Multivariable mixed-effects Cox proportional hazards regression models in two cohorts of idiopathic pulmonary fibrosis (IPF) patients
| Disease severity and emphysema variables in the models | Patients for emphysema thresholds | C-index | Hazard ratio | p-value | 95% CI |
| Cohort 1 | |||||
| VILDemph | Emphysema=0%: n=79 | 0.76 | 1.05 | <0.0001 | (1.04–1.06) |
| Binary visual emphysema (threshold: 0%) | Emphysema >0%: n=141 | 1.03 | 0.931 | (0.55–1.94) | |
| VILDemph | Emphysema ≤5%: n=123 | 0.76 | 1.05 | <0.0001 | (1.04–1.07) |
| Binary visual emphysema (threshold: 5%) | Emphysema >5%: n=97 | 0.62 | 0.090 | (0.35–1.08) | |
| VILDemph | Emphysema ≤10%: n=155 | 0.76 | 1.05 | <0.0001 | (1.04–1.07) |
| Binary visual emphysema (threshold: 10%) | Emphysema >10%: n=65 | 0.75 | 0.273 | (0.44–1.26) | |
| VILDemph | Emphysema ≤15%: n=179 | 0.76 | 1.05 | <0.0001 | (1.04–1.07) |
| Binary visual emphysema (threshold: 15%) | Emphysema >15%: n=41 | 0.71 | 0.244 | (0.39–1.27) | |
| Cohort 2 | |||||
| VILDemph | Emphysema=0%: n=126 | 0.70 | 1.03 | <0.0001 | (1.02–1.04) |
| Binary visual emphysema (threshold: 0%) | Emphysema >0%: n=184 | 0.81 | 0.234 | (0.58–1.14) | |
| VILDemph | Emphysema ≤5%: n=202 | 0.71 | 1.03 | <0.0001 | (1.02–1.04) |
| Binary visual emphysema (threshold: 5%) | Emphysema >5%: n=108 | 0.79 | 0.231 | (0.55–1.16) | |
| VILDemph | Emphysema ≤10%: n=248 | 0.70 | 1.03 | <0.0001 | (1.02–1.04) |
| Binary visual emphysema (threshold: 10%) | Emphysema >10%: n=62 | 0.76 | 0.225 | (0.49–1.18) | |
| VILDemph | Emphysema ≤15%: n=271 | 0.70 | 1.03 | <0.0001 | (1.02–1.04) |
| Binary visual emphysema (threshold: 15%) | Emphysema >15%: n=39 | 0.68 | 0.160 | (0.40–1.16) |
To evaluate whether the combined pulmonary fibrosis and emphysema (CPFE) phenotype had an additive impact on outcome of IPF patients after adjusting for patient age, sex, smoking status (never versus ever), antifibrotic use (never versus ever) and baseline disease severity, multivariable mixed-effects Cox regression models were used to analyse two independent cohorts of IPF patients. Separate centres/countries within cohort 1 and cohort 2 were modelled as multilevel with random effects between centres/countries (a random intercept per centre/country). Baseline disease severity was quantified as the sum of average lobar visual CT extents of emphysema and ILD: VILDemph. The CPFE phenotype was separately characterised by a binary visual emphysema variable using thresholds of emphysema extent including: 0%, 5%, 10% or 15%, which were reported in various definitions of CPFE in the literature. The impact of the four binary visual emphysema thresholds were separately analysed in multivariable models. C-index: concordance index.