This longitudinal cohort study identified distinct biomarkers for BMD changes at each of three anatomical sites. We demonstrated that the predictors of BMD in the lumbar spine were serum homocysteine, whereas the predictors of BMD were ACPA in the proximal femur and serum TRACP-5b in the distal forearm, respectively, along with osteoporosis drugs BP and VitD (summarized in Table 5). To the best of our knowledge, this is the first prospective, longitudinal study showing the distinct differences of biomarkers in different anatomical sites predicting the changes of BMD in patients with RA.
The rationale for and the mechanisms underlying these differences in bone biomarkers at each anatomical site has been little studied. Souza-Faloni et al. demonstrated that osteoclasts from different bone sites appear to differ in many respects. They reported that bone marrow cells from different places in the skeleton have different dynamics of osteoclast genesis and that these differences are mainly related to differences in the cellular conformation of the site-specific bone marrow [28]. Alternatively, Fehérvári demonstrated a body site-specific link between the severity of atherosclerosis and osteoporosis in patients with peripheral artery disease [29]. Further, de Carvalho et al. argued that long-lasting kidney disease, which is another disease causing secondary osteoporosis, is associated with poor BMD at the hip but not at the spine [30]. Therefore, the differences in risk factors and predictive biomarkers that we observed are reasonable from the perspective of bone biology.
This study identified serum homocysteine as a predictive biomarker for the change in BMD in the lumbar spine. A previous animal study showed that high homocysteine levels induce bone loss [31], while another report suggested that high serum homocysteine might influence bone mineral density, bone turnover, bone blood flow, and collagen cross-linking [32]. Homocysteine is known to be associated with inflammatory processes [33], but, in this study, CRP was not a significant predictor of homocysteine level, suggesting that the influence of systemic inflammation may not be directly associated with the homocysteine level. Indeed, Bahtiri et al. reported that serum homocysteine levels were inversely related to lumbar spine BMD and femur neck BMD in women with osteoporosis [34]. The reason why homocysteine was identified as a predictive biomarker of BMD change only in the lumbar spine remains to be investigated, but it is possible that homocysteine tends to accumulate in the spine, which has less cortical bone and more cancellous bone than the other two sites [35].
Another interesting finding of this study was that ACPA was a significant predictor of BMD change in the proximal femur. ACPA is known as a risk factor not only of joint destruction but of bone loss in RA patients [36]. Moreover, a few reports showed significant associations between ACPA and atherosclerosis or ischemic heart disease in RA patients [37]. Taken together, these findings suggest that the proximal femur may be particularly affected by vascular conditions or may be strongly shared in common pathophysiology between bone and vascular metabolism, but these notions should be further investigated.
In contrast, the predictive biomarkers for BMD in the distal forearm remain largely unknown. This study demonstrated that RA patients with higher TRACP-5b tended to lose BMD in the distal forearm in the 2-year period. It is well known that TRACP-5b is predominantly expressed in bone by osteoclasts [38]. Janckila et al. reported that the mean level of TRACP-5b protein was elevated in RA patients compared with healthy controls and other disease groups [39]. They suggested that TRACP-5b activity is a marker of osteoclast number and local or systemic bone destruction, which suggests the hypothesis that osteoclast activity induced by local and/or systemic inflammation might strongly influence bone metabolism, particularly in the distal forearm of RA patients.
One of the noteworthy results of this study is that the biomarkers identified were more potent than other known predictors of BMD changes such as age plus menopause, DM, and the use of steroids, possibly because this study was conducted relatively in a short term, and because the participants were predominantly women with RA. Nonetheless, in the current clinical setting where patients are treated using a T2T strategy, it can be argued that the biomarkers would be a powerful tool to predict the changes in BMD of patients with RA.
One question that remains is how differently each osteoporosis drug affects BMD at each anatomical site. Few studies have investigated the differences between the effects of different osteoporosis drugs on different bones. This study revealed that any drugs did not sufficiently affect the change of BMD in the lumber spine, whereas both BP and VitD affected that in the proximal femur and in the distal forearm. Moreover, the potency of BP may differ among the three sites because t value of multiple regression analysis in the reduced model was higher in the proximal femur (t = 3.19) than that in the distal forearm (t = 2.06). Golub et al. previously demonstrated that skeletal biodistribution of bisphosphonate is anatomic site-dependent in a rat model [40]. While the earlier studies for primary osteoporosis show that BP increases BMD in lumbar spine but does not significantly increase BMD in forearm [41], another previous report demonstrated that bone loss and bone turnover at the distal radius were significantly faster in RA patients than the general population [42]. In addition, the hip and the distal forearm are distinguished from the lumbar spine with more cortical bones than the lumbar spine and are mechanically related to the joints: the bone turnover of the two sites may be enhanced by periarticular osteoporosis compared with lumbar spine [43, 44]. Therefore, the difference of drug effect might reflect differences of drug distribution, the ratios of the cortical and the cancellous bone, and mechanical burdens at each anatomical site. Additional investigation would identify differences in the predictive biomarkers for each anatomical site and may reveal differences in the effects of different osteoporosis drugs.
This study has several limitations. First, the number of participants may not be large enough to identify more significant biomarkers for each anatomical site. Indeed, this study did not include all of biomarkers available due to practical reasons, and studies with more samples and biomarkers may reveal a different set of biomarkers for these anatomical sites. However, the biomarkers distinguished in this study remained significant in each analysis and should be considered reliable for prediction of BMD changes. Second, as reported elsewhere, a decrease in BMD may not necessarily indicate actual bone fragility. However, there is consensus that BMD determined by DEA is one of the most reliable and usable surrogate markers for assessing the risk of fracture in osteoporosis patients. Third, this was a single-center study, which could have led to some selection bias. Fourth, the follow-up period of this study was relatively short. However, the longer the patients are followed, the more confounding factors, such as changes in medication, must be taken into count, which in turn requires greater sample numbers and more complex statistical analyses. Fifth, this study did not include newer osteoporosis drugs such as parathyroid hormone, selective estrogen receptor modulator, and anti-receptor activator of nuclear factor-kappaB antibody for analyses. The influence of these drugs on selecting biomarkers remains to be investigated. Lastly, although we investigated six bone metabolism biomarkers plus general and RA-related biomarkers such as ACPA, other reported biomarkers may be more potent than those assessed in the current study. However, we selected established biomarkers that were representative of different aspects of bone metabolism, inflammation, and serological aspects of RA, which would have covered crucial aspects of pathophysiology of osteoporosis and RA. Which biomarkers are best in a particular clinical setting should be investigated in another study.
Despite the above limitations, no previous longitudinal study has evaluated the relationship of six types of bone metabolism markers and RA-specific parameters with changes in BMD at three body sites in a large RA cohort. In addition, we applied well-known influential factors such as menopause, DM, BMI, steroids, osteoporotic drugs, and bDMARDs as explanatory parameters in the full multiple regression model, which should lead to reliable results. Indeed, the results of the two regression models extracted the similar significant markers in each analysis. Therefore, our results suggest that the biomarkers identified in this study should be considered useful for BMD management in patients with RA treated by the current T2T strategy and osteoporosis drugs.