Bone metabolism and inflammation drive structural damage in rheumatoid arthritis: clustering and multivariable analysis

Abstract

Background:

Rheumatoid arthritis (RA) is characterized by synovial inflammation leading to joint damage, periarticular bone loss, and systemic osteoporosis. While inflammation is a primary driver of structural damage, dysregulation of the Wnt signaling pathway, particularly through inhibitors such as Dickkopf-1 (Dkk1) and sclerostin, has been implicated in RA-associated bone loss.

Objectives:

Our study investigated factors associated with erosive RA, focusing on bone turnover markers and modulators of the Wnt system.

Design:

We performed a cross-sectional study of stable conventional synthetic disease-modifying anti-rheumatic drug (csDMARDs) in RA patients naïve to biologic DMARDs.

Methods:

Clinical, radiographic, and bone mineral density (BMD) data were collected. Serum markers of bone turnover, including Dkk1, sclerostin, C-terminal telopeptide of type I collagen (CTX), procollagen type 1 N-terminal propeptide (P1NP), parathyroid hormone, and vitamin D, were analyzed. Principal component analysis (PCA) and k-means clustering were applied to identify variable associations, and regression models were used to assess their cross-sectional associations with radiographic damage.

Results:

Sixty-two RA patients were included in the study. The Sharp van der Heijde score (SvdHS) was positively correlated with measures of disease activity, glucocorticoid use, anti-citrullinated protein antibodies (ACPA) titer, rheumatoid factor, C-reactive protein, Dkk1 levels, and CTX. P1NP was inversely associated with SvdHS. PCA identified three clusters related to disease activity measures, BMD, and markers of bone metabolism. Dkk1 was linked to ACPAs and osteoclastic activity, suggesting a role in bone loss.

Conclusion:

Our findings confirm the role of inflammation and autoantibodies in RA-related joint damage. We found that BMD and markers of bone metabolism were additional contributors. There is a complex interplay between inflammation, bone metabolism, and structural deterioration in RA.

Keywords

bone metabolism inflammation bone mineral density bone turnover markers erosions rheumatoid arthritis

Introduction

Rheumatoid arthritis (RA) is characterized by prominent synovial inflammation, leading to articular damage and functional impairment.¹ Joint damage and periarticular bone loss are more severe in patients with seropositive RA, defined as positivity for serum Rheumatoid Factor (RF) and or anti-citrullinated protein antibodies (ACPAs), especially in patients positive for the latter.² RA is also associated with systemic bone loss and fractures.³ In this context, glucocorticoids may play a central role, but other metabolic features also significantly contribute to bone damage.^4,5

Bone remodeling is controlled by several signaling pathways, with the Wnt signaling pathway being among the master regulators. As a consequence, disruption of this pathway has been linked to bone loss in RA.^6,7 Specifically, in the RA-affected synovium, synovial fibroblasts and activated T cells are the primary sources of Receptor Activator Of Nuclear Factor Kappa-B Ligand (RANKL), a key driver of osteoclastogenesis, whose expression is finely regulated by the Wnt system.⁸ In addition, sclerostin and Dickkopf-1 (Dkk1), inhibitors of the Wnt pathway, are dysregulated in RA patients and are associated with erosive disease and systemic bone loss.^9–11

However, the specific mechanism leading to systemic and bone loss remains unclear. For example, periarticular bone loss has been observed in the preclinical phase of RA in ACPA+ individuals.¹² In addition, subclinical synovitis, which is detectable only via ultrasonography, has been associated with a greater risk of bone erosion.¹³

The objective of the present study was to investigate the factors associated with erosive RA, with a special focus on markers of bone turnover and modulators of the Wnt system.

Design

We conducted a cross-sectional analysis of seropositive RA patients with inadequate response to first-level disease-modifying antirheumatic drugs taken continuously for at least 3 months, including methotrexate, leflunomide, sulfasalazine, and hydroxychloroquine. In accordance with standard clinical practice, these patients were subsequently considered candidates for a second-line treatment with bDMARD. This is because this population has active disease with an ongoing risk of structural damage.

