307. Long-Term Survival in Systemic Sclerosis Patients: Development of a Predictive Model

Abstract

Background: SSc is associated with increased mortality but few studies have focused on long-term survival predictors. We explore outcome in a large cohort of SSc patients followed for up to 15 years. Methods: First available information on symptoms, signs and test results was recorded and data available within the first 3 years from disease onset were used in the analysis. A prediction model was developed in a cohort of 398 patients with disease onset between 1995 and 1999. Temporal internal validation used a separate cohort of 279 patients with onset between 2000 and 2003. Results: Although diffuse cutaneous subset (dcSSc) was associated with significantly worse survival compared with the limited cutaneous subset (lcSSc) with 5, 10 and 15 year survival of 86%, 72% and 55% v 94%, 82% and 69% respectively, P=0.017, presence of organ-based complications was associated with a substantial increase in mortality in both subsets. In fact, survival rates were very similar for lcSSc and dcSSc patients with organ involvement (88% and 86% at 5, 67% and 68% at 10 and 47% and 43% at 15 years respectively). Although dcSSc patients had similar mortality rates over the first 5 years of disease (86%), whether or not major organ involvement had been identified, later mortality rates decreased and by the end of follow-up their survival was similar to that of lcSSc patients with no organ complications (76% and 81% at 15 years). Univariable Cox regression demonstrated that skin score (HR 1.02), dcSSc subset (HR 1.5), lung function tests-FVC, DLCO and KCO (HR 0.99, 0.97 and 0.99), age at disease onset (HR 1.05), haemoglobin (HR 0.76), ESR (HR 1.02) and creatinine (HR 1.003) strongly predicted survival. In addition, cardiac SSc, pulmonary hypertension, clinically significant pulmonary fibrosis and renal crisis development within 3 years from onset were associated with HRs of 5.4, 6.9, 2.4 and 2.4 respectively. Although SSc hallmark autoantibodies strongly associate with organ disease, the only specificity that significantly predicted survival was anticentromere (HR 0.6), which lost its significance when included into a multivariable model. The model that was initially derived in the 1995- 1999 cohort was subsequently tested and updated in the 2000-2003 cohort. The covariates that remained significant independent predictors of survival were age at disease onset>60 years (HR 3.1), diffuse subset (HR 1.5), DLCO<65% (HR 2) and presence of PH (HR 4.4) or renal crisis (HR 1.9) within the first 3 years of disease. Model calibration was excellent, although discriminatory performance remained moderately good (AUC 0.72). Conclusion: Although SSc specific antibodies predict pattern of organ disease in SSc patients, they do not associate with survival independently. Subset and organ-based complications are better predictors of mortality and this may reflect disease severity.