacin exist: immediate-release (IR), sustained-release/long-acting (SR/LA), and extended-release (ER), which is currently available only as Niaspan.¹ Published incidence of niacin-induced hepatotoxicity varies according to the definition of hepatotoxicity, with a 0% to 46% rate of elevated hepatic enzymes. Hepatotoxicity includes mild liver enzyme elevations, steatosis, hepatitis, abnormal liver biopsies, or fulminant hepatic failure.^2,3 Between 1982 and 1992, 11 case reports have linked IR nicotinic acid to a wide range of hepatotoxicities. For patients taking LA/SR niacin doses ≥3 g/d or switching from the IR to the LA product, 21 case reports have linked LA/SR niacin with adverse outcomes.^3,4 In several of the LA/SR cases, patients were rechal-lenged with IR formulations with no recurrent hepatocellular damage.^3,4 In these case reports, onset of hepatotoxicity ranged from 2 days to 18 months. In a retrospective cohort of 969 veterans taking LA/SR niacin, those who developed hepatotoxicity had onset between 1 and 28 months of initiating treatment.² Studies evaluating the risk of hepatotoxicity with niacin alone and in combination with statins are summarized in the Table .

Because LA/SR niacin has an active metabolite (nicotinamide), hepatotoxicity is more likely to occur with the LA/SR formulation than with IR niacin.³ In a small prospective comparative study of IR and LA/SR niacin (n=46), 0/23 patients taking IR niacin exhibited hepatic toxicity, compared with 12/23 (52%) of patients taking the LA/SR formulation.⁵ In this study, patients receiving 1 g/d of LA/SR niacin had increases in transaminases similar to those of patients on 3 g/d of IR niacin. It is therefore recommended that if a patient cannot tolerate IR niacin and is switched to the LA/SR form, the dosage be reduced by 50% to 70%.⁵ At doses >2 g/d of LA/SR niacin, mean transaminases approached 3 times the upper limit of normal (ULN), supporting recommendations not to exceed this dose for LA/SR niacin.⁵

Several LA/SR products exist, and their differing pharmacologic and clinical properties necessitate monitoring as though starting anew when changing from one LA/SR formulation to another.¹ Because of the unfavorable risk-benefit ratio of LA/SR formulations compared with other niacin formulations, production and marketing of many LA/SR niacin brands has ceased. The ER formulation (Niaspan), only available by prescription, has a balanced metabolism resulting in less hepatotoxicity (<1%).^1,6 Expert opinion mandates continued annual monitoring of liver function tests (LFT) for all patients, including those on a stable ER niacin dose, no new risk factors for hepatotoxicity, and a series of normal LFTs.⁷

TABLE
Studies of niacin toxicity

Recommendations from others

Elevated hepatic enzymes <3 times the ULN may occur but usually resolve with continued therapy or reduced doses. Enzymes >3 times the ULN require discontinuation of therapy.⁸ The American Society of Health-System Pharmacists (ASHP) recommends screening at baseline, every 2 to 3 months for the first year and every 6 to 12 months there-after.⁸ The ASHP also recommends that patients be started on IR niacin products, with consideration of ER products only when IR products are not tolerated or alternative products are ineffective. ASHP makes no mention of LA/SR products in their recommendations.⁸ They recommend more frequent monitoring for high-risk patients—risks include doses >2 g/d for LA/SR and >3 g/d for IR; LA/SR formulations; switching between formulations; taking concomitant drugs that interact (ie, sulfonylureas); excessive alcohol use (undefined); and preexisting liver disease (based on a bivariate analysis of factors associated with increased risk of hepatic toxicity from a single retrospective cohort study)⁵ — and for patients who demonstrate signs/symptoms of toxicity (nausea, vomiting, malaise, loss of appetite, right upper quadrant pain, jaundice, and dark urine).⁸ The National Cholesterol Education Program Expert Panel update in 2004 recommended obtaining ALT/AST initially, 6 to 8 weeks after reaching a daily dose of 1500 mg, 6 to 8 weeks after reaching the maximum daily dose, then annually or more frequently if indicated.⁷

Evidence summary

Two randomized controlled trials have compared the efficacy of oral metronidazole and oral vancomycin for treatment of CDAD.^1,2 Both studies demonstrated statistically equivalent cure rates exceeding 90%, with relapse rates of 10% to 20% for each drug. These small trials lacked the power to detect small but potentially significant differences in treatment response.

No published data exist indicating that vancomycin is more effective than metronidazole in any clinical setting. A dose-range study showed that 125 mg of oral vancomycin 4 times a day is as effective as higher doses.³ Patients who cannot take medication by mouth should receive intravenous metronidazole, 500 mg 4 times per day. Unlike vancomycin, metronidazole achieves potentially effective concentrations in the intestinal lumen following intravenous administration.⁴

Treatment of first recurrences of infection with metronidazole or vancomycin produces response rates similar to treatment of initial infections.⁵ A minority of patients suffers multiple relapses of infection, and there are few data to guide therapy in this setting.

A randomized, double-blinded, placebocontrolled study evaluated the impact of adding the probiotic agent Saccharomyces boulardii. to either metronidazole or vancomycin.⁶ For persons with recurrent infection, addition of S boulardii. led to a 30% decrease in the absolute risk of relapse (64% relapse vs 34%; number needed to treat=3; P.<.05). There was also a nonsignificant trend toward reduced recurrences in the treatment of primary infections. The 2 minor side effects noted with this treatment were dry mouth (number needed to harm [NNH]=11) and constipation (NNH=9). S boulardii. capsules are available from health food stores and via the Internet. Several published case series describe various additional approaches to therapy of recurrent CDAD (Table).

TABLE
Medical treatment of C difficile–.associated diarrhea

Recommendations from others

The American College of Gastroenterology and the American College of Physicians treatment guidelines for CDAD both call for treatment with oral metronidazole 250 mg 4 times daily or 500 mg 3 times daily.^7,8 The American College of Gastroenterology recommends vancomycin (125 mg orally 4 times daily) when there is an intolerance or confirmed resistance to metronidazole, failure of response, when the patient is pregnant, breast feeding, or under 10 years of age, critically ill from colitis, or when the diarrhea could be related to Staphylococcus aureus.. In milder cases, treatment may involve only discontinuation of antibiotics and supportive therapy with observation. Opiates and antispasmodics should be avoided. These guidelines do not recommend any treatment over another for therapy of multiple recurrences.

Evidence summary

Two randomized controlled trials have compared the efficacy of oral metronidazole and oral vancomycin for treatment of CDAD.^1,2 Both studies demonstrated statistically equivalent cure rates exceeding 90%, with relapse rates of 10% to 20% for each drug. These small trials lacked the power to detect small but potentially significant differences in treatment response.

No published data exist indicating that vancomycin is more effective than metronidazole in any clinical setting. A dose-range study showed that 125 mg of oral vancomycin 4 times a day is as effective as higher doses.³ Patients who cannot take medication by mouth should receive intravenous metronidazole, 500 mg 4 times per day. Unlike vancomycin, metronidazole achieves potentially effective concentrations in the intestinal lumen following intravenous administration.⁴

Treatment of first recurrences of infection with metronidazole or vancomycin produces response rates similar to treatment of initial infections.⁵ A minority of patients suffers multiple relapses of infection, and there are few data to guide therapy in this setting.

A randomized, double-blinded, placebocontrolled study evaluated the impact of adding the probiotic agent Saccharomyces boulardii. to either metronidazole or vancomycin.⁶ For persons with recurrent infection, addition of S boulardii. led to a 30% decrease in the absolute risk of relapse (64% relapse vs 34%; number needed to treat=3; P.<.05). There was also a nonsignificant trend toward reduced recurrences in the treatment of primary infections. The 2 minor side effects noted with this treatment were dry mouth (number needed to harm [NNH]=11) and constipation (NNH=9). S boulardii. capsules are available from health food stores and via the Internet. Several published case series describe various additional approaches to therapy of recurrent CDAD (Table).

