Advertisement
Methodological Developments in Survival Analytic Methods: to Inform Cost-Effectiveness Models| Volume 25, ISSUE 1, SUPPLEMENT , S10, January 2022

P45 The Use of Historical Clinical Trial Data to Inform Survival Extrapolation

      Objectives

      Standard parametric distributions are commonly used for the extrapolation of survival data in cost-effectiveness analyses. However, survival data is often immature and uncertainty remains around the survival extrapolations. Mature historical data can be used to better predict survival beyond trial data. This study assessed two methods to incorporate historical data in the extrapolation of immature survival data.

      Methods

      Immature data of a breast cancer trial comparing pertuzumab+trastuzumab+docetaxel versus trastuzumab+docetaxel (follow-up time 38 months; data-cut 2015) was extrapolated and mature survival data (follow-up time 120 months; data-cut 2020) from the same trial was used to validate the extrapolations. The historical data was from a previous breast cancer trial including mature survival data of trastuzumab+docetaxel (follow-up time 50 months; data-cut 2005). Two methods to quantitatively inform the extrapolation of immature survival data with historical data were compared to standard parametric distributions: 1) historical shape parameter as informative prior for the shape of the immature data; 2) historical data as a third arm. Predictions were assessed with delta area under the curve (AUC) values based on the mature survival data.

      Results

      Without priors, the delta AUC was 7.59, 1.62, 13.15, 8.32, 25.15, and with the historical arm the delta AUCs were 9.65, 4.38, 6.79, 8.26, 21.81, for Weibull, loglogistic, lognormal, exponential, and Gompertz, respectively. With priors, the delta AUC were 8.43, 3.37, 9.11, 23.68, for Weibull, loglogistic, lognormal, and Gompertz, respectively (as for exponential there is no shape parameter). The loglogistic distribution without priors predicted the immature data the best. For three out of five distributions, the extrapolations with a historical arm resulted in better predictions compared to the extrapolations without prior.

      Conclusions

      The impact of external data on clinically plausible survival extrapolations can further be improved by using historical data with longer follow-up with treatment patterns similar to the current standard of care.