Abstract
This was a phase-IV double-blind equivalence trial designed to assess the efficacy and tolerability of two doses of flunarizine (10 mg o.d. = FLU 10 mg and 5 mg o.d. = FLU 5 mg) in the prophylaxis of migraine, in comparison with slow-release propranolol (160 mg o.d.). A total of 808 subjects were treated in a treatment period of 16 weeks. 142 subjects discontinued the trial prematurely, mainly because of adverse events (n = 58). The mean attack frequency in the double-blind period was 2.0 for the FLU 5 mg group, 1.9 for the FLU 10 mg group, and 1.9 for the propranolol group. The mean attack frequency in the last 28 days of the double-blind period was 1.8 for FLU 5 mg, 1.6 for FLU 10 mg, and 1.7 for propranolol. Both flunarizine groups were at least as effective as propranolol (P< 0.001 in one-sided test). The percentage of responders (defined as subjects for whom attack frequency decreased by at least 50% compared to run-in) in the last 28 days of the double-blind period was 46% (118/259) for FLU 5 mg, 53% (141/264) for FLU 10 mg, and 48% (125/258) for propranolol. Statistical analysis showed that FLU 10 mg is at least as effective as propranolol (P< 0.001) and showed a trend for noninferiority of FLU5 and propranolol (P = 0.053). No statistically significant differences between the treatment groups were found for any of the secondary parameters. Overall, 190 subjects reported one or more adverse events during the run-in phase: 54 (20.5%) in the FLU 5 mg group, 76 (27.7%) in the FLU 10 mg group and 60 (22.3%) in the propranolol group. The results of this equivalence trial show that 10 mg flunarizine daily with a drug-free weekend is at least as effective as 160 mg propranolol in the prophylaxis of migraine for all evaluated parameters (one-sided equivalence tests) after 16 weeks of treatment. In addition, 5 mg flunarizine proves to be at least as effective as 160 mg propranolol when looking at the mean attack frequency for both the whole double-blind period and the last 28 days of treatment. However, in the analysis of responders, 160 mg propranolol seems to be slightly better than 5 mg flunarizine. In addition, no significant differences between the three treatments were found with regard to safety: all three treatments were generally well-tolerated and safe.
Introduction
Prophylactic treatment for migraine is used in cases where frequency and severity of attacks warrant such an intervention. Flunarizine is a calcium overload blocker, which has been widely used in the prevention of migraine attacks and has been shown to reduce attack frequency (1–22). The evidence for an effect on attack severity is less clear. The most common adverse reactions to flunarizine are drowsiness and weight gain, whereas in rare instances, depression (mostly in patients with a history of depressive illness) and extrapyramidal symptoms (mostly in the elderly) have been reported (23–25). There could be a relationship between the number of adverse reactions and the flunarizine plasma level during chronic treatment.
Propranolol, a beta-blocking agent has also frequently been used for migraine prophylaxis (14, 26–35) and is well known for its safety. It is, however, contra-indicated in patients with specific cardiovascular disorders or patients with asthma or diabetes mellitus. An alternative first-line drug for migraine prophylaxis is therefore often required.
In comparative trials in migraine, flunarizine (10 mg) and propranolol were found to have a similar efficacy (9, 10, 13, 14, 21, 21, 36, 37). The aim of this trial was to investigate whether the treatment of migraine could be improved by enhancement of the tolerability of flunarizine without compromizing its efficacy as already established in comparison with propranolol. The protocol therefore specified two ways of decreasing the dose of flunarizine: (1) introduction of a drug-free weekend and (2) halving the conventional dose.
Subjects and methods
Overall trial design and plan
The trial was a phase-IV comparative trial characterized by a placebo run-in phase followed by a double-blind period during which the subjects received either flunarizine (in two different dose regimens) or propranolol. The trial was approved by the ethics committees of all participating centres. All patients were informed about the aims of the study and gave written consent prior to entering the study.
The trial started with a single-blind placebo run-in phase of 4 weeks. At the end of this screening phase, subjects were randomized to one of three groups, flunarizine 5 mg, flunarizine 10 mg, or propranolol 160 mg, and the double-blind treatment phase was started. The double-blind phase lasted for 16 weeks during which visits were scheduled after every 4 weeks. These visits could deviate from the scheduled day by at most 7 days on the early or late side. Subjects who entered the run-in phase but were not randomized at Visit 1 could be replaced. Sufficient spare boxes containing placebo were provided to the investigator for these replacements. Those who were withdrawn during the double-blind period (i.e. after Visit 1) could not be replaced.
