Aetion is thrilled to share that the final results of the RCT-DUPLICATE study were published today. This publication demonstrates the ability of #RWE studies to reach similar conclusions as randomized clinical trials when specific designs and measurements are replicated. Through efforts like RCT-DUPLICATE, we are proud to help lead the charge in assessing the uses, benefits, and limitations of #RWD. Congratulations to all who were involved in this milestone publication. Learn more here: [https://hubs.li/Q01MHszl0]. #clinicaltrials
Professor of Medicine and Epidemiology, Harvard Medical School | Department of Medicine Brigham and Women's Hospital
RCT-DUPLICATE has published 32 real-world evidence emulations of RCTs. In settings were an RWE study could closely emulate the RCT design and its key measurements we came to highly similar conclusions (r=0.93). In the absence of RCT evidence we have an obligation to produce the best evidence possible using an abundance of digital healthcare data to close many evidence gaps. https://lnkd.in/eN42VMJB Many thanks for @Shirley Wang leadership and the outstanding collaboration with @FDA and @Aetion. Some context: 1) RCT evidence and real-world evidence complements each other – our work is NOT about replacing RCTs! In fact, we wish we had more RCT evidence to answer all relevant clinical questions. 2) The motivation was to understand the validity of RWE studies using claims data to estimate drug effects in clinical practice. Are there circumstances when RWE performed particularly well and can we pin-point predictable areas of concern? 3) The authors’ conclusion is that RWE studies come to the same conclusion as RCTs in this select sample of RCTs that could emulate their design and key measurements. 4) Design emulation issues: Highly controlled trials for regulatory submission may have design aspects that are not observable in clinical practice, e.g. run-in periods with selective drop out before randomization; removal of baseline medications that patients tolerated; longer adherence than observed in practice. 5) Measurement emulation issues: Insurance claims data have many strengths and equally many shortcomings that make it difficulat to study certain scenarios, e.g. claims data do not record in-hospital medications which makes it difficult to study the effectiveness of antiplatelet medications started after an acute coronary event. 6) For this activity we had to assume that a single RCT prepared for regulatory submission is the standard against which we calibrate. We know that a single trial may be contradicted by another trial of the same question even an identically designed ‘twin trial’; IF ANYBODY HAS A BETTER IDEA FOR A BENCHMARK, please let us know. 7) We planned to emulate the trial designs with all inclusion/exclusion criteria, not the resulting study population. We purposefully abstained from post-hoc reweighting population characteristics. Secondary analyses suggest that age-sex reweighting will not change findings. 8) For this activity, we limited ourself to insurance claims data. RCT-DUPLICATE expands in several directions using electronic health records linked with claims data across multiple therapeutic areas, including cardiometabolic and oncology.