Impact of vitamin D levels on disease activity, function and health over 2 years in radiographic axial spondyloarthritis: data from GErman SPondyloarthritis Inception Cohort (GESPIC)

Abstract

Objectives

To investigate the impact of vitamin D on disease activity, function and health in patients with radiographic axial spondyloarthritis (r-axSpA) undergoing biological disease-modifying antirheumatic drug (bDMARD) therapy.

Methods

Patients with r-axSpA and active disease at baseline initiating a bDMARD were included in this analysis. Vitamin D (25-hydroxyvitamin D) was measured at baseline and every 6 months until year 2; deficiency was defined as <20 ng/mL. The outcomes were AxSpA Disease Activity Score (ASDAS), Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), Patient Global Assessment (PGA), Bath Ankylosing Spondylitis Functional Index (BASFI), Assessment of SpondyloArthritis international Society Health Index (ASAS-HI) and C reactive protein (CRP). Longitudinal associations were evaluated using generalised estimating equations, with sensitivity analyses accounting for time-varying confounders.

Results

We analysed 122 patients (mean age 36.6 (10.3) years; 66.4% male), of whom 53.3% had vitamin D deficiency at baseline. In longitudinal models, normal vitamin D levels were associated with lower ASDAS (β per 1 ng/mL increase: −0.010; 95% CI −0.017 to −0.003), BASDAI (β: −0.021; 95% CI −0.036 to −0.006), PGA (β: −0.021; 95% CI −0.039 to −0.003) and BASFI (β: −0.015; 95% CI −0.032 to 0.001). Effect estimates of ASDAS, BASDAI, BASFI, CRP (and their corresponding categorical outcomes) were attenuated in sensitivity analyses but remained directionally consistent. Positive associations with PGA and ASAS-HI were also observed though the estimates were not consistent through all models.

Conclusions

Normal vitamin D levels were modestly associated with lower disease activity in r-axSpA treated with bDMARDs. Monitoring and optimising vitamin D status may support better disease control.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Low serum vitamin D levels are common in axial spondyloarthritis (axSpA) and have been associated in cross-sectional studies with higher disease activity. However, longitudinal evidence is lacking.

WHAT THIS STUDY ADDS

• Higher serum vitamin D levels are associated with lower disease activity and better functional outcomes over 2 years in patients with radiographic axSpA under biological therapy.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings underscore the potential role of vitamin D as a biomarker of disease course in axSpA. Further studies should investigate whether correcting vitamin D deficiency and/or supplementation improves clinical outcomes.

Introduction

Axial spondyloarthritis (axSpA) is a chronic immune-mediated inflammatory disease that predominantly affects the axial skeleton, characterised by inflammation in the spine and sacroiliac joints and resulting in chronic back pain, limiting physical function and reduction of quality of life.¹ Advanced therapies, including targeted synthetic and biological disease-modifying antirheumatic drugs (ts/bDMARDs), have been shown to control disease activity in many patients; however, a substantial proportion of patients still fail to achieve an adequate clinical response.² Identifying modifiable factors that could potentially influence the clinical outcomes represents an important step forward in improving the management of axSpA.

Vitamin D is a steroid hormone well known for its role in calcium regulation and bone homeostasis. Beyond these classical functions, growing evidence has drawn attention to its involvement in modulating both innate and adaptive immune responses, which are key drivers of inflammation.³ These findings support the idea that a deficiency in vitamin D may contribute to promote immune activation in rheumatic diseases, and some research showed that low levels of vitamin D have been associated with increased disease activity and severity in various inflammatory rheumatic diseases such as rheumatoid arthritis and systemic lupus erythematosus.^{4 5}

However, current evidence regarding axSpA remains inconclusive. While some studies report an inverse association between vitamin D levels and disease activity markers, such as Axial Spondyloarthritis Disease Activity Score (ASDAS), the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) and C reactive protein (CRP), others do not confirm this association.^{6 7} Moreover, most of these studies are either cross-sectional or measured vitamin D levels at one timepoint,^{8 9} so the association during the follow-up remains scarce.

