(C08) Using Historical RCT Controls to Contextualize Necessity of Unblinding of Serious Adverse Events and Adverse Events of Special Interest – Conceptual Framework and Database Development
Background: Serious adverse events (SAE) & adverse of special interest (AESIs) occurring during the blinded period of randomized controlled trials (RCT) often present challenges in making decisions about the existence of potential safety signals. Context of the expected frequency of individual SAEs are invaluable in drug development, but information is limited because of sample sizes in comparator arms of RCTs or the absence of an active comparator arm. For the scenario of RCTs with an active comparator, there is the added risk to avoid of unnecessary unblinding that would impact statistical integrity. Though individual SAEs may be rare, 90% of RCTs experience SAEs.1 We evaluated options to mitigate this challenge by utilizing existing data from historical RCTs.
Objectives: To develop a framework for estimating incidence rates and time to AE onset (TTAE) of SAEs from placebo or standard of care subjects among AbbVie historical RCTs. This proof of concept may improve potential signal identification process by leveraging AE rates as context for ongoing, unblinded RCT SAEs.
Methods: AbbVie’s Research and Development Convergence Hub (ARCH) was developed as a data platform to host real-world and RCT data for enterprise use, and allows for incidence and TTAE analyses of SAE/AESIs of interest from placebo subjects to be estimated with the Adverse Events in Placebo groups of Abbvie Clinical Trials (AEPACT) database. Incidence and TTAE (95% CI) can be generated for the overall placebo cohort, a parameterized cohort across all placebo subjects to replicate the ongoing RCT and a parameterized cohort for the unique subject according to demographic and comorbidity profile. Assessments of expectedness across the entire RCT of interest based on anticipated person-time can then be generated for additional context as to the expected rates of the SAEs.
Results: To date AEPACT comprises 91 RCTs (35% CNS, 15% immunology, and others including dermatology, ophthalmology, musculoskeletal) that capture placebo subjects experiencing an AE during the trial (n=5758 with 25314 events). RCTs and AEs span the period 2003-'18, average subject age was 47.5, ranging from 5-90, males constituted 47.7% of AEs. Mild AEs were 63.5% of overall AEs, moderate 32.0% and severe 4.5%. Assessment of relatedness to study drug by severity, 24.5% of Mild AEs were deemed possibly or probably related, moderate 23.2% and severe 16.7%.
Conclusions: RCT controls with SAEs pose substantial risk for unnecessary unblinding. Analyses leveraging AE rates observed in subjects receiving placebo treatment from historical trials hold great potential for informing drug development and single-arm trials. As more RCT data and placebo subjects without an AE are incorporated into AEPACT, assessment methods will continue to be developed.
