Early changes in bone mineral density measured by digital X-ray radiogrammetry predict up to 20 years radiological outcome in rheumatoid arthritis

Rheumatoid arthritis (RA) is an inflammatory disease characterized by chronic synovial inflammation commonly associated with destruction in cartilage and bone tissue. Changes in bone metabolism during the course of RA are usually divided into periarticular bone loss with or without focal articular bone erosion and generalized osteopenia manifested by loss of both trabecular and cortical bone. Periarticular bone loss occurs early and often before erosion is apparent. This is linked to the inflammatory process and may be driven by locally released inflammatory mediators such as the receptor activator of nuclear factor (NF)-κB and its ligand RANKL and the decoy receptor of RANKL, osteoprotegerin. These findings have been extensively investigated and reviewed [1–9]. In RA, periarticular bone loss is seen predominantly in the hands and feet, which are affected early in the disease course. Thus, bone loss may be a marker of disease activity, and it has been suggested that it may also relate to future joint damage [4, 5, 8].

Early loss of bone mineral density (BMD) in the hand has been shown to covariate with radiographic joint damage as measured by the Sharp-van der Heide (SvdH) or Larsen score in RA patients [4–11]. However, outcome measures do not necessarily agree. The two more established radiological scores in RA, the Larsen and the SvdH, demonstrate only a modest correlation. This may partly be explained by the fact that the Larsen score is more global, while the SvdH score measures specific joint regions [12–14].

The bone mineral density measured by digital X-ray radiogrammetry (DXR-BMD) determination has the advantage of being well standardized and not subject to the interpretation and measurement errors that are inherent in both the Larsen and the SvdH scoring systems. Furthermore, DXR-BMD is reliable in quantifying demineralization and/or osteoporosis, which have been shown to be only imprecisely verified with conventional radiography. The radiographic scoring methods and DXR-BMD can be performed on historical radiographs, provided that these are of sufficient quality and taken in a standardized fashion. This is in contrast to other methods of BMD measurements such as ultrasound, dual-emission X-ray absorptiometry and peripheral quantitative computed tomography, all of which have as a focus the measurement of BMD, are not retrospectively applicable to standard radiographs of the hands and sometimes focus on generalized osteopenia [6–8, 15–18].

In Lund, Sweden, we have prospectively monitored a cohort of early RA patients since 1985 and have more than 20 years of follow-up information on radiographic and clinical outcomes [19, 20]. In the present study, we have examined the relationship between early DXR-BMD changes and short- and long-term outcomes as measured radiographically by the Larsen score of the hands and feet. Furthermore, we have examined the relationship between baseline BMD, Larsen score and early Larsen progression and long-term radiographic outcome.

In the present report, we have demonstrated that early bone mineral loss in the hands of patients with RA predict joint damage as measured by the radiographic Larsen score over a 20-year observation period. The difference was seen as early as 1 year after baseline all the way to 20-year follow-up.

We could not verify that DXR-BMD loss is more marked in the early phase of the disease, which has been reported by others, but our study had more limited power [4–11]. Overall our results are consistent with those of other studies. A longitudinal observational study including early RA patients with disease duration <1 year showed that changes in DXR-BMD of the second to fourth metacarpal bones at year 1 were specific and sensitive in identifying both patients who developed erosions scored using both the Larsen and SvdH methods and those whose existing erosion had progressed at year 4 [4]. Similarly, in RA patients with disease duration up to 4 years, hand bone loss measured by DXR-BMD at year 1 was predictive of subsequent radiographic damage scored by the SvdH method at 5 and 10 years [7].

We did not attempt to study the implication of late DXR-BMD loss in this study, since we aimed to find early predictors of outcome. On the other hand, the clinical value of early bone mineral loss as a predictor of future joint damage as measured by Larsen score should not be overemphasized. The relationship is well validated on the group level, but cannot be recommended for use as the only predictor as illustrated by the relatively low correlations. On the other hand, early progression in Larsen score also shows a clear predictive value for later joint damage, but the low correlations revealed that this information also must be confined to the group level. Early bone loss is a risk factor, in addition to other known risk factors. Thus, although statistically significant predictive value for RA prognosis can be found for several early findings such as the presence of rheumatoid factor, anticyclic citrullinated peptide antibody, elevated ESR level and cartilage oligomeric matrix protein level, their value in clinical decisions regarding treatment of the individual patient must not be overemphasized [22–29]. Treatment decisions rely heavily on several aspects not covered by conventional predictive analyses, such as perceived pain, working situation, deteriorating function and so on.

