The Impact of Seropositivity on Bone Mineral Density in Rheumatoid Arthritis - A 3-year Prospective Observational Study

To explore the impact of seropositivity on systemic bone loss in rheumatoid arthritis (RA). presence (ACPA). impact number of antibodies on BMD sub-divided according the antibodies present The changes in BMD were for each at and 3 signicant predictor of △ BMD ≧ –5% at TH (odds ratio 1.85, 95% condence interval 1.03-3.33, p = 0.039). information such as the age at diagnosis, disease duration, disease activity measured by erythrocyte (ESR), protein, and the disease activity score-28 joint-ESR (DAS28-ESR), medications including glucocorticoids (GC) and biological and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs), and presence/absence of ACPA and RF. Lifestyle, evidence of previous fragility fracture (history or radiographic), and risk factors of fragility fracture on the Fracture Risk Assessment Tool (FRAX®) were also recorded. The 10-year probabilities of major and hip fractures in each patient, calculated using the FRAX® with BMD (Taiwan version), were collected. Seropositivity was dened as any value >15 IU/mL for RF and >7 IU/mL for ACPA. The BMD of each patient was measured at enrolment and after the 3-year follow-up using a dual-energy X-ray absorptiometry scanner (Delphi A; Hologic Corp., Waltham, MA, USA) for femoral neck (FN), total hip (TH), and 1st - 4th lumbar vertebrae (L1–4). For postmenopausal women and men aged 50 years and older, osteoporosis was dened as a T-score of –2.5, or less at the FN based on the normal reference database for young white females [15, 16]. We calculated the percentage change in BMD (% △ BMD) for each participant as follows: [(second BMD – baseline BMD) / baseline BMD] × 100 between assessments. In this study, we de ﬁ ned that △ BMD ≧ –5% at three sites as excessive bone loss, based upon the previous studies regarding smallest detectable difference of BMD measurement [17, 18]. We dened a new incident fracture as any symptomatic non-traumatic fracture, including the forearm, hip, pelvic, and humerus fractures or asymptomatic morphometric vertebral fracture. A morphometric fracture was assessed on lateral radiographs of the lumbar spine according to Genant’s semiquantitative ordinal logistic regression analysis performed. After adjusting for the confounding factors (comorbidity, baseline DAS28, osteoporosis, FRAX score (hip), baseline BMD of FN, baseline BMD of TH, GC use), treatment with b/tsDMARDs (OR 1.96, 95% CI 1.08-3.57, p = 0.028) and △ BMD% at TH (OR 0.97, 95% CI 0.93-0.99, p = 0.027) were found to be related to a greater number of autoantibodies. and RF on systemic bone loss in RA, which has not been investigated before. Because of the 3-year observation period, we were able to not only investigate the link between the presence of autoantibodies and BMD at baseline, but also to determine the impact of these autoantibodies on BMD changes over time. This observational study has several limitations. The monocentric observational design used to diagnose RA patients allows us to draw a conclusion from our data about associations, but not about causal relationships. Our patients did not represent an inception cohort; we presented the data at the time of study initiation, not at the time of RA diagnosis. Some of the measurements, including 25(OH) vitamin D and parathyroid hormone, were performed at baseline only (data not shown); serial measurements might be more suitable for exploring the associations. In addition, we did not analyze whether or not seropositivity is an independent factor of systemic bone loss in RA because of the small sample size and short observation period. Finally, we did not check bone markers throughout the study, which hinders the understanding of the pathogenesis of antibodies on systemic bone loss in our cohort. should to and double antibody-positive further investigations explore association between and future


Introduction
Rheumatoid arthritis (RA) is a chronic systemic disease that can lead to local bone erosion and generalized osteoporosis. Rheumatoid factor (RF) and anti-citrullinated protein antibodies (ACPA) are the two most notable autoantibody systems commonly used as aids for diagnosing or classifying RA as well as for providing a variety of clinical and pathophysiological information [1]. Patients positive for ACPA and/or RF may be labeled together as "seropositive" RA (SPRA) and compose approximately 50% to 80 % of the RA population [1]. Several studies have indicated that SPRA has a greater severity in terms of disease activity and function impairment, as well as increased mortality over time [2,3] and poor response to treatment compared to seronegative RA (SNRA) [4]. Although both SPRA and SNRA ful ll the 1987 American College of Rheumatology (ACR) revised classi cation criteria [5] or the 2010 ACR/European League Against Rheumatism (EULAR) classi cation criteria [6], it seems that their clinical manifestations, courses, and prognostic features are distinctive.
