The Idiopathic Pulmonary Fibrosis Prospective Outcomes (IPF-PRO) Registry (ClinicalTrials.gov Identifier: NCT01915511) is a multicenter observational registry of patients with IPF [14]. Patients eligible to participate in the registry were ≥ 40 years of age with IPF that was diagnosed or confirmed at the enrolling center in the past 6 months, did not have malignancy other than skin cancer within the past 5 years, were not on a lung transplant waiting list, and were not participating in a randomized clinical trial. Blood samples were collected at enrollment, and RNA was extracted from whole blood and sequenced as previously described [8]. Participants eligible for this analysis were enrolled between June 2014 and February 2017, had longitudinal outcomes ascertained through December 2019, had blood samples available for sequencing, and had total RNA sequencing that met quality control criteria.

At enrollment, clinical data from the prior 12 months were collected from patient records, and patients provided blood samples and completed patient-reported outcome questionnaires. Patients were considered as using an antifibrotic drug (nintedanib or pirfenidone) at enrollment if they were taking antifibrotic therapy on the day that the enrollment blood sample was taken. Participants were followed prospectively while receiving usual care, with follow-up data (including vital status, lung transplantation, and pulmonary function) collected approximately every six months.

Composite outcomes included death; death or lung transplant; death, lung transplant, or forced vital capacity (FVC) decline; and death, lung transplant, or diffusing capacity of the lungs for carbon monoxide (DLco) decline. FVC decline and DLco decline were defined as the first instance of an absolute decline in the % predicted value of ≥ 10% or ≥ 15%, respectively, compared to the value at enrollment.

As described [9], the 52-gene signature contains 7 increased and 45 decreased genes. Thus, for each patient, the ‘increased ratio’ is calculated by dividing the number of genes with an increased (relative to the cohort geometric mean for the gene) expression by 7, and the ‘decreased ratio’ by dividing the number of genes with decreased expression by 45. Then, an “up score” is derived by obtaining the sum of normalized expression values (subtracted from the cohort geometric mean) of genes with increased expression and multiplying by the increased ratio, and a “down score” derived in a similar manner using the expression values of decreased genes and the decreased ratio. Patients with up scores greater than or equal to and with down scores less than or equal to the cohort median are categorized as at “high risk” of lung transplant or death; other patients are classified as at “low risk.”

Patient characteristics at enrollment were summarized as median (Q1, Q3) and the number of events during follow-up was reported as n (%). Cox proportional hazards models adjusted for age, sex, FVC % predicted, and DLco % predicted at enrollment were used to compare the risk of each clinically meaningful composite outcome among patients with a low-risk compared to a high-risk signature. The Kaplan–Meier method was used to generate survival curves to visualize the event-free survival time in the low-risk and high-risk groups. Time to event was compared between patients with a high-risk versus low-risk signature using the log-rank test. Analyses were performed in all patients and in the subgroup of patients using antifibrotic therapy (nintedanib or pirfenidone) at enrollment.