Usability and relevance of specialized population-based cancer registry data in externally controlled trials

Randomized clinical trials are becoming very challenging to perform in oncology, both in rare cancers and in more common cancers, fragmented according to molecular characteristics. In this proof of concept, we tested the feasibility of using data from a specialized digestive cancer registry as an external control arm in 2 phase III randomized clinical trials.

To emulate PETACC-8, the generated hybrid control arm comprised a constrained proportion of external patients. PRODIGE-13 was emulated through a fully external control arm and through a hybrid control arm containing 50% of external patients. Propensity scores were used either for inverse probability of treatment weighting or for patient matching. Absolute standardized mean differences were used to test the balance between arms. Overall survival was estimated using the Kaplan-Meier estimator and compared using the log-rank test.

For both trials, registry data were considered fit-for-use despite different settings, and patient baseline characteristics were balanced between the emulated trial arms. In the emulated PETACC-8 trial, the overall survival of the hybrid control arms were not different from that of the control and experimental arms. In the emulation of PRODIGE-13, survival analyses showed that the fully external control arm mimicked the control arm and that the hybrid arm with 50% external data mimicked both control and experimental arms. Pending appropriate patient selection, the data from the Registry allowed the generation of external control arms mimicking control arms.

This pilot study provides promising insights for improving the costs and duration of clinical trials by combining experimental arms with specialized cancer registry data.