Analyses

Patient demographics and clinical characteristics at baseline were reported for the pirfenidone 2403 mg/day and placebo groups. A post hoc, blinded review of treatment-emergent MACE-plus and bleeding AEs that occurred up to 28 days post-treatment was conducted. MACE-plus events were adjudicated by an external, academic cardiologist who was blinded to the assigned treatment.

AE preferred terms were used to identify occurrences of MACE-plus and bleeding events. MACE-plus included any CV death [due to myocardial infarction, sudden cardiac death, heart failure, stroke, CV procedures (immediate), and other CV causes (e.g., pulmonary embolism, peripheral arterial disease)], non-fatal myocardial infarction, non-fatal stroke, and acute coronary syndrome [22–24]. Incidences of MACE-plus and bleeding AEs were summarized by treatment group; P values from the Fisher exact test were determined for descriptive purposes. Kaplan-Meier curves were used to graphically display time to first MACE-plus or bleeding event; P values from an unadjusted log-rank test were determined for descriptive purposes.

Subgroup analyses were performed to examine whether the pirfenidone benefit observed in the pooled analyses of the phase III trials persisted when patients also received concomitant cardiovascular medications. Clinical outcomes up to month 12 were compared between treatment arms in each medication subgroup of patients with or without concomitant use of selected CV medication classes during the first year (start date on or before day 365) to assess the impact of these medications on the benefit of pirfenidone. The following outcomes were evaluated: death from any cause; any respiratory-related hospitalization; ≥ 10% absolute decline in percent predicted forced vital capacity (% predicted FVC) or death from any cause; and ≥ 10% absolute decline in % predicted FVC or any respiratory-related hospitalization or death from any cause. To be consistent with the primary efficacy analysis from the individual studies, patients with a missing % predicted FVC value for a reason other than death had the value imputed using the sum of squared differences (SSD) method [20, 21]. The percentage of patients with data imputed using the SSD method was low (month 12: 6.9% pirfenidone, 5.4% placebo); sensitivity analyses were performed to examine the primary analysis results using other imputation methods and the findings remained consistent [20, 21, 25]. For cases when the % predicted FVC was imputed, the corresponding FVC assessment date was imputed as the targeted visit date or the last assessment date, whichever came first. Medication classes included any antithrombotic agent, any anticoagulant agent, warfarin, heparin (including low-molecular-weight heparins), any antiplatelet agent, acetylsalicylic acid, and any statin. Within each medication class, the hazard ratio (95% confidence interval) for the time to first event between pirfenidone and placebo groups was estimated using the Cox proportional hazards model, with individual studies as a stratification factor and no additional covariate adjustments. The proportional hazards assumption was evaluated using the supremum test [26].

As a reference, real-world CV comorbidity rates were also provided based on data from the prospective, real-world pirfenidone registry, PASSPORT, and similar registries of patients with IPF, INSIGHTS-IPF and IPF-PRO [27–29].