Partial bone erosion repair is independently associated with functional improvement in early rheumatoid arthritis: a 2-year prospective high-resolution peripheral quantitative CT (HR-pQCT) longitudinal study

Abstract

Objective

To determine whether change in bone erosion volume detected by high-resolution peripheral quantitative CT (HR-pQCT) is independently associated with functional outcome in early rheumatoid arthritis (ERA).

Methods

In this prospective study, 101 patients with ERA received 2-year tight-control treatment aiming at Simplified Disease Activity Index (SDAI) remission (SDAI ≤3.3). HR-pQCT of the second to fourth metacarpophalangeal joints was performed at baseline and at month 24 to quantify erosion parameters. Partial erosion repair was defined as a reduction of at least one erosion count or a decrease in total erosion volume exceeding the least significant change; progression was defined conversely. Functional status was assessed using the Health Assessment Questionnaire-Disability Index (HAQ-DI).

Results

After 2 years, disease activity and HAQ-DI improved significantly (all p<0.001) in paired analyses (n=98). Sixteen (16.3%) patients showed partial erosion repair and 38 (38.8%) erosion progression. In the partial erosion repair group, reduction in total erosion volume correlated strongly with HAQ-DI improvement (ρ=0.719, p=0.003). This association remained significant in linear regression analyses (β=0.607, p=0.003) independent of the contribution from change in SDAI (β=0.453, p=0.018). By contrast, among erosion progressors, change in HAQ-DI was primarily driven by improvement in SDAI (β=0.755, p<0.001). Baseline swollen joint count was associated with erosion progression in univariate analysis (OR 1.177, 95% CI 1.026 to 1.349, p=0.020) and after adjustment for sex (OR 1.171, 95% CI 1.019 to 1.346, p=0.026).

Conclusion

HR-pQCT-detected partial repair of bone erosion was independently associated with functional improvement beyond inflammation control.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Functional impairment in rheumatoid arthritis (RA) is related to synovial inflammation and joint destruction, with their relationships varying by disease phase.

• High-resolution peripheral quantitative CT (HR-pQCT) enables reliable, three-dimensional assessment of bone erosions and has been increasingly applied in early rheumatoid arthritis (ERA).

• Bone erosion repair in RA is typically partial and can occur following effective suppression of inflammation, although its relationship with functional improvement remains unclear.

• Despite effective inflammation control, bone damage continues to progress in some patients with ERA.

WHAT THIS STUDY ADDS

• HR-pQCT-measured partial erosion repair over 2 years is independently associated with functional improvement in Health Assessment Questionnaire-Disability Index beyond inflammation control and treatment exposure in ERA.

• This study links HR-pQCT-detected structural repair to functional improvement, extending previous work that primarily described erosion repair as an imaging outcome.

• In ERA, higher baseline swollen joint count is associated with both existing and advancing bone erosion severity.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Help support the use of HR-pQCT in research settings to distinguish inflammation-driven and structure-driven contributors to functional outcome in ERA.

• HR-pQCT-derived measures of erosion repair may provide sensitive imaging endpoints to complement clinical assessment of functional improvement in early rheumatoid disease.

Introduction

Rheumatoid arthritis (RA) is a chronic inflammatory disease characterised by persistent synovitis, systemic inflammation and progressive joint destruction. Joint destruction, manifesting as bone erosion and cartilage loss, leads to irreversible disability and loss of physical function. The relationship between disease activity, joint destruction and functional capacity evolves over the course of RA.¹ In established RA, radiographic progression is associated with worse functional outcome, as reflected by increases in Health Assessment Questionnaire-Disability Index (HAQ-DI) scores.² However, this relationship is less pronounced in early RA (ERA), where joint damage is often minimal at baseline and functional impairment is predominantly driven by reversible, disease activity-related factors.^3–5 Early diagnosis and intervention are critical in RA, as it reduces joint damage and disability and improves long-term outcome—particularly early intervention when structural preservation is still feasible.