All subjects who entered the double-blind phase and prematurely discontinued the trial, regardless of the reason, had to be seen for a final evaluation. The Subject Completion/Discontinuation form had to be filled in. Subjects refusing to come in for final evaluation had to be asked in writing to come for a visit and to return all unused medication.
A placebo run-in phase was included in this trial to familiarize patients with trial procedures, establish a subject's competence to correctly fill in the diary, i.e. the capacity to discriminate between migraine attacks and interval headaches and establish a subject's eligibility for participating in the trial (occurrence of 2–6 migraine attacks during the run-in phase). A total of 260 subjects per group were expected to be needed to prove that each of the dosages of flunarizine (5 mg or 10 mg) was at least as effective as propranolol (160 mg) (see below).
Inclusion criteria were as follows:
Male or female.
Age 18–65 years.
Having experienced two to six migraine attacks every month for the preceding two months. The preceding two months (reference period) had to have been adequately documented with respect to the frequency of migraine attacks. If this was not the case, the subject was followed for two months prior to inclusion in the trial.
Migraine present for at least one year.
Migraine with aura (classic) or without aura (common) as defined by the International Headache Society (38).
Occurrence of interval headaches: permitted only if these attacks were well-recognized by the subject and if they did not occur more frequently than 6 days per month. The subject had to be able to differentiate migraine from an interval headache by the quality of pain (one-sided, pulsating, moderate to severe intensity) and/or by associated symptoms (nausea, discomfort to light or sound, visual symptoms or other aura).
Exclusion criteria were:
Use of prophylactic migraine therapy in the two preceding months (reference period).
Previous adequate (i.e. 160 mg propranolol or 10 mg flunarizine per day for at least 2 months) prophylactic use of propranolol or flunarizine without success.
History of depressive illness.
Extrapyramidal disorders.
Chronic obstruction airways disease, bronchospasm or asthma.
Heart failure, sinus bradycardia (<45 bpm), 2nd degree AV block, hypotension or peripheral vascular disease.
Serious diseases (diabetes, serious hepatic, renal, cardiovascular, respiratory or malignant illness).
Alcohol or drug dependence (documented or suspected).
Pregnancy, lactation or child-bearing potential without adequate contraception.
Absence of well-defined pain-free intervals of at least 24 h between headache attacks.
Absence of 2–6 migraine attacks during the run-in phase.
Poor compliance during the run-in period (<75% of trial medication taken).
Treatment
Placebo capsules were given to all subjects during the 4-week run-in. Thereafter, two groups of subjects received either lower-dose or higher-dose flunarizine capsules and one group of subjects received propranolol capsules for 16 weeks. The trial medication was provided by the Janssen Research Foundation (JRF). The treatment consisted of capsules, which were identical in appearance, and which contained 5 mg or 10 mg flunarizine, or 80 mg or 160 mg propranolol, or no active ingredient (placebo). After the placebo run-in, all subjects were assigned to one of three treatment groups according to a computer-generated randomization code prepared by the JRF. Subjects were assigned per centre in consecutive order, starting from the lowest number available. Throughout the trial and regardless of trial medication, all subjects took one capsule daily in the evening. During the 4-week run-in all subjects received placebo. During the 16-week double-blind period the flunarizine subjects received either 10 mg or 5 mg capsules. The 10 mg capsules were replaced by placebo capsules on 2 consecutive days out of 7 (‘drug-free weekends’) in weeks 9 through 16. The propranolol group was treated with 80 mg capsules (slow release) during the first week and with 160 mg capsules from the second week on. These dosing schemes are summarized in Table 1. All medication was randomized in blocks of 6 (2 propranolol+2 flunarizine 5 mg +2 flunarizine 10 mg). Each investigator was provided with 6 (or a multiple of 6) sequentially numbered boxes of trial medication. Prophylactic medication for migraine (such as beta-blockers, MAO-inhibitors, tricyclic antidepressants, methysergide) and regular use (>10 days/month) of symptomatic medication (ergotamine preparations, NSAIDs, psychotropics, alpha-agonists, gastrokinetics) was prohibited. Administration of hormonal treatment in women (including contraceptive pill) had to be kept constant throughout the trial. Rescue analgesic, ergotamine or sumatriptan medication for treating migraine attacks was allowed, but their use had to be adequately documented. Other concomitant treatments were allowed, provided the dose was kept constant throughout the trial whenever possible.
