Estimating long-term survival in progressive fibrosing interstitial lung disease (PF-ILD) other than IPF using matched IPF data

Abstract

<b>Introduction:</b> Current survival data from patients with PF-ILD other than IPF in the Phase III INBUILD trial of nintedanib vs placebo are limited due to trial duration (1-year survival: 95%), and extrapolation to long-term survival estimates will have a high degree of uncertainty. Since IPF has a similar disease course to PF-ILD, data from long-term IPF studies may be suitable for informing estimates of PF-ILD survival. <b>Aim:</b> To improve models of long-term survival in patients with PF-ILD in the INBUILD trial using propensity score-matched long-term IPF data. <b>Methods:</b> Patients in INBUILD (n=332 nintedanib; n=331 placebo) were propensity score-matched (by age, gender, race, FVC, DL_CO and smoking status) against patients with IPF in nintedanib trials: TOMORROW, INPULSIS 1/2, and the long-term extension INPULSIS-ON (total n=726 nintedanib; n=513 placebo). The matched IPF data were used to inform a Bayesian survival analysis of the PF-ILD data. Several survival models were examined. <b>Results:</b> Log-logistic, gamma and Weibull survival distributions were the best fit for the matched IPF data and produced consistent survival estimates for patients with PF-ILD in the Bayesian analysis (Table). <b>Conclusions:</b> Our analysis provides robust 5-year survival estimates for patients with PF-ILD other than IPF treated with nintedanib (59–60%) or placebo (21–32%) that can be used to inform future decision-making in the absence of long-term PF-ILD data. <b>Table:</b> Survival estimates for patients with PF-ILD in INBUILD