Neuromuscular dose-response studies: determining sample size

Background. Investigators planning dose-response studies of neuromuscular blockers have rarely used a priori power analysis to determine the minimal sample size their protocols require. Institutional Review Boards and peer-reviewed journals now generally ask for this information. This study outlines a proposed method for meeting these requirements. Methods. The slopes of the dose-response relationships of eight neuromuscular blocking agents were determined using regression analysis. These values were substituted for g in the Hill equation. When this is done, the coefficient of variation (COV) around the mean value of the ED50 for each drug is easily calculated. Using these values, we performed an a priori one-sample two-tailed t-test of the means to determine the required sample size when the allowable error in the ED50 was varied from +/- 10-20%. Results. The COV averaged 22% (range 15-27%). We used a COV value of 25% in determining the sample size. If the allowable error in finding the mean ED50 is +/- 15%, a sample size of 24 is needed to achieve a power of 80%. Increasing 'accuracy' beyond this point requires increasing greater sample sizes (e. g. an 'n' of 37 for a +/- 12% error). Conclusions. On the basis of the results of this retrospective analysis, a total sample size of not less than 24 subjects should be adequate for determining a neuromuscular blocking drug's clinical potency with a reasonable degree of assurance.