To determine in the elderly the effect of osteoarthritis on bone mineral density (BMD) and on diagnosis of osteoporosis, lumbar spine and hip were radiographed and BMD measured by dual-energy X-ray absorptiometry (DXA) in 120 men and 314 women, aged 60-99 years. Prevalence and severity of osteoarthritis were scored on osteophytes, joint space narrowing and bone sclerosis. Ultrasound measurements were also made at the heel to examine whether osteoarthritis at hip or lumbar spine influence bone at this remote site. Osteophytes were the commonest feature, with men having a higher prevalence than women, and lumbar spine having more disease than hip. Lumbar spine osteophytes affected 75% of men and 61.1% of women, and hip osteophytes affected 31.7% of men and 27.4% of women. Stepwise multiple regression analysis using age, weight, height, osteophytes, sclerosis and joint space narrowing indicated that lumbar osteophytes explained 16.6% of variation in lumbar spine BMD in women, and 22.4% in men. Hip osteophytes had a minimal effect on hip BMD, accounting for only 2.2% of variation in women, and none in men. Sclerosis and joint narrowing had little effect on BMD at lumbar spine or hip. Indirect effects of osteoarthritis on BMD were small and inconsistent across genders. Lumbar spine osteophytes in men explained 3.1% of hip BMD variation and 6% of variation in speed of sound at the heel, whereas hip osteophytes in women explained 2.2% of lumbar spine BMD variation. Osteoporosis at the hip, defined as BMD < 2.5 SD of the young normal mean, was present in 33.1% of women and 25.8% of men, whereas, at the lumbar spine it was present in only 24.2% of women and 4.2% of men. However, in women and men free of spinal osteoarthritis, 37.7% of women and 10% of men had osteoporosis. We conclude that lumbar spine osteophytes affect most subjects over the age of 60 years, and contribute substantially to lumbar spine BMD measured in the anteroposterior position by DXA. The effect is largely direct by virtue of osteophytes being included in the BMD measurement. However, a small indirect effect on remote skeletal sites is also present. Diagnosis of osteoporosis and assessment of osteoporotic fracture risk in the elderly should be based on hip BMD and not on anteroposterior lumbar spine, unless spinal osteoarthritis has been excluded.