Relationships of Serum 25-Hydroxyvitamin D Concentrations, Diabetes, Genetic Susceptibility, and New-Onset Chronic Kidney Disease

Vitamin D, a fat-soluble nutrient, plays a major role in bone health (1). Recently, the high prevalence of vitamin D deficiency (2,3) and its possible associations with multiple health outcomes have attracted much attention (4).

Chronic kidney disease (CKD), with a global prevalence of 9.1% in 2017, is an important risk factor for cardiovascular disease and premature death and has become a major public health problem worldwide (5,6). Two previous studies have found that insufficient serum 25-hydroxyvitamin D (25OHD), a major circulating form of vitamin D, was associated with the change in estimated glomerular filtration rate (eGFR) (7) or development of end-stage kidney disease (ESKD) (8). However, both studies (7,8) included a subset of patients with CKD at baseline, and due to the limited sample size, neither study could draw accurate conclusions about the relationship between 25OHD concentration and new-onset CKD in the population without CKD at baseline. Therefore, to date, the prospective relation of serum 25OHD with new-onset CKD remains uncertain.

Furthermore, previous studies have shown that glucose metabolism disorders significantly affect vitamin D metabolism (1,9). As such, diabetes and insulin resistance were associated with a higher prevalence of vitamin D deficiency (9,10). Moreover, diabetes is one of the most important risk factors for CKD (5,11), and diabetes and vitamin D deficiency seem to share some common pathogenic mechanisms in the development of CKD, such as the activation of the renin-angiotensin-aldosterone system (RAAS) (12,13). Therefore, we speculated that diabetes status might influence the association between 25OHD concentration and new-onset CKD. However, no studies have evaluated this hypothesis.

In addition, genetic factors could significantly affect kidney function (14). However, whether the genetic predisposition may modify the relation of serum 25OHD with the risk of CKD has not been fully examined in previous studies.

To fill the aforementioned knowledge gaps, using data of 348,243 adults from the UK Biobank, the current study aimed to examine the prospective association of serum 25OHD with the risk of new-onset CKD and further explore the modification effect of diabetes status on the association. Moreover, we also evaluated whether the genetic risk of kidney function may modify the association between serum 25OHD and new-onset CKD.

The UK Biobank is a nationwide, prospective, population-based cohort, consisting of 500,000 residents in the United Kingdom, aged between 40 and 69 years at the time of recruitment from 2006 to 2010. Participants were asked to complete a touch screen questionnaire, a face-to-face nurse interview, and a series of physical measurements, as well as provide biological samples for genotype and biomarker analysis. Details of the study design have been described in the official website (https://www.ukbiobank.ac.uk/) and the previous studies (15). All participants provided written informed consent, and the study was approved by the North West Multi-Center Research Ethics Committee (Manchester, U.K.).

Data of 502,414 participants in the UK Biobank were available in the present analysis. Participants with missing data on serum 25OHD, eGFR, urinary albumin-to-creatinine ratio (ACR), glucose, or glycated hemoglobin (HbA_1c; n = 127,207), or those with CKD (hospital-diagnosed CKD, n = 751; eGFR <60 mL/min/1.73 m², or ACR ≥30 mg/g, n = 26,092; self-reported CKD, n = 121) at baseline (n = 26,964) were excluded. After that, 348,243 participants were included in the final analysis (Supplementary Fig. 1).

Serum concentration of 25OHD (nmol/L) was measured by chemiluminescent immunoassay method (DiaSorin Liaison XL, DiaSorin S.p.A.) with an analytical range from 10 nmol/L to 375 nmol/L. Quality control was conducted by the UK Biobank. For internal quality control, coefficient variation of the sample level in low, median, and high for each biomarker was provided. The average within-laboratory total coefficient variation from serum 25-OHD was 5.04–6.14%, and its external quality (using Randox International Quality Assessment Scheme [RIQAS] Immunoassay Specialty 1 scheme) assurance was 100%.

Serum creatinine (µmol/L), urinary creatinine (µmol/L), and glucose (mmol/L) were both detected using an enzymatic method (Beckman Coulter AU5800, Buckinghamshire, U.K.), and the eGFR was calculated based on serum creatinine using the Chronic Kidney Disease Epidemiology Collaboration equation (16). Urinary albumin (mg/L) was measured by an immunoturbidimetric method (Randox Biosciences, Belfast, U.K.), and the urinary ACR (mg/g) was calculated. The HbA_1c (mmol/mol) assay was performed using a high-performance liquid chromatography method (Bio-Rad Variant II Turbo analyzer, Bio-Rad Laboratories). For external quality assurance, the Weqas Mainline Chemistry scheme was used for serum creatinine and glucose, the Weqas General Urine Chemistry External Quality Assessment scheme was used for urinary albumin and urinary creatinine, and the Glycated Haemoglobins External Quality Assessment scheme was used for HbA_1c. The detailed information on the biochemistry assay has been reported previously (https://biobank.ndph.ox.ac.uk/showcase/field.cgi?id=30890).