Bone erosions can occur early and are a central feature of RA reflecting disease severity and poor functional outcome.⁶ A recent ERA radiographic-based study showed that erosion score, but not joint space narrowing, was strongly associated with HAQ-DI at baseline, 2 and 5 years.⁷ However, radiography lacks sensitivity for revealing early, subtle erosions and capturing longitudinal changes in erosion outcomes such as repair or progression. High-resolution peripheral quantitative CT (HR-pQCT) addresses this limitation by providing higher resolution and three-dimensional (3D) depictions of bone erosions.^{8 9} One HR-pQCT-based study showed that achieving sustained Simplified Disease Activity Index (SDAI) remission in patients with ERA was associated with partial erosion repair over 1 year, indicating that structural recovery may occur when inflammation is effectively controlled.¹⁰ Bone erosion repair is thought to reflect restoration of bone formation following suppression of inflammation and is typically partial rather than complete. While previous HR-pQCT studies have primarily described erosion repair as an imaging outcome, its clinical significance, particularly its relationship with functional improvement, remains unclear. Also, baseline erosion scores correlate with baseline functional disability in RA, supporting a link between HR-pQCT-measured erosion and impaired function.¹¹ It remains unclear to what extent HR-pQCT-detected changes in erosion outcomes, including partial repair and progression, are associated with changes in functional outcomes beyond inflammatory control and treatment. Clarifying this relationship may help determine whether HR-pQCT-based structural monitoring can provide clinically relevant information complementary to conventional disease activity assessment.

We hypothesise that, in patients with ERA treated with a treat-to-target strategy over 2 years, HR-pQCT-detected change in erosion volume is associated with changes in HAQ-DI.

Methods

Study design

Participants from a 1-year prospective multicentre cohort study evaluating the impact of achieving SDAI remission on HR-pQCT and assessing joint space changes in ERA¹² were invited to continue follow-up for 2 years in the present study. The primary outcome was the association between 2-year change in HR-pQCT-detected erosion volume and change in HAQ-DI in ERA. Secondary outcomes included (1) association between change in erosion volume and change in HAQ-DI in the partial erosion repair group and the erosion progression group, independent of changes in disease activity and (2) association between baseline and changes in clinical characteristics with erosion progression.

Subjects

All new case referrals to the rheumatology clinic of the Prince of Wales Hospital (PWH) and four other regional hospitals in Hong Kong were screened by a research assistant. Patients referred as RA, possible RA or undifferentiated arthritis were contacted by telephone. Patients with at least one tender and swollen joint without previous use of disease-modifying antirheumatic drugs (DMARDs) were invited to a screening visit at the early arthritis clinic in PWH within 2 weeks of referral.

Patients were eligible if they (1) fulfilled the 2010 American College of Rheumatology/European Alliance of Associations for Rheumatology (EULAR) classification criteria for RA and were aged over 18 years, (2) had symptom onset <2 years and (3) had active disease (SDAI >3.3).¹³ Patients were excluded if they (1) had severe deformity, which precluded a reliable HR-pQCT examination and (2) were pregnant or breastfeeding.

Evaluations and treatment

All patients received treatment according to an intensive treatment protocol aiming at SDAI remission (SDAI ≤3.3)¹⁴ and were followed at PWH by experienced rheumatologists (HS, L-ST), blinded to baseline HR-pQCT results. Assessment of efficacy was performed every 3 months in the first year and every 6 months in the second year. The following clinical and laboratory variables were assessed at each visit: erythrocyte sedimentation rate, C reactive protein, number of tender and swollen joints, visual analogue scale for pain, patient’s global assessment and physician’s global assessment (PGA) (0–10 cm=worst score, scale 0–10). SDAI was calculated at each visit. The number of damaged joints was assessed yearly. Rheumatoid factor and anticyclic citrullinated peptide antibodies status was determined at baseline. Functional disability was assessed by HAQ-DI scores. Additional visits were arranged if clinically indicated. If a patient did not achieve SDAI remission, therapy would be changed according to local RA treatment guidelines (online supplemental figure 1), unless the patient declined or a toxic effect precluded this approach. In short, patients were treated with methotrexate and then other conventional synthetic DMARDs (csDMARDs) followed by biologic/targeted synthetic DMARDs (b/tsDMARDs) if they failed to achieve SDAI remission. Prednisolone at a dose ≤10 mg/day was allowed during the study period. HR-pQCT was performed at baseline, month 6, month 12 and at 24 months.

