Review Reports

Reviewer 1

The manuscript entitled "Vitamin D Levels in a Clinical Population in Germany: Associations with Patient Characteristics and Season" reports the results of a retrospective cohort study performed on rheumatology patients focusing on vitamin D status characterized using serum 25-OH vitamin D2 and 25-OH vitamin D3 concentrations. The manuscript presents the seasonal fluctuation and the association between vitamin D levels and age, BMI, and eGFR.

The rationale behind the aims of this study are not clearly presented. The authors claim that vitamin D deficiency is common and that rheumatology patients may be at particular risk, but it does not explicitly articulate what is unknown (specific references) about vitamin D status in this population or how this study will fill that gap. The transition from national prevalence data (lines 69–77) to your clinic cohort is helpful, but it would be stronger if you briefly highlighted how clinic populations may differ (e.g., higher glucocorticoid exposure, reduced mobility/sun exposure) and why national averages cannot be assumed to apply.

We thank the reviewer for the insightful comments. We have modified the Introduction to emphasize that there is currently limited literature on vitamin D levels in a “real world” population of patients presenting to a rheumatology clinic (see following reply for modified sentences).

Conversely, because rheumatology patients typically receive diverse pharmacologic regimens, often take vitamin supplements, and frequently experience chronic inflammation, numerous confounders affect their vitamin D levels. Consequently, the findings from this cohort cannot be generalized to the wider population—despite the authors’ stated aim to do so (lines 87-89). Moreover, lines 105-107 reiterate the aims of this study, but this is in discordance with the lines 87-89.

We completely agree that these factors will differ in the population we evaluated and have now tried to clarify this further. In particular, we note that: “In addition to sun exposure, vitamin D levels are influenced by multiple additional factors, including age, sex, use of vitamin D supplements, BMI, skin pigmentation, meta-bolic differences, sunscreen use, and chronic illnesses, particularly liver or kidney diseases [11,14]. Many of these factors are likely to be different in a clinical population compared with the overall population. In particular, medications, pre-existing illnesses, reduced mobility/sun exposure, and autoimmune/inflammatory manifestations are likely to vary between a population-wide and clinical sample. To gain further insights into vitamin D status in the rheumatology clinic population, we conducted a retrospective analysis of vitamin D levels in a large cohort of patients seen at a rheumatology center for the evaluation of rheumatologic complaints.”

In addition, in order to avoid confusion we have deleted the objective presented in the Methods section and reworded the objective at the end of the Introduction to: “Although several studies have been published on vitamin D supplementation in defined rheumatology populations, such as patients with rheumatoid arthritis [15], there is limited information on vitamin D levels in patients who were referred for the evaluation of rheumatologic complaints during routine clinical care. The objective of the study was to fill this gap in the literature by providing contemporary data on the vitamin D status of a large clinical population presenting to a rheumatology clinic in Germany and to investigate associations between vitamin D levels, patient variables, and season.”

In addition, the Introduction should cite the most widely accepted guidelines for deficiency threshold of <20 ng/mL. Furthermore, your decision of not using  “insufficiency” (20–30 ng/mL) and using custom cut-off values must be argued and supportive information be added. Stating all thresholds at first mention—and acknowledging the ongoing debate around them—would make later results easier to interpret.

We agree that the exact thresholds are an ongoing area of disagreement. We discuss this in section 2.3 and provide a reference that covers this complex topic in detail, as a comprehensive discussion of this point is beyond the scope of this manuscript. Table 1 is provided so that the reader can easily refer to the categories used in this study. The Limitations paragraph in the Discussion also mentions variability in vitamin D cut-offs and the possible impact this might have on interpreting our results.

Methods: The handling of missing data was not described, although, it is known that retrospective analysis of laboratory datasets often confronts with incomplete vitamin-D or covariate data.

As noted in section 2.1, only patients with vitamin D assay results were included in the study. The structure of the database requires entries for the other covariates presented in this study, so we are happy to state that there were no missing data for the characteristics presented in Table 2.

The authors presumed normally distributed vitamin-D levels (t-test, Pearson r, ANOVA) but does not report normality test results. Vitamin-D data are frequently right-skewed. Report Shapiro–Wilk or Kolmogorov–Smirnov tests.