TABLE
Medical treatment of C difficile–.associated diarrhea

Recommendations from others

The American College of Gastroenterology and the American College of Physicians treatment guidelines for CDAD both call for treatment with oral metronidazole 250 mg 4 times daily or 500 mg 3 times daily.^7,8 The American College of Gastroenterology recommends vancomycin (125 mg orally 4 times daily) when there is an intolerance or confirmed resistance to metronidazole, failure of response, when the patient is pregnant, breast feeding, or under 10 years of age, critically ill from colitis, or when the diarrhea could be related to Staphylococcus aureus.. In milder cases, treatment may involve only discontinuation of antibiotics and supportive therapy with observation. Opiates and antispasmodics should be avoided. These guidelines do not recommend any treatment over another for therapy of multiple recurrences.

Evidence summary

Two randomized controlled trials have compared the efficacy of oral metronidazole and oral vancomycin for treatment of CDAD.^1,2 Both studies demonstrated statistically equivalent cure rates exceeding 90%, with relapse rates of 10% to 20% for each drug. These small trials lacked the power to detect small but potentially significant differences in treatment response.

No published data exist indicating that vancomycin is more effective than metronidazole in any clinical setting. A dose-range study showed that 125 mg of oral vancomycin 4 times a day is as effective as higher doses.³ Patients who cannot take medication by mouth should receive intravenous metronidazole, 500 mg 4 times per day. Unlike vancomycin, metronidazole achieves potentially effective concentrations in the intestinal lumen following intravenous administration.⁴

Treatment of first recurrences of infection with metronidazole or vancomycin produces response rates similar to treatment of initial infections.⁵ A minority of patients suffers multiple relapses of infection, and there are few data to guide therapy in this setting.

A randomized, double-blinded, placebocontrolled study evaluated the impact of adding the probiotic agent Saccharomyces boulardii. to either metronidazole or vancomycin.⁶ For persons with recurrent infection, addition of S boulardii. led to a 30% decrease in the absolute risk of relapse (64% relapse vs 34%; number needed to treat=3; P.<.05). There was also a nonsignificant trend toward reduced recurrences in the treatment of primary infections. The 2 minor side effects noted with this treatment were dry mouth (number needed to harm [NNH]=11) and constipation (NNH=9). S boulardii. capsules are available from health food stores and via the Internet. Several published case series describe various additional approaches to therapy of recurrent CDAD (Table).

TABLE
Medical treatment of C difficile–.associated diarrhea

Recommendations from others

The American College of Gastroenterology and the American College of Physicians treatment guidelines for CDAD both call for treatment with oral metronidazole 250 mg 4 times daily or 500 mg 3 times daily.^7,8 The American College of Gastroenterology recommends vancomycin (125 mg orally 4 times daily) when there is an intolerance or confirmed resistance to metronidazole, failure of response, when the patient is pregnant, breast feeding, or under 10 years of age, critically ill from colitis, or when the diarrhea could be related to Staphylococcus aureus.. In milder cases, treatment may involve only discontinuation of antibiotics and supportive therapy with observation. Opiates and antispasmodics should be avoided. These guidelines do not recommend any treatment over another for therapy of multiple recurrences.

Evidence summary

Prevalence and incidence. In general, studies are limited by variations in analytic techniques and differences in defining the lower limit for normal serum magnesium.¹ Estimates of the prevalence of hypomagnesemia in the general population range from 2.5% to 15%. A study of 11,000 white urban Americans aged 45 to 64 years (probability sampling) found 2.5% with magnesium <0.7 mmol/L and 5% with magnesium <0.75 mmol/L; rates for 4000 African Americans were twice as high.²

Some authors have proposed a higher range for normal serum magnesium, asserting that dietary magnesium deficiency is endemic in developed countries where acid rain reduces the magnesium content of crops and food processing causes further large reductions in the magnesium content of the diet.¹ Moreover, common diseases are associated with hypomagnesemia and likely contaminate studies of “normal” populations. Thus, a study of 16,000 German subjects (including blood donors, outpatients, and children) found a 14.5% prevalence of hypomagnesemia using a lower limit of 0.76 mmol/L¹; however, applying the more commonly cited lower limit of 0.70 mmol/L (1.7 mg/dL) to the same data yielded aprevalence of 2%.

Numerous studies agree that the prevalence of hypomagnesemia is much higher (10%–65%) in subpopulations defined by severity of illness (hospitalization, in intensive care unit [ICU] or pediatric ICU), increasing age (elderly/in nursing home), or specific diseases. For example, of 94 consecutive patients admitted to the ICU, 65% had hypomagnesemia.³ Likewise, for 127 consecutive patients admitted with a diagnosis of alcoholism, the prevalence was 30%.⁴

Because of limitations noted above, as well as the lack of control groups, the relative prevalence in these groups (compared with the general population) is uncertain, but the studies do identify high-risk populations. A single study, which included a control group, demonstrated an 11% prevalence of hypomagnesemia among 621 randomly selected hospitalized patients compared with 2.5% among 341 hospital employees.⁵ Other diseases associated with a high prevalence of hypomagnesemia include cardiovascular disease (hypertension, congestive heart failure, coronary artery disease), diabetes, diarrhea, diuretics use, hypokalemia, hypocalcemia, and malabsorption.^6-9

Common causes. We found no high-quality studies to establish the relative probabilities of various causes in the general population or any subpopulation.¹⁰ The most common causes of significant hypomagnesemia in developed countries are said to be diabetes, alcoholism, and the use of diuretics. In a group of 5100 consecutive patients (predominantly outpatient, middle-aged, and female) presenting to a diagnostic lab, the most common diagnoses associated with hypomagnesemia were diabetes (20% of cases) and diuretic use (14% of cases); however, other potential causes, including alcoholism, were not identified.¹¹ A complete list of causes is in the Table.

Serious causes. A critical serum magnesium level is less than 0.5 mmol/L and is associated with seizures and life-threatening arrhythmias.⁶ Very low magnesium levels typically result when an acute problem is superimposed on chronic depletion. For example, critical levels can occur among patients with diabetes during correction of ketoacidosis or alcoholics who develop vomiting, diarrhea, or pancreatitis.

Magnesium in the 0.5 to 0.7 mmol/L range may be life-threatening in certain disease contexts, such as acute myocardial infarction or congestive heart failure, where there is already a risk of fatal arrhythmia.⁸

Impact. The impact of hypomagnesemia is underestimated largely because clinicians fail to measure magnesium.¹² Since magnesium is a cofactor for more than 300 enzymes and is involved in numerous transport mechanisms, it is not surprising that hypomagnesemia is associated with significant morbidity.

For example, in a study of 381 consecutive admissions at an inner-city hospital,¹³ approximately half the admissions went to ICUs and half to regular wards. Despite similar Acute Physiology and Chronic Health Evaluator (APACHE) scores at admission, hospital mortality was twice as high for hypomagnesemic patients in both care settings.

TABLE
Causes of hypomagnesemia

Recommendations from others

Several review articles include a comprehensive differential diagnosis for causes of magnesium deficiency based on physiologic principles as listed in the Table, but none provide data on the relative frequency of the various causes in the general population or specific subgroups.^6-9

Evidence summary

Prevalence and incidence. In general, studies are limited by variations in analytic techniques and differences in defining the lower limit for normal serum magnesium.¹ Estimates of the prevalence of hypomagnesemia in the general population range from 2.5% to 15%. A study of 11,000 white urban Americans aged 45 to 64 years (probability sampling) found 2.5% with magnesium <0.7 mmol/L and 5% with magnesium <0.75 mmol/L; rates for 4000 African Americans were twice as high.²

Some authors have proposed a higher range for normal serum magnesium, asserting that dietary magnesium deficiency is endemic in developed countries where acid rain reduces the magnesium content of crops and food processing causes further large reductions in the magnesium content of the diet.¹ Moreover, common diseases are associated with hypomagnesemia and likely contaminate studies of “normal” populations. Thus, a study of 16,000 German subjects (including blood donors, outpatients, and children) found a 14.5% prevalence of hypomagnesemia using a lower limit of 0.76 mmol/L¹; however, applying the more commonly cited lower limit of 0.70 mmol/L (1.7 mg/dL) to the same data yielded aprevalence of 2%.