Details on the genotype array coverage, laboratory processes, and quality control checks are provided in the official website (https://www.ukbiobank.ac.uk/enable-your-research/about-our-data/genetic-data). The genetic risks for CKD, ACR, and serum 25OHD were developed by a genetic risk score (GRS) for eGFR, ACR, and 25OHD, respectively. The CKD-GRS, ACR-GRS, and 25OHD-GRS were calculated using 263 single nucleotide polymorphisms (SNPs) of eGFR (17), 59 SNPs of ACR (18) and 35 SNPs of 25OHD (19), respectively. The number of risk alleles was multiplied by a weighted risk estimate (natural logarithm of the published odds ratio) for each genetic variant. The GRS was summed across all variants (20). A higher CKD-GRS or a lower ACR-GRS indicated a lower genetic predisposition to CKD, and a higher 25OHD-GRS indicated a higher genetic predisposition to serum 25OHD concentrations.

Diabetes status at baseline was defined as history of diabetes, glucose ≥7.0 mmol/L, or HbA_1c ≥6.5% (21,22).

Further information on demographic, lifestyle factors, and medications use were collected through a touch screen questionnaire. The following covariates were included: age, sex, ethnicity, centers, height, weight, education levels, household income, Townsend deprivation index, smoking status, alcohol consumption status, diet, physical activity, self-reported diseases (hypertension, hypercholesterolemia), and use of drugs (insulin, antihypertensive drugs, lipid-lowering drugs). BMI was calculated by dividing weight (kg) by square of standing height (m). Moderate or vigorous physical activity was defined as 150 min/week moderate intensity or 75 min/week vigorous intensity or combination, according to global recommendation on physical activity for cardiovascular health (23). The healthy diet score consisted of a total of 10 dietary goals for ideal cardiovascular health (23,24). Participants who achieved 1 of 10 dietary goals, got 1 point. The range of healthy diet score is 0–10.

New-onset CKD was ascertained from ICD-10 codes N12, N13.1, N13.2, N18.0, N18.3, N18.4, N18.5, N18.8, and N18.9 as well as International Classification of Diseases-Ninth Revision code 585, and Office of Population Censuses and Surveys Classification of Interventions and Procedures, version 4 (OPCS-4) M01, which were recorded across all participants’ hospital inpatient records in the primary or secondary position.

Baseline characteristics of study participants are summarized as mean (SD) or proportions for continuous and categorical variables, respectively. Comparisons of characteristics by serum 25OHD categories (severely deficient: <25 nmol/L, moderately deficient: 25–50 nmol/L, insufficient and above: ≥50 nmol/L) (25) among participants with and without diabetes were performed by χ² tests for categorical variables and ANOVA tests for continuous variables.

Variables that are known to be traditional or suspected risk factors for CKD or variables that showed significant differences among different serum 25OHD categories were chosen as the covariates in this analysis. Cox proportional hazards models were used to estimate the relationship of baseline serum 25OHD with new-onset CKD in participants with and without diabetes. Models included the adjustments for age, sex (male, female), ethnicity (White, Black, Asian, others), assessment centers, BMI, education levels, household income (<18,000, 18,000–30,999, 31,000–51,999, 52,000–100,000, >100,000 £/year), Townsend deprivation index, smoking status (never, previous, current), alcohol consumption (daily or almost daily, 1–4 times a week, 1–3 times a month, never or special occasions only), healthy diet scores, supplement of vitamin D, eGFR, ACR, C-reactive protein, self-reported hypertension, self-reported hypercholesterolemia, use of antihypertensive drugs, use of cholesterol-lowering drugs, and blood collection seasons (spring, summer, fall, winter). Potential modifying effects of diabetes on the association between serum 25OHD on new-onset CKD was examined by interaction test with the use of likelihood ratio testing. Restricted cubic splines with four knots (20%, 40%, 60%, and 80% of serum 25OHD) were also used to visualize possible relationships of serum 25OHD with new-onset CKD among patients with and without diabetes.

Moreover, stratified analyses were performed to explore potential modifications of CKD-GRS, ACR-GRS, and other covariates on the association of 25OHD and new-onset CKD among participants with diabetes.

We conducted all analysis using R 4.0.1 software. For all statistical tests, the two-sided P value <0.05 was considered to be statistically significant.