Toxicity monitoring

To monitor for DMARD side effects, complete blood count, liver and renal function tests were performed at every visit. The treating physician recorded all adverse events and, if necessary, made appropriate treatment adjustments in accordance with the protocol.

High-resolution peripheral quantitative CT

Image acquisition

Metacarpal bone erosion was assessed at metacarpophalangeal joints (MCPJ) 2–4 of the most affected hand or dominant hand when both hands were equally affected, using an HR-pQCT system (XtremeCT II; Scanco Medical, Brüttisellen, Switzerland) by a single investigator, blinded to all the clinical information. A standard stabilisation platform was created to secure the MCPJ 2–4 extended to a flat position within the immobiliser cast. A single examination was performed for measurements of MCPJ 2–4. An anteroposterior scout view was used to define the region of interest. The scan region was 107 slices distal and 325 slices proximal to the distal margin of the third MCPJ head. The same MCPJ were examined at baseline, 6, 12 and 24 months.

Erosion identification, measurement and evaluation of partial erosion repair and progression

Erosions were defined as sharply marginated bone lesions with juxta-articular localisation with a cortical break evident in at least two adjacent slices, and often accompanied by loss of adjacent trabecular bone according to the Study rouP for x-trEme Computed Tomography in Rheumatoid Arthritis (SPECTRA) definition.¹⁵ Erosion volume was automatically calculated by the Bone Analysis Modules of an open-source image analysis tool (3D slicer).¹⁶ 3D registration was applied to obtain consistent segmentation of the cortical periosteal surface in the vicinity of the erosion, which was highly important for accurate quantification of erosion volume. One experienced reader (QW) trained in HR-pQCT image acquisition and erosion scoring, unaware of the patients’ treatment, independently identified and measured the bone erosion volumes in paired chronological order to ensure accurate identification and matching of corresponding erosions across serial HR-pQCT scans.

The total erosion volume and numbers of erosions in MCPJ 2–4 for each patient were calculated for analyses. Erosions of volume <1 mm³ were excluded.¹⁷ Patients were classified into one of four groups according to the erosion status over 24 months: erosion-free, partial repair, progression or stable. Partial erosion repair was defined at the patient level as either a decrease in total erosion counts by ≥1 or a reduction in total erosion volume exceeding the least significant change (LSC). Progression was defined conversely.¹¹ Erosions with no change in either count or volume within LSC were defined as ‘stable’. To determine the LSC, 20 randomly selected erosions from the study cohort were scored twice by the reader (QW). LSC for erosion volume was then calculated using the formula: LSC=±2.77×$\sqrt{\frac{\sum {\left(a-b\right)}^2}{2n}}$, where a and b are the scores of the first and second measurements, respectively, and n is the number of observations rescored.¹⁸ The LSC for erosion volume was 0.5 mm³.

Reliability of erosion volume and image motion assessment

Intraclass correlation coefficient (ICC) was used to assess the intra-observer and inter-observer reliability of erosion volume measurement. A two-way random-effects model with absolute agreement for single measurements was used. Intra-observer reliability was evaluated because all study measurements were performed by a single blinded reader. Inter-observer reliability was additionally assessed to examine the reproducibility of the erosion identification and measurement protocol across independent trained observers. Twenty randomly selected erosions from the study cohort were evaluated twice by the same observer, and once independently by a second observer with extensive experience in HR-pQCT image analysis. The sampled erosions originated from different patients and MCPJ and included both small and large erosions. All images acquired were selected for motion grade assessment. Image quality of each acquisition was graded according to the manufacturer-suggested image grading system.¹⁹ The images of good quality (grade 1–3) were used for further analysis.

