Investigating the Inflammatory Link Between Vitamin D and Hidradenitis Suppurativa: A Systematic Review and Causal Inference Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors investigated the inflammatory association between vitamin D and hidradenitis suppurativa. The paper is well-written and fascinating. My comments and questions are as follows.

1. Although this is presented as a systematic review, no quantitative meta-analysis is performed despite multiple studies reporting serum 25(OH)D levels. A pooled estimate (random-effects) or at least a justification for not conducting meta-analysis should be provided.
2. The Discussion extensively speculates on IL-1β/TH17, NOD2, Notch, and follicular signaling pathways. However, most of these mechanisms are inferred from other inflammatory diseases rather than directly demonstrated in HS in relation to vitamin D. This section should be condensed or clearly framed as hypothetical.
3. Different thresholds (10, 20, 30 ng/mL) are used across studies, yet pooled descriptive analyses are presented. This introduces classification bias and should be more critically addressed, possibly by standardizing categories post hoc.

Author Response

The authors investigated the inflammatory association between vitamin D and hidradenitis suppurativa. The paper is well-written and fascinating. My comments and questions are as follows.

Comment: Although this is presented as a systematic review, no quantitative meta-analysis is performed despite multiple studies reporting serum 25(OH)D levels. A pooled estimate (random-effects) or at least a justification for not conducting meta-analysis should be provided.

Response: We thank the Reviewer for this important point and agree that a quantitative meta-analysis should be considered when outcome reporting is sufficiently homogeneous. We evaluated the feasibility of pooling serum vitamin D values across the included studies; however, a formal meta-analysis was not methodologically appropriate for the following reasons:

First, the included studies reported serum vitamin D using non-uniform summary statistics, including means with standard deviations, means with 95% confidence intervals, medians with ranges, and medians with interquartile ranges. In several studies, the information required to derive a common effect estimate and its variance (for example, standard deviation/standard error, exact distributional assumptions, or sufficient sample-level detail) was not available. Because individual participant data were not accessible, reliable transformation of all studies to a common metric suitable for inverse-variance weighting was not possible without introducing substantial assumptions.

Second, there was clinical and measurement heterogeneity across studies, including variation in patient populations (general HS cohorts versus specific subgroups such as severe HS, post-bariatric surgery patients, syndromic HS, and treatment-refractory cohorts), disease severity definitions, and comparator structures. This heterogeneity limits the interpretability of a pooled estimate even under a random-effects framework.

Third, in a subset of studies, the vitamin D analyte/metabolite and/or assay methodology was not clearly specified (particularly in retrospective chart-based or genomic studies). Since the review focuses on serum vitamin D levels and their clinical correlations, uncertainty regarding the measured analyte reduces comparability and risks combining non-equivalent measures.

For these reasons, we chose a structured qualitative synthesis with descriptive aggregation (including standardized unit conversion where possible and presentation of study-level values) rather than a potentially misleading pooled meta-analytic estimate. We have now clarified this rationale in the revised manuscript (Methods/Limitations section).

The Discussion extensively speculates on IL-1β/TH17, NOD2, Notch, and follicular signaling pathways. However, most of these mechanisms are inferred from other inflammatory diseases rather than directly demonstrated in HS in relation to vitamin D. This section should be condensed or clearly framed as hypothetical.

 

We thank the Reviewer for this observation and agree that the evidentiary basis for several of the mechanistic pathways discussed is indirect in the context of hidradenitis suppurativa (HS) and vitamin D. Our intention was to provide a biologically informed framework to contextualise the clinical associations identified in the reviewed studies, rather than to imply that these mechanisms have been definitively established in HS.

In response, we have retained the mechanistic discussion but revised the framing throughout to make its status explicitly hypothesis-generating. Specifically, we now state that the IL-1β/TH17, NOD2, Notch, and follicular signaling pathways are plausible candidate mechanisms, informed by broader immunologic and epithelial biology and by evidence from related inflammatory conditions, but not directly demonstrated in HS specifically in relation to vitamin D in the studies included in this review. Section 3.2 has been restructured to:

1. Title changed to Plausible Mechanistic Pathways Linking Vitamin D and Hidradenitis Suppurativa
2. Added the following to the opening paragraph: While direct mechanistic evidence linking vitamin D to HS remains limited, existing immunologic and epithelial biology supports several plausible hypotheses. In particular, vitamin D may influence HS-relevant processes involving IL-1β/TH17 signaling, innate immune sensing (including NOD2-related pathways), and follicular differentiation networks. We next review potential mechanistic hypothesis mechanistic hypotheses linking vitamin D status to HS-relevant immune and epithelial processes.
3. Added the following text: ‘Vitamin D may thus influence inflammatory pathways relevant to HS through effects on innate and adaptive immune responses, although direct HS-specific mechanistic evidence remains limited.’
4. Added the following text: ‘Plausibly, NOD2-linked innate immune dysregulation represents a potential pathway for vitamin D-related inflammatory variation in HS, but this remains speculative and hypothesis-generating.’
5. Added the folliwing text: ‘These mechanistic pathways should be interpreted as hypothesis-generating models that may help contextualize the observed clinical associations between vitamin D status and HS, rather than as directly causal mechanisms.’

