We studied causes of death (CoDs) between 1952 and 1991 assessed by a clinician before autopsy and then determined at autopsy by a pathologist in 369 subjects with rheumatoid arthritis (RA) and 370 subjects without RA (non-RA). We analysed clinical data for RA subjects between 1973 and 1991. In RA subjects, leading autopsy-based CoDs were RA, cardiovascular diseases and infections. Between diagnoses of CoDs by the clinician and those determined by the pathologist, RA subjects had lower agreement than did the non-RA regarding coronary deaths (Kappa reliability measure: 0.33 vs. 0.46). In non-RA subjects, autopsy-based coronary deaths showed a decline since the 1970s with no such decline in RA. Between subjects treated at any time during RA with disease-modifying anti-rheumatic drugs and those without, autopsy-based CoDs were similar. Coronary death being less accurately diagnosed in RA subjects may indicate that coronary heart disease in RA patients often remains unrecognized.