The purpose of this study was to determine the cumulative incidence of lupus nephritis (LN) and the factors predictive of its occurrence in a multiethnic systemic lupus erythematosus (SLE) cohort. We studied 353 SLE patients as defined by the American College of Rheumatology (ACR) criteria (65 Hispanics, 93 African-Americans and 91 Caucasians). First, we determined the cumulative incidence of LN in all patients. Next, we determined the predictors for LN in those with nephritis occurring after diagnosis. The dependent variable, LN, was defined by: (1) A renal biopsy demonstrating World Health Organization (WHO), class II-V histopathology; and/or (2) proteinuria > or = 0.5 g/24 h or 3+ proteinuria attributable to SLE; and/or (3) one of the following features also attributable to SLE and present on two or more visits, which were performed at least 6 months apart--proteinuria > or = 2+, serum creatinine > or = 1.4 mg/dl, creatinine clearance < or = 79 ml/min, > or = 10 RBCs or WBCs per high power field (hpf), or > or = 3 granular or cellular casts per hpf. Independent variables assessed at diagnosis, and if absent, at baseline, were from four domains: sociodemographic, clinical, immunologic and immunogenetic (including the complete antibody profile and MHC class II alleles), and health habits. Variables with P < 0.05 by chi square analyses were entered into domain-specific stepwise logistic regression analyses controlling for disease duration, with LN as the dependent variable. Significant domain-specific regression variables (P < or = 0.1) were then entered into an overall model. The cumulative incidence of LN was 54.3% in all patients, and 35.3% for those developing LN after diagnosis. LN after diagnosis occurred in 43.1% of 65 Hispanics, 50.5% of 93 African-Americans, and 14.3% of 91 Caucasians, P < 0.0001. The duration of follow-up for those with LN after diagnosis was 5.5+/-2.4 vs 4.0+/-2.9 years for those without LN. Hispanic (odds ratio (OR) = 2.71, 95% confidence limits (CL) = 1.07-6.87, P < 0.04) and African-American ethnicities (OR = 3.13, 95% CL = 1.21-8.09, P < 0.02), not married or living together (OR = 3.45, 95% CL = 1.69-7.69, P < 0.0003), higher SLAM score (OR = 1.11, 95% CL = 1.02-1.19, P < 0.007), anti-dsDNA (OR = 3.14, 95% CL = 1.50-6.57, P < 0.0001) and anti-RNP (OR = 4.24, CL = 1.98-9.07, P < 0.0001) antibodies were shown to be significant predictors of the occurrence of LN. Repeated analyses excluding the patients with missing HLA data showed that absence of HLA-DQB1*0201 was also a significant predictor for the occurrence of LN (OR = 2.34, CL = 1.13-5.26, P < 0.04). In conclusion, LN occurred significantly more often in Hispanics and African-Americans with SLE. Sociodemographic, clinical and immunologic/immunogenetic factors seem to be predictive of LN occurring after the diagnosis of SLE has been made.