Propranolol in the Management of Recurrent Migraine: A Meta-analytic Review
Kenneth A. Holroyd, Ph.D., Donald B. Penzien, Ph.D. and Gary E. Cordingley, M.D., Ph.D.
From the Institute of Health and Behavioral Sciences, Ohio University (Dr. Hurloid), University of Mississippi Medical Center (Dr. Penzien); and Ohio University College of Osteopathic Medicine (Dr. Cordingley). Reprint requests to: K. A. Hurloid, Ph.D, Porter Hall 235, Institute of Health and Behavioral Sciences, Ohio University, Athens, Ohio 45701-2979. Accepted for Publication: April 9, 1991, SYNOPSIS
We used meta-analytic statistical techniques to synthesize findings from studies that evaluated propranolol HCI for the prevention of recurrent migraine (2,403 treated patients). The modal migraine sufferer treated in these studies was a female, about 37 years of age, who suffered from common (rather than classical) migraines and reported a 17-year history of problem migraines. The modal treatment was 160 mg. propranolol per day. Meta-analysis revealed that, on average, propranolol yielded a 44% reduction in migraine activity when daily headache recordings were used to assess treatment outcome, and a 65% reduction in migraine activity when less conservative measures (e.g., clinical ratings of improvement, global patient reports) were used. Meta-analysis thus revealed substantial support for short-term effectiveness of propranolol. However, little information was available concerning the long-term effectiveness of propranolol. Key words: propranolol, migraine, meta-analysis. (Headache 31:333-340, 1991 ) Rabkin and colleagues first observed the beneficial effect of propranolol HCI on migraine headaches when evaluating this medication in the management of angina pectoris.1 In the subsequent two decades, more than four-dozen studies evaluated propranolol as a treatment for migraine, and propranolol has become the most widely prescribed prophylactic medication for this condition. However, reviews have restricted their focus to just a half-dozen of the available trials. To our knowledge, no comprehensive summary of available clinical trials has been published. We used meta-analytic review techniques to examine results from studies evaluating propranolol for the prevention of recurrent migraine. In meta-analysis, findings from mulitiple studies are aggregated and subjected to statistical analysis.2,3,4,5 This process derives quantatative information from a larger sample than any single study could provide. Relationships between study characteristics (e.g., characteristics of patients, treatment protocols and research designs) and treatment effects also can be examined. We used meta-analysis to generate information on: (a) characteristics of available studies, b) changes in migraine activity observed across studies with propranolol and with placebo, and (c) relationships between study characteristics and propranolol effects. One study characteristic, the type of measure used to assess treatment outcome, was of particular interest because of evidence showing hat the results obtained in clinical trials varies with he type of measure that is used to assess improve-Rent.6,7 METHODS
Selection of Studies. For a study to be included n the meta-analysis, the following minimal criteria had to be met: (a) identification of patients as adult migraine sufferers, (b) treatment of no fewer than 5 migraine patients with propranolol, and (c) description of dependent measure(s) that could be used to derive a migraine improvement score (see below). n order to obtain an adequate number of placebo improvement scores for analysis, placebo improvement scores were obtained not only from clinical trials of propranolol, but also from clinical trials evaluating other prophylactic migraine medications (e.g., calcium channel antagonists). No attempt was made, however, to locate all existing studies evaluating any prophylactic medication that included a placebo condition. Fifty-three studies reporting usable data for 73 propranolol or placebo treatment groups met the above criteria for inclusion in the meta-analysis. Studies were located through the following literature search methods: (1) Computer searches of the Medline and Medlars II data bases. (2) Recent volumes of 10 journals that publish relevant articles were searched. (3) The reference section of each article located via the above search was examined, along with the reference sections of numerous journal articles and book chapters on the topic of headache.