“Non-response” was defined according to routine clinical practice as persistent moderate or high disease activity (DAS28-CRP >3.2) despite ⩾3 months of continuous csDMARD therapy, consistent with EULAR recommendations.¹⁴

Methods

Anamnestic, clinical, radiographic (both hand and foot X-ray), and laboratory parameters were collected. Participants were assessed at baseline for bone mineral density (BMD) at both the femoral neck and lumbar spine areas with dual energy X-ray absorptiometry (DXA), via the QDR Hologic Delphi system (Hologic Inc., Bedford, MA, USA). Measurements followed standard DXA procedures: for the lumbar region, scans were acquired from L1 to L4 with the patient in the supine position and legs supported to reduce lumbar lordosis, and for the femoral neck, the non-dominant hip (preferably) was scanned with the limb internally rotated approximately 15–20° to optimize visualization. The coefficient of variation was set at 1% for the vertebral site and 1.2% for the femoral neck. Standard X-rays of the hands and feet were taken, and the Sharp van der Heijde score (SvdHS) was calculated by two independent readers (G.A., F.P.), both blinded to clinical and laboratory data with intraclass correlation coefficient (ICC, two-way mixed model, absolute agreement) for total SvdHS of 0.92 (95% CI: 0.87–0.95), indicating good agreement; discrepancies were resolved by consensus; this is a validated tool to assess the structural joint damage in RA. It evaluates erosions (scored 0–5 per joint) and joint space narrowing (scored 0–4 per joint) in specific joints of the hands, wrists, and feet, for a total possible score of 0–448. Serum samples were collected, fasting, in the morning and at baseline. The serum samples were aliquoted and stored at −80°C until they were assayed in a single batch. The analysis included the following biomarkers: RANKL, C-Terminal Telopeptide Of Type I Collagen (CTX, indicative of bone resorption), Bone Alkaline Phosphatase (BAP, a marker for bone formation), Procollagen I Intact N-Terminal Peptide (P1NP, another marker for bone formation), Dkk1 (a Wnt signaling inhibitor), Sclerostin (another Wnt inhibitor), Osteoprotegerin (OPG), 25-OH-vitamin D (VitD), and Parathyroid Hormone (PTH). CTX, BAP, and P1NP were measured using the IDS-ISYS Multi-Discipline Automated Analyzer, utilizing chemiluminescence technology. The intra-assay variation was 3.0% for P1NP, 4.0% for BAP, and 2.0% for CTX. Dkk1 and sclerostin levels were determined via ELISA kits with sensitivities of 0.89 and 8.9 pmol/L, respectively, and intra-assay variation coefficients of 7.8% for Dkk1 and 5.6% for sclerostin. Their inter-assay variations were 8.2% for Dkk1 and 6.9% for sclerostin. PTH was measured using ELISA with an intra-assay variability of 6% and an inter-assay variability of 7%. 25OHVitD was measured using the LIAISON 25OHVitD assay, with an intra-assay variability of 8% and an inter-assay variability of 12%. RANKL and OPG were measured using ELISA kits, with intra-assay coefficients of variation of 8.1% and 6.3%. The inter-assay variabilities were 9.1% and 7.2% for RANKL and OPG, respectively. To minimize inter-assay variability, all samples were measured in a single batch.

We included patients aged 18 years or older who provided signed informed consent and had an RA diagnosis satisfying the ACR/EULAR 2010 criteria with stable csDMARD for at least 3 months and were naïve to bDMARDs. The exclusion criteria were patients with a rheumatic diagnosis other than RA; who were suffering from bone conditions other than osteoporosis (e.g., Paget’s disease of bone); or who were receiving treatments with bisphosphonates (with a grace period of up to 12 months for oral bisphosphonates and up to 24 months for zoledronic acid), strontium ranelate, teriparatide, selective estrogen receptor modulators, or denosumab. We also excluded patients with kidney disease (eGRF <30 mL/min), advanced liver disease (Child–Pugh grade B or C), untreated endocrine disorders, and those receiving intra-articular injections in the MCPs or MTPs within 3 months prior to the study, as well as people who were pregnant or breastfeeding. No specific selection criteria (other than inclusion/exclusion criteria) were applied. Patients were consecutively enrolled from May 2023 to December 2024.

The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.¹⁵ The completed STROBE checklist is provided as Supplemental File 1.

Statistical analysis

Descriptive statistics are presented as means and standard deviations (SDs) or medians and interquartile ranges (IQRs), as appropriate. Correlations between continuous variables were assessed through Spearman’s correlation coefficients. To account for multiple comparisons, the two-stage step-up method by Benjamini, Krieger, and Yekutieli was applied, with the false discovery rate controlled at 5% (Q = 0.05).