Numerous studies agree that the prevalence of hypomagnesemia is much higher (10%–65%) in subpopulations defined by severity of illness (hospitalization, in intensive care unit [ICU] or pediatric ICU), increasing age (elderly/in nursing home), or specific diseases. For example, of 94 consecutive patients admitted to the ICU, 65% had hypomagnesemia.³ Likewise, for 127 consecutive patients admitted with a diagnosis of alcoholism, the prevalence was 30%.⁴

Because of limitations noted above, as well as the lack of control groups, the relative prevalence in these groups (compared with the general population) is uncertain, but the studies do identify high-risk populations. A single study, which included a control group, demonstrated an 11% prevalence of hypomagnesemia among 621 randomly selected hospitalized patients compared with 2.5% among 341 hospital employees.⁵ Other diseases associated with a high prevalence of hypomagnesemia include cardiovascular disease (hypertension, congestive heart failure, coronary artery disease), diabetes, diarrhea, diuretics use, hypokalemia, hypocalcemia, and malabsorption.^6-9

Common causes. We found no high-quality studies to establish the relative probabilities of various causes in the general population or any subpopulation.¹⁰ The most common causes of significant hypomagnesemia in developed countries are said to be diabetes, alcoholism, and the use of diuretics. In a group of 5100 consecutive patients (predominantly outpatient, middle-aged, and female) presenting to a diagnostic lab, the most common diagnoses associated with hypomagnesemia were diabetes (20% of cases) and diuretic use (14% of cases); however, other potential causes, including alcoholism, were not identified.¹¹ A complete list of causes is in the Table.

Serious causes. A critical serum magnesium level is less than 0.5 mmol/L and is associated with seizures and life-threatening arrhythmias.⁶ Very low magnesium levels typically result when an acute problem is superimposed on chronic depletion. For example, critical levels can occur among patients with diabetes during correction of ketoacidosis or alcoholics who develop vomiting, diarrhea, or pancreatitis.

Magnesium in the 0.5 to 0.7 mmol/L range may be life-threatening in certain disease contexts, such as acute myocardial infarction or congestive heart failure, where there is already a risk of fatal arrhythmia.⁸

Impact. The impact of hypomagnesemia is underestimated largely because clinicians fail to measure magnesium.¹² Since magnesium is a cofactor for more than 300 enzymes and is involved in numerous transport mechanisms, it is not surprising that hypomagnesemia is associated with significant morbidity.

For example, in a study of 381 consecutive admissions at an inner-city hospital,¹³ approximately half the admissions went to ICUs and half to regular wards. Despite similar Acute Physiology and Chronic Health Evaluator (APACHE) scores at admission, hospital mortality was twice as high for hypomagnesemic patients in both care settings.

TABLE
Causes of hypomagnesemia

Recommendations from others

Several review articles include a comprehensive differential diagnosis for causes of magnesium deficiency based on physiologic principles as listed in the Table, but none provide data on the relative frequency of the various causes in the general population or specific subgroups.^6-9

Evidence summary

Prevalence and incidence. In general, studies are limited by variations in analytic techniques and differences in defining the lower limit for normal serum magnesium.¹ Estimates of the prevalence of hypomagnesemia in the general population range from 2.5% to 15%. A study of 11,000 white urban Americans aged 45 to 64 years (probability sampling) found 2.5% with magnesium <0.7 mmol/L and 5% with magnesium <0.75 mmol/L; rates for 4000 African Americans were twice as high.²

Some authors have proposed a higher range for normal serum magnesium, asserting that dietary magnesium deficiency is endemic in developed countries where acid rain reduces the magnesium content of crops and food processing causes further large reductions in the magnesium content of the diet.¹ Moreover, common diseases are associated with hypomagnesemia and likely contaminate studies of “normal” populations. Thus, a study of 16,000 German subjects (including blood donors, outpatients, and children) found a 14.5% prevalence of hypomagnesemia using a lower limit of 0.76 mmol/L¹; however, applying the more commonly cited lower limit of 0.70 mmol/L (1.7 mg/dL) to the same data yielded aprevalence of 2%.

Numerous studies agree that the prevalence of hypomagnesemia is much higher (10%–65%) in subpopulations defined by severity of illness (hospitalization, in intensive care unit [ICU] or pediatric ICU), increasing age (elderly/in nursing home), or specific diseases. For example, of 94 consecutive patients admitted to the ICU, 65% had hypomagnesemia.³ Likewise, for 127 consecutive patients admitted with a diagnosis of alcoholism, the prevalence was 30%.⁴

Because of limitations noted above, as well as the lack of control groups, the relative prevalence in these groups (compared with the general population) is uncertain, but the studies do identify high-risk populations. A single study, which included a control group, demonstrated an 11% prevalence of hypomagnesemia among 621 randomly selected hospitalized patients compared with 2.5% among 341 hospital employees.⁵ Other diseases associated with a high prevalence of hypomagnesemia include cardiovascular disease (hypertension, congestive heart failure, coronary artery disease), diabetes, diarrhea, diuretics use, hypokalemia, hypocalcemia, and malabsorption.^6-9

Common causes. We found no high-quality studies to establish the relative probabilities of various causes in the general population or any subpopulation.¹⁰ The most common causes of significant hypomagnesemia in developed countries are said to be diabetes, alcoholism, and the use of diuretics. In a group of 5100 consecutive patients (predominantly outpatient, middle-aged, and female) presenting to a diagnostic lab, the most common diagnoses associated with hypomagnesemia were diabetes (20% of cases) and diuretic use (14% of cases); however, other potential causes, including alcoholism, were not identified.¹¹ A complete list of causes is in the Table.

Serious causes. A critical serum magnesium level is less than 0.5 mmol/L and is associated with seizures and life-threatening arrhythmias.⁶ Very low magnesium levels typically result when an acute problem is superimposed on chronic depletion. For example, critical levels can occur among patients with diabetes during correction of ketoacidosis or alcoholics who develop vomiting, diarrhea, or pancreatitis.

Magnesium in the 0.5 to 0.7 mmol/L range may be life-threatening in certain disease contexts, such as acute myocardial infarction or congestive heart failure, where there is already a risk of fatal arrhythmia.⁸

Impact. The impact of hypomagnesemia is underestimated largely because clinicians fail to measure magnesium.¹² Since magnesium is a cofactor for more than 300 enzymes and is involved in numerous transport mechanisms, it is not surprising that hypomagnesemia is associated with significant morbidity.

For example, in a study of 381 consecutive admissions at an inner-city hospital,¹³ approximately half the admissions went to ICUs and half to regular wards. Despite similar Acute Physiology and Chronic Health Evaluator (APACHE) scores at admission, hospital mortality was twice as high for hypomagnesemic patients in both care settings.

TABLE
Causes of hypomagnesemia

Recommendations from others

Several review articles include a comprehensive differential diagnosis for causes of magnesium deficiency based on physiologic principles as listed in the Table, but none provide data on the relative frequency of the various causes in the general population or specific subgroups.^6-9

TABLE
Approach to monitoring of INR for long-term anticoagulation

Evidence summary

Under- or over-treatment with warfarin can result in life-threatening complications. Limited research exists to guide the selection of an interval for monitoring anticoagulation in stabilized patients. One RCT compared INR monitoring in an anticoagulation clinic at 6 weeks and 4 weeks among 124 patients with a prosthetic heart valve on stable oral anticoagulant treatment and found no difference in thromboembolic or hemorrhagic events.¹ A study in the United Kingdom used a 14-week interval for selected patients, but it used no comparison group.² Kent et al developed a computer-based model to compute the optimum interval for monitoring anticoagulation that considers the variability of the patient’s previous levels and costs associated with testing and potential complications. This model achieved a maximum interval of 11 weeks for very stable patients.³

More frequent testing results in higher time in therapeutic range, particularly when patients selfmonitor. A German study of 200 patients with prosthetic heart valves found that they tested within a therapeutic range 48% of the time when monitored by their physician “as usual” (average interval 24 days), and 64% of the time when the interval was increased to 2 weeks.⁴ When the same patients then went to self-monitoring every 8, 4, and 2 days, they achieved therapeutic levels 76%, 89%, and 90% of the time, respectively. Bleeding and thromboembolic complications were not reported, but have been demonstrated elsewhere to be lower among patients who self-test frequently (eg, twice weekly) when compared with usual care (average interval 19 days) (4.49% and 0.9% vs 10.9% and 3.6%; number needed to treat [NNT]=15.6 for bleeding, NNT=37 for thromboembolism).⁵

Recommendations from others

The American College of Chest Physicians (ACCP) recommends individualizing management as the optimal frequency of INR monitoring varies according to patient compliance, dosing decisions, duration of therapy and changes in health, diet, or medications.⁶ The ACCP, the American Heart Association,⁷ Micromedex DrugPoints System,⁸ Goodman and Gilman’s Pharmacological Basis of Therapeutics.,⁹ and Cecil’s Textbook of Medicine.¹⁰ all recommend monthly monitoring once stable. The Institute for Clinical Systems Improvement’s Anticoagulation Therapy Supplement Management.¹¹ and Managing Oral Anticoagulation Therapy Clinical and Operation al Guidelines.¹² also recommend monthly monitoring for stable patients, but suggest that the interval can be increased to 6 weeks for selected stable patients.

