Simulation of survival with first- and second-line non-small cell lung cancer (NSCLC) therapy using a public domain drug-disease modeling framework

Abstract

8087 Background: Modeling and simulation approaches are advocated to support drug development decisions. In this study, we evaluated the ability of a drug-disease model recently developed by scientists at the FDA using data from pivotal studies in NSCLC (Wang et al. http://www.fda.gov/ohrms/dockets/ac/08/briefing/2008–4351b1–00-index.htm ) to simulate survival and tumor response for approved therapies in NSCLC. Methods: The modeling framework comprises a longitudinal tumor size model and a survival model relating change in tumor size at first visit (week 8) and patient characteristics (tumor size and ECOG performance status at baseline) to survival time. The tumor size and survival models were used to simulate change in tumor size at first visit and expected survival in the carboplatin/paclitaxel (CP) arm of the E4599 study (first-line NSCLC, N Eng J Med. 2006;355:2542–2550) and in the erlotinib arm in the BR.21 study (second-line NSCLC, N Eng J Med. 2005;353:123–132). The predictive distributions (95% prediction interval [PI]) of survival times were derived from multiple replicates (500) of 400 CP patients and 500 erlotinib patients with similar characteristics to patients in the original studies. Results: There was a high level of concordance between the results of the simulation and the observed results in the two arms, indicating that the modeling framework successfully predicted survival and tumor response. Expected median survival was 9.8 (95% PI, 8.2–11.8) months (vs 10.3 months observed) for the CP arm and 6.8 (95% PI, 5.3–8.9) months (vs 6.7 months observed) for the erlotinib arm. The median change in relative tumor size from baseline at week 8 predicted by the model was 12.1% for the CP arm and 9.5% for the erlotinib arm. Conclusions: The modeling framework can be used to perform simulations of survival of approved treatments. These results suggest that the model could be used to simulate survival for investigational treatments based on tumor shrinkage data observed in early clinical studies (e.g. phase II) to support end-of-phase II decisions and the design of phase III studies. [Table: see text]