RESUMO
Background: Organ involvement often occurs in early systemic sclerosis and has been related to premature death. Identifying patients at diagnosis at risk of developing early organ involvement would be useful to optimize screening and management strategies. Objective: To develop prediction models for the 5-year development of interstitial lung disease, pulmonary arterial hypertension and death. Methods: A European multicentre inception cohort was created. For modelling, predefined clinical variables with known predictive value at diagnosis were used. Univariate and multivariate regression analysis were done to select baseline predictors and build the prediction models. The models were tested using the area under the receiver operating characteristic curve comparing observed and expected frequencies. Results: Of 735 patients, 23% developed interstitial lung disease, 8% developed pulmonary arterial hypertension 12% died. The interstitial lung disease model included diffuse cutaneous systemic sclerosis (OR = 1.8), systemic sclerosis disease duration < 3 years (OR = 1.4), puffy fingers (OR = 1.6), and anti-topoisomerase-I-antibodies (OR = 1.8). The pulmonary arterial hypertension model included age > 65 years (OR = 3.2), forced vital capacity < 70% (OR = 2.5) and diffusing capacity of the lung for carbon monoxide < 55% (OR = 1.9). Death was predicted best by age > 65 years (OR = 4.1), male gender (OR = 1.9), no anti-centromere antibodies (OR = 0.5), proteinuria (OR = 1.9), forced vital capacity < 70% (OR = 1.8) and pulmonary arterial hypertension at diagnosis (OR = 10.1). The area under the receiver operating characteristic was 0.66 (95% CI 0.64-0.67), 0.66 (95% CI 0.64-0.68) and 0.70 (95% CI 0.69-0.72), respectively. Conclusion: We have shown that it is possible to predict interstitial lung disease, pulmonary arterial hypertension and death using established variables already available at the moment of systemic sclerosis diagnosis. Discriminatory performance of the models was suboptimal. Further research including new variables is necessary to improve performance.
