Poster Presentation COSA-IPOS Joint Scientific Meeting 2012

Projecting long term overall survival benefit in metastatic melanoma patients treated with ipilimumab using extrapolation methods (#739)

Hansoo Kim 1 , Greg Cook 1 , Philip McCloud 2
  1. Bristol-Myers Squibb, Noble Park North, VIC, Australia
  2. McCloud Consulting Group, Sydney, Australia

Ipilimumab is an immuno-oncological medicine registered in Australia for the treatment of patients with metastatic melanoma who have failed or are intolerant to prior therapy. Ipilimumab specifically blocks the inhibitory signal of CTLA-4, resulting in T-cell activation, proliferation, and lymphocyte infiltration into tumours, leading to tumour cell death. Treatment consists of a regimen of four infusions of ipilimumab at a dose of 3mg/kg every three weeks. As a proportion of patients treated with ipilimumab have previously been demonstrated to experience long term overall survival benefit, this study sought to project overall survival of patients beyond the 2 year study period of the pivotal ipilimumab phase III study (MDX010-020). Additionally, we sought to validate the extrapolation method selected by comparing to the observed survival rates from two phase II ipilimumab studies where long term survival data is available for patients with 4 and 5 years follow-up. Traditional parametric curves such as Weibull, exponential and log-normal produced a poor fit to the Kaplan-Meier survival curve. Investigation of the hazard function revealed that the survival curve could be characterized as a mixture of two different curves - as the magnitude of the hazards changed approximately 2 years (turning point) after treatment initiation. A Weibull curve was fitted to the survival curve after the turning point and projections were derived. The projected overall survival rate at 5 years was 16.1%. This method projected the overall survival rates at 4 and 5-years to within a margin of less than 1% difference when compared to the observed rates from two phase II ipilimumab trials. This comparison confirms the hypothesis that overall survival for ipilimumab does not follow traditional parametric curves but rather a mixture of distributions. Applying this extrapolation approach allows one to project the long term effect of ipilimumab on overall survival with increased certainty.