Analysis of statistical tests to compare doses of analgesics among groups

Abstract

Background: A common type of anesthesiologic study determines the effect of an intervention on the doses of analgesics given to groups of patients. The goal of this study was to evaluate the effectiveness of several appropriate statistical tests to examine the results of such studies. Methods: Total doses of morphine received in a postanesthesia care unit were recorded for patients undergoing vitrectomy (244), laparoscopic cholecystectomy (104), or abdominal hysterectomy (143). Multiple simulated samples were then drawn from these data. These simulated samples were used in computer simulations of clinical trials comparing doses among groups. Assessments were made of the relative performance of the t test; analysis of variance; randomization; and the Mann-Whitney, Kruskal-Wallis, and chi- square tests for comparing doses of analgesics given to two or three groups of patients. Results: The distributions were not normal. No statistical test incorrectly suggested that a difference existed among groups when there was no difference. A randomization test to compare two groups' means overestimated actual P values. A chi-square test which compared fractions of patients in each group who received no morphine, had approximately the same power as the t or analysis of variance tests to detect differences among groups. Mann-Whitney or Kruskal-Wallis tests, for differences between two or more than two groups' medians, respectively, had the greatest power to detect difference(s) among groups. Conclusions: Mann-Whitney and Kruskal-Wallis are the best tests to find differences among doses of analgesics given to groups of patients.