TABLE
Approach to monitoring of INR for long-term anticoagulation

Evidence summary

Under- or over-treatment with warfarin can result in life-threatening complications. Limited research exists to guide the selection of an interval for monitoring anticoagulation in stabilized patients. One RCT compared INR monitoring in an anticoagulation clinic at 6 weeks and 4 weeks among 124 patients with a prosthetic heart valve on stable oral anticoagulant treatment and found no difference in thromboembolic or hemorrhagic events.¹ A study in the United Kingdom used a 14-week interval for selected patients, but it used no comparison group.² Kent et al developed a computer-based model to compute the optimum interval for monitoring anticoagulation that considers the variability of the patient’s previous levels and costs associated with testing and potential complications. This model achieved a maximum interval of 11 weeks for very stable patients.³

More frequent testing results in higher time in therapeutic range, particularly when patients selfmonitor. A German study of 200 patients with prosthetic heart valves found that they tested within a therapeutic range 48% of the time when monitored by their physician “as usual” (average interval 24 days), and 64% of the time when the interval was increased to 2 weeks.⁴ When the same patients then went to self-monitoring every 8, 4, and 2 days, they achieved therapeutic levels 76%, 89%, and 90% of the time, respectively. Bleeding and thromboembolic complications were not reported, but have been demonstrated elsewhere to be lower among patients who self-test frequently (eg, twice weekly) when compared with usual care (average interval 19 days) (4.49% and 0.9% vs 10.9% and 3.6%; number needed to treat [NNT]=15.6 for bleeding, NNT=37 for thromboembolism).⁵

Recommendations from others

The American College of Chest Physicians (ACCP) recommends individualizing management as the optimal frequency of INR monitoring varies according to patient compliance, dosing decisions, duration of therapy and changes in health, diet, or medications.⁶ The ACCP, the American Heart Association,⁷ Micromedex DrugPoints System,⁸ Goodman and Gilman’s Pharmacological Basis of Therapeutics.,⁹ and Cecil’s Textbook of Medicine.¹⁰ all recommend monthly monitoring once stable. The Institute for Clinical Systems Improvement’s Anticoagulation Therapy Supplement Management.¹¹ and Managing Oral Anticoagulation Therapy Clinical and Operation al Guidelines.¹² also recommend monthly monitoring for stable patients, but suggest that the interval can be increased to 6 weeks for selected stable patients.

TABLE
Approach to monitoring of INR for long-term anticoagulation

Evidence summary

Under- or over-treatment with warfarin can result in life-threatening complications. Limited research exists to guide the selection of an interval for monitoring anticoagulation in stabilized patients. One RCT compared INR monitoring in an anticoagulation clinic at 6 weeks and 4 weeks among 124 patients with a prosthetic heart valve on stable oral anticoagulant treatment and found no difference in thromboembolic or hemorrhagic events.¹ A study in the United Kingdom used a 14-week interval for selected patients, but it used no comparison group.² Kent et al developed a computer-based model to compute the optimum interval for monitoring anticoagulation that considers the variability of the patient’s previous levels and costs associated with testing and potential complications. This model achieved a maximum interval of 11 weeks for very stable patients.³

More frequent testing results in higher time in therapeutic range, particularly when patients selfmonitor. A German study of 200 patients with prosthetic heart valves found that they tested within a therapeutic range 48% of the time when monitored by their physician “as usual” (average interval 24 days), and 64% of the time when the interval was increased to 2 weeks.⁴ When the same patients then went to self-monitoring every 8, 4, and 2 days, they achieved therapeutic levels 76%, 89%, and 90% of the time, respectively. Bleeding and thromboembolic complications were not reported, but have been demonstrated elsewhere to be lower among patients who self-test frequently (eg, twice weekly) when compared with usual care (average interval 19 days) (4.49% and 0.9% vs 10.9% and 3.6%; number needed to treat [NNT]=15.6 for bleeding, NNT=37 for thromboembolism).⁵

Recommendations from others

The American College of Chest Physicians (ACCP) recommends individualizing management as the optimal frequency of INR monitoring varies according to patient compliance, dosing decisions, duration of therapy and changes in health, diet, or medications.⁶ The ACCP, the American Heart Association,⁷ Micromedex DrugPoints System,⁸ Goodman and Gilman’s Pharmacological Basis of Therapeutics.,⁹ and Cecil’s Textbook of Medicine.¹⁰ all recommend monthly monitoring once stable. The Institute for Clinical Systems Improvement’s Anticoagulation Therapy Supplement Management.¹¹ and Managing Oral Anticoagulation Therapy Clinical and Operation al Guidelines.¹² also recommend monthly monitoring for stable patients, but suggest that the interval can be increased to 6 weeks for selected stable patients.

Evidence summary

In numerous systematic reviews, RCTs, and metaanalyses, 70% of children responded to stimulant medications with short-term improvements in ADHD symptoms (inattention and hyperactivity/ impulsivity) and academic achievement. A fortyyear review looked at 135 trials and 413 RCTs of methylphenidate in over 19,000 children with an average age of 8.8 years (range, 8.3–9.4 years) for an average duration of 6 weeks (range, 3.3–8.0 weeks).^1-3

Study groups included mostly elementary school–aged male children, with few minorities represented. Comorbid conditions, present in 65% of children with ADHD, were often poorly controlled. Outcome measures varied among studies.³

The effect size from stimulant medication in these studies averaged 0.8 for symptom relief and between 0.4 and 0.5 for academic achievement. (Effect size is the difference between the means of the experimental and control groups expressed in standard deviations. An effect size of 0.2 is considered small, 0.5 is medium, and 0.8 is considered moderate to large.)

A large randomized trial of 579 children with ADHD (20% girls) aged 7 to 9.9 years compared outcomes of 4 treatment strategies: stimulant medication, intensive behavioral treatment, combined stimulant medication and behavioral interventions, and standard community care.⁴ All children met the DSM-IV. criteria for ADHD Combined Type (the most common type of ADHD in this age group). The stimulant medication strategy included an initial dose titration period followed by monthly 30-minute visits. Intensive behavioral treatment involved child, parent, and school personnel components of therapy. Combination therapy added the regimens for medication and behavioral treatment together. Standard community care consisted of usual (nonsystematic) care, evaluated at 6 different sites.

After 14 months of treatment, children in the medication group and the combined treatment groups showed more improvement in ADHD symptoms than children given intensive behavioral treatment or those who received standard community care. When combined with medication, those treated with behavioral therapy showed slight improvement in social skills, anxiety, aggression, oppositional behavior, and academic achievement over medication alone. At the conclusion of the study, 74% of the 212 children on medication were successfully maintained on methylphenidate alone, 10% required dextroamphetamine, and no children required more than one medication. This study found that higher doses of medication with more frequent office follow-up and regular school contact were important features of successful treatment. Only 40% of families were able to complete the intensive behavioral therapy.