In this study, we employed principal component analysis (PCA) to reduce the dimensionality of the dataset and identify clusters of variables that conveyed overlapping or redundant information. In the obtained plot, variables with longer vectors have a stronger influence on a component; vectors forming a small angle indicate variables that are positively correlated; an angle of about 90° suggests little or no correlation, whereas large angles, close to 180°, indicate a negative correlation between the variables. Principal components (PCs) were selected on the basis of eigenvalues, with components explaining more than 75% of the total variance retained for further analysis. This threshold ensures that the majority of the data’s underlying structure is preserved while minimizing noise. Principal component regression (PCR) was subsequently applied to derive SvdHS, leveraging the PCs as input variables. This approach facilitated the development of a robust model by capturing the essential variance within the dataset and mitigating the influence of multicollinearity. PCA is a widely used dimensionality reduction technique that transforms a large set of variables into a smaller set of uncorrelated components, known as PCs, which retain most of the original data’s variance. This method is particularly valuable for identifying patterns in data, minimizing information loss, and mitigating issues related to collinearity among variables. PCR integrates PCA with regression analysis, addressing multicollinearity and improving the reliability of regression coefficients. PCR was used as a multivariable technique, with all retained PCs included as predictors with coefficients of PCs not directly interpretable in terms of the original variables. In summary, we applied PCA to reduce the dimensionality of the dataset and facilitate visualization, and then used PCR to assess the multivariable associations between the resulting components and SvdHS. By using the PCs derived from PCA as covariates in the regression model, PCR improves model stability and mitigates multicollinearity among highly correlated variables.

Following PCA, k-means clustering was applied to the same dataset to partition the data into distinct groups, allowing the identification of clusters with shared characteristics. The k-means algorithm iteratively assigns data points to k clusters by minimizing the within-cluster variance. The number of clusters (k) was determined through iterative optimization and visual inspection of the elbow plot, which graphically represented the sum of squared distances from points to their assigned cluster centroids. The optimal k was selected at the point where additional clusters yielded diminishing returns in variance reduction. Clusters, derived from k-means, were supported by PCA visualization, following the PC1 and PC2 scatter plots to explore how the data were distributed across the primary dimensions identified via PCA. All differences were considered significant when p value was less than 0.05.

Nevertheless, given the relatively small sample size, we assessed the adequacy of the correlation structure before interpreting PCA results. An exploratory factor analysis using principal axis factoring (unrotated) indicated biologically coherent groupings of variables, which closely mirrored the clusters identified by k-means. Bartlett’s test of sphericity was significant (χ² = 422, df = 253, p < 0.001), supporting the suitability of the correlation matrix for dimensionality reduction. Goodness-of-fit indices suggested acceptable model fit (χ² non-significant, RMSEA = 0.043), though the global KMO indicated limited sampling adequacy. While formal stability validation was not feasible, these analyses support the exploratory robustness of the observed patterns

All the statistical analyses were performed with SPSS Version 26 (SPSS, Inc., Chicago, IL, USA) and the PCA package in GraphPad Prism version 10.4.1 (GraphPad Software, San Diego, CA, USA).

Results

Sixty-two RA patients aged 57.2 years (SD 12.1) were consecutively enrolled in the study. The mean DAS28-CRP was 4.17 (SD 1.27), and the median SvdHS was 24 (IQR 12–53). Table 1 shows the characteristics of the population.

Table 1.

Characteristics of the study population.

Characteristics	n = 62
Sex, female (%)	52 (83.9)
Age, years (mean ± SD)	57.2 ± 12.1
Weight, kg (mean ± SD)	73.8 ± 14.7
Height, cm (mean ± SD)	165 ± 7.8
Seropositive either RF or ACPA, n (%)	62 (100%)
Disease duration, years (mean ± SD)	8.75 ± 7.91
Swollen joints, n (median ± IQR)	4 (8)
Tender joints, n (median ± IQR)	4 (6)
PGA (mean ± SD)	6.34 ± 2.02
PhGA (mean ± SD)	6.72 ± 1.31
CRP serum levels, mg/L (median ± IQR)	2 (7.5)
DAS28-CRP (mean ± SD)	4.17 ± 1.28
mSvdH score (median ± IQR)	24 (41)
Treatment with csDMARD, n (%)
Hydroxychloroquine	5 (8.1%)
Methotrexate	36 (58.1%)
Leflunomide	18 (29.0%)
Sulfasalazine	3 (4.8%)
Vitamin D supplements taking, n (%)	62 (100%)
GC taking, n (%)	34 (54.8%)
GC daily dose at enrollment, mg (median ± IQR)	2.5 (5)
GC cumulative dose prior to enrollment, mg (median ± IQR)	256 (3283)
ACPA titer, IU (median ± IQR)	1591 (1158)
RF titer, IU (mean ± SD)	94.1 ± 136
Smoking habit, n (%)	28 (45.2%)
No	34 (54.8%)
Yes	17 (27.4%)
Former	11 (17.7%)

