The predictive value of a number of clinical and laboratory variables for the mortality of 148 patients with systemic lupus erythematosus (SLE) with a mean observation period of 8 years and a 10-year-survival of 80 per cent was calculated by means of differentiated survival rate analyses and stepwise regression analyses. The predictive power of several variables increased if the calculations were based on deaths caused by SLE rather than on the total mortality rate. The survival rate decreased after 1973 because a diagnosis of SLE was made in some patients with terminal disease who would have remained without a diagnosis before that time. The causes of death and the treatment were identical before and after 1973. The presence of a high number of diagnostic ARA criteria within the first year of observation was a predictor of decreased survival. Severe but non-fatal infections (meningitis, septicemia, pneumonia) significantly reduced the survival rate. Patients with proteinuria and azotemia, within the first 2 years of observation, had a 10-year-survival of 70 per cent. The survival of patients with CNS manifestations was not significantly reduced. The butterfly rash and the presence of lymphopenia were predictors of decreased survival, whereas the presence of DNA antibodies had no predictive value for survival.