In this study, we aimed to investigate the association between serum vitamin D levels and disease activity, function and health in patients with axSpA receiving bDMARD therapy, taking into account repeated vitamin D measurements over a 2-year follow-up period rather than relying on baseline vitamin D status alone.

Methods

Study population

We conducted a longitudinal cohort study using data from the German Spondyloarthritis Inception Cohort-Ankylosing Spondylitis (GESPIC-AS), a prospective longitudinal subcohort of the GESPIC. The GESPIC-AS cohort enrolled patients with active radiographic axSpA (r-axSpA) initiating bDMARD therapy at the rheumatology outpatient clinic of Charité – Universitätsmedizin Berlin between 2015 and 2019.

Included patients: (1) had a clinically confirmed diagnosis of r-axSpA; (2) met the modified New York criteria¹⁰; (3) had no prior exposure to ts/bDMARDs or a minimum 3-month discontinuation before enrolment and (4) were eligible for initiation of advanced therapy due to high disease activity according to local guidelines and regulations, with no restriction on other therapies. For the present analysis, only patients initiating bDMARD therapy were included; patients receiving tsDMARDs, including JAK inhibitors, were not part of the analysed cohort.

At baseline, all patients initiated a bDMARD prescribed at the discretion of the treating rheumatologist. No patients treated with tsDMARDs were included in this analysis. No patient was receiving a supplement of vitamin D at baseline. For the present analyses, we used data from follow-up visits that occurred every 6 months for a maximum duration of 2 years, corresponding to baseline, month 6, month 12, month 18 and month 24 assessments.

Collected data: covariates

We collected data using a standardised case report form at each study visit. At baseline, we recorded the patients’ demographic information and HLA-B27 status. During each visit, we systematically documented lifestyle factors, such as body mass index (BMI), smoking status and alcohol consumption, and clinical features, including musculoskeletal and extramusculoskeletal manifestations, as well as spinal mobility assessment (Bath Ankylosing Spondylitis Mobility Index). Furthermore, we recorded concurrent treatment regimens, including non-steroidal anti-inflammatory drugs (NSAIDs), high-dose corticosteroids (≥7.5 mg/day prednisone equivalent) and conventional DMARDs intake. We also examined laboratory test results, including CRP levels in mg/L. The quantification of NSAID use was facilitated by the NSAID intake score.

Vitamin D measurement: exposure

We measured serum 25-hydroxyvitamin D [25(OH)D] at each visit using stored serum samples. Measurements were performed using a competitive chemiluminescence immunoassay with the LIAISON 25-OH Vitamin D Assay (DiaSorin) on the LIAISON XL analyser. The assay had an analytical measuring range of 4.0–150 ng/mL. We defined low vitamin D levels (vitamin D deficiency) as serum concentrations below 20 ng/mL (equivalent to <50 nmol/L).

Disease activity and outcomes

We assessed disease activity using the ASDAS, the BASDAI, patient global assessment (PGA, 0–10 Numeric Rating Scale -NRS), and CRP levels. We evaluated physical function with the Bath Ankylosing Spondylitis Functional Index (BASFI) and overall health status using the Assessment of SpondyloArthritis International Society Health Index (ASAS-HI). We defined inactive disease for ASDAS as scores <1.3 (ASDAS-ID) and low disease activity as scores <2.1 (ASDAS-LDA), which also includes ASDAS-ID. We categorised LDA for BASDAI and PGA as scores ≤4 points, and normal CRP as levels <5 mg/L.

Statistical analysis

We reported the baseline characteristics using means and SD for continuous variables and counts with percentages for categorical variables.

We selected covariates for the multivariable models based on existing literature, clinical expert knowledge and the temporal ordering of repeated measurements. Where relevant, we distinguished prior (‘t–1’) and current (‘t’) values in the sensitivity analyses. We then analysed the longitudinal associations between vitamin D levels and outcomes using generalised estimating equations (GEE), including all available repeated measurements collected over the 24-month follow-up period. To evaluate potential effect modification by season, we initially specified models including an interaction term between vitamin D and season (vitamin D×season). For continuous vitamin D, effect modification was assessed using Wald tests of the single interaction coefficient; for categorical vitamin D, we performed global Wald tests jointly evaluating all category-by-season interaction terms. Because no relevant effect modification by season was identified, we present the primary models without interaction terms. The working correlation structure was selected by comparing independence, exchangeable and first-order autoregressive structures using the quasi-likelihood under the independence-model criterion, with independence selected for the primary analyses. Given the presence of time-varying exposures and confounders, we conducted sensitivity analyses using sequential conditional mean models. The aim was to mitigate potential bias arising from standard regression approaches. These models incorporated adjustments for prior outcomes and prior exposure to estimate the total effect of vitamin D on disease activity and functional outcomes.