Statistical analyses

Non-normally distributed data were reported as median (IQRs). Changes in continuous outcome measures (eg, erosion volume, HAQ-DI) over 2 years were compared using the Mann-Whitney U test, while categorical outcomes (eg, medication use) were analysed using the χ² test, as appropriate. Correlation analyses were used to assess associations between HR-pQCT parameters and clinical/functional outcomes at baseline and over 2 years, both in the overall cohort and within erosion subgroups (partial repair and progression). Spearman’s rank correlation was used as the primary method due to non-normal distributions of HR-pQCT parameters and functional outcomes. Multiple linear regression was used to investigate the independent association between change in erosion volume (independent variable) and change in HAQ-DI (dependent variable), adjusting for a priori confounders influencing both domains: age, sex, baseline erosion volume, change in disease activity (SDAI) and b/tsDMARD use during months 18–24, which was included to reflect treatment status immediately preceding the month 24 assessment. Normality of residuals was checked using normal probability plots and Kolmogorov-Smirnov test. To enhance the robustness of regression analyses against skewed distributions and potential outliers, the primary regression analyses were conducted using rank-transformed variables. Binary logistic regression was performed to determine whether baseline clinical and demographic characteristics and changes over 2 years could predict erosion progression. Covariates (p≤0.10 in univariate analyses) were selected for inclusion. All tests were two-sided with α=0.05. Analyses were conducted in SPSS Statistics, V.29 (IBM, Armonk, New York, USA).

Results

Patient characteristics

One hundred and one patients with ERA were recruited (figure 1). Paired HR-pQCT analyses were available for 98 patients (three patients were excluded due to motion artefact on HR-pQCT). Seventy-seven (78.6%) patients were women, with a mean age of 58±12 years at baseline. After 2 years of protocolised treatment (online supplemental figure 1), disease activity parameters improved significantly (all p<0.001). Median SDAI scores declined from 25.4 at baseline to 4.1 at month 24, with 45.9% and 33.7% of patients achieving remission and low disease activity, respectively (table 1 and online supplemental figure 2B). Functional outcome improved concurrently, with median HAQ-DI decreasing from 1.0 at baseline to 0.25 at month 24 (p<0.001, table 1 and online supplemental figure 2A). The use of csDMARDs and b/tsDMARDs increased significantly, while non-steroidal anti-inflammatory drug use remained stable (table 1, online supplemental figure 2C).

Figure 1. Participant flow chart. The final study population consisted of 101 participants; all of whom had participated in the baseline visit and 2-year follow-up. Complete paired HR-pQCT erosion analyses were performed in 98 patients. HR-pQCT, high-resolution peripheral quantitative CT; PI, principal investigator; RA, rheumatoid arthritis.

Table 1. Clinical, demographic and erosion parameters at baseline and 2 years.

