Abstract Background Continued access to study medications in both active and placebo-treated subjects is frequently advocated after randomized clinical trial (RCT) completion. There is uncertainty how such OLE trials should be analyzed as all subjects are receiving the active therapy, yet some therapies have a legacy effect on outcomes. We sought to explore different analytical approaches to adjust for the lack of a control arm in an OLE phase clinical trial. Purpose Explore analytical approaches to evaluate CV outcomes in the OLE phase after a RCT trial using a counterfactual derived placebo arm when treatment effects are delayed. Methods We analyzed data from the FOURIER RCT (N=27,564) who were followed for a median of 2.2 years and the FOURIER-OLE (N=6,635) for a median of 5 years. In the RCT, half of patients received evolocumab (EVO) and half received placebo. After the RCT subjects entered the OLE phase where all patients received EVO. The primary cardiovascular (CV) composite endpoint PEP (CV death, MI, stroke, hospitalization for UA, or coronary revascularization) was analyzed using Cox model and generating hazard ratio (HR), and 95% CI with prespecified stratification factors at initial randomization and FOURIER-OLE study protocol. In an additional analysis, we used rank preserving structural failure time (RPSFT) that modelled the treatment effect in the cross-over of subjects randomized to placebo during FOURIER who received EVO in the OLE phase. RPSFT estimates the counterfactual time-to-event for FOURIER placebo subjects throughout the OLE period by assuming the FOURIER observed treatment effect within the OLE period. This method was evaluated in the FOURIER + OLE trial for PEP. Results Using the Cox-model analysis patients initially randomized to EVO in the RCT continued to have lower incidence of PEP in the OLE; HR 0.85 95% CI 0.75-0.96, p=0.008 compared to those receiving EVO after 2.2 years of placebo. A delayed treatment effect was evident using observed data (by comparing early initiated EVO group versus late initiated EVO group). Using RPSFT analysis, strong long-term treatment benefit of EVO was estimated for PEP, a HR = 0.81 (95% CI: 0.76, 0.87) in the overall FOURIER + OLE cohort between the counterfactual placebo arm and observed long-term EVO arm. Results were robust to sensitivity analyses. Data displayed for the combined analysis for PEP (figure 1) and the individual endpoint of MI (figure 2) HR = 0.66 (95% CI: 0.59-0.74). Conclusions Open label extension designs have limitations for analytical purposes given the absence of a control arm. We propose an analytical counterfactual framework (RPSFT) that could be used to explore long-term efficacy endpoints in OLE trials. The counterfactual analysis, including both the RCT and OLE part, shows a large treatment difference and the results based on observed data only may potentially underestimate the true long-term cardiovascular effects.
Walsh et al. (Sat,) studied this question.