Fifty four coded sera, 38 from eight patients with systemic lupus erythematosus (SLE), four from one patient with systemic vasculitis, one from one patient with polyarthritis and 11 normal controls were tested for anti-dsDNA antibodies using seven commercial enzyme linked immunosorbent assays (ELISA) and the radioimmunoassay method (RIA) routinely used in our unit. Sensitivity, specificity and predictive values were tested for both SLE diagnosis and disease activity. Using anti-dsDNA antibodies as a diagnostic test for SLE there were differences in sensitivity (from 66% to 95%), specificity (from 75% to 100%), predictive positive values (from 89% to 100%) and predictive negative values (from 50% to 87%) among ELISA kits. The RIA method was either more specific or equal to ELISA kits. Using 'equivocal' values as positive values an increase in sensitivity was observed but at the expense of specificity. Similar differences and trends were observed when the results were used as a measure for disease activity. This suggests that there are differences in sensitivity, specificity and predictive values among ELISA kits both in the diagnosis of SLE as well as in the determination of disease activity.