Safety Outcomes in Meta-analyses of Phase 2 vs Phase 3 Randomized Trials

Abstract

Recent studies have reported disagreement between meta-analysis of small trials and subsequent large trials addressing the same question. However, disagreement for uncommon but serious adverse safety outcomes has not been examined. To explore disagreement for serious adverse safety (intracranial hemorrhage [ICH]) and efficacy outcomes between meta-analysis of phase 2 (small) vs meta-analysis of phase 3 (large) randomized controlled trials comparing the efficacy of bolus thrombolytic therapy with infusion for acute myocardial infarction (AMI). Electronic databases (MEDLINE, Cochrane Database of Clinical Trials) between January 1980 and December 1999 using the search terms thrombolysis, thrombolytic therapy, and myocardial infarction; conference proceedings; and reference lists. Fifteen randomized trials comparing thrombolytic agents administered by bolus injection with standard infusion therapy in patients with AMI. Data on ICH, other causes of stroke, total mortality, and reinfarction were independently extracted from each study by 2 observers. Meta-analysis of 9 phase 2 trials (n = 3956) revealed a lower risk of ICH with bolus thrombolytic therapy (odds ratio [OR], 0.53; 95% confidence interval [CI], 0.27-1.01), which was not statistically significant. Meta-analysis of 6 phase 3 trials (n = 62 673) indicated a significant increase in risk of ICH (OR, 1.25; 95% CI, 1.06-1.49). These results were significantly different (P =.01). There was no disagreement for efficacy outcomes. Phase 2 trials included younger and heavier patients with lower baseline blood pressures, and were more often open-label. Subgroup analyses suggested that each of these factors was associated with a lower estimate of risk of ICH with bolus agents. Our results suggest that when therapeutic interventions are associated with a potential for uncommon but serious adverse safety outcomes, there may be differences between small phase 2 and large phase 3 trials that result in their disagreement for safety but not necessarily efficacy outcomes. Further investigation of the frequency and causes of disagreement between small and large trials for safety outcomes is warranted.