Re‐evaluation of randomized control trials of lithium monotherapy: a cohort effect

Abstract

The reported reduction of lithium's efficacy in the prophylaxis of bipolar illness has been attributed to various factors, including diagnostic changes and heterogeneous study designs. We attempted to quantify the impact of pre-randomization enrichment designs and diagnostic drift on randomized controlled trials (RCTs) of lithium maintenance therapy. Using the Cochrane RCT search filter, MEDLINE, EMBASE, and PSYCHINFO were searched (1966 to June 2004) for all available randomized studies using the text word 'lithium'. Studies of 1 year minimum duration in bipolar disorder involving lithium and placebo arms were identified. Superiority trials without a placebo arm, discontinuation and mirror image studies were excluded. Standardized scales were used to assess randomization and allocation concealment. Nine RCTs enrolling 1432 bipolar I and II patients, randomizing 341 to lithium and 386 to placebo were identified, with 705 reported pre-randomization dropouts. The pooled odds of remaining recurrence free in two non-enriched RCTS using Research Diagnostic Criteria (RDC) or Feighner criteria were 3.2:1 (95% CI 0.65--15.46) trending in favor of lithium over placebo, and 22.0:1 (95% CI 7.0--68.7) for three trials using lithium enrichment and excluding atypical bipolar disorder. The odds of remaining recurrence free using DSM-IV criteria and lamotrigine enrichment were 1.9:1 (95% CI 1.2-2.8). Lithium maintenance RCTs differ in patient selection, design, and outcome. A cohort effect can be associated with the use of pre-randomization enrichment phases and, to a lesser extent, with diagnostic drift, compromising straightforward comparisons across three decades of lithium monotherapy in bipolar illness.