Prediction of five-year survival following presentation with scleroderma: development of a simple model using three...

To identify combinations of easily acquired clinical variables, at first presentation with scleroderma, that would predict subsequent mortality.

In this prospective study of all new patients at one major scleroderma center, 280 patients with definite scleroderma (according to the American College of Rheumatology criteria) whose disease onset occurred from 1982 to 1991 and who were followed up to the end of 1996 were identified. Standardized data collection was performed at entry to obtain data on major clinical and laboratory variables. Vital status was determined by linkage to the National Death Index.

At 5 years, 55 (26%) of 215 women and 21 (32%) of 65 men had died. Univariate analysis showed that older age, diffuse skin disease, higher skin score, low carbon monoxide diffusing capacity, abnormal electrocardiogram findings, proteinuria, hematuria, low hemoglobin level, elevated erythrocyte sedimentation rate (ESR), and presence of antitopoisomerase antibody were all associated with increased mortality. A logistic regression model, validated by Monte Carlo simulation, identified 3 factors, proteinuria, elevated ESR, and low carbon monoxide diffusing capacity, that in combination, had an accuracy of >80% in predicting mortality. The absence of these 3 factors was associated with 93% survival.

A simple model has been developed that appears to accurately predict mortality over 5 years in a cohort of patients newly presenting with scleroderma.