	Baseline	Month 24	Changes over 2 years	P value
Clinical and demographic parameters (n=98)
Age (years)	58±12	59±12
Gender (female), n (%)	77 (78.6)
Disease duration (months)	1.3±2.2
Rheumatoid factor (positive, %)	86 (87.8)
ACCP (positive, %)	86 (87.8)
Disease activity parameters
Patient VAS pain	5 (3, 7)	2 (0, 3)	−3 (−5, −1)	<0.001
Patient’s global assessment	6 (4, 8)	2 (1, 3)	−4 (−5, −2)	<0.001
Physician’s global assessment	6 (4, 8)	1 (0, 2)	−5 (−7, −2)	<0.001
Tender joint count	8 (4, 13)	0 (0, 2)	−7 (−11, −4)	<0.001
Swollen joint count	4 (2, 6)	0 (0, 1)	−3 (−5, −2)	<0.001
ESR (mm/hour)	43 (26, 80)	26 (14, 45)	−14 (−28, −3)	<0.001
CRP (mg/L)	8.5 (2.5, 25.0)	1.5 (0.7, 4.1)	−3.4 (−21.4, 0)	<0.001
SDAI score	25.4 (16.4, 33.4)	4.1 (1.4, 8.5)	−20.3 (−28.6, −11.5)	<0.001
Disease activity states according to SDAI, n (%)
Remission	0	45 (45.9)
Low disease activity	10 (10.2)	33 (33.7)
Moderate disease activity	41 (41.8)	16 (16.3)
High disease activity	47 (48.0)	4 (4)
HAQ-DI (0–3)	1 (0.5, 1.5)	0.25 (0, 0.625)	−0.625 (−1.125, −0.2)	<0.001
Current treatment, n (%)
Prednisolone	25 (25.5)	6 (6.1)		0.053
NSAIDs	58 (59.2)	41 (41.8)		0.075
csDMARDs	26 (26.5)	93 (94.9)		<0.001
Methotrexate	24 (24.5)	86 (87.8)
Sulphasalazine	4 (4.0)	22 (22.4)
Leflunomide	0	18 (18.4)
Hydroxychloroquine	2 (2.0)	15 (15.3)
b/tsDMARDs	2 (2.0)	24 (24.5)		<0.001
Adalimumab	1 (1.0)	15 (15.3)
Tocilizumab	0	1 (1.0)
Sarilumab	0	1 (1.0)
Abatacept	0	1 (1.0)
Tofacitinib	1 (1.0)	0
Upadacitinib	0	6 (6.1)
HR-pQCT erosion parameters (n=98)
Total number of erosions, n	132	186
Total number of new erosions across 2 years, n		58
Number of patients with erosions, n (%)
Never		37 (37.8)
Ever	54 (55.1)	61 (62.2)
Measurements at patient level (with erosions)
Total erosion count per patient	2 (1, 3)	3 (2, 4)	1 (0, 1)	<0.001
Total erosion volume per patient (mm3)	4.7 (1.6, 19.1)	9.3 (4.9, 25.5)	1.2 (−1.8, 7.9)	0.04
Total erosion volume (mm3) by erosion outcome
Repair (n=16, 16.3%)	19.1 (9.7, 39.2)	14.1 (8.0, 19.5)	−8.2 (−15.2, −1.9)
Progression (n=38, 38.8%)	4.1 (1.8, 10.4)	11.6 (4.7, 38.8)	7.1 (2.2, 23.4)
Stable (n=7, 7.1%)	3.3 (1.5, 5.4)	3.3 (1.4, 5.4)	−0.06 (−0.26, 0.12)

Longitudinal changes in disease activity, functional status, treatment exposure and HR-pQCT-derived erosion parameters over the 2-year follow-up. Data were reported as mean±SD, median (IQR) or number (%). All clinical disease activity parameters improved over 2 years while the total erosion count increased. P-value < 0.05 means significance between baseline and month-24 parameters using related-samples Mann-Whitney U test.

ACCP, anticyclic citrullinated peptide antibodies; b/tsDMARDs, biologic and targeted synthetic disease-modifying antirheumatic drugs; CRP, C reactive protein; csDMARDs, conventional synthetic disease-modifying antirheumatic drugs; ESR, erythrocyte sedimentation rate; HAQ-DI, Health Assessment Questionnaire-Disability Index; HR-pQCT, high-resolution peripheral quantitative CT; NSAIDs, non-steroidal anti-inflammatory drugs; SDAI, Simplified Disease Activity Index; VAS, visual analogue scale.

HR-pQCT measurements

Changes in erosion parameters

At baseline, 54 patients (55.1% of 98 patients) had a total of 132 erosions. Over 2 years, 37 patients remained erosion-free throughout follow-up, while the numbers of patients with detectable erosions increased from 54 at baseline to 61 at month 24, with 58 new erosions detected. Of the 61 patients with erosion throughout the follow-up, 16 patients demonstrated partial erosion repair (‘partial erosion repair group’), 38 patients showed progression (‘erosion progression group’) and seven patients remained stable (‘stable group’) (table 1). The total erosion count and total erosion volume for each patient with erosion both increased significantly at month 24 (table 1). Representative HR-pQCT images illustrating 2-year changes in erosion size (partial repair or progression) were shown in online supplemental figure 3. The intra-observer and inter-observer reliability of erosion volume measurements was excellent, with intra-observer and inter-observer ICCs of 0.999 and 0.989, respectively.

Correlation between erosion volume and baseline disease activity parameters

Among patients with erosion, baseline swollen joint count (SJC) was positively correlated with erosion volume across all groups at baseline (figure 2A, C and E), and at month 24 in the entire erosive group (partial repair, progression and stable) and in the erosion progression group (figure 2B, D and F), but not with change in erosion volume (data not shown). No correlation was observed between erosion volume and baseline tender joint count or SDAI (data not shown).

