OBJECTIVES: Longer-term follow-up data can offer additional certainty around long-term projections of efficacy outcomes from randomized controlled trials. This study analyzed the impact of an additional >20 months of follow-up on long-term predictions of DFS, the primary outcome from CheckMate-274.

METHODS: Patient-level DFS data from the two successive database-locks (DBLs) with 11.0-month and 31.6-month of minimum follow-up for the intention-to-treat population were used. Nivolumab and surveillance DFS were modelled using corresponding Kaplan-Meier (KM) curves from CheckMate-274 (up to year 3) and the control arm from EORTC-30994 (7 year median follow-up) from 3 to 5 years [KM-based or using parametric extrapolation (1-knot spline odds)]. Patients who are disease-free by year 5 were assumed as functionally cured. Cured patients experienced no recurrence and died only from non-disease related causes in the model. Age- and sex-adjusted non-disease related mortality rates were derived from UK lifetables. The performance measure compared between the two DBLs was the restricted mean (RM) DFS at 5, 10 and 30 years.

RESULTS: Results are presented for non-parametric (KM-based) and parametric models. For surveillance both DBLs resulted in similar 30-yr RM DFS [77-months (76-months) for the earlier versus 77-months (74-months) the later DBL]. For nivolumab the earlier DBL predictions resulted in lower 30-yr RM DFS versus the later DBL [mean DFS increased from 97 to 99-months (91 to 96-months)]. Non-parametric and parametric models estimated similar RM DFS at 5-years for both arms [31-months for nivolumab versus 25-months for surveillance in both DBLs]. The parametric models resulted in lower 10-year RM DFS than the non-parametric approach in both DBLs for both arms [52-months (50-months) and 53-months (52-months) for nivolumab versus 42-months (42-months) and 42-months (41-months) for surveillance].

CONCLUSIONS: Across a life-time horizon, DFS projections obtained from the later DBL confirmed that the predictions for nivolumab from the earlier DBL were marginally more conservative.