Several short-term reviews and meta-analyses show that side effects from stimulant medications are mild and have short duration.⁵ More long-term studies are required to evaluate effects on growth. RCTs have limited power to detect rare adverse events that may be better detected by large observational studies.⁶

Atomoxetine, a specific norepinephrine reuptake inhibitor, is an FDA-approved alternative to stimulants for ADHD treatment in children and adolescents. Based on 3 RCTs⁷ of 588 children between the ages of 7 and 18 years, atomoxetine showed dose-related improvement in ADHD rating scales. Side effects of atomoxetine are similar to stimulants and include mild but significant increases in blood pressure and pulse.⁷

A meta-analysis of 11 non-randomized trials using clonidine for ADHD showed a smaller effect size compared with stimulants.⁸ One RCT of 136 children with ADHD and tics showed improvement of both problems with the use of methylphenidate and clonidine, particularly in combination.⁹ Second-line medications such as clonidine, pemoline (Cylert), and tricyclic antidepressants have more potential serious side effects and are not well studied.¹⁰

Recommendations from others

The American Academy of Pediatrics recommends that clinicians: 1) manage ADHD as a chronic illness, 2) collaborate with parents, the child, and school personnel to define specific desired outcomes, 3) use stimulant or behavioral therapy to improve these outcomes; if one stimulant is not effective at the highest feasible dose, try another, 4) reevaluate the diagnosis, treatment options, adherence, and possible coexisting conditions if treatment is not achieving the desired outcomes, and 5) follow-up regularly with parents, child, and teachers to monitor for progress and adverse effects.¹¹

TABLE
Commonly used medications for ADHD

Evidence summary

In numerous systematic reviews, RCTs, and metaanalyses, 70% of children responded to stimulant medications with short-term improvements in ADHD symptoms (inattention and hyperactivity/ impulsivity) and academic achievement. A fortyyear review looked at 135 trials and 413 RCTs of methylphenidate in over 19,000 children with an average age of 8.8 years (range, 8.3–9.4 years) for an average duration of 6 weeks (range, 3.3–8.0 weeks).^1-3

Study groups included mostly elementary school–aged male children, with few minorities represented. Comorbid conditions, present in 65% of children with ADHD, were often poorly controlled. Outcome measures varied among studies.³

The effect size from stimulant medication in these studies averaged 0.8 for symptom relief and between 0.4 and 0.5 for academic achievement. (Effect size is the difference between the means of the experimental and control groups expressed in standard deviations. An effect size of 0.2 is considered small, 0.5 is medium, and 0.8 is considered moderate to large.)

A large randomized trial of 579 children with ADHD (20% girls) aged 7 to 9.9 years compared outcomes of 4 treatment strategies: stimulant medication, intensive behavioral treatment, combined stimulant medication and behavioral interventions, and standard community care.⁴ All children met the DSM-IV. criteria for ADHD Combined Type (the most common type of ADHD in this age group). The stimulant medication strategy included an initial dose titration period followed by monthly 30-minute visits. Intensive behavioral treatment involved child, parent, and school personnel components of therapy. Combination therapy added the regimens for medication and behavioral treatment together. Standard community care consisted of usual (nonsystematic) care, evaluated at 6 different sites.

After 14 months of treatment, children in the medication group and the combined treatment groups showed more improvement in ADHD symptoms than children given intensive behavioral treatment or those who received standard community care. When combined with medication, those treated with behavioral therapy showed slight improvement in social skills, anxiety, aggression, oppositional behavior, and academic achievement over medication alone. At the conclusion of the study, 74% of the 212 children on medication were successfully maintained on methylphenidate alone, 10% required dextroamphetamine, and no children required more than one medication. This study found that higher doses of medication with more frequent office follow-up and regular school contact were important features of successful treatment. Only 40% of families were able to complete the intensive behavioral therapy.

Several short-term reviews and meta-analyses show that side effects from stimulant medications are mild and have short duration.⁵ More long-term studies are required to evaluate effects on growth. RCTs have limited power to detect rare adverse events that may be better detected by large observational studies.⁶

Atomoxetine, a specific norepinephrine reuptake inhibitor, is an FDA-approved alternative to stimulants for ADHD treatment in children and adolescents. Based on 3 RCTs⁷ of 588 children between the ages of 7 and 18 years, atomoxetine showed dose-related improvement in ADHD rating scales. Side effects of atomoxetine are similar to stimulants and include mild but significant increases in blood pressure and pulse.⁷

A meta-analysis of 11 non-randomized trials using clonidine for ADHD showed a smaller effect size compared with stimulants.⁸ One RCT of 136 children with ADHD and tics showed improvement of both problems with the use of methylphenidate and clonidine, particularly in combination.⁹ Second-line medications such as clonidine, pemoline (Cylert), and tricyclic antidepressants have more potential serious side effects and are not well studied.¹⁰

Recommendations from others

The American Academy of Pediatrics recommends that clinicians: 1) manage ADHD as a chronic illness, 2) collaborate with parents, the child, and school personnel to define specific desired outcomes, 3) use stimulant or behavioral therapy to improve these outcomes; if one stimulant is not effective at the highest feasible dose, try another, 4) reevaluate the diagnosis, treatment options, adherence, and possible coexisting conditions if treatment is not achieving the desired outcomes, and 5) follow-up regularly with parents, child, and teachers to monitor for progress and adverse effects.¹¹

TABLE
Commonly used medications for ADHD

Evidence summary

In numerous systematic reviews, RCTs, and metaanalyses, 70% of children responded to stimulant medications with short-term improvements in ADHD symptoms (inattention and hyperactivity/ impulsivity) and academic achievement. A fortyyear review looked at 135 trials and 413 RCTs of methylphenidate in over 19,000 children with an average age of 8.8 years (range, 8.3–9.4 years) for an average duration of 6 weeks (range, 3.3–8.0 weeks).^1-3

Study groups included mostly elementary school–aged male children, with few minorities represented. Comorbid conditions, present in 65% of children with ADHD, were often poorly controlled. Outcome measures varied among studies.³

The effect size from stimulant medication in these studies averaged 0.8 for symptom relief and between 0.4 and 0.5 for academic achievement. (Effect size is the difference between the means of the experimental and control groups expressed in standard deviations. An effect size of 0.2 is considered small, 0.5 is medium, and 0.8 is considered moderate to large.)

A large randomized trial of 579 children with ADHD (20% girls) aged 7 to 9.9 years compared outcomes of 4 treatment strategies: stimulant medication, intensive behavioral treatment, combined stimulant medication and behavioral interventions, and standard community care.⁴ All children met the DSM-IV. criteria for ADHD Combined Type (the most common type of ADHD in this age group). The stimulant medication strategy included an initial dose titration period followed by monthly 30-minute visits. Intensive behavioral treatment involved child, parent, and school personnel components of therapy. Combination therapy added the regimens for medication and behavioral treatment together. Standard community care consisted of usual (nonsystematic) care, evaluated at 6 different sites.

After 14 months of treatment, children in the medication group and the combined treatment groups showed more improvement in ADHD symptoms than children given intensive behavioral treatment or those who received standard community care. When combined with medication, those treated with behavioral therapy showed slight improvement in social skills, anxiety, aggression, oppositional behavior, and academic achievement over medication alone. At the conclusion of the study, 74% of the 212 children on medication were successfully maintained on methylphenidate alone, 10% required dextroamphetamine, and no children required more than one medication. This study found that higher doses of medication with more frequent office follow-up and regular school contact were important features of successful treatment. Only 40% of families were able to complete the intensive behavioral therapy.

Several short-term reviews and meta-analyses show that side effects from stimulant medications are mild and have short duration.⁵ More long-term studies are required to evaluate effects on growth. RCTs have limited power to detect rare adverse events that may be better detected by large observational studies.⁶

Atomoxetine, a specific norepinephrine reuptake inhibitor, is an FDA-approved alternative to stimulants for ADHD treatment in children and adolescents. Based on 3 RCTs⁷ of 588 children between the ages of 7 and 18 years, atomoxetine showed dose-related improvement in ADHD rating scales. Side effects of atomoxetine are similar to stimulants and include mild but significant increases in blood pressure and pulse.⁷

A meta-analysis of 11 non-randomized trials using clonidine for ADHD showed a smaller effect size compared with stimulants.⁸ One RCT of 136 children with ADHD and tics showed improvement of both problems with the use of methylphenidate and clonidine, particularly in combination.⁹ Second-line medications such as clonidine, pemoline (Cylert), and tricyclic antidepressants have more potential serious side effects and are not well studied.¹⁰

Recommendations from others

The American Academy of Pediatrics recommends that clinicians: 1) manage ADHD as a chronic illness, 2) collaborate with parents, the child, and school personnel to define specific desired outcomes, 3) use stimulant or behavioral therapy to improve these outcomes; if one stimulant is not effective at the highest feasible dose, try another, 4) reevaluate the diagnosis, treatment options, adherence, and possible coexisting conditions if treatment is not achieving the desired outcomes, and 5) follow-up regularly with parents, child, and teachers to monitor for progress and adverse effects.¹¹

TABLE
Commonly used medications for ADHD

Evidence summary

Oppositional and defiant behaviors include noncompliance, temper tantrums, arguing, and mild aggression. Children exhibiting these behaviors may have a diagnosis of ODD. Importantly, this review does not examine treatments for children diagnosed with conduct disorder or those exhibiting more deviant behaviors such as serious aggression and delinquency.