Figure 2. Scatter plots showing correlations between baseline swollen joint count (SJC) and erosion volume (mm³) at baseline (A, C, E) and month 24 (B, D, F). Higher SJCs at baseline are associated with greater erosion volume at baseline. Spearman’s rank-order correlation assessed the association between two parameters; statistical significance was defined as p≤0.05. Correlation coefficients (ρ, Rho) and p values were reported when p≤0.05.

Correlation between changes in erosion volume and function

HAQ-DI scores improved over 2 years in all patient groups (erosion-free, repair, progression and stable) and with no significant between-group differences detected at any time point (online supplemental figure 2B). Among patients with erosion, no association was observed between change in erosion volume and change in HAQ-DI (figure 3A). In subgroup analysis, reduction in total erosion volume correlated with HAQ-DI score improvements in the partial erosion repair group (n=16, ρ=0.719, p=0.003; figure 3B), that is, partial erosion repair was associated with functional improvement. No association between change in erosion volume and HAQ-DI was observed in the erosion progression or erosion stable groups (figure 3C,D).

Figure 3. Scatter plots showing correlations between change in erosion volume (mm³) and change in Health Assessment Questionnaire-Disability Index (HAQ-DI) for the (A) entire group, (B) partial repair group, (C) progression group and (D) stable group. Partial erosion repair was associated with improvement in HAQ-DI. Δ: changes in the variable from baseline to month 24. Spearman’s rank-order correlation assessed the association between two parameters; statistical significance was defined as p≤0.05. Correlation coefficients (ρ, Rho) and p values were reported when p<0.05.

Separately, univariate analysis showed that change in HAQ-DI was associated with change in SDAI (online supplemental table 1). Patients who maintained b/tsDMARD use during months 18–24 experienced greater reductions in HAQ-DI (−1 (−1.875, −0.75)) compared with non-users (−0.625 (−1, −0.1275), p=0.008).

In multivariable models, in the partial erosion repair group (n=16), both reduction in volume and improvement in SDAI were significantly and independently associated with functional improvement. After adjusting for change in SDAI, partial erosion repair remained independently associated with HAQ-DI improvement (coefficient β=0.607, p=0.003; table 2), alongside the significant contribution of change in SDAI (coefficient β=0.453, p=0.018; table 2). This association also persisted after accounting for biologic exposure (coefficient β=0.656, p=0.006; table 2). Among erosion progressors, HAQ-DI change was associated only with change in SDAI (β=0.755, p<0.001; table 2). Associations of HAQ-DI change with b/tsDMARD use were not statistically significant.

Table 2. Regression analysis of 2-year change in HAQ-DI versus erosion volume change.

	Linear regression
	95% CI for B	Coefficient β	P value
Entire erosion group (n=61)
Adjusted for Δ SDAI
Δ volume	0.256 (–0.104, 0.616)	0.150	0.160
Δ SDAI	0.610 (0.384, 0.837)	0.570	<0.001
Adjusted for b/tsDMARDs use
Δ volume	0.173 (–0.146, 0.737)	0.295	0.186
b/tsDMARDs use	−0.062 (–24.254, 14.878)	−4.688	0.633
Partial repair (n=16)
Adjusted for Δ SDAI
Δ volume	2.709 (1.099, 4.319)	0.607	0.003
Δ SDAI	0.515 (–0.009, 0.345)	0.453	0.018
Adjusted for b/tsDMARDs use
Δ volume	2.930 (0.977, 4.883)	0.656	0.006
b/tsDMARDs use	9.886 (–16.820, 36.592)	0.162	0.438
Progression (n=38)
Adjusted for Δ SDAI
Δ volume	0.195 (–0.096, 1.044)	0.474	0.100
Δ SDAI	0.680 (0.494, 1.015)	0.755	<0.001
Adjusted for b/tsDMARDs use
Δ volume	0.711 (–0.080, 1.502)	0.293	0.077
b/tsDMARDs use	−10.451 (–0.081, 16.665)	−0.126	0.439