Dose scheme for the three treatment groups
At the start of the trial (Visit 0, baseline evaluation, start of week − 4), a complete medical history and specific migraine history was recorded and a general physical and neurological examination were performed. The migraine was classified as with or without aura. The subject was given a diary in which all migraine attacks and interval headaches had to be recorded. In addition, the severity of every migraine attack (mild-moderate-severe), the duration (hours) and any accompanying symptoms (nausea, vomiting, photophobia, phonophobia) had to be noted and their severity indicated. All symptomatic medication had to be recorded in the diary. From Visit 1 (start of double-blind period=start of week 1) through Visit 4 (end of week 12), the subject's diary was checked and collected. Subjects not having had two to six migraine attacks or well-defined pain-free intervals of at least 24 h during the baseline period were not included in the trial.
The primary efficacy parameters were the mean migraine attack frequency per 4 weeks, based on the number of attacks during the entire double-blind period, the mean attacks frequency during the last 28 days before the last dose, the percentage of responders (defined as subjects for whom the attack frequency decreased by ≥50% compared to run-in) per 4 weeks in the double-blind period, and the percentage of responders in the last 28 days of the double-blind period. For the double-blind period, a migraine attack which ended or was interrupted by sleep and then relapsed within 24 h had to be recorded as one attack and not two.
The secondary efficacy parameters were: mean duration of migraine attacks; severity (severe and/or moderate) of the attacks; total number of hours with migraine; percentage of migraine with accompanying symptoms (nausea, vomiting, photophobia, phonophobia) and their severity (severe and/or moderate); global treatment evaluation by the subject (poor-moderate-good-excellent) at endpoint; number of days with interval headaches; number of migraine attacks per 4 weeks treated with symptomatic treatment and mean time between two consecutive migraine attacks.
From Visit 1 on, the subject was asked whether he/she had experienced any adverse events. Special attention was paid to the occurrence of sedation, weight gain, extrapyramidal symptoms, arterial hypotension, bradycardia and depression. Systolic and diastolic blood pressure were measured. Pulse and weight were also recorded. Vital signs were recorded at visits 1–4 (inclusive). The trial was monitored according to the JRF standard operating procedure for the monitoring of clinical trials.
Statistics
If a 50% decrease in migraine frequency is considered to be a clinically relevant response, and a response rate of 70% was assumed in each treatment group (allowing for a clinically relevant difference of 10%), 260 subjects per group were expected to be needed to prove that flunarizine was at least as effective as propranolol at the 5%-level, with one-tailed testing and a power of 80%. One-tailed tests were used for the primary efficacy parameters and two-tailed tests were used for the secondary efficacy parameters with a significance level of 5%. For each subject, the endpoint value of a parameter was the last observed value for that subject on that parameter. The endpoint value was always analysed and printed on data listings. Week 6, 10, 14 were not planned in the protocol, and therefore not analysed, only listed. For efficacy analysis the following populations were defined:
Intention-to-treat (ITT): All randomized subjects with at least one intake of the trial medication and one postbaseline efficacy observation period of at least 1 month (with a minimum postbaseline duration of 21 days).
Per protocol: All ITT-included subjects excluding the protocol violators. For safety analysis all randomized subjects with at least one intake of the trial medication were considered.
The primary parameters were compared for different subgroups of the ITT population: Migraine with/without aura; migraine with/without interval headaches: subjects were allocated to the ‘with interval’ group if at least one of the attacks recorded in the diary was marked as being an interval headache; Completers: subjects completing the trial (16 weeks), defined as having a visit to the doctor after day 97; Subjects that had taken the concomitant medication sumatriptan.
