PP47 Modelling Non-small Cell Lung Cancer Treatment: Predicted and Observed Impact Of Immunotherapy In The Netherlands

Le résumé fourni par la source

Introduction Patients treated with immunotherapy are divided into two subgroups: (i) long-term survivors (LTS) and (ii) moderate survivors. Nevertheless, clinical trials (RCTs) report only average treatment effects such as hazard rate (HRs). Health economic-models often only input average treatment effects, even though it has been shown that accounting for the LTS subgroup is crucial for accurate projection of long-term survival under immunotherapy. We investigated the incorporation of a statistical mixture cure model (MCM) in a health-economic model for lung cancer as a way to account for LTS while incorporating reported average RCT-based treatment effects. Methods We developed a microsimulation model describing disease progression under three treatment lines in advanced lung cancer using Dutch real-world data of chemotherapies treated patients. Here we focus on first-line treatment, for which we used gompertz distribution to simulate time-to-progression. To simulate the impact of immunotherapy, we adjusted base-model assuming MCM for first-line treatment, where the LTS subgroup was not at risk to progress, but instead die from background mortality. The subgroup of moderate survivors on the other hand are at risk to progress with adjusted progression-free HR (PF-HR). We simulated the model with size of LTS (prop_LTS) ranging from 14-34 percent (keynote-001 five-year overall survival [OS], 95% confidence interval) while fixing average RCT PF-HR at 0.5. Model predictions under the different prop_LTS were compared to real-world Dutch OS as well as the long-term RCT five-year OS. Results With respect to observed short-term survival outcomes, model predictions were insensitive to assumptions regarding the size of the LTS subgroup. However, to match the five-year RCT OS rate reported (32%), the prop_LTS had to be equal to 34 percent. Under this latter setting for the prop_LTS, the progression HR in the subgroup of moderate survivors was calibrated to be 1.1. Conclusions The use of a mixture cure model improves long-term model-based projections with the implicit assumption that moderate survivors have little or no treatment benefit.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.