We have followed the reviewer’s suggestion and now report the results of the normality test within the Methods section of the manuscript. The performed Kolmogorov-Smirnov test indicated a rejection of normality for our vitamin D data/levels (statistic: 0.096; df 4979; p < 0.001). However, we believe that the application of the t-test remains appropriate in our context due to the interval-scaled nature of the data and the exceptionally large sample size. Consequently, switching to a nonparametric alternative (e.g., Mann-Whitney U-test) is not considered necessary.

Specifically, with a sample size of almost 5,000 cases, t-tests and ANOVA analyses are generally robust to potential violations of the normality assumption, primarily due to the Central Limit Theorem (CLT). The pertinent distribution for the t-test is not the distribution of the raw data itself, but rather the sampling distribution of the sample means (or the differences in means). According to the CLT, this distribution approaches normality with large sample sizes.

Furthermore, it is common for normality tests, such as the Kolmogorov-Smirnov or Shapiro-Wilk test, to yield statistically significant results in very large samples, even when deviations from perfect normality are clinically or practically irrelevant. This high sensitivity allows these tests to detect even minor departures from a perfectly normal distribution. Therefore, a significant result from the Kolmogorov-Smirnov test with N=4979 indicates that the data are not perfectly normally distributed, but it does not invalidate the application of the t-test. Correspondent arguments regarding robustness and the Central Limit Theorem similarly apply to ANOVA analyses.

The post-hoc test used for monthly (Bonferroni post-hoc) comparison must be mentioned here.

We thank the reviewer for catching this omission and have added the information to Section 2.4.

Results: First of all, your statement regarding the analysis of associations between diagnosis and vitamin D levels can be interpreted as data splicing. Beware of this practice.

We agree that data splicing should be avoided and have tried to modify this section to correct this.

Discussion: The results regarding seasonal variation are overinterpreted. The authors should focus on other influencing factors found as significant, such as sex differences and BMI.

Given the large sample size and the 2.7-fold difference in vitamin D levels between the lowest and highest months, we believe it is justified to give seasonal variation appropriate attention in the presentation and interpretation of the results. As another reviewer noted, the r values (correlation) of our other observations (sex, BMI) are quite low, so we would prefer not to focus further on those data other than the remarks already in the manuscript.

The authors should be aware that comparisons with general population are not relevant. In addition, specific information on assay harmonization and potential selection bias (clinic population vs. community sample) must be added.

We appreciate this point and have added the following sentence to the end of the 2^nd paragraph in the Discussion: “It should be noted, however, that comparisons between community samples and other clinical populations may be influenced by selection bias and differences in vitamin D assays.”

The authors acknowledge the limitations of the current study, such as the absence of nutrition/supplement data, but they attribute lower deficiency rates partly to possible supplement use in their clinic cohort. In addition, other possible confounders (sun-exposure behaviour, disease activity, glucocorticoid dose, ethnicity/skin tone) should also be added. These may partly explain demographic differences in vitamin D concentrations.

We appreciate this comment and have added this information to the first sentence in the third paragraph.

Overall, this manuscript focuses on rheumatology patients, so readers may expect analysis of vitamin D status vs. disease type (RA, SLE, etc.) or glucocorticoid dose. Without these data I cannot support the publication of this manuscript.

We fully agree that an analysis of vitamin D levels by disease subtype has the potential to be very interesting and these evaluations are ongoing. However, because of the multiple different types of specific rheumatology diseases as well as the overall distinction between inflammatory and non-inflammatory disorders, these analyses will generate a large quantity of data. Accordingly, we made the decision to publish the overall epidemiologic results separately and to create another publication evaluating vitamin D by disease type and category, thereby allowing us to build on the associations observed in the overall clinic cohort. As the vast majority of patients included in the present study had no inflammatory condition, we are convinced that comparisons with the general population are justifiable.

We note our intention to publish information on specific diagnoses and their associations with vitamin D levels separately in Section 3.1 (line above Table 2).

Reviewer 2

This is an interesting research article with quite adequate novelty. Some points should be addressed.

We thank the reviewer for their favorable comments and have tried to address the points raised.

•  The authors should report in the title and the abstract that their study population are patients with rheumatoid diseases.