Baseline between-group comparability with respect to demographic variables and efficacy parameters were assessed. Descriptive statistics for each treatment separately and for the total population were provided. The Kruskall-Wallis statistic was applied to investigate treatment comparability with respect to continuous variables. Whenever P-values <0.1 occurred, the Mann–Whitney-U statistic was applied to test which treatment groups differed from one another. The Chi-square test was used to test for intergroup differences in categorical variables. The results of the primary efficacy parameters are described by one-tailed tests, and the results of the secondary efficacy parameters are described by two-tailed tests. Furthermore, the only planned primary efficacy parameters were the mean number of migraine attacks per 4 weeks, based on the number of attacks during the entire double-blind period and the number of responders. Additionally, these parameters (mean number of migraine attacks and the number of responders) were also analysed for the last 28 days before last intake (since efficacy of the evaluated treatments was expected to gradually increase in the course of treatment). Finally, an exploratory subgroup analysis was performed of the primary efficacy parameters, which was not planned in the protocol. The subgroups were: migraine with vs. migraine without aura, migraine with vs. migraine without interval headaches, completers vs. noncompleters, and the use of sumatriptan. For all variables, the following rule applied: If the time between two subsequent attacks of the same type (interval headache/migraine) did not exceed 24 h, the attacks were counted as one. The worst score was taken for attack severity and symptom severity. The duration of the attack was the sum of the different durations. The start date was used to determine to which period an attack belonged. Only observations in the diary with a start date between the start date of the selection phase and the date of the last dose were used in the analysis. If no date of last dose was available, the last visit date was used to define the cut off point. Descriptive statistics for each treatment separately per interval were performed. The Hauck-Anderson procedure was performed to study the equivalence of the treatment groups for the frequency of migraine attacks per 4 weeks (‘number of attacks during the last 28 days’). A difference of 1 attack was considered as clinically relevant. The Chi-square test was used to test intergroup differences in the number of responders. The Blackwelder procedure was used to test the equivalence of the response rate observed for the two groups at the end of the trial.
Propranolol was compared with each of the dose groups of flunarizine separately, and the two doses of flunarizine were also compared with one another. If the assumptions were fulfilled, the difference between treatments was evaluated using an analysis of covariance (
Results
Patient disposition
The trial was run from 13 April 1992 to 29 March 1996. Overall 130 investigators in 8 countries participated in the trial (see Appendix). In total, 826 subjects were recruited; of the 810 randomized, 263 were assigned to flunarizine 5 mg (FLU 5 mg), 275 to flunarizine 10 mg (FLU 10 mg; with a drug-free weekend), and 270 to propranolol. Another 2 subjects did not receive treatment (in total 808 subjects had at least one intake of trial medication). Of the 808 patients who had at least one intake of trial medication 25 patients (4 in the FLU 5 mg group; 10 in the FLU 10 mg group with drug-free weekend; 11 in the Prop group) had no. 1-month postbaseline efficacy observation and were not included in the ITT analysis. Therefore the ITT population includes 783 patients. One investigator was suspected of fraud. Therefore, a second analysis was performed of the efficacy parameters excluding the subjects of this investigator. However, since the fraud was only suspected (no prove was available) and also because the results of the primary efficacy parameters were not altered when the subjects of this investigator were excluded, the results of this second analysis are not shown in this publication.
Of the 808 subjects treated, 142 discontinued prematurely: 44 subjects from the FLU 5 mg group, 53 from the FLU 10 mg group and 45 from the propranolol group (Table 2). The main reasons for premature discontinuation were adverse events (58/142, 40.8%), lost to follow up (29/142, 20.4%), withdrawal of consent (20/142, 14.1%), uncooperative (16/142, 11.3%) and insufficient response (15/142, 10.6%). No statistically significant intergroup difference in the number of discontinuations was observed (P=0.660). Apart from premature discontinuations, described above, major protocol deviations (baseline disease conditions out of limits; intercurrent treatment; inconsistencies in CRF/diary) were noted in 18 subjects.
Main features of the subject sample and patient disposition
The demographic data of the 808 subjects randomized and treated were as follows: overall, 81.4% of the subjects were female, and the median ages in the three groups was 37 years with minima and maxima ranging from 17 to 66 years. The majority of subjects was Caucasian (97.9%). Demographic data were not significantly different between the three groups. Overall, 71.8% of the subjects were diagnosed as having migraine without aura, and the median number of years with migraine was 10.0 years in all groups with minima and maxima ranging from 0.8 to 57 years. The median duration of an attack was 15.5 h, with minima and maxima ranging from 1 to 168 h. Provoking factors could be identified in about half of the subjects (49.8%). The median number of attacks per month based on historical recall was 4 in all groups, with minima and maxima ranging from 2.0 to 20.0 attacks. The median number (95% CI) of attacks during the run-in period was also comparable between the treatment groups: 2.9 (2.6–2.9) for the FLU 5 mg, 2.9 (2.5–2.9) for the FLU 10 mg and 2.8 (2.5–2.9) for the propranolol group.