It has been reported that the annual rate of bone loss in patients with active RA ranges between 5.5% and 10% [7]. Meanwhile, RF and/or ACPA are associated with juxta-articular osteoporosis, erosions, and generalized bone loss [8,9]. Several studies have shown that ACPAs and RF synergize to promote RA-associated in ammation, disease activity, and clinical onset [10,11] and can be associated with bone erosion and structural damage of RA [12]. The combination of ACPA and RF could predict the therapeutic responses to rituximab and abatacept [13].
However, the clinical predictive value of each of the antibodies (singly or in combination) on systemic bone loss has not been well studied.
In this study, we aimed to explore long-term bone mineral density (BMD) changes in patients with SPRA and SNRA, and investigate the association between the number of antibodies and systemic bone loss in patients with RA.

Study population and design
The inclusion criteria for participants and the methods used in this study were the same as those reported previously [14]. Study participants enrolled in a RA registry at Kaohsiung Chang Gung Memorial Hospital starting from 1 September 2014. A total of 651 patients who satis ed the 1987 revised ACR criteria for RA [5] or the 2010 ACR/EULAR classi cation criteria for RA [6], were enrolled in this study. Relevant information was recorded, including demographic characteristics such as age, sex, comorbidities, and body mass index, as well as disease-specific information such as the age at diagnosis, disease duration, disease activity measured by erythrocyte sedimentation rate (ESR), C-reactive protein, and the disease activity score-28 joint-ESR (DAS28-ESR), medications including glucocorticoids (GC) and biological and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs), and presence/absence of ACPA and RF. Lifestyle, evidence of previous fragility fracture (history or radiographic), and risk factors of fragility fracture on the Fracture Risk Assessment Tool (FRAX®) were also recorded. The 10-year probabilities of major and hip fractures in each patient, calculated using the FRAX® with BMD (Taiwan version), were collected. Seropositivity was de ned as any value >15 IU/mL for RF and >7 IU/mL for ACPA.
The BMD of each patient was measured at enrolment and after the 3-year follow-up using a dual-energy X-ray absorptiometry scanner (Delphi A; Hologic Corp., Waltham, MA, USA) for femoral neck (FN), total hip (TH), and 1st -4th lumbar vertebrae (L1-4). For postmenopausal women and men aged 50 years and older, osteoporosis was de ned as a T-score of -2.5, or less at the FN based on the normal reference database for young white females [15,16]. We calculated the percentage change in BMD (%△BMD) for each participant as follows: [(second BMDbaseline BMD) / baseline BMD] × 100 between assessments. In this study, we defined that △BMD≧-5% at three sites as excessive bone loss, based upon the previous studies regarding smallest detectable difference of BMD measurement [17,18]. We de ned a new incident fracture as any symptomatic non-traumatic fracture, including the forearm, hip, pelvic, and humerus fractures or asymptomatic morphometric vertebral fracture. A morphometric fracture was assessed on lateral radiographs of the lumbar spine according to Genant's semiquantitative assessment of vertebral fractures [19]. An independent radiologist assessed the evidence of morphometric vertebral compression fractures at enrolment and subsequently on an as-needed basis, with follow-up spinal radiographs during the 3-year observation period and at the end of the study.
The local Institutional Review Board of Chang Gung Memorial Hospital approved the study (104-3530B, 201901054B0), which was performed according to the principles of the Declaration of Helsinki. Informed consent was obtained from all participants.
Patients who were positive for RF or ACPA were grouped into the SPRA group, while those who were negative were grouped as SNRA. The participants were sub-grouped into A and B after matching for age and sex. Finally, the matched group was sub-grouped according to the number of antibodies present (0, group I; 1, group II; 2, group III).

