Heterogeneity in the progression of idiopathic pulmonary fibrosis (IPF) might reflect diversity in underlying pathobiology, and represents a major challenge in the prediction of clinical progression and treatment benefit. Previous studies have found peripheral blood concentrations of several protein biomarkers to be prognostic for overall survival duration in patients with IPF, but these findings have generally not been directly compared and replicated between cohorts. We aimed to use the pivotal trials for pirfenidone to evaluate prognostic and predictive properties of biomarkers across multiple endpoints, and whether they are modulated by pirfenidone treatment. We did post-hoc analyses of test and replication cohorts from CAPACITY 004 (NCT00287716), CAPACITY 006 (NCT00287729), and ASCEND (NCT01366209) trials for the plasma proteins CCL13, CCL17, CCL18, CXCL13, CXCL14, COMP, interleukin 13, MMP3, MMP7, osteopontin, periostin, and YKL40. Eligible participants had IPF and received pirfenidone 2403 mg/day or placebo in CAPACITY (test cohort) or ASCEND (replication cohort), were aged 40-80 years, and without missing biomarker data at baseline. To identify biomarkers that were consistently prognostic for clinical outcome measures, the primary analysis was the association between biomarker concentrations at baseline and absolute change in percentage of predicted forced vital capacity (FVC%pred) at 12 months (CAPACITY week 48, ASCEND week 52) in the placebo group. Biomarkers within the test cohort that met predefined success criteria of a prognostic p value less than 0·10 from multivariate analysis were further assessed in the replication cohort. Furthermore, the predictive effect size (ie, biomarkers that were predictive for benefit from pirfenidone) was calculated as the difference in FVC%pred treatment effect (pirfenidone in relation to placebo) between high versus low biomarker subgroups at week 48 (test cohort) or week 52 (replication cohort). Several baseline biomarkers (CCL13, CCL18, COMP, CXCL13, CXCL14, periostin, and YKL40) were prognostic for progression outcomes in the placebo groups of the test cohort. However, only CCL18 was consistently prognostic for absolute change in percentage of FVC%pred in both the test (p=0·032) and replication (p=0·004) cohorts. Pirfenidone treatment benefit was consistent regardless of baseline biomarker concentration. Blood CCL18 concentrations were the most consistent predictor of disease progression across IPF cohorts with potential to inform new target discovery and clinical trial design. Future validation of these findings in prospective studies is warranted. Genentech Inc.