For continuous outcomes, we used linear GEE models; for binary outcomes, we used logistic GEE models. We reported effect estimates as regression coefficients (β) or ORs with 95% CIs. Because previous biologic exposure may influence treatment response and longitudinal disease outcomes, we additionally performed a subgroup analysis restricted to patients who were bDMARD-naive at baseline. No imputation was performed for the main exposure or outcome variables. The primary GEE analyses used available observations across visits. The only exception was BMI, which was included as a time-varying covariate using a filled longitudinal BMI variable; when BMI was missing at a visit, the most recently observed BMI value was carried forward for the same patient. We performed main statistical analyses using the ‘geepack’ package in R V.4.3.3 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Patient clinical characteristics

The analysed cohort comprised only patients initiating bDMARD therapy; no patients receiving tsDMARDs were included. Of 127 patients assessed at baseline, one was excluded due to duplicate baseline records. Among the remaining 126 patients, 122 had available baseline serum vitamin D measurements and were included in the analysis (figure 1). The flow diagram also depicts patient availability and vitamin D measurement status across follow-up visits from baseline to 24 months. Overall, follow-up retention was high, with 112/122 patients (91.8%) attending the month 6 visit, 112/122 (91.8%) the month 12 visit, 104/122 (85.2%) the month 18 visit and 114/122 (93.4%) the month 24 visit. Most patients contributed data to most scheduled follow-up visits, although missed visits were not always from the same individuals across time points.

Figure 1. Participant inclusion and follow-up with vitamin D (Vit D) assessments (baseline to 24 months). Reasons for non-participation or missing visits during follow-up were not systematically recorded in the cohort database.

Among 122 patients with available vitamin D serum levels at baseline, 65 (53.3%) had low vitamin D levels (table 1). The mean age of the overall population was 36.6 (10.3) years, and 81 (66.4%) were male. Most patients (88.5%) were HLA-B27-positive, had a mean symptom duration of 10.6 (9.2) years and a mean BMI of 25.0 (4.3) kg/m². At baseline, 47 patients (38.5%) were current smokers. Patients showed high disease activity, with a mean CRP of 13.5 (17.7) mg/L, an ASDAS of 3.4 (0.8) and a BASDAI of 5.5 (1.3). Most were receiving NSAIDs (78.7%) with a mean NSAID intake score of 67.2 (42.7). Only 1.6% received high-dose corticosteroids. At baseline, 4 (3.3%) of participants were taking conventional synthetic DMARDs and 26 (21.3%) had prior exposure to bDMARDs. To explore possible attrition bias, we compared baseline characteristics between patients who completed month 24 follow-up and those who did not. Baseline characteristics were broadly similar between the groups, although the number of non-completers was small (online supplemental table 1).

Table 1. Baseline demographic and clinical characteristics.