Eight well-done systematic reviews examined the effectiveness of parent training programs. Parent training is typically conducted by clinical child psychologists but may also be available through certified parenting educators (see the National Parenting Education Network web page for links to state organizations, at www.ces.ncsu.edu/depts/fcs/npen/). Parent training strategies are also described for parents in books such as Your Defiant Child.¹

The most rigorous of the reviews looked at 16 randomized controlled trials that examined the effectiveness of training programs for children between the ages of 3 and 10 years who had “externalizing problems,” including temper tantrums, aggression, and noncompliance.² All studies included in the review compared a group-based parent training program with a no-treatment wait-list control group and assessed outcomes using a standardized measure of behavior. In studies where sufficient data were provided, effect sizes ranged from 0.6 to 2.9. This indicates that, on a standardized child behavioral measure, parental report of children’s externalizing problems decreased by 0.6 to 2.9 standard deviations from pre- to posttreatment (an effect size of >0.8 is considered large). In the 2 studies that included independent observations of child behavior, the benefits reported by parents were confirmed by these observations.

Although parent training has the strongest evidence as a treatment for oppositional and defiant behavior, other psychosocial treatment interventions have been found by multiple randomized controlled trials to be superior to no treatment or wait-list controls (Table).

In treating oppositional behaviors among children with ADHD and comorbid oppositional defiant disorder or conduct disorder, a meta-analysis identified 28 studies of children age 7 to 15 years that addressed oppositional/aggression-related behaviors within the context of ADHD.⁸ The analysis found that stimulants are efficacious. The overall weighted effect size (a measure of improvement representing the average effects across all reporters) was 0.89. This indicates that raters saw a change in oppositional behaviors—noncompliance, irritability, and temper tantrums—that corresponded to a drop in scores of approximately 1 standard deviation.

TABLE
Additional ODD treatments supported by randomized controlled trials

Recommendations from others

Two parent training interventions meet the American Psychological Association’s criteria for well-established treatments.⁹ These include programs based on Patterson and Gullion’s Living with Children, a short-term, behavioral parent training program, and programs based on WebsterStratton’s Videotape Modeling parent training program. Two additional treatments, Anger Coping Therapy and Problem Solving Skills Training, meet the criteria for “probably efficacious.”

According to the International Consensus Statement on ADHD and Disruptive Behavior Disorders, “pharmacological treatment of pure ODD should not be considered except in cases where aggression is a significant, persistent problem.”¹⁰

Evidence summary

Oppositional and defiant behaviors include noncompliance, temper tantrums, arguing, and mild aggression. Children exhibiting these behaviors may have a diagnosis of ODD. Importantly, this review does not examine treatments for children diagnosed with conduct disorder or those exhibiting more deviant behaviors such as serious aggression and delinquency.

Eight well-done systematic reviews examined the effectiveness of parent training programs. Parent training is typically conducted by clinical child psychologists but may also be available through certified parenting educators (see the National Parenting Education Network web page for links to state organizations, at www.ces.ncsu.edu/depts/fcs/npen/). Parent training strategies are also described for parents in books such as Your Defiant Child.¹

The most rigorous of the reviews looked at 16 randomized controlled trials that examined the effectiveness of training programs for children between the ages of 3 and 10 years who had “externalizing problems,” including temper tantrums, aggression, and noncompliance.² All studies included in the review compared a group-based parent training program with a no-treatment wait-list control group and assessed outcomes using a standardized measure of behavior. In studies where sufficient data were provided, effect sizes ranged from 0.6 to 2.9. This indicates that, on a standardized child behavioral measure, parental report of children’s externalizing problems decreased by 0.6 to 2.9 standard deviations from pre- to posttreatment (an effect size of >0.8 is considered large). In the 2 studies that included independent observations of child behavior, the benefits reported by parents were confirmed by these observations.

Although parent training has the strongest evidence as a treatment for oppositional and defiant behavior, other psychosocial treatment interventions have been found by multiple randomized controlled trials to be superior to no treatment or wait-list controls (Table).

In treating oppositional behaviors among children with ADHD and comorbid oppositional defiant disorder or conduct disorder, a meta-analysis identified 28 studies of children age 7 to 15 years that addressed oppositional/aggression-related behaviors within the context of ADHD.⁸ The analysis found that stimulants are efficacious. The overall weighted effect size (a measure of improvement representing the average effects across all reporters) was 0.89. This indicates that raters saw a change in oppositional behaviors—noncompliance, irritability, and temper tantrums—that corresponded to a drop in scores of approximately 1 standard deviation.

TABLE
Additional ODD treatments supported by randomized controlled trials

Recommendations from others

Two parent training interventions meet the American Psychological Association’s criteria for well-established treatments.⁹ These include programs based on Patterson and Gullion’s Living with Children, a short-term, behavioral parent training program, and programs based on WebsterStratton’s Videotape Modeling parent training program. Two additional treatments, Anger Coping Therapy and Problem Solving Skills Training, meet the criteria for “probably efficacious.”

According to the International Consensus Statement on ADHD and Disruptive Behavior Disorders, “pharmacological treatment of pure ODD should not be considered except in cases where aggression is a significant, persistent problem.”¹⁰

Evidence summary

Oppositional and defiant behaviors include noncompliance, temper tantrums, arguing, and mild aggression. Children exhibiting these behaviors may have a diagnosis of ODD. Importantly, this review does not examine treatments for children diagnosed with conduct disorder or those exhibiting more deviant behaviors such as serious aggression and delinquency.

Eight well-done systematic reviews examined the effectiveness of parent training programs. Parent training is typically conducted by clinical child psychologists but may also be available through certified parenting educators (see the National Parenting Education Network web page for links to state organizations, at www.ces.ncsu.edu/depts/fcs/npen/). Parent training strategies are also described for parents in books such as Your Defiant Child.¹

The most rigorous of the reviews looked at 16 randomized controlled trials that examined the effectiveness of training programs for children between the ages of 3 and 10 years who had “externalizing problems,” including temper tantrums, aggression, and noncompliance.² All studies included in the review compared a group-based parent training program with a no-treatment wait-list control group and assessed outcomes using a standardized measure of behavior. In studies where sufficient data were provided, effect sizes ranged from 0.6 to 2.9. This indicates that, on a standardized child behavioral measure, parental report of children’s externalizing problems decreased by 0.6 to 2.9 standard deviations from pre- to posttreatment (an effect size of >0.8 is considered large). In the 2 studies that included independent observations of child behavior, the benefits reported by parents were confirmed by these observations.

Although parent training has the strongest evidence as a treatment for oppositional and defiant behavior, other psychosocial treatment interventions have been found by multiple randomized controlled trials to be superior to no treatment or wait-list controls (Table).

In treating oppositional behaviors among children with ADHD and comorbid oppositional defiant disorder or conduct disorder, a meta-analysis identified 28 studies of children age 7 to 15 years that addressed oppositional/aggression-related behaviors within the context of ADHD.⁸ The analysis found that stimulants are efficacious. The overall weighted effect size (a measure of improvement representing the average effects across all reporters) was 0.89. This indicates that raters saw a change in oppositional behaviors—noncompliance, irritability, and temper tantrums—that corresponded to a drop in scores of approximately 1 standard deviation.

TABLE
Additional ODD treatments supported by randomized controlled trials

Recommendations from others

Two parent training interventions meet the American Psychological Association’s criteria for well-established treatments.⁹ These include programs based on Patterson and Gullion’s Living with Children, a short-term, behavioral parent training program, and programs based on WebsterStratton’s Videotape Modeling parent training program. Two additional treatments, Anger Coping Therapy and Problem Solving Skills Training, meet the criteria for “probably efficacious.”

