OBJECTIVES: Idiopathic pulmonary fibrosis (IPF) is a chronic and debilitating disease with poor long-term prognosis and survival. When assessing the long-term effectiveness and cost-effectiveness of IPF treatments, it is necessary to model long-term survival since most patients are alive at the end of clinical trials. This study assessed the feasibility of modeling long-term survival in IPF using patient-level simulation based on IPF-specific mortality risk prediction models.

METHODS: We developed a patient-level simulation model to predict the natural history disease trajectories for patients with IPF based on published estimates of disease progression over time. The model produced patient-level trajectories for key clinical parameters and mortality risk predictors, including percent predicted forced vital capacity, percent predicted diffusing capacity of the lungs for carbon monoxide, 6-minute walk distance, and history of respiratory hospitalization, in 6-month increments over a lifetime time horizon. Mortality was estimated from patients’ disease trajectories using three different mortality risk prediction models: Gender-Age-Physiology (GAP), longitudinal GAP, and du Bois et al. Survival estimates were compared to long-term survival data from published trials and observational studies.

RESULTS: The predicted 2-, 5-, and 10-year predicted survival rates were 49%-86%, 14%-62%, and 1%-31%, respectively, with the highest estimates produced by the longitudinal GAP model and the lowest by the du Bois et al. model. Results also varied across the different GAP and longitudinal GAP model specifications (calculator, categorical or point index versions). Despite the considerable variation in long-term survival predictions, estimates lie within the wide range of long-term survival data from published clinical and observational studies.

CONCLUSIONS: Patient-level simulation informed by disease progression trajectories and mortality risk prediction models is a viable approach for predicting long-term survival in IPF. Further investigation and validation of long-term disease progression and mortality risk predictions are needed to increase the confidence in this modeling approach for use in health technology assessment.