Effect of a Run-In Period on Estimated Treatment Effects in Cardiovascular Randomized Clinical Trials: A Meta-Analytic Review

Abstract

Background A run-in period may increase adherence to intervention and reduce loss to follow-up. Whether use of a run-in period affects the magnitude of treatment effects is unknown. Methods and Results We conducted a meta-analysis comparing treatment effects from 11 systematic reviews of cardiovascular prevention trials using a run-in period with matched trials not using a run-in period. We matched run-in with non-run-in trials by population, intervention, control, and outcome. We calculated a ratio of relative risks (RRRs) using a random-effects meta-analysis. Our primary outcome was a composite of cardiovascular events, and the primary analysis was a matched comparison of clinical trials using a run-in period versus without a run-in period. We identified 66 run-in trials and 111 non-run-in trials (n=668 901). On meta-analysis there was no statistically significant difference in the magnitude of treatment effect between run-in trials (relative risk [RR], 0.83 [95% CI, 0.80-0.87]) compared with non-run-in trials (RR, 0.88 [95% CI, 0.84-0.91]; RRR, 0.95 [95% CI, 0.90-1.01]). There was no significant difference in the RRR for secondary outcomes of all-cause mortality (RRR, 0.97 [95% CI, 0.91-1.03]) or medication discontinuation because of adverse events (RRR, 1.05 [95% CI, 0.85-1.21]). Post hoc exploratory univariate meta-regression showed that on average a run-in period is associated with a statistically significant difference in treatment effect (RRR, 0.94 [95% CI, 0.90-0.99]) for cardiovascular composite outcome, but this was not statistically significant on multivariable meta-regression analysis (RRR, 0.95 [95% CI, 0.90-1.0]). Conclusions The use of a run-in period was not associated with a difference in the magnitude of treatment effect among cardiovascular prevention trials.

Keywords: cardiovascular prevention; meta‐analysis; run‐in; trial methodology.

Figure 1. Pooled ratio of relative risks (RRR) and 95% CIs for run‐in versus non–run‐in randomized clinical trials (RCTs): cardiovascular composite.

Single best population, intervention, control, and outcome score analysis. Pooled ratio of relative risks (RRR) for matched run‐in and non–run‐in RCTs that reported a cardiovascular composite outcome. “No. of Studies” refers to the number of non–run‐in trials that have been matched to each individual run‐in trial. The size of the data markers indicates the weight of the comparison in the meta‐analysis. A summary estimate <1 represents an exaggerated treatment effect of run‐in. df indicates degrees of freedom; I², degree of heterogeneity; Q, Cochran’s Q; RE, random‐effects meta‐analysis; and RR, relative risk.

Figure 2. Pooled ratio of relative risks and 95% CIs for run‐in versus non–run‐in randomized clinical trials (RCTs): all‐cause mortality.

Single best population, intervention, control, and outcome score analysis. Pooled ratio of relative risks for matched run‐in and non–run‐in RCTs that reported mortality outcomes. “No of Studies” refers to the number of non–run‐in trials that have been matched to each individual run‐in trial. The size of the data markers indicates the weight of the comparison in the meta‐analysis. A summary estimate <1 represents an exaggerated treatment effect of run‐in. df indicates degrees of freedom; I², degree of heterogeneity; Q, Cochran’s Q; RE, random‐effects meta‐analysis; and RR, relative risk.

Figure 3. Pooled ratio of relative risks and 95% CIs for run‐in versus non–run‐in randomized clinical trials (RCTs): adverse events leading to medication discontinuation.

Single best population, intervention, control, and outcome score analysis. Pooled ratio of relative risks for matched run‐in and non–run‐in RCTs that reported adverse events leading to discontinuation of study medication. “No. of Studies” refers to the number of non–run‐in trials that have been matched to each individual run‐in trial. The size of the data markers indicates the weight of the comparison in the meta‐analysis. A summary estimate <1 represents an exaggerated treatment effect of run‐in. df indicates degrees of freedom; I², degree of heterogeneity; Q, Cochran’s Q; RE, random‐effects meta‐analysis; and RR, relative risk.