Propranolol and Placebo Conditions. Propranolol. For data to be included in the propranolol condition, propranolol had to be administered specifically for the treatment of recurrent migraines. Studies could report treatment outcome in 1 of 2 ways: (1) "propranolol vs. pretreatment" -improvement relative to a pretreatment baseline period when neither prophylactic headache medication nor placebo was administered, or (2) "propranolol vs. placebo" - improvement relative to a placebo treatment period when placebo was administered. Improvement scores of both types were included, but these 2 different improvement scores were treated separately in analyses of treatment outcome. Placebo. For data to be included in the placebo condition, placebo had to be administered on the same schedule as the active prophylactic medication, and the study had to permit the calculation of the improvement observed with placebo. Treatment Effects. The preferred measure of migraine activity was the headache index (H.I.). This measure is a composite score which takes both intensity and duration into account. When no headache index was reported, a composite score was formed by taking an average of the available measures (e.g., frequency, duration, intensity). When the study presented data that allowed the calculation of the improvement achieved with propranolol relative to a pretreatment baseline (when neither prophylactic medication nor placebo was administered), percent improvement in migraine activity was calculated using the following formula: Pr opranolol
pretreatment H . I. − propranolol treatment H . I . = − − − − − − − − − − − − − − − − − − − − − − − − × 100
vs. Pr etreatment
Pr eatment H . I.
The propranolol treatment headache index was assessed following the final propranolol dose adjustment. Where the study presented data that allowed calculation of the improvement acheived with propranolol relative to a placebo baseline (where matched placebo was administered), percent improvement in migraine activity was calculated using the following formula: pretreatment H . I . − propranolol treatment H . I.
Pr opranolol vs.
= − − − − − − − − − − − − − − − − − − − − − − − − × 100
placebo treatment H .. I.
Percent improvement in migraine activity achieved with placebo was calculated using the following formula: Placebo vs. Pr etreatment
pretreatment H. I. − placebo treatment H. I. = − − − − − − − − − − − − − − − − − − − − − − − − ×100 Pr etreatment H. I.
The placebo treatment headache index was assessed following the final placebo dose adjustment. When data obtained from patients' daily headache recordings were presented, this information was used to calculate the above improvement scores. When only information from physician or patient global ratings of improvement was available, or improvement was reported in terms of the percentage of patients clinically improved, these alternate measures of improvement were used. However, the type of outcome measure (percent improvement score calculated from daily headache recording vs. alternate measures of improvement) was coded so we could determine if results reported with propranolol varied systematically with the type of measure. Other Study Variables.* Characteristics of the patient sample, propranolol treatment protocol, and other relevant study characteristics were also coded. This information allowed us to characterize available clinical trials of propranolol as wall as to examine relationships between study characteristics and treatment outcome. Data were coded separately for propranolol and placebo treatment groups wherever possible; where data was not available for individual treatment groups, data for the entire study sample was substituted. Seven characteristics of patient samples were coded: (a) Number of Patients - number of patients completing treatment; (b) Gender - percentage of males in the treatment sample; (c) Age-the mean age of the treatment sample; (d) Migraine Diagnosis- percentage of patients with a diagnosis of classic migraine; (e) Chronicity - mean chronicity of headache problems (in years); (f) Dropout Rate- percentage of patients who failed to complete the treatment protocol; (g) Dropout Rate Due to Treatment Side Effects - percentage of patients identified as failing to complete the treatment protocol specifically because of drug side-effects. Two characteristics of the treatment protocol were also coded: (a) Propranolol Dose - the final propranolol dose (mg./day); in studies where dose varied across patients, the mean dose was used; (b) Blind to Drug Condition - whether double-blind, single-blind, or open-drug administration protocol was used. Finally, 2 additional study characteristics were coded: (a) Year - year a study was published or presented; (b) Type of Outcome Measure - percentage reduction in migraine activity calculated from patient daily headache recordings was distinguished from other outcome measures (e.g., physician ratings of improvement, patient global ratings of improvement). Each article was independently reviewed and its data encoded by the 2 senior authors. Encoding disagreements were identified and discussed until resolved. RESULTS
Characteristics of Studies. Patient samples. Characteristics of propranolol and placebo treatment
groups are presented in Table 1. It can be seen that patient samples have been predominantly female with a mean age in the late thirties. Patients treated with propranolol averaged a 17-year history of migraines. About a quarter of these patients were diagnosed as suffering from classical (rather than common) migraines. Approximately 1 out of 6 patients discontinued propranolol therapy before completion of the treatment protocol. The authors identified about a third of these dropouts as discontinuing treatment because of drug side effects. Although overall drop-out rates for propranolol and placebo treatment did not differ significantly, patients were significantly more likely to discontinue propranolol than placebo because of drug side effects. (t(52) = 2.0, p