ACPA, anti-citrullinated protein antibodies; CRP, C-reactive protein; csDMARD, conventional synthetic disease-modifying anti-rheumatic drug; DAS28-CRP, disease activity score in 28 joints using CRP; GC, glucocorticoids; IQR, interquartile range; mSvdH, modified Sharp/van der Heijde score; PGA, Patient Global Assessment; PhGA, Physician Global Assessment; RF, rheumatoid factor; SD, standard deviation.

Univariate analysis

We found a weak-moderate positive association between Dkk1 levels and the ACPA titer (ρ 0.31, p 0.03) and a weak positive association between Dkk1 and PTH (ρ 0.33, p 0.02). We found a negative association between RANKL and age (ρ −0.28, p 0.04).

We found a positive correlation between SvdHS and age (ρ 0.37, p 0.01), RF titer (ρ 0.31, p 0.03), BAP (ρ 0.29, p 0.04), CRP (ρ 0.37, p 0.01), and PTH (ρ 0.29, p 0.05).

The bone turnover markers were highly positively correlated with each other (P1nP and BAP ρ 0.61, p < 0.01; P1nP and CTX ρ 0.52, p < 0.01; CTX and BAP ρ 0.49, p < 0.01). Glucocorticoids daily dose was negatively correlated with markers of bone formation, BAP, and P1NP (ρ 0.42, p < 0.01 and ρ 0.31, p 0.01, respectively), but we did not find a significant correlation with CTX levels. Figure 1 shows the correlation matrix for the univariate analysis.

Figure 1.

Correlation matrix showing the correlations among univariate predictors of the Sharp van der Heijde score. Panel (a) shows the Spearman’s correlation coefficients. Panel (b) shows the p values below 0.05.

Principal component analysis

Figure 2 shows the loading plot of the PCA, representing the variables of interest: PC1 representing inflammatory and disease activity variables, PC2 is dominated by BMD measures, and PC3 captures bone turnover markers, particularly those involved in Wnt pathway regulation, in which specifically Dkk1 shows strong positive loadings alongside CTX and sclerostin. We then applied k-means clustering and identified three patient-clusters (Figure 3): the BMD cluster, the bone marker cluster, and the disease activity/treatment cluster.

Figure 2.

Loading plot of principal components: variable distribution across principal components. Vectors with a small angle indicate positive correlations between variables. A 90° angle suggests no correlation, whereas angles near 180° reflect a negative correlation.

Figure 3.

K-means cluster visualization on PC1 and PC2: distribution of data points across principal components.

Overall, PCA was used to visualize the main patterns among variables, while k-means clustering confirmed these patterns, supporting the robustness and interpretability of the observed groupings. Despite the limited sample size, the concordance between methods and the biologically coherent grouping of variables suggests the robustness of the observed structure.

Table 2 shows the results of the PCR associated with SvdHS. We found that age, GC treatment, ACPA titer, RF titer, CRP level, erythrocyte sedimentation rate (ESR), CTX serum level, and Dkk1 serum level were significantly positively correlated with SvdHS, whereas P1NP serum level and Patient Global Assessment score were negatively correlated with SvdHS.

Table 2.

Principal component regression results associated with the SvdHS.

Variable	Estimate	Standard error	95% CI (asymptotic)	p Value
Intercept	−13.85	26.92	−69.82 to 42.12	0.61
Age	0.55	0.18	0.17 to 0.93	0.01
Weight	0.00	0.13	−0.27 to 0.28	0.97
Disease duration	0.53	0.19	0.16 to 0.99	0.04
GC daily dose	2.05	0.69	0.62 to 3.49	<0.01
GC cumulative dose	0.02	0.00	0.00 to 0.03	0.04
TJ	−0.23	0.30	−0.84 to 0.39	0.46
SJ	−0.06	0.34	−0.77 to 0.65	0.85
PGA	−2.09	0.84	−3.84 to −0.34	0.02
PhGA	−0.45	1.04	−2.62 to 1.71	0.67
ACPA	0.01	0.00	0.00 to 0.01	0.03
RF	0.10	0.04	0.02 to 0.18	<0.01
CRP	0.40	0.15	0.07 to 0.72	0.02
ESR	0.44	0.13	0.17 to 0.71	<0.01
Hb	−0.14	0.17	−0.49 to 0.21	0.41
CTX	30.05	13.62	1.72 to 58.38	0.04
P1NP	−0.21	0.08	−0.38 to −0.04	0.02
Dkk1	0.33	0.14	0.03 to 0.62	0.03
SOST	0.18	0.28	−0.41 to 0.77	0.53
OPG	−0.39	0.72	−1.89 to 1.11	0.59
RANKL	−12.86	13.49	−40.92 to 15.20	0.35
PTH	−0.29	0.17	−0.64 to 0.06	0.10
VitD	0.13	0.20	−0.28 to 0.54	0.52
BMD LS Ts	1.16	1.69	−2.36 to 4.68	0.50
BMD Neck Ts	−0.90	1.58	−4.19 to 2.40	0.58
BMD Tot Ts	−0.73	1.49	−3.83 to 2.38	0.63