The majority of attacks were moderate (40.5%) or severe (58.6%). Nausea was moderate in 46.8% of the subjects and severe in 26.4%. Vomiting was absent in 35.0% of the subjects; if present, it was of moderate severity in 28.8% of the subjects and mild in 24.4%. Photophobia was often present (93.3%) and mainly of moderate (41.8%) or severe (30.4%) severity. Phonophobia was moderate in 36.9% of the subjects, severe in 26.2% and mild in 24.6%. No statistically significant intergroup differences were observed. Overall, less than 10% of the subjects had previously used prophylactic treatments for their migraine. Those who previously did use prophylactic treatment, had good (38.7%) or moderate (51.6%) results with flunarizine, good (43.8%) or moderate (50.0%) results with propranolol, mainly moderate (60.0%) results with alpha-agonists, mainly poor (50.0%) results with methysergide and moderate (45.3%) to poor (33.3%) results with other treatments. Concomitant treatments were not different in the three treatment groups. 19.6–26.5% of patients treated their migraine attcks with sumatriptan.
Efficacy
Overall, 783 subjects were included in the intent-to-treat analysis: 259 in the FLU 5 mg group, 265 in the FLU 10 mg group, and 259 in the propranolol group. A total of 142 subjects discontinued from the trial prematurely (44 in the FLU 5 mg group, 53 in the FLU 10 mg group with drug-free weekends, 45 in the Prop treatment group). Table 3 and Fig. 1 summarize the mean number of attacks per 4 weeks during the double-blind period. Statistical testing (one-sided) showed that both flunarizine treatments were at least as effective as propranolol with respect to frequency of migraine attacks per 4 weeks during the double-blind period (P<0.001). Table 3 also summarizes the mean number of attacks during the last 28 days before the last dose intake. Statistical testing showed that both flunarizine treatments were at least as effective as propranolol with respect to frequency of migraine attacks over the last 28 days before last dose intake (P<0.001). Analysis of the number of responders (defined as subjects for whom the attack frequency in the last 28 days before the last dose intake decreased by ≤50% compared to run-in) showed that in the FLU 5 mg group 118/259 (46%) subjects were responders, 141/264 (53%) in the FLU 10 mg group and 125/258 (48%) in the propranolol group (Table 3, Fig. 2). Two patients (one in the FLU 10 mg group with drug-free weekend and one in the Prop treatment group) were excluded from this analysis because of missing information about the attack frequency in the 28 day-period preceding the last drug intake. Statistical testing (Blackwelder test with a maximum clinical allowable difference of 10%) showed that FLU 10 mg was at least as effective as propranolol (P<0.001 in one-sided test), and showed a trend for noninferiority of FLU 5 and propranolol (P=0.053 in one-sided test). The observed response rate of approximately 50% was lower than the expected 70% on which the power calculations were based, reducing the test's power. Analysis of the number of responders (defined as subjects for whom the attack frequency decreased by ≤50% compared to run-in) per 4 weeks in the double-blind period (ITT) showed that in the FLU 5 mg group 92/259 (36%) subjects were responders, 116/264 (44%) in the FLU 10 mg group and 113/258 (44%) in the propranolol group. Statistical testing (Blackwelder test with a maximum clinical allowable difference of 10%) showed that FLU 10 mg was at least as effective as propranolol (P=0.010), but that FLU 5 mg is inferior to propranolol (P=0.344 in one-sided test).
Efficacy parameters in the intention-to-treat population (ITT)
∗Responders: decrease in attack frequency with at least 50% as compared to run-in.
†Results demonstrated significant equivalence for both flunarizine groups with the propranolol group (P < 0.001 in one-sided equivalence test).
‡Statistical analysis showed that FLU 10 mg is equivalent to propranolol (P = 0.010 in one-sided equivalence test), and that FLU 5 mg is not equivalent to propranolol (P = 0.344 in one-sided equivalence test).
§Statistical analysis showed that FLU 10 mg is equivalent to propranolol (P < 0.001 in one-sided equivalence test), and showed a trend for equivalence of FLU 5 mg and propranolol (P = 0.053 in one-sided equivalence test).
Mean number of migraine attacks in the baseline period (□), per 4 weeks during the double-blind period () and the last 28 days of treatment (▪). Mean values and 95% confidence intervals.
Percentage of responders (>50% reduction in migraine frequency) during the double blind treatment period (□) and the last 28 days of treatment (▪).