According to the International Consensus Statement on ADHD and Disruptive Behavior Disorders, “pharmacological treatment of pure ODD should not be considered except in cases where aggression is a significant, persistent problem.”¹⁰

Evidence Summary

A common female endocrinopathy, PCOS affects 5% to 10% of women. Characterized by anovulation and hyperandrogenism, it often manifests as infertility and irregular menstruation. Metformin and thiazolidinediones are likely effective treatments for these expressions of insulin resistance, but study limitations restrict our ability to clearly define their role.

The most influential systematic review was a meta-analysis that reviewed 13 RCTs including 543 women to determine the effects of metformin on ovarian function in PCOS.^1,2By selecting RCTs, performing precise statistical analysis according to the Cochrane protocols, and clearly stating limitations, this review gives good evidence that metformin modestly increases the odds of ovulation for women with PCOS (odds ratio [OR]=3.88; 95% confidence interval [CI], 2.25–6.69 for metformin vs placebo) and that metformin with clomiphene (Clomid) effectively increases ovulation (OR=4.41; 95% CI, 2.37–8.22) and pregnancy rates (OR=4.40; 95% CI, 1.96–9.85) when compared with clomiphene use alone. When metformin is used as a sole agent, ovulation is achieved in 46% of recipients compared with 24% in the placebo arm (number needed to treat [NNT]=4.4). When metformin and clomiphene are used in combination, 76% of recipients ovulate compared with 42% receiving clomiphene alone (NNT=3.0).

Several problems with recommending metformin as first-line therapy exist: (1) equal or better ovulation rates have been described by using lifestyle interventions to achieve weight loss, (2) there are no long-term studies of the effects of metformin in PCOS patients, and (3) we cannot assess the clinically important outcome of pregnancy rates because the trials did not control for other infertility factors and did not define live births as a primary outcome. In addition, there are no head-to-head trials of metformin vs clomiphene, the standard first-line therapy for ovulation induction. Only 1 study addressed menstrual patterns specifically; they were improved with metformin (OR=12.88; 95% CI, 1.85–89.61).

An additional meta-analysis reports similar results.³ Eight RCTs addressing the use of metformin or clomiphene for treatment of PCOS were reviewed for ovulation and pregnancy rates. Metformin is 50% better than placebo for ovulation induction among infertile PCOS patients (relative risk [RR]=1.50; 95% CI, 1.31–1.99), but this benefit is not necessarily improved with longer duration (>3 months) of therapy (RR=1.37; 95% CI, 1.05–1.79). Also, metformin is beneficial in regulating cycles for fertile PCOS patients with irregular menses (RR=1.45; 95% CI, 1.11–1.90).

The conclusions regarding pregnancy rates and combined therapy with metformin and clomiphene are limited due to small samples, short follow-up time (2–6 months), and study design. An ongoing randomized trial (Pregnancy in Polycystic Ovarian Syndrome: PPOS study) of 768 infertile PCOS patients is investigating effects of metformin vs clomiphene on ovulation induction and achievement of singleton pregnancies. These outcomes should clarify remaining uncertainties regarding appropriate use of metformin.

Finally, a review of 7 RCTs describes the evidence accumulated by well-designed trials and its clinical relevance.⁴ Metformin improves ovulation and menstrual cyclicity but these improvements were variable and modest. On average, 1 additional ovulation is attained in every 5-month interval with metformin treatment; specifically, the baseline of 1 ovulation per 5-month interval increased to 2 ovulations per 5-month interval. Spontaneous ovulation and normal menstruation are achieved rapidly (within 3 months of the start of therapy). These data corroborate the benefits of metformin but place its clinical significance in perspective. For PCOS patients seeking cycle regulation but not pregnancy, oral contraceptives may remain better therapy because metformin does not normalize menses.

Less information exists on the role of TZDs and ovarian function in PCOS. Studies of the most researched drug in the class, troglitazone (Rezulin), report improvements in ovulation rates and metabolic markers of PCOS.^5,6 Troglitazone has been taken off of the market due to hepatotoxicity, but results from a RCT of 40 patients with PCOS reported that the use of pioglitazone (Actos) for 3 months increased normal regular cycles and ovulations over placebo (41.2% vs 5.6%; P<.02).⁷ No liver effects were noted, but caution must be taken since these drugs are pregnancy class C. Two small RCTs studied the use of rosiglitazone (Avandia) in combination with clomiphene and reported improvements in menstrual regularity⁸ (92% with combination therapy achieved improved menstrual cycles vs 68% with rosiglitazone alone; OR=0.185) and both spontaneous and clomiphene-induced ovulation rates (52% of clomiphene-resistant women ovulated after rosiglitazone therapy and 77% vs 33% ovulated with combination therapy vs rosiglitazone alone, P=.04).⁹ Further research is needed to determine the clinical effects of the thiazolidinediones.

Recommendations from Others

The American College of Obstetricians and Gynecologists guideline on diagnosis and management of PCOS reports that interventions that improve insulin sensitivity, including weight loss, use of metformin, and use of TZDs are useful for improving ovulatory frequency for women with PCOS.¹⁰ The recommendation is based on good and consistent scientific evidence (SOR: A). They also note that insulin-sensitizing agents may improve many risk factors for diabetes and cardiovascular disease, but this recommendation is based on limited evidence (SOR: B). Finally, they recommend, based on expert opinion (SOR: C), that caution be used with these agents because their effects on early pregnancy are unknown, even though metformin appears to be safe.

The American Association of Clinical Endocrinologists recommends using metformin 850 mg twice daily to treat the hyperandrogenic state of PCOS.¹¹ The use of TZDs is less clear due to limited evidence and risks of teratogenicity.

Evidence Summary

A common female endocrinopathy, PCOS affects 5% to 10% of women. Characterized by anovulation and hyperandrogenism, it often manifests as infertility and irregular menstruation. Metformin and thiazolidinediones are likely effective treatments for these expressions of insulin resistance, but study limitations restrict our ability to clearly define their role.

The most influential systematic review was a meta-analysis that reviewed 13 RCTs including 543 women to determine the effects of metformin on ovarian function in PCOS.^1,2By selecting RCTs, performing precise statistical analysis according to the Cochrane protocols, and clearly stating limitations, this review gives good evidence that metformin modestly increases the odds of ovulation for women with PCOS (odds ratio [OR]=3.88; 95% confidence interval [CI], 2.25–6.69 for metformin vs placebo) and that metformin with clomiphene (Clomid) effectively increases ovulation (OR=4.41; 95% CI, 2.37–8.22) and pregnancy rates (OR=4.40; 95% CI, 1.96–9.85) when compared with clomiphene use alone. When metformin is used as a sole agent, ovulation is achieved in 46% of recipients compared with 24% in the placebo arm (number needed to treat [NNT]=4.4). When metformin and clomiphene are used in combination, 76% of recipients ovulate compared with 42% receiving clomiphene alone (NNT=3.0).

Several problems with recommending metformin as first-line therapy exist: (1) equal or better ovulation rates have been described by using lifestyle interventions to achieve weight loss, (2) there are no long-term studies of the effects of metformin in PCOS patients, and (3) we cannot assess the clinically important outcome of pregnancy rates because the trials did not control for other infertility factors and did not define live births as a primary outcome. In addition, there are no head-to-head trials of metformin vs clomiphene, the standard first-line therapy for ovulation induction. Only 1 study addressed menstrual patterns specifically; they were improved with metformin (OR=12.88; 95% CI, 1.85–89.61).

An additional meta-analysis reports similar results.³ Eight RCTs addressing the use of metformin or clomiphene for treatment of PCOS were reviewed for ovulation and pregnancy rates. Metformin is 50% better than placebo for ovulation induction among infertile PCOS patients (relative risk [RR]=1.50; 95% CI, 1.31–1.99), but this benefit is not necessarily improved with longer duration (>3 months) of therapy (RR=1.37; 95% CI, 1.05–1.79). Also, metformin is beneficial in regulating cycles for fertile PCOS patients with irregular menses (RR=1.45; 95% CI, 1.11–1.90).