These are initial consultations with patients who were referred for the evaluation of rheumatologic complaints. Only the minority of individuals in our database had an inflammatory rheumatologic condition. We have therefore changed the title to “Vitamin D Levels in Patients Presenting to a Rheumatology Clinic in Germany: Associations with Patient Characteristics and Season.” We have also added “presenting to a rheumatology clinic” to the abstract.

•  The sentence in lines 58-63 "Although vitamin D supplementation ..... and known seasonal variations in vitamin D levels [1,8]." is quite complex and needs revisions to be more easily readable.

The reviewer has raised an excellent point and we have rewritten this sentence accordingly.

•  A paragraph describing the rheumatoid diseases concerning its epidemiological data, its pathophysiological mechanisms and its risk factor should be added in the Introduction section.

We thank the reviewer for this suggestion and have added some information on rheumatologic disorders to the introduction. However, we prefer not to expand on this topic in detail as we are planning to publish an analysis of vitamin D levels by specific rheumatologic condition separately.

•  The authors should emphasize at the end of the introduction section the literature gap that their study aims to cover.

We appreciate this insight and now specify that “Although several studies have been published on vitamin D supplementation in defined rheumatology populations, such as patients with rheumatoid arthritis [15], there is limited information on vitamin D levels in patients who were referred for the evaluation of rheumatologic complaints in routine clinical care.”

• In the statistical analysis section, the normality distribution test should be reported.

We have followed the reviewer’s suggestion and now report the results of the normality test in Section 2.4 (Statistical analysis). The performed Kolmogorov-Smirnov test indicated a rejection of normality for our vitamin D data/levels (statistic: 0.096; df 4979; p < 0.001).

• The size of the letters of the words of Figure 1 should be increased.

We agree and have increased the size of the lettering as requested.

• The resolution of Figure 2 should be improved

We have improved the figure resolution.

•The 3rd and 5th paragraphs of the discussion section are too long and should be split into two smaller paragraphs each of them.

We thank the reviewer for this observation and have divided the paragraphs as suggested.

Reviewer 3

1. The study relies solely on univariate statistics (t-tests and Pearson correlations). Without multivariable regression, potential confounding (e.g., age-BMI interactions) is not controlled. Please include a multiple linear regression (or logistic regression for deficiency as outcome) to model vitamin D levels with simultaneous adjustment for age, sex, BMI, renal function, and season.

We agree that multivariable analysis is important for comprehensively understanding the associations with vitamin D levels, controlling for potential confounding factors.

As suggested, we have now performed a multiple linear regression analysis with vitamin D levels as the dependent variable. This model simultaneously adjusts for age, sex, body mass index (BMI), renal function (eGFR categories), and season (categorized by month).
The regression analysis revealed that BMI and GFR are negatively associated with vitamin D levels, while being male also predicts lower values. Several summer and autumn months (May–October) are associated with significantly higher vitamin D levels compared to January. Overall, the model explains a small but significant portion of the variance in vitamin D. The regression model explains about 7.5% of the variance in vitamin D levels (R² = 0.075, Adjusted R² = 0.072), indicating that the included predictors have a modest but statistically significant effect

We added the following to the Results section of the manuscript (new section 3.5):
“To comprehensively assess the relationships between vitamin D levels and key variables, we conducted a multiple linear regression analysis with serum vitamin D levels as the dependent variable and the independent variables of age, BMI, eGFR, sex, and survey month. The regression analysis confirmed that BMI (p<0.001) and eGFR (p<0.001) are negatively associated with vitamin D levels and that male sex (p<0.001) also predicted lower values. Several summer and autumn months were associated with significantly higher vitamin D levels compared with January (May–October; all p≤0.002 with most p<0.001). Overall, the model explained a small but significant portion (7.5%; R2=0.075) of the variance in vitamin D levels (F=26.64; df=15; p<0.001).”

2. While the study mentions that 72% had non-inflammatory diagnoses, it omits any subgroup analysis by disease type (e.g., RA vs OA vs fibromyalgia). Please perform a brief stratification or summary data by diagnostic groups, in order to enhance the interpretability, especially in a rheumatology-focused population.

