Abstract Background Adaptive designs have been used in therapeutics’ development to increase the efficiency of RCTs. However, trials to validate novel diagnostics such as myocardial perfusion imaging (MPI) tracers have a number of additional challenges. The prevalence of the disease ranges in the enrolled population, depending on sites and patients mix, which can substantially impact the sample size representing a challenge to trial’s feasibility. Purpose The objective is to design an adaptive trial to assess the diagnostic performance of a novel tracer for positron emission tomography (PET) MPI in patients referred for invasive coronary angiography (ICA) due to suspected coronary artery disease (CAD). Methods/Results The Phase 3 prospective, multicentre, open-label study is designed to assess diagnostic efficacy of a novel 18F-labelled tracer for PET imaging in patients with suspected CAD. The primary endpoints are sensitivity and specificity against ICA followed by invasive fractional flow reserve (FFR) or instantaneous wave-free ratio (iwFR) measurements when appropriate (40-90% stenosis in major branch in ICA). Sensitivity and specificity of PET MPI will be tested against 60% at a one-sided significance level alpha of 2.5% with an exact binomial test for proportions. The prevalence of CAD in published trials of patients referred to ICA ranges from less than 20% to 50%. Power analysis (Figure 1) shows the substantial impact of CAD prevalence in the study population on the sample size. An additional uncertainty is the true sensitivity and specificity. To improve the efficiency of the study and to avoid exposing large number of patients to study procedures, an adaptive design with blinded monitoring of CAD prevalence and 3 unblinded interim analyses is implemented to mitigate uncertainties regarding prevalence, true sensitivity and specificity (Figure 2). Population enrichment may be imposed based on the blinded monitoring of CAD prevalence before the first unblinded interim analysis. Unblinded interim analyses (IA1, IA2, and IA3) allow for early futility and efficacy decisions as well as a sample size re-estimation at IA3. Operating characteristics of the proposed design were established by simulations. The overall type I error is controlled at 2.5% one-sided. The proposed adaptive design allows to initiate the study with the planned sample size of 330 patients, IA with possible early efficacy stop at ≈160 patients and an option to increase sample size up to 802 based on results of IAs. Conclusions Proposed adaptive design with blinded monitoring of CAD prevalence and 3 unblinded interim analyses allowing for early futility and efficacy decisions as well as a sample size re-estimation can improve the efficiency of Phase 3 study evaluating a novel diagnostic method while maintaining operating characteristics required for the registrational trial.
Muehlemann et al. (Sat,) studied this question.