The scientific basis of medication choice in symptomatic migraine treatment.

With the rapid advances in the treatment of acute attacks of migraine in the last few years and a number of new treatments, has come the practical clinical problem of comparing emerging acute attack therapies alone and with regard to current treatments. Acute migraine therapies can usefully be regarded as non-specific and specific, from the perspective of migraine, since some medicines, such as aspirin or paracetamol, are used to treat pain more broadly. In this review I will compare both non-specific and specific compounds. To some extent the introduction into trial then clinical use of sumatriptan, the first of the 5HT1B/1D agonists or triptans, brought new standards in both clinical trial design, and execution and clinical outcome. Thus sumatriptan has become the de facto gold standard and will be thus employed here. To be practical the discussion of the new triptans will be limited to those available widely, naratriptan, rizatriptan and zolmitriptan. There are two broad issues when comparing treatments: what end-point should be considered and then, how can different compounds be compared with respect to that end-point. In terms of end-points those used here relate to pain relief because they have been collected robustly in the clinical studies and, fortunately, rapid pain relief is what patients questioned in population-based studies rate highest in an acute attack medicine. Headache pain has been rated on a scale of nil, mild, moderate and severe and success rated as either a response, nil or mild pain, or headache free, nil pain, at two or four hours. The ideal comparison of the triptans would be a randomized controlled clinical trial directly comparing the medicines in each case. Given that these are not available for all the compounds and the well characterised placebo response in acute migraine studies, summary measures have been developed to express the differences between compounds to try and adjust for the varying placebo effect. The two most widely used are the therapeutic gain, response on active medication minus response on placebo, and the number-needed-to-treat (NNT). The NNT is the reciprocal of the therapeutic gain as a proportion. The strengths and weaknesses of this approach will be discussed, including the importance of the calculation of confidence intervals. It can be concluded that our current instruments are rather blunt and patient preference needs much greater study.