Data from hospital-clinic based studies suggest that patients with rheumatoid arthritis (RA) have a reduction in areal bone mineral density (BMD_a) at both the hip and spine compared to non-RA controls [1–4]. Such studies of patients with established disease will tend to exclude subjects with less marked disease who may be managed in primary care, either from the outset or discharged from specialist care early during the disease course. We studied a sample of women ten years prospectively, recruited following the onset of inflammatory arthritis presenting to primary care, to determine the relative influence of both arthritis and non-arthritis related factors on BMD_a at the spine and femoral neck 10 years after disease onset.

It also included questions about age at menarche and menopause and use of both the oral contraceptive pill and hormone replacement therapy. The revised questionnaire was not formally validated. Height and weight were measured using standard equipment and body mass index (BMI) calculated. Bone densitometry (Hologic, QDR 4000) was performed at the spine and femoral neck sites. All aspects of the study were in compliance with the Helsinki Declaration and approved by the Norfolk Research Ethics Committee.

In this inception cohort of women presenting to primary care with inflammatory polyarthritis there were several interesting findings relating to BMD_a assessed after 10 years. Standard osteoporosis risk factors apart from age and body mass had relatively little impact. Of the disease factors considered those relating to activity and disability were not predictive of bone loss, whereas measures of joint damage were strongly predictive, especially at the femoral neck.

There are several limitations which need to be considered in interpreting the results. The subjects were deliberately selected to include all cases of inflammatory polyarthritis and not restrict inclusion to those with RA. Indeed we have shown that assigning criteria for RA is unstable in this setting during the first 5 years of disease [9]. As shown above however the results are stable independent of RA status. In this study we looked at BMD_a in a subset of subjects attending their 10^th anniversary visit. It is possible that those who had BMD_a performed may have differed from those who did not. However, those scanned did not differ in terms of joint count (swollen joint count mean = 3.9 vs 3.1; p = 0.11), HAQ score (mean = 1.02 vs 1.04; p = 0.82), or CRP level (mean = 12.3 mg/L vs 13.3; p = 0.59), from those who did not providing some reassurance against selection bias though we can not exclude this. Hand radiographs at the 10^th anniversary were only available on a proportion of the subjects. In order to have a 10^th anniversary radiograph, erosive change had to be present on the 5^th anniversary radiograph, so it is likely that these subjects had more severe disease. Indeed we found these subjects to be slightly older, have more joint involvement, higher CRP levels and HAQ score compared to those who did not have a 10^th anniversary hand radiograph. We did not have data on current smoking or OCP use at the time of bone density assessment or during the 10 year period of the study, so the influence of these could not be assessed. Finally the sample size was relatively small and the study had limited power to detect weak associations.

The lack of influence of disease activity on bone loss might be explained by measurement imprecision in relation to the joint counts. In an attempt to minimise this all of the assessments were undertaken by trained research nurses; annual comparative assessments were conducted to ensure standardisation of assessment. Four patients (with active joints) were invited to take part in the assessment process. Each nurse assessed each patient together with an independent observer. The results were compared, discrepancies discussed and repeat examinations performed as necessary to ensure agreement. Further any such imprecision should have applied equally to deformed joint counts which are perhaps harder to standardise, yet which showed a strong influence on BMD_a.

Data from some though not all studies of patients with RA have suggested an association between disease activity assessed using erythrocyte sedimentation rate (ESR) or CRP and reduced bone mass [4, 10, 11]. In our long term follow up study, we observed a trend towards a positive association between CRP and bone mass, though the association did not attain statistical significance. This positive association has been reported previously though the mechanism is unclear [4].

In contrast to some though not all studies [2, 10, 12–15], we found no influence of corticosteroid use. We looked, however, only at ever use of steroids during the follow up period. There was insufficient information about dose of steroids used during the follow up period to allow assessment of the impact of current or cumulative dose in the study.

Previous studies have shown an inverse association with joint damage in RA as expressed by the Larsen score and BMD_a by dual energy x-ray absorptiometry (DXA) [16–18]. In our study disease damage as determined both by the deformed joint count and erosive joint count at follow up was linked with increased bone loss. Such measures presumably represent a better marker of cumulative disease activity resulting over a long period in joint damage - though additional factors including immobility related bone loss may also be a factor. Many of the lifestyle factors investigated, such as smoking, exercise and hormonal use, had no significant influence on BMD_a in this group. It is possible that associations with these variables were masked given the presence of longstanding arthritis. As expected based on studies of women without inflammatory arthritis, increasing age at menarche, weight and body mass index were linked with increased BMD_a [19–21].

In summary, in this group of women with longstanding inflammatory polyarthritis, cumulative disease damage was the strongest predictor of reduced bone density. Other disease and lifestyle factors had only a modest influence.