ACPA, anti-citrullinated protein antibodies; BMD, bone mineral density; CRP, C-reactive protein; CTX, C-terminal telopeptide of type I collagen; Dkk1, Dickkopf-1; ESR, erythrocyte sedimentation rate; GC, glucocorticoids; Hb, hemoglobin; LS, lumbar spine; OPG, osteoprotegerin; P1NP, procollagen type 1 N-terminal propeptide; PGA, Patient Global Assessment; PhGA, Physician Global Assessment; PTH, parathyroid hormone; RANKL, receptor activator of nuclear factor kappa-B ligand; RF, rheumatoid factor; SJ, swollen joints; SOST, sclerostin; SvdHS, Sharp van der Heijde score; TJ, tender joints; Tot, total hip; Ts, T-score; VitD, Vitamin D.

Bold values indicate statistically significant results (p < 0.05).

Discussion

Herein, we investigated the relationships among disease activity, bone metabolism, and structural damage in RA patients with inadequate response to first-level csDMARDs.

Overall, we found that SvdHS was associated with markers of disease activity, such as age, RF, ACPA titer, CRP, and ESR. These findings add to the literature, further confirming the central role of inflammation and autoantibodies in structural joint damage. Interestingly, and seemingly counterintuitively, GC treatment was positively associated with SvdHS. This finding might be easily explained by confounding by indications, with GC-treated patients being those with more severe disease. However, the positive association between GC dose and SvdHS was maintained in the PCR multivariate analysis, which controlled for many other covariates, possibly attenuating confounding bias. Indeed, GCs cause bone loss, which can create a fertile substrate for joint erosions.^16,17 Several studies indeed demonstrated a link between osteoporosis and more erosive RA.^18,19

Wnt signaling is a major regulator of bone and joint remodeling. Recently, Dkk1 levels were correlated with the levels of CTX, a marker of bone resorption, and with disease activity in RA patients.²⁰ In our study, Dkk1 was positively associated with SvdHS, again indicating its role in joint damage through bone resorption and suppressed osteoblast activity.¹⁰ In line with this observation, CTX, a bone resorption marker, was correlated with SvdHS, and P1NP, a bone formation marker, was negatively associated with SvdHS. Interestingly, we found that the levels of ACPAs were positively correlated with Dkk1. We and others in the literature previously reported that the ACPA titer was independently associated with bone loss and fracture risk in RA patients.^21–25 These findings suggest that Dkk1 might mediate the association between ACPAs and bone loss in RA. Interestingly, FcyR-expressing cells are a major source of Dkk1, which has been shown to directly regulate osteoclastic and osteoblastic activity in erosive RA.^26–30 However, our findings, given the observational nature of the study, cannot infer causality.