The conclusions regarding pregnancy rates and combined therapy with metformin and clomiphene are limited due to small samples, short follow-up time (2–6 months), and study design. An ongoing randomized trial (Pregnancy in Polycystic Ovarian Syndrome: PPOS study) of 768 infertile PCOS patients is investigating effects of metformin vs clomiphene on ovulation induction and achievement of singleton pregnancies. These outcomes should clarify remaining uncertainties regarding appropriate use of metformin.

Finally, a review of 7 RCTs describes the evidence accumulated by well-designed trials and its clinical relevance.⁴ Metformin improves ovulation and menstrual cyclicity but these improvements were variable and modest. On average, 1 additional ovulation is attained in every 5-month interval with metformin treatment; specifically, the baseline of 1 ovulation per 5-month interval increased to 2 ovulations per 5-month interval. Spontaneous ovulation and normal menstruation are achieved rapidly (within 3 months of the start of therapy). These data corroborate the benefits of metformin but place its clinical significance in perspective. For PCOS patients seeking cycle regulation but not pregnancy, oral contraceptives may remain better therapy because metformin does not normalize menses.

Less information exists on the role of TZDs and ovarian function in PCOS. Studies of the most researched drug in the class, troglitazone (Rezulin), report improvements in ovulation rates and metabolic markers of PCOS.^5,6 Troglitazone has been taken off of the market due to hepatotoxicity, but results from a RCT of 40 patients with PCOS reported that the use of pioglitazone (Actos) for 3 months increased normal regular cycles and ovulations over placebo (41.2% vs 5.6%; P<.02).⁷ No liver effects were noted, but caution must be taken since these drugs are pregnancy class C. Two small RCTs studied the use of rosiglitazone (Avandia) in combination with clomiphene and reported improvements in menstrual regularity⁸ (92% with combination therapy achieved improved menstrual cycles vs 68% with rosiglitazone alone; OR=0.185) and both spontaneous and clomiphene-induced ovulation rates (52% of clomiphene-resistant women ovulated after rosiglitazone therapy and 77% vs 33% ovulated with combination therapy vs rosiglitazone alone, P=.04).⁹ Further research is needed to determine the clinical effects of the thiazolidinediones.

Recommendations from Others

The American College of Obstetricians and Gynecologists guideline on diagnosis and management of PCOS reports that interventions that improve insulin sensitivity, including weight loss, use of metformin, and use of TZDs are useful for improving ovulatory frequency for women with PCOS.¹⁰ The recommendation is based on good and consistent scientific evidence (SOR: A). They also note that insulin-sensitizing agents may improve many risk factors for diabetes and cardiovascular disease, but this recommendation is based on limited evidence (SOR: B). Finally, they recommend, based on expert opinion (SOR: C), that caution be used with these agents because their effects on early pregnancy are unknown, even though metformin appears to be safe.

The American Association of Clinical Endocrinologists recommends using metformin 850 mg twice daily to treat the hyperandrogenic state of PCOS.¹¹ The use of TZDs is less clear due to limited evidence and risks of teratogenicity.

Evidence Summary

A common female endocrinopathy, PCOS affects 5% to 10% of women. Characterized by anovulation and hyperandrogenism, it often manifests as infertility and irregular menstruation. Metformin and thiazolidinediones are likely effective treatments for these expressions of insulin resistance, but study limitations restrict our ability to clearly define their role.

The most influential systematic review was a meta-analysis that reviewed 13 RCTs including 543 women to determine the effects of metformin on ovarian function in PCOS.^1,2By selecting RCTs, performing precise statistical analysis according to the Cochrane protocols, and clearly stating limitations, this review gives good evidence that metformin modestly increases the odds of ovulation for women with PCOS (odds ratio [OR]=3.88; 95% confidence interval [CI], 2.25–6.69 for metformin vs placebo) and that metformin with clomiphene (Clomid) effectively increases ovulation (OR=4.41; 95% CI, 2.37–8.22) and pregnancy rates (OR=4.40; 95% CI, 1.96–9.85) when compared with clomiphene use alone. When metformin is used as a sole agent, ovulation is achieved in 46% of recipients compared with 24% in the placebo arm (number needed to treat [NNT]=4.4). When metformin and clomiphene are used in combination, 76% of recipients ovulate compared with 42% receiving clomiphene alone (NNT=3.0).

Several problems with recommending metformin as first-line therapy exist: (1) equal or better ovulation rates have been described by using lifestyle interventions to achieve weight loss, (2) there are no long-term studies of the effects of metformin in PCOS patients, and (3) we cannot assess the clinically important outcome of pregnancy rates because the trials did not control for other infertility factors and did not define live births as a primary outcome. In addition, there are no head-to-head trials of metformin vs clomiphene, the standard first-line therapy for ovulation induction. Only 1 study addressed menstrual patterns specifically; they were improved with metformin (OR=12.88; 95% CI, 1.85–89.61).

An additional meta-analysis reports similar results.³ Eight RCTs addressing the use of metformin or clomiphene for treatment of PCOS were reviewed for ovulation and pregnancy rates. Metformin is 50% better than placebo for ovulation induction among infertile PCOS patients (relative risk [RR]=1.50; 95% CI, 1.31–1.99), but this benefit is not necessarily improved with longer duration (>3 months) of therapy (RR=1.37; 95% CI, 1.05–1.79). Also, metformin is beneficial in regulating cycles for fertile PCOS patients with irregular menses (RR=1.45; 95% CI, 1.11–1.90).

The conclusions regarding pregnancy rates and combined therapy with metformin and clomiphene are limited due to small samples, short follow-up time (2–6 months), and study design. An ongoing randomized trial (Pregnancy in Polycystic Ovarian Syndrome: PPOS study) of 768 infertile PCOS patients is investigating effects of metformin vs clomiphene on ovulation induction and achievement of singleton pregnancies. These outcomes should clarify remaining uncertainties regarding appropriate use of metformin.

Finally, a review of 7 RCTs describes the evidence accumulated by well-designed trials and its clinical relevance.⁴ Metformin improves ovulation and menstrual cyclicity but these improvements were variable and modest. On average, 1 additional ovulation is attained in every 5-month interval with metformin treatment; specifically, the baseline of 1 ovulation per 5-month interval increased to 2 ovulations per 5-month interval. Spontaneous ovulation and normal menstruation are achieved rapidly (within 3 months of the start of therapy). These data corroborate the benefits of metformin but place its clinical significance in perspective. For PCOS patients seeking cycle regulation but not pregnancy, oral contraceptives may remain better therapy because metformin does not normalize menses.

Less information exists on the role of TZDs and ovarian function in PCOS. Studies of the most researched drug in the class, troglitazone (Rezulin), report improvements in ovulation rates and metabolic markers of PCOS.^5,6 Troglitazone has been taken off of the market due to hepatotoxicity, but results from a RCT of 40 patients with PCOS reported that the use of pioglitazone (Actos) for 3 months increased normal regular cycles and ovulations over placebo (41.2% vs 5.6%; P<.02).⁷ No liver effects were noted, but caution must be taken since these drugs are pregnancy class C. Two small RCTs studied the use of rosiglitazone (Avandia) in combination with clomiphene and reported improvements in menstrual regularity⁸ (92% with combination therapy achieved improved menstrual cycles vs 68% with rosiglitazone alone; OR=0.185) and both spontaneous and clomiphene-induced ovulation rates (52% of clomiphene-resistant women ovulated after rosiglitazone therapy and 77% vs 33% ovulated with combination therapy vs rosiglitazone alone, P=.04).⁹ Further research is needed to determine the clinical effects of the thiazolidinediones.

Recommendations from Others

The American College of Obstetricians and Gynecologists guideline on diagnosis and management of PCOS reports that interventions that improve insulin sensitivity, including weight loss, use of metformin, and use of TZDs are useful for improving ovulatory frequency for women with PCOS.¹⁰ The recommendation is based on good and consistent scientific evidence (SOR: A). They also note that insulin-sensitizing agents may improve many risk factors for diabetes and cardiovascular disease, but this recommendation is based on limited evidence (SOR: B). Finally, they recommend, based on expert opinion (SOR: C), that caution be used with these agents because their effects on early pregnancy are unknown, even though metformin appears to be safe.

The American Association of Clinical Endocrinologists recommends using metformin 850 mg twice daily to treat the hyperandrogenic state of PCOS.¹¹ The use of TZDs is less clear due to limited evidence and risks of teratogenicity.