Secondary efficacy variables
Mean duration of migraine attacks: no statistically significant intergroup differences were observed regarding the mean duration of migraine attacks for any of the time intervals analysed and only for the FLU 5 mg group a statistically significant time effect was observed (P=0.004). The Page test for decreasing scores confirmed the significant decrease in mean duration of migraine attacks over time for the FLU 5 mg group (P<0.001). For the changes in duration of migraine attacks when compared to run-in, no statistically significant intergroup differences were noted, and no time effect was observed in any of the treatment groups. These results clearly show that onset of action was comparable between the three treatments. The mean number of severe migraine attacks did not differ statistically significantly between the treatment groups at any point in time and for all three treatment groups a clear time effect was observed (P=0.001): the number of severe attacks decreased in the course of the treatment. At run-in, the mean number of severe attacks was 1.1 (±0.08) for the FLU 5 mg group, 1.1 (±0.07) for the FLU 10 mg group, and 1.0 (±0.07) for the propranolol group. At endpoint these numbers were 0.5 (±0.08), 0.4 (±0.06) and 0.4 (±0.06), respectively. For changes in the number of severe attacks when compared with run-in, no intergroup differences were observed at any of the time points, and a time effect was significant for both flunarizine treatments but not for propranolol. No significant intergroup differences in the number of hours with migraine were observed: the mean number of hours ranged from 45 : 35±1 : 36 h at run-in to 18 : 52±1 : 36 h at endpoint. Intergroup comparison of changes from run in the number of hours with migraine at all time points again showed no statistically significant differences between the treatments. No significant intergroup differences in the percent of migraine attacks with nausea were observed, except for a difference between FLU 10 mg (mean percentage 72.9±2.21) and propranolol (mean percentage 66.5±2.38) at run in. Analysis of differences with run-in demonstrated statistically significant intergroup differences at 60 days between FLU 10 mg and propranolol (P=0.008), and at 90 days between FLU 10 mg and propranolol (P=0.040). No significant intergroup differences in the percent of migraine attacks with vomiting were observed, nor did analysis of differences with run-in demonstrate statistically significant intergroup differences. No significant intergroup differences in the percent of migraine attacks with photophobia or phonophobia were observed, nor did analysis of differences with run-in demonstrate statistically significant intergroup differences.
No significant intergroup differences were observed for the other secondary endpoints (for details see Table 4). No statistically significant intergroup differences were observed regarding the global evaluation by the subject at endpoint. Overall, treatment was evaluated as excellent by 209/740 (28%) of the subjects, as good by 101/740 (14%) of the subjects, as moderate by 338/740 (46%) of the subjects and as poor by 92/740 (12%) of the subjects. Table 5 summarizes the results per treatment group.
Secondary efficacy parameters in the intention-to treat (ITT) population
ns, no significant differences.
Global evaluation of the prophylactic therapy by the subject at endpoint
No significant intergroup differences in the number of days with interval headaches were observed, except for a statistically significant difference between both flunarizine treatments at 90 days (P=0.047). Analysis of differences with run-in demonstrated no statistically significant intergroup differences. No significant intergroup differences in the number of migraine attacks per 4 weeks treated with symptomatic treatment were observed except at 30 days between FLU 5 mg and propranolol (P=0.041) and between FLU 10 mg and propranolol (P=0.033). Analysis of differences with run-in demonstrated no statistically significant intergroup differences. No significant intergroup differences in the mean time between two consecutive migraine attacks were observed.
