Patient characteristics at baseline and at follow-up are presented in Table . The 215 examined patients in this study had shorter disease duration (9 years versus 15 years, P < 0.01), lower disease activity measured by the DAS28 (4.00 versus 4.62, P < 0.01), lower global assessment (19 versus 30, P < 0.01) and used less prednisolone (37% versus 54%, P < 0.01) compared with those who were not included (n = 151) from the original cohort (n = 366). The two groups were similar regarding age, gender, body mass index, smoking habits, rheumatoid factor, age of disease onset, erythrocyte sedimentation rate, menopause in women and use of DMARD and AOT.
Change in bone mineral density
In the entire group, a significant loss in hand BMD was seen at 2 years as measured by DXR-BMD (-0.90%) and DXR-MCI (-1.18%), but not as measured in the DXA-BMD hand (0.00%) (Figure ). A significant bone loss was also observed for the DXA-BMD in the total hip (-0.72%) and in the spine L2–L4 (-0.78%) (Figure ).
Figure 1 Bone loss in 215 rheumatoid arthritis patients followed for 2 years. Bone loss assessed by digital X-ray radiogrammetry (DXR) bone mineral density (BMD) and metacarpal cortical index (MCI) of the hand, and by dual-energy X-ray absorptiometry (DXA) BMD (more ...)
The correlation (r value) between the DXR-BMD hand and the DXA-BMD hand was 0.88 (P < 0.001) for baseline values and was 0.35 (P < 0.001) for 2-year BMD changes. Correlations between the change in the DXA hand and in the DXA total hip and spine were 0.35 (P < 0.001) and 0.18 (P = 0.01), whereas correlations between the change in the DXR hand and DXA total hip and spine were 0.23 (P = 0.001) and 0.10 (P = 0.16), respectively. The DXR-MCI was highly correlated with the DXR-BMD both at baseline (r = 0.86, P < 0.001) and as the percentage change over 2 years (r = 0.94, P < 0.001).
Association between disease duration and bone loss
At baseline 37 patients had a disease duration of 3 years or less and 178 patients had a disease duration longer than 3 years. DXA-BMD hand bone loss was only observed in patients with disease duration less than 3 years and not in patients with longer disease duration (-0.96% versus 0.24%, P < 0.01) (Table ), whereas loss of DXR-BMD (-0.46% versus -0.93%, P = 0.76) as well as loss of DXR-MCI (-0.89 versus -1.29, P = 0.66), of the DXA-BMD total hip (-0.26% versus -0.76%, P = 0.51) and of the DXA-BMD spine (-0.71% versus -0.82%, P = 0.64) occurred independent of disease duration. The changes in BMD in the subgroups (according to disease duration) were all significant except for the DXA-BMD hand patients with disease duration longer than 3 years (P = 0.26) and the DXA-BMD spine (P = 0.60) and DXA-BMD total hip patients with disease duration less than 3 years (borderline significant, P = 0.06).
Comparison of the baseline and the change in hand bone mineral density
The patients with short and long disease duration were comparable with regard to demographic variables, disease activity and treatment with DMARD and corticosteroids, but AOT was used less frequently by patients with short disease duration than by patients with long disease duration (16.1% versus 35.5%, P = 0.04). The difference in DXA hand bone loss across patients with short and long disease duration, however, was also significant in the subgroup not using AOT (-1.41% versus 0.11%, P = 0.02). These findings are consistent in a linear regression model adjusted for other variables that may influence hand bone loss (Table ). The analysis was performed both with and without the outlier, with the same results.
Risk factors for hand bone loss in a multivariate linear regression model
Association between disease activity score and hand bone loss
At baseline 55 patients had low disease activity, 103 patients had moderate disease activity and 44 patients had high disease activity. Bone loss changes, as measured by DXR-BMD, differed across patients with different levels of disease activity (low, -0.29%; moderate, -1.13%; and high, -1.03%; P = 0.03), and were borderline significant for DXR-MCI (-0.76, -1.34 and -1.13, P = 0.06) (Table ). No significant difference in DXA-measured hand BMD change was found for the low, moderate and high levels of disease activity (-0.40% versus 0.26% versus 0.04%, respectively; P = 0.40). Hand BMD baseline values, however, were significantly lower in the group with high disease activity in both the DXR-BMD and the DXA-BMD (Table ).
The correlation (r value) between the DAS28 at baseline (continuous scale) and the hand DXR-BMD change was -0.19 (P = 0.01), between the DAS28 and the DXR-MCI change was -0.16 (P = 0.03), and between the DAS28 and the hand DXA-BMD change was 0.08 (P = 0.27). Patients in the group with high disease activity were significant older than the group with lowest disease activity. In a multivariate model, however, disease activity was independently associated with the percentage change in DXR-BMD (B = -0.47, P < 0.01) (Table ) and with the DXR-MCI (B = -0.47, P < 0.01), after adjusting for other variables that could influence hand bone change as well as age.
Association between functional disability (MHAQ score) and hand bone loss
At baseline, 102 patients had a MHAQ score less than 1.50, 78 patients a score between 1.50 and 1.99, and 34 patients had a MHAQ score of two or more. The patient with highest MHAQ score was older than patients with lower MHAQ scores. Regarding correlation between the MHAQ score at baseline and the change in hand DXR-BMD, the DXR-MCI hand and the DXA-BMD hand were nonsignificant both for continuous values (r = 0.00, P = 0.96; r = 0.03, P = 0.70; and r = -0.05, P = 0.51) and for groups (r = 0.02, P = 0.82; r = 0.05, P = 0.47; and r = -0.02, P = 0.82) for the MHAQ score ranges <1.5, 1.50–1.99 and ≥2, respectively. There were no differences in the change in hand BMD dependent on the MHAQ group either in the DXR-BMD hand, the DXR-MCI hand or the DXA-BMD hand. Baseline values, however, were significantly higher in the group with the lowest MHAQ score with regards to DXR-BMD and DXR-MCI (Table ). No such findings were seen regarding DXA measurements.
Associations between treatment and hand bone loss
At follow-up 33% of the patients were current users of AOT (88% used hormone replacement therapy and 12% used bisphosphonates) and 44% were ever users. A significant difference in DXA-BMD hand change was found between users and nonusers of AOT (0.44% versus 0%, P = 0.04). No such difference was seen for DXR-BMD (-1.01% versus -0.66%, P = 0.54) or DXR-MCI (-1.14 versus -1.19, P = 0.60) in users versus nonusers of AOT. Use of AOT, however, was not significantly associated with the change in DXA-BMD in the multivariate analyses (Table ).
No significant difference in hand bone change was seen between ever users (83%) and never users (17%) of DMARD regarding the DXR-BMD hand (-0.90% versus -0.85%, P = 0.29), the DXR-MCI hand (-1.19 versus -0.78, P = 0.17) or the DXA-BMD hand (0.27% versus -0.34%, P = 0.22). During the 2-year follow-up 45% of patients had used prednisolone and 41% were current users at follow-up with a mean dose of 5.7 mg. No significant difference in change of hand BMD was observed between users and nonusers of prednisolone regarding DXR-BMD (-0.94% versus -0.66%, P = 0.19) or DXA-BMD (0.62% versus 0%, P = 0.17), but a group difference between users and nonusers was observed for DXR-MCI (-1.42 versus -0.98, P = 0.05). Prednisolone users, however, had a significantly higher disease activity than nonusers (data not shown) and the significant association between prednisolone and the change in DXR-MCI disappeared in the multivariate analysis (Table ).