Variable	Available N	Overall N=122	Normal vitamin D (≥20 ng/mL) N=57	Low vitamin D (<20 ng/mL) N=65
Male sex, n (%)	122	81 (66.4)	36 (63.2)	45 (69.2)
Age, years	122	36.6 (10.3)	36.5 (9.9)	36.6 (10.7)
Symptom duration, years	122	10.6 (9.2)	10.3 (7.2)	10.9 (10.6)
Disease duration, years	122	4.7 (6.9)	4.6 (6.1)	4.7 (7.6)
Current smoker, n (%)	122	47 (38.5)	15 (26.3)	32 (49.2)
Ever smoker, n (%)	122	68 (55.7)	29 (50.9)	39 (60.0)
Current alcohol use, n (%)	122	41 (33.6)	21 (36.8)	20 (30.8)
HLA-B27-positive, n (%)	122	108 (88.5)	48 (84.2)	60 (92.3)
Family history of SpA, n (%)	122	48 (39.3)	19 (33.3)	29 (44.6)
Body mass index, kg/m²	122	25.0 (4.3)	23.9 (3.8)	26.0 (4.5)
Current peripheral arthritis, n (%)	122	22 (18.0)	8 (14.0)	14 (21.5)
Tender joint count	122	0.4 (1.2)	0.4 (1.5)	0.4 (0.9)
Swollen joint count	122	0.4 (1.2)	0.4 (1.5)	0.4 (0.9)
Current enthesitis, n (%)	122	38 (31.1)	20 (35.1)	18 (27.7)
Enthesitis sites count, MASES	122	0.9 (2.1)	1.1 (2.3)	0.8 (1.9)
Ever dactylitis, n (%)	122	6 (4.9)	3 (5.3)	3 (4.6)
Current dactylitis, n (%)	122	2 (1.6)	0 (0.0)	2 (3.1)
Ever uveitis, n (%)	122	27 (22.1)	11 (19.3)	16 (24.6)
Ever psoriasis, n (%)	122	18 (14.8)	7 (12.3)	11 (16.9)
Ever IBD, n (%)	122	9 (7.4)	7 (12.3)	2 (3.1)
CRP, mg/L	118	13.5 (17.7)	14.1 (15.6)	13.1 (19.4)
Elevated CRP (>5 mg/L), n (%)	118	73 (61.9)	33 (62.3)	40 (61.5)
ESR, mm/hour	121	23.9 (20.9)	25.9 (20.1)	22.1 (21.5)
Pain (VAS)	121	6.8 (1.6)	6.9 (1.5)	6.7 (1.6)
Patient Global Assessment	122	6.8 (1.6)	6.8 (1.7)	6.8 (1.6)
Physician Global Assessment	122	5.8 (1.3)	5.9 (1.4)	5.8 (1.2)
ASDAS	118	3.4 (0.8)	3.5 (0.8)	3.4 (0.8)
BASDAI	122	5.5 (1.3)	5.5 (1.3)	5.6 (1.3)
BASFI	122	4.4 (2.0)	4.4 (1.9)	4.5 (2.2)
ASAS-HI score	122	7.6 (2.9)	7.9 (2.7)	7.3 (3.1)
BASMI	121	3.0 (1.6)	2.8 (1.5)	3.2 (1.7)
NSAID intake, n (%)	122	96 (78.7)	44 (77.2)	52 (80.0)
NSAID intake score	122	67.2 (42.7)	64.7 (44.2)	69.4 (41.4)
Any steroid intake, n (%)	122	7 (5.7)	2 (3.5)	5 (7.7)
High-dose (>7.5 mg) steroid intake, n (%)	122	2 (1.6)	0 (0.0)	2 (3.1)
csDMARDs intake, n (%)	122	4 (3.3)	2 (3.5)	2 (3.1)
Prior bDMARD intake, n (%)	122	26 (21.3)	12 (21.1)	14 (21.5)

Values are presented as mean (SD), or as n (%), unless otherwise indicated.

ASAS-HI, Assessment of SpondyloArthritis international Society Health Index; ASDAS, Axial Spondyloarthritis Disease Activity Score; BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; BASFI, Bath Ankylosing Spondylitis Functional Index; BASMI, Bath Ankylosing Spondylitis Mobility Index; bDMARD, biologic disease-modifying antirheumatic drug; CRP, C reactive protein; csDMARDs, conventional synthetic DMARDs; ESR, erythrocyte sedimentation rate; HLA, human leucocyte antigen; IBD, inflammatory bowel disease; MASES, Maastricht Ankylosing Spondylitis Enthesitis Score; NSAID, non-steroidal anti-inflammatory drug; SpA, spondyloarthritis; VAS, Visual Analogue Scale.