Safety and tolerability
Overall, 190 subjects reported one or more adverse events during the placebo run-in phase (safety data related to the run in period was missing for two patients: one in the FLU 10 mg with drug-free weekend and one in the Prop treatment group): 54/263 (20.5%) in the FLU 5 mg group, 76/274 (27.7%) in the FLU 10 mg group and 60/269 (22.3%) in the propranolol group. During the double-blind phase a total of 264 subjects reported adverse events: 88 subjects in each treatment group. Table 6 summarizes the most frequent adverse events during the run-in and double-blind phase, by severity and relation to trial medication. The most frequent adverse event during the run-in phase were weight increase, migraine, and fatigue, all reported by at least 3% of the subjects in any treatment group. During the double-blind phase, the most frequent adverse events were weight increase, fatigue, dizziness, injury, rhinitis, depression, somnolence, abdominal pain, and hypotension. During the placebo run-in period a remarkable high incidence of weight increase was reported. This adverse event was more frequently reported during the run-in than during the double-blind phase. During the double-blind period, weight increase was more frequently reported in the FLU 5 mg group than in the FLU 10 mg group, which may indicate that this adverse event is not dose related. The difference was not significant. Depression occurred in all treatment groups during the double-blind period, it was reported in 2 subjects (0.7%) in FLU 10 mg, in 7 subjects (2.7%) in FLU 5 mg group, and in 5 subjects (1.9%) in the propranolol group. No extrapyramidal symptoms were observed in any of the subjects treated in this trial. The majority of adverse events were mild or moderate. A large proportion of the most frequent adverse events were assessed as possibly or definitely drug-related in some or all three treatment groups, e.g. weight increase, fatigue, dizziness, depression, abdominal pain, nausea, somnolence, hypotension, and headache. In the run-in phase, only one subject reported a serious adverse event: an injury was noted in the propranolol group. In the double-blind phase 1 subject in the FLU 5 mg group reported 2 serious adverse events: malaise and vertigo. In the FLU 10 mg group, 5 subjects reported a serious adverse event: urinary incontinence (n=1), injury (n=1), cholelithiasis (n=1), breast neoplasm (n=1) and depression (n=1). In the propranolol group 2 subjects reported a serious adverse event: one injury and 1 menstrual disorder. Overall, 58 subjects discontinued prematurely due to adverse events: 21 in the FLU 5 mg group, 19 in the FLU 10 mg group and 18 in the propranolol group. During the run-in phase weight increase (n=5), fatigue (n=3) and somnolence (n=3) were the most frequent reasons for treatment discontinuation. During the double-blind phase fatigue (n=8), weight increase (n=7), depression (n=5), nausea (n=4) and headache (n=3) were most frequently noted as reasons for discontinuations.
Saftey data from the trial
∗ n, number of subjects with data.
Vital signs and physical findings
Details on systolic blood pressure, diastolic blood pressure and heart rate are presented in Table 7. The changes in systolic and diastolic blood pressure and heart rate were clinically irrelevant. No consistent or clinically relevant changes occurred in any of the treatment groups with respect to the subject's physical appearance or neurological examination. For all treatment groups, mean weight increased significantly (P=0.001) during the trial: at baseline mean (±SE) weight for subjects from the FLU 5 mg group was 66.7 (±0.78) kg, 66.6 (±0.74) kg for those from the FLU 10 mg group and 65.2 (±0.72) kg for propranolol subjects. At endpoint these values were 68.2 (±0.77) kg, 68.6 (±0.74) kg, and 66.1 (±0.71) kg, respectively. Clinically relevant changes in weight were defined as increase or decrease by at least 5% when compared to baseline. At endpoint, in a total of 17 subjects weight had significantly decreased (6/263 on FLU 5 mg, 4/275 on FLU 10 mg and 7/270 on propranolol). In 167 (21%) subjects weight had increased significantly at endpoint in comparison to baseline: in 60/263 (23%) FLU 5 mg subjects, 67/275 (24%) FLU 10 mg subjects, and in 40/270 (15%) propranolol subjects.
Additional safety observations
∗ n, number of subjects with data.
†Clinically relevant change: change with at least 20% when compared to baseline for systolic blood pressure, and 10% for diastolic blood pressure and heart rate.
‡Clinically relevant change: weight increased or decreased with at least 5% when compared to baseline.
Discussion
This was a phase-IV double-blind equivalence trial designed to assess the efficacy and tolerability of two doses of flunarizine (10 mg o.d.+drug-free weekend, and 5 mg o.d.) in the prophylaxis of migraine, in comparison with the frequently used drug propranolol (160 mg o.d.). A total of 808 subjects were treated (treatment period was 16 weeks) and 142 subjects discontinued the trial prematurely, mainly because of adverse events (n=58).
For first two primary efficacy parameters, mean attack frequency per 4 weeks in the entire double-blind period and in the last 28 days of treatment, both 5 mg and 10 mg flunarizine were at least as effective as 160 mg propranolol (P<0.001). Analysis of the other primary efficacy parameters, the number of responders (defined as subjects for whom attack frequency decreased by at least 50% compared to run-in) in the double-blind period and in the last 28 days of treatment, showed that flunarizine 10 mg was at least as effective as 160 mg propranolol. Flunarizine 5 mg was inferior to propranolol for the number of responders in the double-blind period (P=0.344), although a trend for noninferiority of 5 mg flunarizine was observed for the number of responders in the last 28 days of treatment (P=0.053).
