Although antiperinuclear factor (APF) has the same specificity for rheumatoid arthritis (RA) as rheumatoid factor (RF), there is no evidence that this specificity is maintained in patients with positive RF-agglutination tests. Thus, we evaluated the specificity and usefulness of APF for RA diagnosis, regardless of RF titre. APF was tested (1:100 threshold) on 214 sera sent for RF evaluation over a 9-month period. These sera were previously determined to have latex or Rose-Waaler (RW) titres > or = 12 or 4 IU, respectively, but not necessarily above the threshold values of 100 and 32 IU. The APF test was performed blindly, and physicians were not advised of the results. In the patient population (119 RA and 95 non-RA) APF still demonstrated good specificity (0.82) for RA. As expected, APF proved useful for RA diagnosis in 28/33 (85%) RA cases with an RF level below 1:100 for latex and 1:32 for RW, thus reducing the number of 'seronegative' RA from 33/119 to 5/119. However, it also improved the serological positive predictive value for RA, even in cases when RW results were > or = 32 IU. Indeed, the positive predictive value for RA when both tests were positive was 0.94 (68/72), whereas concordant results (either positive or negative) for both APF and RW tests allowed correct classification (RA or non-RA) in 94% of cases.