Patients with low vitamin D levels had higher BMI (26.0 (4.5) vs 23.9 (3.8) kg/m²) and a higher rate of active smoking (49.2% vs 26.3%) compared with those with normal levels. Other demographic, clinical and treatment-related variables were similar between the two groups (table 1). Across follow-up visits, the distribution of vitamin D status changed over time (online supplemental table 2). The proportion of patients with normal vitamin D levels increased from 46.7% at baseline to 58.9% at Month 12 and was 54.0% at Month 24. As shown in online supplemental table 3, mixed follow-up patterns were common, with 28.7% of patients with low vitamin D at baseline and 19.7% of those with normal vitamin D at baseline showing both low and normal values during follow-up. In contrast, persistent change from low vitamin D at baseline to only normal values at follow-up was observed in 4.1% of patients. In addition, mean BMI and the proportion of current smokers remained broadly stable over time, without major descriptive shifts during follow-up (online supplemental table 4).

Effects of vitamin D on disease activity and function

Over 2 years of follow-up, we observed inverse associations in the main models between serum vitamin D levels and disease activity (table 2). In interaction analyses, we did not detect relevant effect modification by season; therefore, the primary GEE models are presented without vitamin D-by-season interaction terms (all interaction p values >0.25). In adjusted multivariable GEE model, a 1 ng/mL increase in vitamin D corresponded to a 0.010-point reduction in ASDAS (95% CI −0.017 to −0.003), a 0.021-point reduction in BASDAI (95% CI −0.036 to −0.006) and 0.021-point reduction in PGA (95% CI −0.039 to −0.003). Similarly, we found a 0.015-point reduction in BASFI per 1 ng/mL increase in vitamin D (95% CI −0.032 to 0.001), while estimates for ASAS-HI and CRP remained closer to zero. When categorising vitamin D exposure as normal or low, we observed that normal vitamin D was associated with a 0.187-point lower ASDAS (95% CI –0.1395 to 0.022) compared with low levels. The remaining outcome estimates were mostly in the same negative direction, yet their CIs included null values (table 2).

Table 2. Association between vitamin D (Vit D) levels and clinical outcomes during the 24-month follow-up in patients with axial spondyloarthritis: multivariable GEE analyses with continuous and categorical outcomes.

Exposure	Continuous outcome	Adjusted β (95% CI)	Categorical outcome	OR (95% CI)
Vit D (per 1 ng/mL)	ASDAS	−0.010 (−0.017 to −0.003)	ASDAS Low Disease Activity	1.031 (1.012 to 1.050)
			ASDAS Inactive Disease	1.017 (0.998 to 1.036)
	BASDAI	−0.021 (−0.036 to −0.006)	BASDAI <4	1.025 (1.006 to 1.045)
	PGA	−0.021 (−0.039 to −0.003)	PGA <4	1.018 (1.002 to 1.033)
	BASFI	−0.015 (−0.032 to 0.001)	BASFI <4	1.018 (0.995 to 1.041)
	ASAS-HI	−0.019 (−0.054 to 0.017)	ASAS-HI Good health state	1.011 (0.990 to 1.032)
	CRP	−0.044 (-0.130 to 0.043)	CRP <5 mg/L	1.012 (0.988 to 1.037)
Normal Vit D (≥20 ng/mL) vs low	ASDAS	−0.187 (−0.395 to 0.022)	ASDAS Low Disease Activity	1.616 (1.049 to 2.489)
			ASDAS Inactive Disease	1.684 (0.975 to 2.907)
	BASDAI	−0.326 (−0.770 to 0.119)	BASDAI <4	1.263 (0.823 to 1.940)
	PGA	−0.308 (−0.871 to 0.255)	PGA <4	1.180 (0.766 to 1.818)
	BASFI	−0.296 (−0.738 to 0.146)	BASFI <4	1.525 (0.876 to 2.656)
	ASAS-HI	−0.487 (−1.301 to 0.326)	ASAS-HI good health state	1.109 (0.683 to 1.801)
	CRP	−1.327 (−3.239 to 0.585)	CRP <5 mg/L	1.720 (1.062 to 2.786)

Parameter estimates from the multivariable models adjusted for sex, age, BMI, active smoking, IBD, psoriasis, NSAID intake score, high dose steroid intake and season.