The results were similar for all subgroup analyses: for subjects with migraine with or without aura, with or without interval headaches, completers vs. noncompleters and those who did or did not use sumatriptan. Equivalence between the three treatments was demonstrated in all subgroups of sufficient sample size on the basis of mean attack frequency, but not for responders. No statistically significant differences between the treatment groups were found for any of the secondary parameters: mean duration of migraine attacks, number of severe and number of moderate or severe attacks, total number of hours with migraine, percent of migraine attacks with nausea, with vomiting, with photophobia, and with phonophobia, percent of severe and severe or moderate migraine attacks, percent of migraine attacks with at least one severe symptom and with at least one moderate or severe symptom, global evaluation by the subject at endpoint, number of days with interval headaches, number of migraine attacks treated with symptomatic treatment, and mean time between two consecutive migraine attacks. The equivalence of propranolol and flunarizine is also found, when success rates for the two substances are compared across trials (9, 14, 17, 21, 22, 29). Holroyd et al. (10) published a metaanalysis of clinical trials with flunarizine or propranolol as an abstract. They included data from 31 flunarizine trials and 32 propranol trials with about 3000 treated patients. The modal propranolol dose was 160 mg/day and the modal dose of fluanrizine 10 mg/day. On average, the trials lasted 16 weeks. The mean improvement in migraine frequency (flunarizine − 62%; propranolol − 55%) was similar. Propranolol therapy was associated with more drop outs (18%) compared to flunarizine (13%). The response to flunarizine and propranolol was negatively associated with the chronicity of the migraine disorder. The lower responder rate in the present study was probably due to a low attack frequency at baseline. Our study also had a low patient satisfaction with fewer than 50% rating any of the treatments ‘good’ or ‘excellent’. A major shortcoming of this trial is the lack of a placebo group. This trial was initiated in Germany at a time when some ethics committes refused placebo groups in migraine trials with the argument that effective therapies were available. A number of trials both in acute therapy of migraine attacks and migraine prophylaxis were performed consecutively without placebo groups. In the meantime this has changed and it is again possible to perform placebo-controlled trials.
Overall, 190 subjects reported one or more adverse events during the run-in phase: 54/263 (20.5%) in the FLU 5 mg group, 76/274 (27.7%) in the FLU 10 mg group and 60/269 (22.3%) in the propranolol group. During the double-blind phase a total of 264 subjects reported adverse events: 88 subjects in each treatment group. The most frequent adverse event during the run-in phase were weight gain, fatigue, migraine and headache. Weight increase is of particular interest. This adverse event was more frequently reported during the run-in than during the double-blind phase. Patients knew from the information sheet, that weight gain could occur. We did not ask wether this affected their eating behaviour. Furthermore, a reduction of the flunarizine dose from 10 mg to 5 mg did not result in a lower incidence of weight increase. Finally, the mean weight increases as observed with 5 or 10 mg flunarizine were not significantly different from those observed with propranolol.
Depression occurred in all treatment groups, and reducing the dose of flunarizine did not result in a lower incidence: depression was reported in 0.7% in FLU 10 mg, in 2.7% in FLU 5 mg group, and in 1.9% in the propranolol group. In most of the cases, the depression was reported to be (possibly) trial drug-related. This adverse event is known from earlier trials (39) and led to discontinuation in 2 subjects in the FLU 10 mg group, 1 in the FLU 5 mg group, and 2 in the propranolol group. Overall, the incidence of depression seen in the flunarizine treatment groups was not different from that seen in the propranolol group.
The results of this equivalence trial show that 10 mg flunarizine daily with a drug-free weekend is at least as effective as 160 mg propranolol in the prophylaxis of migraine for all evaluated parameters (one-sided equivalence tests) after 16 weeks of treatment. In addition, 5 mg flunarizine proves to be at least as effective as 160 mg propranolol when looking at the mean attack frequency for both the whole double-blind period and the last 28 days of treatment. However, in the analysis of responders, 160 mg propranolol seems to be slightly better than 5 mg flunarizine. In addition, no significant differences between the three treatments were found with regard to safety: all three treatments were generally well-tolerated and safe. We, however, failed to show, that a lower dose of flunarizine reduces the number of side-effects and has better tolerability.
Footnotes
Acknowledgements
This study was supported by a grant from Janssen Beerse, Belgium. The trial was monitored by BRI International. Ragheno Business Center, Motstraat 54, B-2800 Mechelen, Belgium. The study protocol was developped by the principal investigator (HCD).