 

 

Different thresholds (10, 20, 30 ng/mL) are used across studies, yet pooled descriptive analyses are presented. This introduces classification bias and should be more critically addressed, possibly by standardizing categories post hoc.

 

We thank the Reviewer for this methodological point and agree that the use of different vitamin D thresholds across studies ( <10, <20, and <30 ng/mL) can introduce classification bias and limit direct comparability of deficiency/insufficiency prevalence estimates.

In the revised manuscript, we have strengthened the discussion of this issue and clarified that our pooled descriptive summaries should be interpreted as narrative/descriptive aggregation across heterogeneous definitions, rather than as directly comparable prevalence estimates derived from harmonised criteria. We agree in principle that post hoc standardisation of categories would be desirable; however, this was not consistently feasible because many studies did not provide sufficient granular data (raw individual-level 25(OH)D values or re-tabulated counts across multiple threshold cut-offs) to permit reliable reclassification into a common framework.

Where possible, we retained and reported the original study-defined thresholds to preserve fidelity to the source data and avoid introducing additional assumptions through indirect recategorisation. We have now explicitly identified threshold heterogeneity as a source of classification bias and a limitation of cross-study descriptive comparisons, and we have revised the relevant Results/Discussion text accordingly.

The following changes have been implemented in response to this comment:

1. In methodology, we have clarified by adding the following: ‘Vitamin D deficiency/insufficiency categories were extracted as reported in the original studies; because threshold definitions varied (for example, <10, <20, and <30 ng/mL). Additionally, study-level reporting was insufficient for consistent reclassification thus no post hoc standardisation of categorical prevalence estimates was performed. Cross-study summaries of categorical vitamin D status were treated as descriptive and interpreted with caution’
2. In the results section, this paragraph has been restructured :’ A total of 11 studies reported mean or median serum vitamin D levels in patients with HS. Across these studies, the reported central tendency values were descriptively ag-gregated, yielding an overall weighted mean of 17.90 ± 20.34 ng mL-1. This summary should be interpreted cautiously in light of substantial between-study heterogeneity in re-porting format (mean versus median), dispersion metrics, and study design. Overall, se-rum vitamin D levels were consistently low across HS cohorts, and multiple studies re-ported a high frequency of vitamin D deficiency (Figures 2A and 2B). Categorical vitamin D status, however, was defined using non-uniform thresholds, with variable criteria for deficiency/sufficiency across studies. Accordingly, the propor-tions reported here are presented as a descriptive summary of study-specific classifica-tions rather than directly comparable prevalence estimates, and no harmonized cross-study prevalence inference is implied. For this reason, category counts and propor-tions were retained according to each study’s original cut-off definitions to preserve fidel-ity to the source data and avoid introducing additional assumptions through post hoc re-classification. Koumaki et al. (2025) applied a sufficiency threshold of 20 ng mL-1, result-ing in a greater proportion of patients being classified as sufficient relative to studies using higher thresholds, the mean vitamin D level in their HS cohort remained significantly lower than in controls, consistent with the findings of four other studies.[29,30,34,39,40]’
3. In the discussion section, we have specified in the limitations ‘Furthermore, the reported heterogeneity in vitamin D threshold definitions across studies introduces potential classification bias and constrains direct comparison of deficiency/insufficiency prevalence. In the absence of sufficiently granular published data post hoc harmonization of categories was not consistently possible’

Reviewer 2 Report

Comments and Suggestions for Authors

Reviewed paper present comprehensive state of art considering the role of vitamine D level in the patients with hydradenitis suppurativa. Having known that the pathogenesis of this disease is still not fully understood this kind of research can be helpful to increase the knowledge about it. 

Author Response

Reviewer 2

Reviewed paper present comprehensive state of art considering the role of vitamine D level in the patients with hidradenitis suppurativa. Having known that the pathogenesis of this disease is still not fully understood this kind of research can be helpful to increase the knowledge about it. 

We thank the reviewer for this comment.

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript demonstrates an important and timely question on the relationship between vitamin D and HS, combining a systematic review with a two-sample MR analysis to disentangle correlation from causation. My comments are as follows:

1. The authors conclude that hypovitaminosis D is “likely a marker rather than a direct causative factor” based on the null liner MR estimate. However, the conclusion should be more cautious because the current analyses only test a very specific causal model: a linear, average effect of lifelong, genetically determined 25(OH)D levels on HS risk, in a European ancestry population. From that, the authors can safely say “we did not detect a linear causal effect,” but it is a logical overreach to say “vitamin D is not causally involved at all. Since non-linear or mediated effects (e.g. obesity) were not tested, I recommend a more cautious conclusion.