Interpretation: Continuous and categorical outcomes are presented side by side to improve readability; ASDAS is shown over two rows because two categorical outcomes were analysed (low disease activity and inactive disease).

ASAS-HI, Assessment of SpondyloArthritis international Society Health Index; ASDAS, Axial Spondyloarthritis Disease Activity Score; BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; BASFI, Bath Ankylosing Spondylitis Functional Index; BMI, body mass index; CRP, C reactive protein; GEE, generalised estimating equation; IBD, inflammatory bowel disease; NSAID, non-steroidal anti-inflammatory drugs; PGA, patient global assessment.

To address the potential impact of previous biologic exposure, we performed a subgroup analysis restricted to the 96 patients who were bDMARD-naive at baseline; the associations were broadly similar in direction to those observed in the full cohort, although the CIs were generally wider and several estimates included the null, indicating lower precision in the smaller subgroup (online supplemental table 5). Further sensitivity analyses included alternative adjustment sets and showed a broadly similar directional pattern. The effect sizes attenuated when we adjusted for prior ASDAS or prior vitamin D, but the overall direction of the estimates remained similar (online supplemental table 6). For instance, when adjusting for ASDAS at the previous timepoint, the estimated difference in ASDAS per 1 ng/mL of vitamin D was –0.004 (95% CI –0.012 to 0.004 in model 1). The effect size remained comparable in models that accounted for previous vitamin D levels (–0.006; 95% CI –0.015 to 0.002 in model 2).

In categorical outcome models, we found higher odds of achieving ASDAS-LDA in patients with normal vitamin D (table 2). For each 1 ng/mL increase in vitamin D, the odds of achieving ASDAS-LDA were 3.1% higher (OR 1.031, 95% CI 1.012 to 1.050). Compared with low vitamin D levels, normal vitamin D levels were associated with 61.6% higher odds of achieving ASDAS-LDA (OR 1.616, 95% CI 1.049 to 2.489) and 68.4% higher odds of achieving ASDAS inactive disease (OR 1.684, 95% CI 0.975 to 2.907). Overall, the comparison between the full cohort and the bDMARD-naive subgroup did not indicate a major shift in the main findings (online supplemental table 5). Further, sensitivity analyses, which were adjusted for prior outcomes and exposures (online supplemental table 7), demonstrated attenuated effect estimates. However, the estimates generally pointed in the same direction, although several CIs included the null.

Discussion

Vitamin D deficiency is common in immune-mediated diseases and represents a potentially modifiable nutritional factor, but evidence linking vitamin D status to axSpA outcomes remains uncertain. Here, we investigated the longitudinal association between serum vitamin D levels and disease activity outcomes in patients with r-axSpA starting bDMARD therapy, using biannual measurements over a 2-year period to capture vitamin D status across seasons. We found that higher vitamin D levels, measured both continuously and categorically, using the ≥20 ng/mL threshold, were generally associated with more favourable disease activity and functional outcomes over time. However, several estimates were modest in magnitude, and not all were precise, particularly in sensitivity analyses. Patients with vitamin D levels in the normal range were also more likely to achieve ASDAS low/inactive disease and CRP normalisation than those with vitamin D deficiency.

Previous research has already highlighted the clinical importance of vitamin D in managing inflammatory activity in axSpA.^{9 11} Our findings align with prior reports from DESIR (DEvenir des Spondylarthropathies Indifférenciées Récentes) and ASAS-COMOSPA (COMOrbidities in SPondyloArthritis) cohorts, which have reported the associations between vitamin D deficiency and higher disease activity, increased inflammation and more severe structural progression in axSpA.^{8 9} However, not all published studies agree on these associations. For example, Klingberg et al did not observe a clear relationship between vitamin D status and disease activity in the Swedish cohort of axSpA patients.⁷ Similarly, Gula et al found null associations in a smaller Polish cohort.⁶ These discrepancies across the studies likely reflect methodological differences, including variations in vitamin D measurement frequency, definitions of deficiency and confounding factors such as geographical and seasonal variability.

