Abstract
Dear sir Silberstein and Goadsby (1) recently reviewed preventive medication used in the treatment of migraine. In Table 3 of their paper, preventive drugs were assigned to one of four categories depending on their efficacy. These categories were ‘high efficacy’, ‘low efficacy’, ‘unproven efficacy’, and ‘proven not effective’. The medications in the latter category were ‘proven to have limited or no efficacy’. Based on the results of two studies, carbamazepine was placed in this category. I do not believe this is justified.
In a double-blind cross-over study, carbamazepine was compared with placebo in the prevention of migraine (2). The average number of migraine attacks reported on carbamazepine during a 6-week period was 0.65 per patient and the corresponding average number for placebo was 3.87 per patient (highly statistically significant). According to the International Headache Society, this is the preferred parameter in controlled trials of preventive drugs in migraine (3). Compared with base-line data, 38 (84.4%) of 45 patients improved on carbamazepine, while 13 (27.1%) of 48 patients improved on placebo (P < 0.001). It is possible to perform an intention-to-treat analysis based on the secondary outcome ‘improvement’. This analysis also favours carbamazepine.
The paper by Rompel and Bauermeister (2) can be criticised, mainly because of inadequate reporting (4). The report does not comply with modern standards for how a manuscript on a randomised clinical trial should be presented (5). However, using the validated instrument developed by Jadad et al. (6) to assess the quality of clinical reports on pain relief, the study receives a score of five out of five. This means that the methods of randomization and blinding were appropriate, and withdrawals and drop-outs were sufficiently described. Concealment of the randomization code was adequate. There is empirical evidence that these factors influence the estimates of effects in clinical studies (7, 8).
The second study compared pindolol, clonidine and carbamazepine (9). This study was not blinded or placebo controlled, and it is unclear whether it was randomised. The authors claimed that the response to pindolol was better than the response to carbamazepine, and the response to clonidine was not significantly better than that to carbamazepine. This study receives, at the most, a score of one out of five on the Jadad et al. scale (6).
In the review by Silberstein and Goadsby (1), the studies that included carbamazepine do not prove that this drug is ineffective, nor do they prove that it has little effect in the preventive treatment of migraine. The study with the highest quality indicates that carbamazepine may have large preventive effects, and we need more data to better estimate the magnitude of these effects.