Olanzapine in Acute Bipolar Mania

I read with interest the article by Tohen et al,¹ which has clearly demonstrated the efficacy of olanzapine in the treatment of acute bipolar mania; however, this study¹ has certain methodological limitations. First, the authors mentioned that a minimum baseline total Young–Mania Rating Scale (Y-MRS)² score of at least 20 was a prerequisite on the day of randomization; however, they included a single patient whose baseline total Y-MRS score was 14. No reason has been given for this inclusion in their methodology. Second, although the authors used both the Y-MRS and the 21-item Hamilton Psychiatric Rating scale for Depression (HAMD-21) to evaluate psychopathology, both the definition of improvement and the final analysis (ie, improvement from baseline to end point) were restricted only to Y-MRS scores. Since 42.6% of the study subjects had mixed affective episodes, and the Y-MRS does not evaluate depressive symptoms, both the outcome definition and the final analysis of the Tohen et al study seem less meaningful. The need to consider depressive symptoms in the global outcome definition and in the final analysis is also evident from their Table 3,¹ in which significant group differences were noted only in Y-MRS and Clinical Global Impression mania scores but not in HAMD-21 total and Clinical Global Impression depression scores. Third, in this study,¹ the Fisher exact test was used inappropriately. The Fisher exact test is indicated when sample size is equal to or less than 20 or if the smallest expected frequency is less than 5 when sample size is between 20 and 40. In other situations, a χ² test with or without continuity correction is used.³ Hence, I believe that using the Fisher exact test to analyze variables such as number of patients who completed the study, sex, psychotic features, type of episode, rapid cycling, and lifetime substance use might have reduced the statistical vigorousness of the Tohen et al study. In my view, use of a more appropriate outcome definition, final analysis, and statistical tests would have made the study more meaningful.