Clustering analysis of the study population revealed three distinct groups of variables that roughly corresponded to “disease activity,” “BMD measures,” and “bone metabolism/turnover markers.” Both PCA and k-means consistently identified these clusters, the first for the variables themselves and the latter for the patients, underscoring the robustness of the findings. The first cluster grouped variables associated with disease activity, including GC use (previously analyzed), tender and swollen joint counts, patient and physician global assessments. This result was in line with the vast majority of the literature on this topic.^31–34 Nonetheless, studies have shown that erosion progression is, at least to a certain extent, independent of disease activity.^35–37 Data from the RA BIODAM study revealed that a very strict treatment-to-target strategy did not provide additional benefit on radiographic progression over less strict strategies in RA patients.³⁷ Overall, in the RA BIODAM, approximately 15% of patients in a low disease activity state or even in remission experienced erosion progression, which is in line with other studies.^36–39 This residual erosion progression observed in patients with well-controlled disease activity might be partially explained by factors outside traditional measures of disease activity, such as those represented by the metabolic and BMD clusters. These clusters may reflect underlying skeletal fragility phenotypes, characterized by altered bone turnover, impaired microarchitectural integrity, and systemic metabolic dysregulation, that exert a silent but cumulative impact on structural damage. Consequently, structural progression in “controlled” RA may be a multifactorial phenomenon, in which subtle metabolic alterations, subclinical bone loss, and Wnt pathway dysregulation operate alongside, or even independently from, synovial inflammation, collectively eroding joint integrity despite apparent therapeutic success. In detail, the second cluster comprised the BMD at the femoral neck, total hip, and lumbar spine, along with RANKL, a critical mediator of bone resorption; this is a pathophysiologically coherent axis: as RANKL increases, osteoclastic activity intensifies and drives both local and systemic decreases in BMD.⁴⁰ This coupling strengthens the argument that bone loss in well-controlled disease may arise from intrinsic skeletal regulatory pathways, in addition to the inflammatory burden. Moreover, the BMD of RA patients was significantly lower than that of age-matched psoriatic arthritis patients, despite both conditions being inflammatory in nature.^20,41 This difference may, at least in part, reflect a bias related to the greater cumulative exposure to glucocorticoids typically observed in RA compared with PsA. However, other pathophysiological pathways are also likely to contribute to this disparity. For instance, PsA and RA exhibit similarities in bone destruction, but PsA is also defined by bone formation; thus, the level of DKK1 has been found to be lower in PsA, than in RA.^7,10,20 Indeed, the third cluster comprised markers of bone metabolism, such as sclerostin, Dkk1, PTH, CTX, and P1NP. A large observational study of more than 1000 RA patients demonstrated that the bone metabolism signature of RA patients largely differed from that of age-matched healthy controls.⁴¹ RA patients presented higher levels of Dkk1 and lower levels of P1NP and BAP, indicative of an imbalance favoring bone resorption over formation.⁴¹ In summary, these findings suggest that the dysregulation of bone metabolism in RA is driven by unique inflammatory and metabolic pathways, contributing to a greater risk of structural damage and fragility fractures. The clustering of variables in the present analysis reinforces this notion, providing a framework to better understand the complex interplay between inflammation, bone health, and disease activity in RA.

This study has several strengths. It employs a comprehensive approach by integrating advanced statistical techniques, such as PCA and k-means clustering, to identify and analyze distinct patterns in disease activity, bone density, and bone metabolism. The inclusion of robust measures, including detailed radiographic scoring, bone turnover markers, and DXA for BMD assessment, provides a multidimensional view of the factors contributing to structural damage in RA patients. The enrollment of bDMARD-naïve patients ensures a relatively homogeneous population, reducing the confounding effects of prior advanced treatments. However, the study has several limitations. Its cross-sectional design precludes causal inferences, making establishing temporal relationships between inflammation, bone turnover, and structural damage difficult; moreover, one-time measurements may not capture fluctuations in biomarkers or symptoms, and exposures of varying duration cannot be accurately evaluated, potentially leading to residual confounding. The relatively modest sample size may limit the generalizability of the findings to broader RA populations. Although confounding by indication was addressed through multivariate analysis, residual confounding cannot be entirely excluded. The exclusion of patients with prior or ongoing advanced osteoporosis treatments may underestimate the impact of long-term bone health alterations in RA patients. Importantly, while clustering analyses revealed distinct variable patterns, the functional and clinical implications of these clusters require further longitudinal validation to confirm their relevance and potential utility in guiding patient management strategies.

Conclusion

Our findings confirm the central role of inflammation and autoantibodies in joint damage and suggest that glucocorticoid use and altered bone turnover markers may contribute to erosion progression. The identification of distinct clusters related to disease activity, bone density, and bone metabolism confirms the complexity of RA pathology, suggesting that metabolic and bone-specific factors are potential contributors to structural damage beyond inflammation.

Supplemental Material

sj-doc-1-tab-10.1177_1759720X261433255 – Supplemental material for Bone metabolism and inflammation drive structural damage in rheumatoid arthritis: clustering and multivariable analysis

Supplemental material, sj-doc-1-tab-10.1177_1759720X261433255 for Bone metabolism and inflammation drive structural damage in rheumatoid arthritis: clustering and multivariable analysis by Francesco Pollastri, Ombretta Viapiana, Davide Gatti, Angelo Fassio, Camilla Benini, Carmela Dartizio, Isotta Galvagni, Valeria Messina, Maurizio Rossini and Giovanni Adami in Therapeutic Advances in Musculoskeletal Disease

Footnotes

Acknowledgements

None.

