ISPOR 17th Annual European Congress
Amsterdam, The Netherlands
November, 2014
PRM80
Cancer
Research on Methods (RM)
Modeling Methods (MS)
MODELLING SURVIVAL IN THE PRESENCE OF DIFFERENT MECHANISMS OF ACTION: IPILIMUMAB AND VEMURAFENIB IN ADVANCED MELANOMA
Lee D1, Porter J1, Hertel N2, Hatswell AJ1
1BresMed, Sheffield, UK, 2Bristol Myers Squibb, Uxbridge, UK
OBJECTIVES Traditional indirect treatment comparison methods assume the underlying survival profiles of treatments are similar (i.e. proportional hazards). This assumption is unlikely to hold for the comparison of ipilimumab and vemurafenib: Whereas vemurafenib exhibits improved short-term survival compared with ipilimumab, pooled study data for ipilimumab consistently show that patients achieve durable long-term survival. We present a method to compare across trials with differing survival profiles by accounting for follow-on treatments and different patient baseline characteristics. METHODS Comparative survival estimates for ipilimumab and vemurafenib were produced using patient-level data from trial CA184-024 for ipilimumab and published survival curve fits from BRIM-3 (along with registry data) for vemurafenib. The BRIM-3 vemurafenib overall survival curve was adjusted to account for (a) the effect of second-line ipilimumab (via a tunnel-state methodology) and (b) differences in patient baseline characteristics between BRIM-3 and CA184-024, by means of a model (Korn model), constructed to predict the outcomes for dacarbazine-treated patients. The resulting survival estimates were compared with naïve unadjusted survival curve fits, and estimates produced using a hazard ratio (from an indirect comparison) to the ipilimumab data. RESULTS Estimated survival for ipilimumab was 3.3 years (mean). Predicted survival for vemurafenib, using a naïve comparison, was 3.0 years (mean). Adjusting for second-line ipilimumab and different baseline characteristics resulted in an estimate of 2.8 years for vemurafenib. When a hazard ratio was applied to the ipilimumab data, which underlies the here strong assumption that the vemurafenib overall survival profile is similar to that of ipilimumab, predicted survival for vemurafenib increased to 4.2 years. CONCLUSIONS Depending on the methodology used, the mean predicted survival for vemurafenib varied from 2.8 to 4.2 years. Alternative methods that incorporate the long-term survival profile of ipilimumab (naïve comparison or more sophisticated adjustment methodology) demonstrate a higher number of life years with ipilimumab versus vemurafenib.