Predicting longer-term progression-free survival (PFS) with durvalumab after chemoradiotherapy (CRT) in unresectable stage III NSCLC using a mixture cure model (MCM)

Abstract

Background Five-year results from the placebo-controlled phase 3 PACIFIC trial (NCT02125461) demonstrated robust and sustained overall survival (OS) and durable PFS benefit with durvalumab (given every 2 weeks up to 12 months) in pts with unresectable Stage III non-small-cell lung cancer (NSCLC) and no disease progression after platinum-based concurrent CRT. Given the potential for longer-term PFS gain in this curative-intent setting, we used MCM analysis, a common statistical model in such settings, to predict the potential longer-term PFS benefit with durvalumab. Methods Pt-level data from 5 years follow-up in PACIFIC were used. MCMs were fitted to PFS data (BICR; RECIST v1.1) to estimate the proportion of pts who are ‘long-term survivors’ (the statistical cure fraction, defined as the proportion of pts with no risk of disease progression) in each arm. Treatment benefit for the non-cured proportion was also modelled. Model estimates were then used to predict 10-year PFS probabilities (‘rates’), while adjusting for background mortality. Models assuming no cure were fitted as sensitivity analyses. Results The MCM reported a consistent benefit with durvalumab versus placebo at 10 years with different parametric models and fitted better than non-cure models. The best fitting model used log-normal distribution for pts at risk of progression. Using this model, the estimated statistical cure fraction for PFS was 36.0% (95% CI, 30.2–42.3) with durvalumab and 19.4% (13.7–26.6) with placebo. When accounting for background mortality, the estimated 10-year PFS rate (95% CI) was 28.7% (24.3–33.5) and 15.4% (11.1–21.0), respectively. Similar analysis applying 5-year OS data to both cure and non-cure models also demonstrated longer-term benefit with durvalumab. Conclusions MCMs had a good statistical fit with PACIFIC 5-year PFS data. The best fitting cure model estimated a 10-year PFS rate of >25% with durvalumab, almost twice the rate estimated with best supportive care only (i.e. placebo). These analyses, while modelled predictions based on PFS, may assist oncologists with understanding longer-term outcomes in unresectable, Stage III NSCLC.