Substantial evidence supports that vitamin D actively modulates immune responses by suppressing inflammation and promoting immune balance. Vitamin D, particularly its active form 1,25-dihydroxyvitamin D3, binds to the vitamin D receptor (VDR), a nuclear receptor expressed in many immune cells including T cells, B cells, dendritic cells, monocytes and macrophages. Once activated, the VDR-vitamin D complex translocates to the nucleus and regulates the transcription of several genes involved in the immune function.¹² Vitamin D modulates innate immunity by increasing macrophage differentiation and inhibiting dendritic cell maturation, which thereby limits excessive antigen presentation. In adaptive immunity, vitamin D downregulates Th1 and Th17 responses and reduces the secretion of key pro-inflammatory cytokines such as TNF-alpha, IL-6 and IL-17, while promoting regulatory T cell differentiation and IL-10 production.¹³

There are some limitations that should be acknowledged. We did not include information in our study on vitamin D supplement intake, sunlight exposure or physical activity—all of which may influence serum vitamin D levels. These lifestyle and environmental factors vary between individuals and across seasons, and their absence from the analysis may contribute to measurement variability. To provide additional context for this limitation, we described the longitudinal distribution of vitamin D status across visits and at the patient level. These analyses showed a shift towards normal vitamin D levels during follow-up but also substantial within-patient fluctuations over time (online supplemental tables 2 and 3). This pattern suggests that changes in vitamin D status were not uniform or consistently maintained across the cohort. Although BMI and smoking were measured longitudinally and included as covariates, we did not perform a formal mediation analysis. Such an analysis would require stronger assumptions about the temporal and causal relationships between bDMARD initiation, lifestyle changes, vitamin D status and clinical outcomes than could be supported in the present study. In addition, this was an observational longitudinal cohort study rather than a study designed specifically for a single vitamin D hypothesis, so that no prospective sample size calculation was performed for the present analysis. The relatively limited sample size may therefore have reduced the precision and stability of some estimates. This is reflected by several CIs being wide and including the null, particularly in the sensitivity analyses. We also examined multiple related continuous and categorical outcomes and did not apply a formal correction for multiple testing. Therefore, the findings should be interpreted cautiously and in the context of the overall pattern and direction of effect estimates rather than any isolated significant results. Another important limitation is that only patients with r-axSpA were included; therefore, the findings may not be directly generalisable to the broader axSpA spectrum, particularly to non-r-axSpA. However, the prospective design, biannual measurements across different seasons over five time points and adjustment for time-varying covariates strengthened the longitudinal assessment. Moreover, adjustment for season in the multivariable models helped to reduce the risk of misclassification due to transient or seasonal changes. Additionally, our adjustment for several key demographic and clinical factors strengthened the validity of our findings. It is also important to note that the observed associations, although generally directionally similar across several outcomes, were modest in magnitude and not all effect estimates were precise. Taken together, the findings suggest that while vitamin D may play a modulatory role in disease activity, its effect is likely to be limited. To provide a more clinically interpretable scale, a 10 ng/mL higher vitamin D level corresponded approximately to a 0.10-point lower ASDAS, a 0.21-point lower BASDAI, a 0.21-point lower PGA and a 0.15-point lower BASFI in the main models. These differences indicate a modest average association at the group level rather than a clearly clinically meaningful change at the individual patient level.

Clinically, our findings support routine assessment of serum vitamin D levels and correction of vitamin D deficiency in patients with r-axSpA initiating biological treatment, consistent with the European Society of Clinical and Economical Aspects of Osteoporosis, Osteoarthritis and Musculoskeletal Diseases recommendations.¹⁴ Although higher vitamin D concentrations were associated with more favourable disease activity outcomes, the observational design, potential residual confounding and attenuation of effects in the sensitivity analysis limit the ability to establish a direct effect of vitamin D on clinical response. Importantly, the additional analysis restricted to bDMARD-naive patients showed a broadly similar direction of association, which argues against prior biologic exposure being the sole explanation for the main findings; however, the reduced precision in this subgroup also supports cautious interpretation. Therefore, randomised supplementation clinical trials are needed to define the optimal target concentrations and to determine whether correcting deficiency brings clinically meaningful improvements during biological treatment in patients with axSpA.

Data availability statement

Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.