Predictors of survival were determined in 171 patients with systemic sclerosis by univariate analysis, and the Cox proportional hazards model using both cross sectional data at entry into the follow-up and time-dependent follow-up data. Clinical and laboratory data were evaluated from 1982 to the end of 1993. The presence of diffuse scleroderma, kidney and cardiac involvements were unfavourable prognostic signs in both the univariate analysis, and the Cox proportional hazards models. The Cox model, using the variables detected at study entry, indicated that pericarditis, and anaemia were bad prognostic signs. Analysis with time dependent data has not been reported in systemic sclerosis. The appearance of pigmentation disturbances, anaemia, and respiratory failure during the follow-up also caused a poor prognosis of the disease by the Cox model. In the stepwise selection models, diffuse scleroderma, internal organ manifestations including renal, and cardiac involvements were predominantly selected as the most unfavourable factors for survival. As to the extent of skin involvement and internal organ manifestations, the general behaviour of the disease seems to be similar throughout the world. The early appearance of pericarditis and pigmentation disturbances at study entry are bad prognostic signs.