Effect size of efficacy measures comparing divalproex, lithium and placebo in acute mania

Abstract

Effect size (ES) is a statistical concept that can be used to improve the interpretation of results from psychopharmacological studies. ES may aid interpretation of results when sample size is unbalanced or small or when units or levels of baseline measures differ across items. Usually, an investigator can define a threshold value for a clinically meaningful ES based on published data and clinical judgment or by resorting to conventions, e.g., a medium ES = 0.5 S.D., which can usually be discerned by the trained clinician. In the present study, we apply ES analysis to results from a study comparing the effectiveness of divalproex (DIVAL), lithium (LI), and placebo (PLA) in hospitalized, acutely manic patients. One hundred seventy-six patients were randomly assigned to DIVAL, LI, or PLA in a 2:1:2 ratio, with drug administered in a double-blind, parallel group design for 21 days. The primary efficacy measure was the Mania Rating Scale from the Schedule for Affective Disorders and Schizophrenia, composed of the Manic Syndrome Score (MSS) from items that are relatively specific to the manic state, and the Behavior and Ideation Score (BIS), which reflects severe but nonspecific psychopathology. Improvement of the MSS after 5 days of treatment was difficult to interpret based on percentage change (DIVAL = 19%, LI = 13.5%, PLA = 8.5%). However, the corresponding effect sizes of 0.79, 0.55, and 0.35 indicated a medium to marked ES for DIVAL, a medium ES for LI, and a small ES for PLA at this early point in treatment. Similarly, the ES for change on the MSS at the end of treatment indicated a large, readily observable improvement with both DIVAL (ES = 1.01) and LI (ES = 0.79) vs. an ES of 0.37 for PLA. ES analysis also indicated that the BIS is a less robust indicator of change to either drug. The ES at the end of treatment for the BIS was 0.67 for DIVAL-, 0.62 for LI-, and 0.25 for PLA-treated patients.