The findings that early DXR-BMD loss in the cortical shaft of metacarpal bones is predictive of later progression of joint-related hand Larsen score was not unexpected because of the close vicinity of these structures. We were not able to demonstrate an unequivocal relationship between early DXR-BMD loss in the metacarpal bones and Larsen score in the foot, even though there was a quite clear trend for such a relationship as has been reported for SvdH [11]. However, in the Larsen scoring system, as in the SvdH scoring system, the feet do not have the same number of joints as the hands, thus giving a ceiling effect with reduced possibility of finding differences [12–14]. This could at least partly explain the nonsignificant differences found in the Larsen scores of the feet. Methods of using DXR-BMD have not been developed for the metatarsal bones, thus precluding direct comparisons of hand and foot bone loss as well as a possible relationship between more generalized bone loss and multiple joint damage as measured by the total Larsen score.

Several factors are known to promote osteopenia, such as female sex, older age and severe RA [30, 31]. It was therefore not unexpected to find that the degree of early DXR-BMD loss identified patients with different baseline characteristics. Thus early markers of more severe disease, such as high HAQ score, high DAS44 score and high ESR and CRP levels, were all more common among the patients with higher DXR-BMD loss. Also older age contributed to high early bone loss, while rheumatoid factor status or anticitrullinated peptide antibody positivity and sex did not. The low numbers of patients taking systemic glucocorticoids represent the therapeutic tradition in the 1980s and prevent any meaningful analysis of this cause of osteopenia.

The strength of the DXR-BMD technique is that it is well standardized, can be used to perform a large number of objective measurements and can be done on existing radiographs. However, the radiographs must be taken in a standardized fashion if the results are to be given in grams per square centimeter. DXR-BMD can be somewhat more generalizable if the results are given as percentages, which are sufficient to measure relative changes. Both the Larsen and SvdH scores require experienced and licensed readers and still have a sizable reading error [12], a shortcoming that DXR-BMD has minimized, although the radiographic equipment used can influence the results to a minor degree. DXR-BMD is able to quantify only cortical and not trabecular bone mass, thus limiting its use as a general osteopenia measure. However, periarticular cortical bone loss, in some cases measured by DXR-BMD, has been shown to predict erosive disease in RA [4–11].

The strengths of the current study include its community-based recruitment, its prospective gathering of a large amount of laboratory and clinical information and its almost complete follow-up. The exclusion of the sizable proportion of RA patients who died during follow-up could be argued for. However, they did represent a group of RA patients that presumably had several severity markers in common, and the exclusion of these patients could seriously hamper the results of the study. We also refrained from extrapolating and imputing missing Larsen scores, since this also leads to several assumptions and limitations in this restricted sample size.

The limitations of the present study include the retrospective approach of DXR-BMD measurements on preexisting radiographs. This accounted for the loss of 48 patients. However, their baseline characteristics were largely similar to those of the total group of patients. Also, the total sample size was not very large, but this was one of the very first early RA cohorts recruited, and it represents what one center could bear during a total of >25 years of follow-up. Thus, although some of the finer details may not always be statistically verified, the cohort size permits firm conclusions on several robust sets of data such as those in the present report. It must be remembered that in comparing data with more recent early RA studies, in which new and often more liberal criteria sets are being used, we used the 1958 American Rheumatism Association criteria to establish definite diagnoses of RA in the present study [32]. Also, the introduction of new and potentially bone loss-protective treatment cohorts must be accounted for in more recent early RA studies when comparing them with the Lund early RA cohort treated according to the clinical standards used over 25 years ago.

Early DXR-BMD loss in the present study predicted future joint damage as measured by Larsen score both in the short-term perspective (1 year), which confirms previous studies that used the SvdH score, and for the first time in the very long-term perspective (20 years).