To evaluate the predictive power of changes in levels of antibodies to double-stranded DNA (anti-dsDNA) as a predictor of disease exacerbations in systemic lupus erythematosus (SLE), we performed a prospective study on 72 unselected patients with SLE (mean duration of study 18.5 months, range 6-35 months). Patients were seen at least once every 3 months, and disease activity was scored according to a specific protocol. Plasma samples were obtained at least once every month and were assessed for anti-dsDNA antibody (by the Crithidia luciliae assay, an enzyme-linked immunosorbent assay [ELISA], and the Farr assay) and for complement components C3 and C4. Twenty-seven of 33 disease exacerbations observed during the study period were accompanied by a positive test result for anti-dsDNA antibody (27 by the Farr assay, 19 by the C luciliae assay, and 23 by the ELISA). Twenty-four of these exacerbations were preceded by a significant increase in anti-dsDNA antibody levels (23 by the Farr assay, 12 by the C luciliae assay, and 17 by the ELISA). The first observance of a significant increase in anti-dsDNA antibody levels preceded the exacerbation by 8-10 weeks. Significant increases in anti-dsDNA antibody levels not followed by an exacerbation were observed in 5 cases by the Farr assay, in 7 cases by the C luciliae assay, and in 3 cases by the ELISA; however, in 3 cases, 2 cases, and 1 case, respectively, these increases were followed by an increase in disease activity that did not fulfill the criteria for an exacerbation. Serial measurement of anti-dsDNA antibody levels was more sensitive for predicting exacerbations than was measurement of C3 and/or C4 levels (P less than 0.03). Serial assessment of anti-dsDNA antibody levels, especially by the Farr assay, is a sensitive and reasonably specific method for predicting disease exacerbations in SLE.