Statistics
The data were checked for normality, which showed that the baseline characteristics had a skewed distribution; therefore, they were analyzed using non-parametric methods. Independent two-sample t-test and one-way analysis of variance (ANOVA) test were used to compare continuous variables with normal distribution and expressed as mean ± standard deviation (SD). The Mann-Whitney U test and Kruskal-Wallis test were used to compare continuous variables with skewed distribution; they were expressed as median (interquartile range, IQR). The chisquare or Fisher's exact test was used to assess the association between categorical variables. The intra-and inter-group BMD changes from enrollment to 3 years later were compared with a linear mixed model. Trend analyses were performed using ANOVA for each category of number of antibodies. A binary logistic regression analysis (enter model) was performed using excessive bone loss as the dependent variable. Ordinal regression was performed to compare baseline characteristics between patients harbouring 0-2 autoantibodies and to yield odds ratio (OR) and 95% con dence interval (CI), which signify the increase in association with every additional autoantibody. Data that resulted in p< 0.10 in univariate analysis were studied through logistic regression to access their independence. All statistical analyses were performed using IBM SPSS version 22 software (IBM Co., Armonk, NY, USA). A p-value of < 0.05, was considered signi cant in all analyses.

Results
Demographics and clinical characteristics of SPRA and SNRA participants The participants' disposition is shown in Fig. 1. A total of 651 participants were registered for the study cohort, which would last until the end of April 2020, and a total of 477 participants completed the 3-year observation period. To compare BMD changes in SNRA and SPRA, taking the mutual interference of age and sex into account, we controlled for these two factors. We obtained 312 matched participants, of whom 104 were allocated to group A and 208 to group B (Fig. 1). The baseline DAS28-ESR and ESR were signi cantly higher in group B than in group A (p = 0.002 and p = 0.001, respectively) ( Table 1). The proportion of patients using b/tsDMARDs in group B was higher than that in group A (p = 0.007). No signi cant difference was found between the two groups in terms of history of previous fractures, use of anti-osteoporotic therapy, and baseline GC (Table 1). Table 1 clinical characteristics of participants before and after matching The demographics and clinical characteristics of the participants are presented in Table 2 In pairwise comparisons, group III had signi cantly higher levels of DAS28 (p = 0.005) and ESR (p <0.001), higher rate of osteoporosis (p = 0.011) and 10-year probability of hip fracture (p =0.043), lower BMD at FN (p = 0.007), higher proportion of GC (p = 0.002) and b/tsDMARDs (p = 0.017) than group I; they also had signi cantly higher rates of comorbidity than group II (p = 0.01). Table 2 Clinical characteristics of participants grouped by presence of antibodies

Changes of BMD in groups A and B
Group B had a signi cantly higher proportion of patients with osteoporosis (p = 0.003) and a higher 10-year probability of hip fracture (p = 0.018), but had lower BMD at FN and TH (p = 0.004, p = 0.005) at enrollment, compared to group A (Table 1, right column). Three years later, the BMD at FN, TH, and L1-4 was signi cantly decreased from baseline in group B (p <0.001, p = 0.001, p = 0.006), respectively (Fig. 2a); a signi cant decrease was found only in FN in group A (p = 0.002) (Fig. 2a). Intergroup differences over 3 years were signi cant at FN and TH (p = 0.003 and p = 0.005, respectively), but were insigni cant at L1-4 (p = 0.105) (Fig. 2a). In terms of △BMD% after 3 years, group B demonstrated signi cant bone loss at TH compared to group A (p = 0.027) (Fig. 3a), while there was no obvious △BMD% between groups A and B at FN and L1-4. The difference in the rate of new incident fractures was not obvious between groups A and B over time (  . 2b). Intergroup BMD differences over 3 years were signi cant at FN and TH (p = 0.014, p = 0.020), but not at L1-4 (p = 0.250) (Fig. 2b). Among the three subgroups (groups I-III), there was a signi cant negative antibody numberdependent effect on the △BMD% at TH, 3 years later (p for trend 0.021, Fig. 3b).