We fully agree that an analysis of vitamin D levels by disease subtype has the potential to be very interesting and these evaluations are ongoing. However, because of the multiple different types of specific rheumatology diseases as well as the overall distinction between inflammatory and non-inflammatory disorders, these analyses will generate a large quantity of data. Accordingly, we made the decision to publish the overall epidemiologic results separately and to create another publication evaluating vitamin D by disease type and category, thereby allowing us to build on the associations observed in the overall clinic cohort.

Our intention to publish information on specific diagnoses and their associations with vitamin D levels separately is noted in Section 3.1 (line above Table 2). 

3. The absence of data on supplementation status is a major limitation since it could significantly bias the distribution of vitamin D levels. At minimum, this limitation should be discussed more prominently in the Discussion. If any proxy data (e.g., prescription records) are available, even limited reporting would be helpful.

We agree that data on vitamin D supplementation would have been desirable. However, we have observed that few epidemiological studies have solid data on supplementation, probably because patients are not asked about their practices unless a problem is detected. In Germany most forms of vitamin D supplementation are sold over the counter, so electronic health record data and prescription records do not contain these data. We have modified the 2^nd paragraph to read: “Our database did not contain information on vitamin D supplementation, which is an important limitation of this study. However, it is possible that a clinical population would have higher levels of vitamin D usage than the overall general population.” This point is also raised in the Limitations paragraph.

4. The range of vitamin D levels extends up to 240 ng/mL, which is likely to represent analytical or data entry errors or unreported supplementation/toxicity. Please conduct a sensitivity analysis excluding extreme outliers (e.g., >100 ng/mL) and comment on the implications.

We appreciate the reviewer's comment regarding the range of vitamin D levels and the presence of extreme values. As suggested, we performed a sensitivity analysis by excluding participants with vitamin D levels exceeding 100 ng/mL.

In total, 18 cases (only 0.4% of the total sample) had vitamin D levels >100 ng/mL and were excluded for this sensitivity analysis. Re-running all primary statistical analyses (e.g., t-tests, ANOVAs) on this filtered dataset revealed no substantive changes in our main findings. Specifically, all previously significant effects remained significant, and their effect sizes were comparable to the original analysis. This consistency underscores the robustness of our conclusions, even in the presence of a small number of extremely high vitamin D values in the full dataset.

We have added a comment on this as the last sentence of section 3.3.

5. Many of the observed correlations (e.g., r < 0.1 for age or eGFR) are statistically significant due to sample size but not clinically meaningful. Please emphasize effect sizes and clinical interpretation over p-values in the Results and Discussion sections.

We agree with this comment and direct the reviewer’s attention to the first sentence of paragraph 3 in the Discussion: “In our study, correlations between vitamin D levels and patient characteristics were weak (r <0.2),...”  Based on the reviewer’s suggestion, we have now added: “The statistical significance observed for some results may be due to sample size rather than to a clinically relevant association.”

6. Some portions of the manuscript reiterate well-established background information (e.g., vitamin D metabolism) in excessive detail for the target audience. Consider shortening the Introduction to better focus on the rationale for studying this specific clinical population.

We appreciate this observation. We have removed some of the vitamin D metabolic information from the Introduction and added information on rheumatologic disorders and the relevance of vitamin D levels in this population.

7. Inconsistencies in vitamin D status terminology (e.g., "sufficient" vs. "optimal") could confuse readers. These should be clearly defined and used consistently.

We concur that the terminology applied to vitamin D levels can be confusing, which is why we have supplied Table 1 for readers to refer to. We have added more mentions of Table 1 throughout the text and modified the part of the Discussion on vitamin D guidelines to refer to “target” levels rather than “optimal” levels to help remove some of this confusion.

8. Consider increasing the font size in figures and adding colorblind-friendly contrasts in Figure 3’.

We have increased the font size as suggested and changed the colors in Figure 3 to align with those recommended by the Netherlands Cancer Institute for color-blind-friendly graphs.

9. Most references are recent and relevant; however, ensure the guidelines cited (e.g., German Osteology Association, Endocrine Society) are accurately contextualized with respect to their target populations.

We agree that the references mentioned by the reviewer refer to very specific populations and can confirm that we have accurately contextualized these references within the guidelines provided by these organizations.