We analyzed 566 patients (515 females and 51 males) with systemic lupus erythematosus who were treated in Okinawa, Japan, from 1972 to 1991 and followed until April 1993. One hundred four patients (95 females and 9 males) died, and 51 were considered to have end-stage lupus nephritis. The annual incidence and prevalence, per million population in each sex, were increased from 16.0 and 66.0 in 1972 to 46.7 and 683.9 in 1991 in the female patients, and from 4.2 and 8.3 in 1973 to 8.3 and 70.0 in 1991 in the male patients, respectively. Cox proportional hazard analysis was done to determine the effects of several covariates on patients and renal survival. The patients' survival rate improved, as the hazard ratio (HR) was decreased to 0.69 (year of diagnosis, 1982 to 1986) and to 0.48 (year of diagnosis, 1987 to 1991) when the HR in patients diagnosed before 1981 was taken as 1.00. Similarly, we examined renal survival by using the Cox proportional model. For this analysis, the date of start of dialysis therapy was regarded as the time of renal death. Male patients had significantly poor renal survival; the HR was 3.64 (95% confidence interval, 1.89 to 6.98) when the HR in the females was taken as 1.00. However, age at diagnosis and year of diagnosis did not significantly affect renal survival. The cumulated incidence of end-stage lupus nephritis from the time of diagnosis of systemic lupus erythematosus was 3.1% at 5 years, 9.4% at 10 years, 15.5% at 15 years, and 21.0% at 20 years.(ABSTRACT TRUNCATED AT 250 WORDS)