Declarations

ORCID iDs

Francesco Pollastri

Maurizio Rossini

Giovanni Adami

Supplemental material

Supplemental material for this article is available online.

References

Aletaha

Smolen

JS.

Diagnosis and management of rheumatoid arthritis: a review. JAMA 2018; 320: 1360.

Bøyesen

Hoff

Ødegård

, et al. Antibodies to cyclic citrullinated protein and erythrocyte sedimentation rate predict hand bone loss in patients with rheumatoid arthritis of short duration: a longitudinal study. Arthritis Res Ther 2009; 11: R103.

Adami

Saag

KG.

Osteoporosis pathophysiology, epidemiology, and screening in rheumatoid arthritis. Curr Rheumatol Rep 2019; 21: 34.

Adami

Regulation of bone mass in inflammatory diseases. Best Pract Res Clin Endocrinol Metab 2022; 36: 101611.

Adami

Saag

KG.

Glucocorticoid-induced osteoporosis update. Curr Opin Rheumatol 2019; 31: 388–393.

Adami

Fassio

Rossini

, et al. Osteoporosis in rheumatic diseases. Int J Mol Sci 2019; 20: 5867.

Fassio

Idolazzi

Viapiana

, et al. In psoriatic arthritis Dkk-1 and PTH are lower than in rheumatoid arthritis and healthy controls. Clin Rheumatol 2017; 36: 2377–2381.

Gupta

Kanwar

Arora

, et al. The interplay of rheumatoid arthritis and osteoporosis: exploring the pathogenesis and pharmacological approaches. Clin Rheumatol 2024; 43: 1421–1433.

Rossini

Viapiana

Adami

, et al. In patients with rheumatoid arthritis, Dickkopf-1 serum levels are correlated with parathyroid hormone, bone erosions and bone mineral density. Clin Exp Rheumatol 2015; 33(1): 77–83.

10.

Diarra

Stolina

Polzer

, et al. Dickkopf-1 is a master regulator of joint remodeling. Nat Med 2007; 13: 156–163.

11.

Zhang

Wang

, et al. The serum level of Dickkopf-1 in patients with rheumatoid arthritis: a systematic review and meta-analysis. Int Immunopharmacol 2018; 59: 227–232.

12.

Kleyer

Finzel

Rech

, et al. Bone loss before the clinical onset of rheumatoid arthritis in subjects with anticitrullinated protein antibodies. Ann Rheum Dis 2014; 73: 854–860.

13.

Brown

Conaghan

Karim

, et al. An explanation for the apparent dissociation between clinical remission and continued structural deterioration in rheumatoid arthritis. Arthritis Rheum 2008; 58: 2958–2967.

14.

Smolen

Landewé

RBM

Bergstra

, et al. EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2022 update. Ann Rheum Dis 2023; 82(1): 3–18.

15.

Von Elm

Altman

Egger

, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008; 61: 344–349.

16.

Adami

Saag

KG.

Glucocorticoid-induced osteoporosis: 2019 concise clinical review. Osteoporos Int 2019; 30: 1145–1156.

17.

Adami

Fassio

Rossini

, et al. Bone loss in inflammatory rheumatic musculoskeletal disease patients treated with low-dose glucocorticoids and prevention by anti-osteoporosis medications. Arthritis Rheumatol 2023; 75: 1762–1769.

18.

Solomon

Finkelstein

Shadick

, et al. The relationship between focal erosions and generalized osteoporosis in postmenopausal women with rheumatoid arthritis. Arthritis Rheum 2009; 60: 1624–1631.

19.

Rossini

Bagnato

Frediani

, et al. Relationship of focal erosions, bone mineral density, and parathyroid hormone in rheumatoid arthritis. J Rheumatol 2011; 38: 997–1002.

20.

Riitano

Spinelli

Manganelli

, et al. Wnt signaling as a translational target in rheumatoid and psoriatic arthritis. J Transl Med 2025; 23: 158.

21.

Cheng

T-T

S-F

F-M

, et al. Anti-CCP-positive patients with RA have a higher 10-year probability of fracture evaluated by FRAX^®: a registry study of RA with osteoporosis/fracture. Arthritis Res Ther 2018; 20: 16.

22.

Sghiri

Ben Hassine

Ghozzi

, et al. Autoantibodies serum level and 10-year risk of fractures evaluated by FRAX^® tool in rheumatoid arthritis patients. Open Access Rheumatol 2024; 16: 137–145.