Reduction in erosion volume is independently associated with functional improvement across different groups stratified by erosion outcome. Linear regression model was used to assess the association between the 2-year changes in HAQ-DI and changes in erosion volume for the entire group and two subgroups: partial erosion repair group (n=16) and erosion progression group (n=38). Model adjusted for change in SDAI, use of b/tsDMARDs during months 18–24. CI for unstandardised B, coefficient (β) and P-value indicating the statistical significance of the association between each predictor and the outcome in the linear regression model was reported. b/tsDMARDs use denoted the use of b/tsDMARDs during months 18–24.

b/tsDMARDs, biologic and targeted synthetic disease-modifying antirheumatic drugs; HAQ-DI, Health Assessment Questionnaire-Disability Index; SDAI, Simplified Disease Activity Index; Δ, change of the variable from baseline to month 24.

Baseline clinical characteristics and changes across erosion outcome groups

Compared with the erosion-free group, patients in the erosion-progression group (n=38) were older (59±11 vs 55±12 years; p=0.019) and had higher baseline SJC (4 (3, 6) vs 2 (2, 4); p=0.010). The erosion-free group also had higher baseline patient global assessment scores than the erosion repair group (7 (4, 8) vs 5 (3, 6); p=0.005) (online supplemental figure 4).

Among changes over time, only SJC differed between the erosion-free and partial repair groups (−2 (−4, −1) vs −4 (−5, −3); p=0.029). No between-group differences were observed in changes in other disease activity parameters or in HAQ-DI across the four groups.

When classifying patients into erosion progressors (n=38) and erosion non-progressors (n=60, comprising erosion-free, partial repair and stable groups), higher baseline SJC was observed in progressors compared with non-progressors (p=0.043, online supplemental table 2). Univariate analysis showed that higher baseline SJC was associated with erosion progression (OR 1.177, 95% CI 1.026 to 1.349, p=0.020; online supplemental table 3). After adjusting for sex, higher baseline SJC remained associated with erosion progression (OR 1.171, 95% CI 1.019 to 1.346, p=0.026; online supplemental table 3). Clinical parameters at month 24 and their changes over 2 years were similar between erosion progressors and non-progressors (data not shown).

Discussion

This is the first 2-year treat-to-target study in ERA to show that, among patients with partial erosion repair, reduction in HR-pQCT-measured erosion volume is correlated with functional improvement, as measured by HAQ-DI, independent of change in SDAI and b/tsDMARD exposure.

Functional impairment in RA is a result of complex interactions involving synovial inflammation, bone destruction and cartilage degradation. Baseline SJC exhibited a positive correlation with erosion volume across all groups of patients with erosive RA. Additionally, it demonstrated a predictive link with erosion progression over time. These findings indicate that SJC, serving as a marker of active synovial inflammation, plays a pivotal role in both existing and advancing bone damage in patients with ERA. These conclusions align with recent findings of radiographic and ultrasound-based studies.^{20 21}

Radiographically evident erosion repair is rare in RA, partly due to active suppression of bone formation by proinflammatory cytokines.⁶ Animal studies show that once synovitis resolves, osteoblasts populate the eroded bone surfaces enabling erosion repair.²² This suggests that when erosion repair is not apparent, residual subclinical inflammation may be present. With effective synovitis control using DMARDs, partial erosion repair appears possible, manifesting as new bone apposition (sclerosis) at the erosion base.^{6 23–28} Our finding that one-sixth of patients with ERA achieved partial erosion repair after 2 years of treatment concurs with previous HR-pQCT studies reporting erosion repair in RA.^{10 12 23 25 29 30}

