OBJECTIVES: The relative effectiveness of sunitinib and everolimus to treat pNET patients has been assessed using matching adjusted indirect comparison (MAIC) based on the RADIANT-3 everolimus trial. We performed an analysis using updated OS data for both treatments using individual patient data (IPD) from sunitinib’s trial (A6181111), offering an opportunity to assess robustness of results to IPD source.

METHODS: Analyses included Bucher-type comparisons using hazard ratios (HRs) for progression-free survival (PFS) and overall survival (OS) vs. best standard of care (BSC), and a MAIC matching on all baseline characteristics available from both studies. An anchored MAIC was performed for PFS, applying balancing weights in A6181111 to derive an adjusted HR for sunitinib vs BSC. The ratio of this and the HR for everolimus yielded a HR for sunitinib vs. everolimus. An unanchored MAIC was performed for OS since survival in the BSC arms may be contaminated differentially by crossover. A Cox model was fit the weighted OS data for sunitinib and virtual IPD for death/censoring times for everolimus derived from its published OS curve.

RESULTS: The effective sample size after matching was 43 (original N=86) for sunitinib and 31 (original N=85) for BSC. The unadjusted PFS HR for sunitinib vs everolimus was 1.20 (0.72-2.01), and 0.85 (0.39-1.89) after adjustment. Similarly, an unmatched OS comparison yielded 1.03(0.75-1.42); this reduced to 0.82 (0.53-1.27) after matching.

CONCLUSIONS: Like the prior MAIC, analyses demonstrate comparable PFS and OS with sunitinib and everolimus, but produced point estimates that differ in direction. For PFS, this is attributable to baseline imbalances in A6181111 favouring BSC that are adjusted for in the current analysis; for OS, the direction may reflect changes in updated survival data. Limitations include uncertainty due to the small size of the sunitinib trial, and possible residual confounding in OS comparisons.

DISCLOSURE: This research was sponsored by Pfizer.