Recent epidemiological studies have suggested that joint risk factors occur for rheumatoid arthritis (RA) and coronary heart disease. We studied serum cholesterol concentration for its association with the incidence or RA in 28 362 men and 24 444 women free from arthritis at baseline. During a mean follow-up of 21 yr, 161 men and 351 women developed RA. Of these incident cases, 119 men and 229 women were rheumatoid factor (RF) positive. The serum cholesterol concentration was directly proportional to the risk of RF-positive RA among women and RF-negative RA among men; the age-adjusted relative risks (95% confidence intervals) per S.D. (1.4 mM/l) of the cholesterol distribution were 1.20 (1.05-1.38) and 1.56 (1.15-2.10), respectively. No association was observed, however, for RF-negative RA among women or RF-positive RA among men. The results suggest that a still unknown factor closely associated with serum cholesterol may be involved in the aetiology of RA, but complex interactions with sex and RF status seem to occur.