Background While clinical trials in systemic sclerosis-related interstitial lung disease (SSc-ILD) have traditionally used FVC% predicted as the primary outcome, combining individual outcomes may lead to a more comprehensive measure of treatment response and minimize the risk of type 1 error from multiple testing of secondary outcomes.
Objectives To develop a composite outcome measure to assess treatment response in patients with SSc-ILD.
Methods Using data from the SLS-I study comparing cyclophosphamide treatment versus placebo in 158 patients with SSc-ILD, we entered the following outcome variables into an univariate analysis to determine which variables had a significant treatment effect at 12 months: FVC% predicted, TLC% predicted, computer-based quantitative lung fibrosis in the zone of maximum fibrosis (QLF-ZM) score from thoracic high-resolution computed tomography scans, transitional dyspnea index (TDI), visual analogue scale for breathing (VAS-B). We subsequently combined the variables with significant treatment effects in a principal component analysis to assess the difference between treatment groups.
Results Of the 158 patients, 83 had complete outcome data and were included in this analysis. The univariate analysis demonstrated significant treatment effects for the following individual outcome variables (estimate [SE]; p-value): QLF-ZM (-10.5 [3.6]; p=0.005), TDI (3.7 [0.8]; p<0.0001). There was a trend for a significant treatment effect using FVC% predicted as the outcome in the univariate analysis (estimate 3.6 [SE 1.8]; p=0.05). The regression model with the first principal component for TDI and QLF-ZM as the composite outcome demonstrated a significant treatment effect favoring cyclophosphamide (Estimate 0.7 [SE 0.2]; p=0.001). Adding FVC% predicted to the composite outcome model did not change the overall treatment effect.
Conclusions The composite outcome comprised of QLF-ZM and TDI demonstrated a significant treatment effect favoring cyclophosphamide for the treatment of SSc-ILD. The treatment effect observed from using the composite outcome was more significant than the effect observed using FVC %predicted. These findings suggest that combining a patient-reported outcome with a structural outcome into a single measure may serve as a more robust measure of treatment response compared with FVC alone. This model requires validation using another dataset.
Acknowledgements The Specialty Training and Advanced Research (STAR) Program at the David Geffen School of Medicine at UCLA supports the first author.
Disclosure of Interest None declared