Tricyclic antidepressants for migraine and tension-type headaches.

In the linked meta-analysis (doi:10.1136/bmj.c5222), Jackson and colleagues assess the efficacy of tricyclic antidepressants in the treatment of migraine, tension headache, and mixed headache .1

Tricyclic antidepressants have a long history in the treatment of headache. In 1964 Lance and Curran reported a better response to amitriptyline (30-75 mg/day) in chronic tension headache (n=280) compared with 11 other commonly used drugs (such as benzodiazepines, vasodilators, and sedatives).2 They also reported a placebo controlled crossover trial of 27 patients that showed clinically significant improvements (≥50% reduction) in headaches in 55% of patients taking amitriptyline, but only 11% of those taking placebo. When the results for amitriptyline were aggregated across the two studies, the effect of amitriptyline became more pronounced over time, was independent of the presence of depression, was evident in patients with continuous as well as episodic headaches, and was most pronounced in older (≥60 year of age) patients. A decade passed before similar positive findings were reported for migraine.3

With replication of these findings tricyclic antidepressants, particularly amitriptyline, have been recommended in textbooks for at least 35 years. With the advent of formal treatment guidelines for headache, tricyclics also have been recommended in clinical guidelines for the treatment of tension-type headache and migraine.4 5 6

The current meta-analysis updates the evidence base for tricyclic antidepressants in the treatment of migraine and tension headache.1 Results from the 37 trials of tricyclic antidepressants are analysed; the 20 placebo controlled trials primarily evaluate amitriptyline (14 trials) or clomipramine (four trials). The meta-analysis largely confirms Lance and Curran’s original observations. Across trials, low dose tricyclic antidepressants (mean amitriptyline dose 80 mg/day) reduced headache by at least 50% compared with placebo (tension: relative risk 1.41, 95% confidence interval 1.02 to 1.89; migraine: 1.80, 1.24 to 2.62). The proportion of people who stopped treatment did not differ significantly between people taking tricyclic antidepressants or placebo. Treatment effects increased over time. Other preventive drugs (topiramate or β blockers) showed no advantage over tricyclic antidepressants. Tricyclic antidepressants were significantly more effective than serotonin reuptake inhibitors (tension (four trials): 1.73, 1.34 to 2.22; migraine (five trials): 1.72, 1.15 to 2.55), although dry mouth, drowsiness, and weight gain were also significantly more common with tricyclics.

Conclusions that can be drawn from meta-analyses depend on the number and quality of available trials. As the authors rightly point out, the number of studies was not large and most were small and of short duration (average 11 weeks). Moreover, most trials would not meet current methodological standards because 80% (16/20) of placebo controlled trials were completed at least 20 years ago.7 8 As a result, convincing evidence is available for only the most general conclusion: amitriptyline is more effective than placebo for migraine and tension headache. Amitriptyline also seems to be more effective than serotonin reuptake inhibitors, although few direct comparisons are available.

After a half century of research the most important clinical questions remain unanswered. Is the observed effect of amitriptyline a true class effect of tricyclic antidepressants or specific to a subset of tricyclics? Are newer selective dual action (serotonin and noradrenaline) antidepressants as effective as tricyclic antidepressants with, potentially, a more favourable side effect profile? How does the effectiveness of tricyclic antidepressants compare with other preventive drugs or with non-drug treatments? Which people are the best candidates for tricyclic antidepressants rather than other preventive drugs or non-drug treatments? Can tricyclic antidepressants be beneficially combined with other preventive drugs or non-drug treatments in people who do not respond to monotherapy? Secondary analyses of available clinical trial data can yield answers to some of these questions.9 However, just a few large properly designed trials might answer many of these questions, as well as other clinically relevant questions.

Unfortunately, a 50 year history indicates that such trials will not be conducted. Incentives are too few and disincentives too substantial for the drug industry to conduct such trials. The establishment of a combination of incentives and requirements to encourage the industry to provide information about comparative effectiveness and other clinically relevant questions during premarket approval or postmarket research is unlikely to occur. Thus, support from government health agencies or from other healthcare players with an interest in identifying cost effective treatments will be needed in addition to existing examples of government funded trials.10 11