In uential factors related to excessive bone loss at three sites In univariate logistic analysis, baseline DAS28 (p = 0.016), mean DAS28 (p = 0.028), ESR (p = 0.011), and BMD at TH (p = 0.017) were significantly correlated with excessive bone loss at TH (Table 3). In addition, the presence of seropositivity also showed a signi cant correlation with excessive bone loss at TH (p = 0.037). In multivariate logistic analysis, along with BMD at TH, the presence of seropositivity increased the potential of excessive bone loss at TH (OR 1.85, 95% CI 1.03-3.33, p = 0.039) ( Table 3). However, the seropositivity was not signi cantly associated with excessive bone loss at FN or spine (data not shown). Table 3 Predictive values for excessive bone loss (△BMD≧-5%) at the total hip In uential factors associated with the number of autoantibodies In an attempt to identify the in uential factors associated with a greater number of autoantibodies, ordinal logistic regression analysis was performed. After adjusting for the confounding factors (comorbidity, baseline DAS28, osteoporosis, FRAX score (hip), baseline BMD of FN, baseline BMD of TH, GC use), treatment with b/tsDMARDs (OR 1.96, 95% CI 1.08-3.57, p = 0.028) and △BMD% at TH (OR 0.97, 95% CI 0.93-0.99, p = 0.027) were found to be related to a greater number of autoantibodies.

Discussion
Our study provides 3-year follow-up results on the impact of seropositivity and the number of autoantibodies on the changes in BMD in patients with RA. We noticed that SPRA patients experienced a signi cant decrease in BMD at the FN, TH, and L1-4 compared with baseline, but not in SNRA at the TH and L1-4. SPRA patients demonstrated clear bone loss at the TH over time compared to SNRA patients, as indicated by the △BMD%. Furthermore, we recognized that the presence of antibodies had a signi cant predictive value for △BMD≧-5% at TH, along with baseline BMD at TH. Finally, we found that patients with higher numbers of antibodies had more systemic bone loss at TH.
It has been well documented that RA patients have a higher risk of osteoporosis and fragility fractures than the general population [18]. RAassociated bone loss is not only related to traditional risk factors of osteoporosis (e.g., aging, female sex) but also to factors related to the disease itself (e.g., disease activity and duration, GC use, and functional disability) [20,21]. However, it is still controversial whether seropositivity affects osteoporosis and BMD changes in RA. Some studies have reported that SNRA patients have a greater chance of developing osteoporosis and lower BMD than SPRA patients [22,23], but other investigations revealed opposite results [18, [24][25][26]. These investigations were subjected to a cross-sectional study, which had a small sample size and confounding factors were not adjusted for. In terms of the in uential factors screened, we ran binary regression analysis and observed that seropositivity was independently associated with excessive bone loss at TH during the follow-up period. In this investigation, we explored the BMD changes in SPRA and SNRA patients in a 3year, longitudinal, real-world observation study with adequate sample size; we adjusted the analysis, factoring in the confounding factors. The results of our study suggest that seropositivity has a detrimental effect on systemic bone loss in RA patients.
It seems that the greater systemic bone loss observed in SPRA patients is multifactorial. Recent investigations have suggested that not only do SPRA patients have distinct genomic backgrounds [27], clinical presentations [28] and treatment responses [29] from SNRA patients, they also have higher mortality [3]. It raises the possibility that SPRA and SNRA are two distinctive disease entities that mediate the different patterns of systemic bone loss seen in these two subtypes of RA [18]. Current investigations revealed that RA-related factors and its treatment, including baseline DAS28-ESR, ESR, and rate of GC and biologics use, were signi cantly higher in SPRA than SNRA, which further suggests that these subtypes of RA are different disease entities. The disease activity of RA [20,30] and GC use [21,31] are two of the determinants for RA-related systemic bone loss, suggesting that higher disease activity and more GC usage are the important determinants of greater systemic bone loss and a higher proportion of osteoporosis in SPRA than SNRA in our cohort.
In addition to the indirect effect of different disease patterns on bone loss between SPRA and SNRA, the direct effect of seropositivity on bone loss has been demonstrated. Harre et al. revealed an association between ACPA and serum markers for osteoclast-mediated bone resorption in RA patients [32]. Later on, Krishnamurthy et al. demonstrated that polyclonal ACPAs enhance osteoclast differentiation through a PADdependent IL-8-mediated autocrine loop [33]. A recent study also illustrated that ACPA-positive RA has a higher expression of CXCL2 (which can mediate the osteoclastogenesis in vitro) than ACPA-negative RA [34]. Monocyte lineage challenge with ACPA or RF autoantibodies can induce production of the cytokines TNF-α, IL-1β, IL-6, and IL-8 which are directly able to enhance osteoclast formation [35]. All of the evidence suggests that ACPA or RF may induce osteoclastogenesis directly, which might partly mediate the systemic bone loss in RA.
