Significance of Efficacy Results in a Randomized Clinical Trial

To the Editor Dr Tzanetakou and colleagues reported on the results of a placebo-controlled clinical trial of anakinra in hidradenitis suppurativa. The primary efficacy end point was based on decreased disease activity scores from baseline to the end of treatment, and the 2 study arms were also compared regarding the Hidradenitis Suppurativa Clinical Response (HiSCR). The Fisher exact test was performed for both the primary analysis and the HiSCR analysis, as stated in the "Statistical Analysis" section. For the primary analysis, the disease activity score was decreased in 20% (2 of 10) of the placebo arm compared with 67% (6 of 9) of the anakinra arm. For the HiSCR analysis, response was achieved in 30% (3 of 10) of the placebo arm and 78% (7 of 9) of the anakinra arm. A P value of .04 was reported for both the primary analysis and the HiSCR analysis. However, our calculations show that the P value for both analyses should be .07. A χ² test would have resulted in a P value of P = .04 for both analyses. However, a χ² test is known to inflate the type I error rate when the sample size is small and is therefore not appropriate for these analyses (see eg, McDonald, pages 88-89).

http://ift.tt/2vFOWQF