The goal of this prospective longitudinal study was to determine the serological profile of early rheumatoid arthritis (RA), and to test whether antikeratin antibody (AKA), antiperinuclear factor (APF), anti-RA33 antibody and antinuclear antibodies (ANA) had an additional diagnostic value when prescribed after rheumatoid factor (RF)-detecting methods. Sixty-nine patients with early polyarthritis suggestive of RA, seen between 1991 and 1993, were included. Five autoantibodies (i.e. RF, AKA, APF, RA33, ANA) were looked for at regular intervals. After 24 months follow-up, patients were classified as having RA (n = 49), unclassified polyarthritis (UP; n = 15) or other rheumatic diseases. Among patients with early RA, the sensitivity of these markers was 40.8% for RF, 36.7% for AKA, 28.6% for APF and 28.6% for anti-RA33. Among RF-negative RA patients, 51.7% were positive for AKA, APF, anti-RA33 antibodies and/or ANA. Positivity of the three recent markers usually persisted throughout follow-up, whereas RF was lost by 58% of patients with early, RF-positive, treated RA. Using multivariate analysis, only latex, RF test and AKA or APF had an independent and statistically significant diagnostic value for early RA. Our data suggest that RF and AKA (or APF) should be concomitantly determined for diagnosis in patients with suspected early RA.