Objective.—Comparison of use of therapeutic gain (TG) and therapeutic ratio (TR) for relative assessments of selective 5HT1B/1D agonist efficacy in the acute treatment of migraine.
Background.—Comparison of selective 5HT1B/1D agonist efficacy in the acute treatment of migraine is complicated by the impracticality of conducting high-quality head-to-head active comparator trials for all permutations of drug, dose, and route of administration; the clinical trials that are available are usually noncontemporaneous, and have fluctuating active and placebo response rates (ARR and PRR, respectively) for the necessarily subjective endpoints. The standard primary endpoint is conversion of headache score 2 (moderate pain) or score 3 (severe pain) into score 0 (no pain) or score 1 (mild pain) at a single timepoint (usually 2 or 4 hours postdose). Two principal methods have been used for comparison of ARR between studies: TG = ARR – PRR and TR = ARR/PRR. Secondary endpoints can be treated in the same way, although the 24-hour remedication rate is the only secondary endpoint that is subject to regulatory authority standardization. Does either TG or TR have an advantage over the other?
Methods.—Efficacy data were collected from published sources and included pain score conversion at 2- and 4-hour, as well as 24-hour remedication rates. TG and TR were calculated for each endpoint, as shown above. The ranges of these variables were found from all observed measures. Correlations between TG and TR were found using least squares methods. The 95% confidence intervals (CI) for TG and TR were then found for all clinical trials, and for the whole range of observed PRR.
Results.—Values for TR and TG generally were well correlated for pain score conversion at 2 hours postdose (n = 51 clinical trials). Deviations from this correlation were greatest for drugs that were evidently very efficacious. When standardized for the observed ranges of PRR and ARR, the 95% CI for TR were narrower than for TG. A subset of these studies also reported 4-hour data (n = 23 clinical trials), but the correlation between TR and TG was again good, with narrower 95% CI for TR than TG. For the 24-hour remedication data (n = 18 trials), PRR was less variable; TR and TG again correlated well, but there was little difference in their relative 95% CI.
Conclusions.—For comparison of drugs in noncontemporaneous studies, when PRR fluctuates, relative ranges for TR are narrower than for TG. For assessing single clinical trial results (ie, point estimates of efficacy) TR will generally be relatively nearer to its mean than TG. For both these reasons, TR is a more robust statistic than TG.