Background: The presence of "anti-DNA antibodies in abnormal titres" is a well established criterion for SLE classification, but there is no agreement on the performance of this test. Objective: To study the correlation between clinical findings and five different solid and solution phase anti-DNA antibody assays. Methods: 158 consecutively collected ANA positive sera were studied in a double blind fashion. Anti-DNA antibodies were determined by different solid phase assays (ssDNA-, dsDNA- specific ELISA, EliA anti-dsDNA assay, Crithidia luciliae assay), and by an experimental solution phase anti-DNA assay using biotinylated pUC18 plasmid, human, calf thymus, and E coli DNA. Antibody affinity was determined by surface plasmon resonance. Clinical data were obtained independently of the laboratory analyses and later related to the anti-dsDNA findings. Results: Anti-dsDNA antibodies were most frequently detected by ELISA, but were not specific for SLE as they were present in up to 30% of other disease groups. Those detected by the Crithidia luciliae assay were predictive for SLE, while antibodies binding in solution phase ELISA using the pUC18 correlated strongly with the Crithidia luciliae assay. Surface plasmon resonance analysis showed that antibody binding to pUC18 was not due to higher relative affinity for dsDNA in general, but apparently to specificity for that plasmid DNA. Serum samples from three patients with lupus nephritis were positive in both pUC18 solution phase and Crithidia luciliae assays. Conclusions: Assay principle selection is decisive for the detection of clinically significant anti-DNA antibodies. Revision of the anti-DNA antibody criterion in the SLE classification may be needed.