The parametric estimation procedure to predict long-term efficacy on survival from publication data of clinical trials for metastatic colorectal cancer (mCRC).

Abstract

541 Background: In comparative clinical trials with time-to-event outcomes, the long-term treatment effect profile is clinically crucial. The conventional non-parametric measures, i.e., the median survival times and the t-year survival rates are informative but sometimes cannot be observed due to the limited study follow-up. For example, in the PEAK trial (Schwartzberg LS et al. J Clin Oncol 2014), the group contrast over 3 years could not be adequately confirmed due to the short follow-up. To predict the subsequent survival prognosis with the limited outcome data, one may estimate the survival curve using a parametric model. Here, we illustrate the parametric estimation procedure to predict long-term efficacy using the results of two randomized controlled trials for the patients with wild-type KRAS mCRC, i.e., PRIME (Douillard JY et al. J Clin Oncol 2010, Douillard JY et al. Ann Oncol 2014) and PEAK trials. Methods: We considered both the primary and updated results for the PRIME to assess the validity of the procedure, and then estimated the long-term efficacy in the two trials. We used the reconstructed overall survival (OS) data obtained by scanning the Kaplan-Meier curves in the literature. Fitting the data to the Weibull distribution, we estimated the parametric group contrast measures including the difference in the 3- and 5-years restricted mean survival times and mean survival times. Results: The extrapolated parametric OS curves from the primary PRIME results fitted well with the observed Kaplan-Meier curves for OS in the updated results. The parametric estimations demonstrated that, in the PRIME trial, panitumumab plus FOLFOX arm increased 2.0 (95% CI: 0.1-4.0), 4.0 (0.5-7.5), and 5.4 (0.3-10.5) months in survival on average over 3-, 5-years and the entire time, respectively, compared to FOLFOX arm. In PEAK trial, compared to bevacizumab plus FOLFOX arm, panitumumab plus FOLFOX arm increased 3.6 (0.7-6.5), 7.9 (2.1-13.7), and 12.8 (0.5-25.1) months in survival on average, respectively. Conclusions: The estimators for parametric survival curve would provide the informative summaries for long-term survival profile to the clinicians and patients.