2.  The cohort is the European population, and it may not generalize to non-European populations. I propose adding a sentence in the discussion to indicate that the results may not apply to non-European populations with different vitamin D distributions and HS risk factors.

3. Please consider adding a brief risk-of-bias assessment or at least a structured description (e.g., selection bias, confounding by BMI/smoking, lack of adjustment) across the 12 studies.

Minor comment:

4. Please ensure that all abbreviations (e.g., IHS4, HS-PGA) are defined in the figure captions and within the text.

Author Response

1. The authors conclude that hypovitaminosis D is “likely a marker rather than a direct causative factor” based on the null liner MR estimate. However, the conclusion should be more cautious because the current analyses only test a very specific causal model: a linear, average effect of lifelong, genetically determined 25(OH)D levels on HS risk, in a European ancestry population. From that, the authors can safely say “we did not detect a linear causal effect,” but it is a logical overreach to say “vitamin D is not causally involved at all. Since non-linear or mediated effects (e.g. obesity) were not tested, I recommend a more cautious conclusion.

 

Thank you for the precise and important methodological clarification. We agree that our original wording was too broad relative to the scope of inference supported by the MR analysis.

 

In response, we have revised the relevant text to state explicitly that our MR findings indicate no evidence of a detectable linear causal effect of genetically predicted 25(OH)D levels on HS risk under the assumptions of the applied MR model, rather than excluding any causal involvement of vitamin D in HS more generally. We also now clarify that the MR analysis addresses a specific causal contrast (a linear, average effect of lifelong genetically influenced 25(OH)D on HS risk in a predominantly European-ancestry framework) and does not evaluate potential non-linear, threshold-dependent, context-specific, or mediated pathways (including pathways involving adiposity or other metabolic/inflammatory intermediates).

 

Accordingly, we have tempered the conclusion and Discussion wording to avoid overinterpretation of the null MR estimate and to frame the findings as narrowing, rather than closing, the range of plausible causal models.

 

In the results section, we have restructured as follows: ‘ Inverse-variance weighted (IVW) analysis does not support a linear causal effect of geneti-cically predicted 25(OH)D on HS (β = -0.390, SE = 0.35, p = 0.270). Concordant results were obtained using other MR methods, as shown in Figures 3A and B. This does not exclude the possibility of non-linear, threshold-dependent, context-specific, or mediated effects that were not evaluated in the present study.’

 

In the abstract we have clarified as follows: ‘A null MR estimate consistent with the absence of a detectable average linear causal effect of lifelong genetically predicted 25(OH)D levels on HS risk in the analyzed population was observed.’

 

The opening section of the discussion has also been updated to reflect this point.

2. The cohort is the European population, and it may not generalize to non-European populations. I propose adding a sentence in the discussion to indicate that the results may not apply to non-European populations with different vitamin D distributions and HS risk factors.

Response:

We agree that the generalisability of our findings is limited by the ancestry composition of the analysed cohorts. We have added a statement in the Discussion clarifying that the MR results were derived in a predominantly European-ancestry population and may not directly extrapolate to non-European populations, in whom vitamin D distributions, genetic architecture, environmental exposures, and HS risk factor profiles may differ.

We have added the following sentence to the discussion: ‘Additionally, the estimated effect may not be directly transferable to non-European populations, where differences in allele frequencies, linkage disequilibrium structure, vitamin D distributions, and HS risk determinants could alter both instrument performance and causal estimates’

The limitation section of the manuscript already addresses this point.

3. Please consider adding a brief risk-of-bias assessment or at least a structured description (e.g., selection bias, confounding by BMI/smoking, lack of adjustment) across the 12 studies.

In the revised manuscript, we have now incorporated a structured methodological appraisal of the included studies using the Mixed Methods Appraisal Tool (MMAT). This is included in the supplementary document. This has been specified in the methodology. In the results, we have incorporated the following text: ‘Structured appraisal using the MMAT indicated that most included studies met the majority of design-specific quality criteria, particularly regarding the appropriateness of vitamin D-related measurements and the use of suitable statistical analyses for the study design. Potential selection bias (single-centre, retrospective, severe HS cohorts), variable control of confounding (BMI, smoking, adiposity/metabolic factors, and comorbidities), and the predominance of cross-sectional designs, which limits temporal and causal inference were identified.’

Minor comment:

4. Please ensure that all abbreviations (e.g., IHS4, HS-PGA) are defined in the figure captions and within the text.

This point has been addressed.

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript has been properly revised.