In this study, we also demonstrated that the concurrence of ACPA and RF was linked to more systemic bone loss over time. The additive effect of ACPA and RF on the production of the pro-in ammatory cytokine TNF-α, which is among the most potent cytokines to stimulate osteoclastogenesis, has been noted [10,36,37]. This may explain why there is a 'dose-dependent' effect of antibody number on systemic bone loss (Fig. 2b), and the concomitant presence of ACPA and RF in RA patients had the most detrimental effect on BMD in the three groups. To afford possible predictors of a greater number of autoantibodies, we carried out ordinal regression analysis and found that increased use of b/tsDMARD and more negative △BMD% at TH were considered expected factors of the concurrence of RF and anti-ACPA.
Interestingly, we found that not only is substantial bone loss consistent compared with baseline at TH either in SPRA or RA patients with more autoantibodies (Fig. 2a, Fig. 2b), △BMD% is also more negative (Fig. 3a, Fig. 3b) over time. This is consistent with the ndings of previous studies suggesting that RA patients had more bone loss occurred at the hip rather than at the spine [38 -40]. Additionally, it is well known that RA predominantly affects the small joints, with early bony erosion in hand joints. Orsolini et al. demonstrated that the independent role of ACPA has a negative titer-dependent effect on baseline BMD at femoral sites, mainly constituted by cortical bone, which is the most interesting in erosion pathogenesis [41]. However, why SPRA patients, especially double seropositive patients, had greater bone loss over time at the hip than the vertebrae remains unknown. We hypothesize, based on the previous studies [38 -42] and the ndings of this one, that the osteoclastogenesis effect of ACPA and/or RF on cortical and cancellous bones is different. However, its pathogenesis remains to be explored.
The FRAX® was launched in 2008 to allow health providers to estimate individual 10-year probabilities of fragility fractures [43], which is a free, reliable, and validated tool that is used globally. Our previous study demonstrated that RA patients with positive ACPA had a higher 10-year probability of major or hip fractures independent of ACPA levels [14]. Additionally, this study found that there was a signi cantly higher FRAX score (hip) in SPRA participants regardless of the presence of a single or double-positive antibody (Tables 1 and 2). Despite the higher FRAX score and greater bone loss in SPRA patients, no signi cant difference was observed in baseline previous fracture and new incident fractures over time between SPRA and SNRA, as well as in groups I-III. This suggests that more RA patients and a longer period of observation are needed to determine whether seropositivity in RA also has a detrimental effect on incident fracture.
The strengths of our study are as follows. As a real-world investigation, we documented as much as possible, the reported clinical variables that may be associated with osteoporosis/fracture in RA patients in order to avoid the possible factors related to BMD changes that were not observed in previous investigations. Aging and sex are two of the most important determinants of BMD. We investigated the impact of ACPA and RF on BMD changes in RA patients by matching age and sex to examine differences between SPRA and SNRA that had not been performed in previous studies. We also explored the additive effect of ACPA and RF on systemic bone loss in RA, which has not been investigated before. Because of the 3-year observation period, we were able to not only investigate the link between the presence of autoantibodies and BMD at baseline, but also to determine the impact of these autoantibodies on BMD changes over time.
This observational study has several limitations. The monocentric observational design used to diagnose RA patients allows us to draw a conclusion from our data about associations, but not about causal relationships. Our patients did not represent an inception cohort; we presented the data at the time of study initiation, not at the time of RA diagnosis. Some of the measurements, including 25(OH) vitamin D and parathyroid hormone, were performed at baseline only (data not shown); serial measurements might be more suitable for exploring the associations. In addition, we did not analyze whether or not seropositivity is an independent factor of systemic bone loss in RA because of the small sample size and short observation period. Finally, we did not check bone markers throughout the study, which hinders the understanding of the pathogenesis of antibodies on systemic bone loss in our cohort.

Conclusion
Compared to SNRA patients, SPRA patients had a higher prevalence of osteoporosis, consistent bone loss at all sites, and a higher 10-year probability of hip fracture. We also found that the presence of more autoantibodies in RA was associated with more detrimental effects on BMD. We therefore, suggest that more attention should be paid to osteoporosis and systemic bone loss in SPRA patients, especially double antibody-positive patients. As for fragility fractures, further investigations are needed to explore the association between seropositivity in RA and future fractures in the long term.