Beyond confirming that partial erosion repair can be detected in ERA, our study further showed that the magnitude of HR-pQCT-measured erosion volume reduction was associated with functional improvement. Bone erosion and cartilage damage play a more significant role in determining functional impairment than inflammation in established RA.^{1 31} In the present ERA cohort, reduction in erosion volume was independently associated with improvement in HAQ-DI among patients with partial erosion repair, even after accounting for improvement in SDAI and b/tsDMARD exposure. These findings suggest that HR-pQCT-detected structural recovery may provide information beyond clinical disease activity alone. Interestingly, at month 24, only PGA remained weakly associated with HAQ-DI. This may reflect the limited variability in disease activity after successful treat-to-target management, while PGA, as a global clinical assessment, may better capture residual disease burden than individual disease activity measures. The mechanism through which partial erosion healing may contribute to improved hand function, such as enhancing joint stability, reducing pain, increasing range of motion or reducing joint stiffness, will require investigation in future studies. Future studies may also benefit from incorporating hand-specific functional outcome measures, such as the Michigan Hand Outcomes Questionnaire or the Cochin Hand Function Scale, to better capture the functional consequences of local structural changes detected by HR-pQCT.

In contrast, among erosion progressors, functional improvement was primarily associated with reduction in overall disease activity, rather than a change in erosion volume, consistent with prior studies.^{1 5 31 32} Bone erosion progressed in over one-third of total 98 patients, suggesting that factors beyond clinically apparent synovial inflammation may contribute to ongoing erosion progression despite overall clinical improvement, not fully captured by clinical disease activity measures alone. Residual inflammation is frequently detected on ultrasound or MRI in joints in clinical remission.³³ Glinatsi et al reported that MRI-detected wrist synovitis, rather than bone damage, was associated with physical impairment in ERA.³⁴ Despite differences in the anatomical sites examined, imaging modalities and structural outcome measures, we also found that changes in the overall erosion score were not associated with changes in HAQ-DI. As HR-pQCT cannot visualise synovitis, tenosynovitis or bone marrow oedema (osteitis), residual subclinical inflammation may have contributed to the associations observed in the present study. While ultrasound and MRI remain important for assessing subclinical inflammation in daily practice, HR-pQCT provides a unique structural perspective by quantifying erosion morphology and longitudinal changes. Although currently restricted to research settings, our results suggest the potential of HR-pQCT to be integrated into broader clinical workflows as ‘imaging remission’ may serve as a useful complement to clinical remission in ERA. To this end, the independent effect of HR-pQCT-assessed erosion progression on functional impairment in ERA should be investigated in larger, long-term studies, ideally during clinical remission so that symptoms do not confound HAQ-DI-based functional assessment.

This study has several strengths, including a prospective 2-year ERA cohort with systematic clinical and imaging assessments; use of HR-pQCT to sensitively quantify erosion volume; rigorous subgroup and multivariable analyses adjusting for disease activity (SDAI) and b/tsDMARD exposure and linkage of structural change to a clinically meaningful outcome (HAQ-DI). Limitations include the small partial-repair subgroup (n=16), which may constrain precision and generalisability; incomplete understanding of the mechanisms by which differing structural trajectories (erosion repair vs progression) translate to functional gains and how these interact with residual inflammation across disease phases and real-world settings and a 2-year follow-up that may be insufficient to capture long-term structural evolution and durability of functional effects, underscoring the need for larger cohorts and extended follow-up. Additionally, HR-pQCT evaluation was confined to a small number of joint regions. Although these joints are functionally important for grip, grasp and fine hand movements, whether the local structural dynamics over MCPJ 2–4 are truly functionally significant or just reflective of more generalised skeletal changes remains uncertain. Finally, although this study was conducted prospectively as an extension of a previously published cohort, no preregistered protocol or prespecified statistical analysis plan was available for the present analyses. This may reduce transparency regarding the prespecification of study outcomes, and the findings should therefore be interpreted with appropriate caution and confirmed in future prospective studies.

In conclusion, this 2-year HR-pQCT study demonstrates that partial erosion repair in early RA is not merely structural but also functionally meaningful. Specifically, HR-pQCT-detected erosion repair was independently associated with improvement in HAQ-DI, a relationship that persisted after accounting for the significant functional benefits of systemic inflammation suppression and beyond treatment exposure. Clinically, these results support maintaining tight control of inflammation to promote structural preservation/repair and maximise functional recovery. Sensitive imaging techniques such as HR-pQCT may provide additional clinically relevant insights into dynamic changes in bone erosion in early disease, complementing conventional disease activity assessment.

Data availability statement

Data are available on reasonable request.