Prediction of Pulmonary Complications and Long‐Term Survival in Systemic Sclerosis

Objective To assess survival and incidence of organ‐based complications in a large single‐center cohort of unselected systemic sclerosis (SSc) patients, and to explore predictors of survival and clinically significant pulmonary fibrosis (PF) and pulmonary hypertension (PH). Methods The study cohort consisted of 398 consecutive SSc patients, followed up for up to 15 years. Cox proportional hazards analysis with demographic, clinical, and laboratory characteristics as predictor variables was used to develop prediction models for pulmonary complications and survival. Results The overall survival estimate at the end of followup was 57% among patients with limited cutaneous SSc (lcSSc) and 50% among patients with diffuse cutaneous SSc (dcSSc) (P = 0.017). We found that greater age at disease onset, dcSSc, lower diffusing capacity for carbon monoxide (DLco), lower hemoglobin levels, higher serum creatinine levels, and the presence of PH or cardiac involvement were independent predictors of worse survival. Over the entire followup period, 42% of dcSSc patients and 22% of lcSSc patients developed clinically significant PF (P < 0.001). The variables that predicted clinically significant PF development were dcSSc, greater age at onset, lower forced vital capacity and DLco, and the presence of anti–topoisomerase I antibody, while the presence of anticentromere antibody was protective. There was no difference in cumulative incidence of PH between the 2 subsets—24% in lcSSc and 18% in dcSSc (P = 0.558). Incidence rates were 1–2% per year. The PH prediction model demonstrated that greater age at onset, increase in serum creatinine levels, lower DLco, and the presence of anti–RNA polymerase III or anti–U3 RNP antibodies were associated with increased risk of PH, while anti–topoisomerase I antibody positivity reduced the hazard. Conclusion Our study provides data on long‐term outcome of SSc and the timing and frequency of major organ complications. The predictive models we present could be used as clinical tools for patient risk stratification and could facilitate cohort enrichment for event‐driven studies.