23.

Orsolini

Caimmi

Viapiana

, et al. Titer-dependent effect of anti-citrullinated protein antibodies on systemic bone mass in rheumatoid arthritis patients. Calcif Tissue Int 2017; 101(1): 17–23.

24.

Bugatti

Bogliolo

Vitolo

, et al. Anti-citrullinated protein antibodies and high levels of rheumatoid factor are associated with systemic bone loss in patients with early untreated rheumatoid arthritis. Arthritis Res Ther 2016; 18: 226.

25.

Bruno

Fedele

Tolusso

, et al. Systemic bone density at disease onset is associated with joint erosion progression in early naive to treatment rheumatoid arthritis: a prospective 12-month follow-up open-label study. Front Med (Lausanne) 2021; 8: 613889.

26.

Lehmann

Thiele

Baschant

, et al. Mice lacking DKK1 in T cells exhibit high bone mass and are protected from estrogen-deficiency-induced bone loss. iScience 2021; 24: 102224.

27.

Juarez

McGettrick

Scheel-Toellner

, et al. DKK1 expression by synovial fibroblasts in very early rheumatoid arthritis associates with lymphocyte adhesion in an in vitro flow co-culture system. Arthritis Res Ther 2016; 18: 14.

28.

Komatsu

Takayanagi

Mechanisms of joint destruction in rheumatoid arthritis—immune cell-fibroblast-bone interactions. Nat Rev Rheumatol 2022; 18: 415–429.

29.

Phillips

NK cells induce a pro-inflammatory phenotype in RA synovial fibroblasts. Nat Rev Rheumatol 2021; 17: 645.

30.

Zhao

Grieshaber-Bouyer

Rao

, et al. Effect of JAK inhibition on the induction of proinflammatory HLA-DR+CD90+ rheumatoid arthritis synovial fibroblasts by interferon-γ. Arthritis Rheumatol 2022; 74: 441–452.

31.

Combe

Dougados

Goupille

, et al. Prognostic factors for radiographic damage in early rheumatoid arthritis: a multiparameter prospective study. Arthritis Rheum 2001; 44: 1736–1743.

32.

Paulus

Di Primeo

Sanda

, et al. Progression of radiographic joint erosion during low dose corticosteroid treatment of rheumatoid arthritis. J Rheumatol 2000; 27: 1632–1637.

33.

Boers

Aletaha

Mela

, et al. Glucocorticoid effect on radiographic progression in placebo arms of rheumatoid arthritis biologics trials. J Rheumatol 2016; 43: 1024–1026.

34.

Carpenter

Norton

Nikiphorou

, et al. Reductions in radiographic progression in early rheumatoid arthritis over twenty-five years: changing contribution from rheumatoid factor in two multicenter UK inception cohorts. Arthritis Care Res (Hoboken) 2017; 69: 1809–1817.

35.

Baker

Østergaard

George

, et al. Greater body mass independently predicts less radiographic progression on X-ray and MRI over 1–2 years. Ann Rheum Dis 2014; 73(11): 1923–1928.

36.

Adami

Fassio

Pistillo

, et al. Factors associated with radiographic progression in rheumatoid arthritis starting biological diseases modifying anti-rheumatic drugs (bDMARDs). Ther Adv Musculoskelet Dis 2023; 15: 1759720X231174534.

37.

Ramiro

Landewé

Heijde

D van der

, et al. Stricter treat-to-target in RA does not result in less radiographic progression: a longitudinal analysis in RA BIODAM. Rheumatology (Oxford) 2023; 62: 2989–2997.

38.

Versteeg

Steunebrink

Vonkeman

, et al. Early radiological progression remains associated with long-term joint damage in real-world rheumatoid arthritis patients treated to the target of remission. Scand J Rheumatol 2022; 51: 87–96.

39.

Syversen

Goll

Heijde

D van der

, et al. Prediction of radiographic progression in rheumatoid arthritis and the role of antibodies against mutated citrullinated vimentin: results from a 10-year prospective study. Ann Rheum Dis 2010; 69: 345–351.

40.

Wang

, et al. Osteoprotegerin and RANKL in the pathogenesis of rheumatoid arthritis-induced osteoporosis. Rheumatol Int 2012; 32: 3397–3403.

41.

Adami

Fassio

Rossini

, et al. Machine learning to characterize bone biomarkers profile in rheumatoid arthritis. Front Immunol 2023; 14: 1291727.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.08 MB

0.00 MB