Data, analytic methods, and study materials that support the findings of this study are available from the corresponding authors on request.

The final analysis included 348,243 participants (Supplementary Fig. 1). Among those, the mean age was 56.3 (SD 8.1) years, and 164,021 (47.1%) were men. The median concentration of serum 25OHD was 47.1 nmol/L (interquartile range 32.6–62.5). A total of 45,728 (13.1%), 145,764 (41.9%), and 156,751 (45.0%) participants had a 25OHD concentration <25 nmol/L (severely deficient group), 25 to <50 nmol/L (moderately deficient group), and ≥50 nmol/L (insufficient and above), respectively.

The prevalence of diabetes was 6.5%. The serum 25OHD concentration in participants with diabetes was significantly lower than in those without diabetes (median 41.6 nmol/L [interquartile range 28.2–57.4] vs. median 47.5 nmol/L [interquartile range 33.0–62.9]; P < 0.001). Moreover, as indicated in Table 1, in participants both with and without diabetes, those with higher 25OHD concentrations were older, thinner, more physical active, and more likely to be nonsmokers, have a healthier diet, and take a vitamin D supplement and lipid-lowering drugs; had higher household incomes and lower glucose, HbA_1c, eGFR, C-reactive protein, and education levels, and consumed more alcohol.

Moreover, in the multivariate regress models, there was a significant inverse association between serum 25OHD and HbA_1c levels among the total population and the participants with and without diabetes (all P values <0.001). In addition, a stronger inverse relation of serum 25OHD with HbA_1c was found in participants with diabetes versus those without diabetes (P for interaction <0.001) (Supplementary Table 1).

During a median follow-up of 12.1 years (interquartile range 11.4–12.8 years), 9,344 cases (2.7%) of new-onset CKD were documented in the total participants. Of those, 7,500 (2.3%) were in participants without diabetes and 1,844 (8.2%) were in those with diabetes.

Overall, there was a significant inverse association between baseline serum 25OHD and new-onset CKD in participants with diabetes (per SD increment, adjusted hazard ratio [HR] 0.91; 95% CI 0.86–0.96), but not in those without diabetes (per SD increment, adjusted HR 0.98; 95% CI 0.96–1.01; P-interaction for serum 25OHD and diabetes = 0.004) (Table 2 and Fig. 1).

Accordingly, among patients with diabetes, compared with those baseline 25OHD <25 nmol/L, a significantly lower risk of new-onset CKD was found in participants with 25OHD ≥50 nmol/L (adjusted HR 0.77; 95% CI 0.67–0.89) (Table 2).

Including those with self-reported CKD at baseline in the analysis (Supplementary Table 2), or further adjustments for CKD-GRS and ACR-GRS (Supplementary Table 3), or using the newer Chronic Kidney Disease Epidemiology Collaboration creatinine equation (2021) (26) to calculate eGFR (Supplementary Table 4), or excluding education and income levels in the adjustments (Supplementary Table 5), did not substantially change the results.

Prevalent CKD was defined as eGFR <60 mL/min/1.73 m², or ACR ≥30 mg/g, or diagnosed by physician, or self-reported CKD. Overall, the association between 25OHD-GRS and CKD seemed to be linear in participants with diabetes (P for nonlinearity = 0.108) or without diabetes (P for nonlinearity = 0.194). When 25OHD-GRS was assessed as quartiles, in participants with diabetes, compared with those in the first quartile, a significant lower prevalence of CKD was found in participants in the 3–4 quartiles (adjusted odds ratio 0.91; 95% CI 0.83–0.99). However, in participants without diabetes, there was no significant association between 25OHD-GRS and prevalent CKD (Supplementary Fig. 2 and Supplementary Table 6).

Stratified analyses were performed to explore potential modifying factors on the association of 25OHD (<25, 25 to <50, ≥50 nmol/L) with new-onset CKD in patients with diabetes. None the following factors, including age, sex, ethnicity, BMI, physical activity, healthy diet score, eGFR, ACR, C-reactive protein, use of antihypertensive drugs, glucose, average time spent outdoors, genetic risk of CKD, and ACR, significantly modified the association between baseline serum 25OHD and new-onset CKD (all P values for interactions >0.05) (Fig. 2).

Of note, according to categories of genetic risk of CKD, compared with participants in the low genetic risk of CKD, significantly higher risks of new-onset CKD were found in those in the high genetic risk of CKD in both participants with and without diabetes and among those with different serum 25OHD concentrations (<25, 25 to <50, and ≥50 nmol/L) (all P values for trend across categories <0.001) (Supplementary Table 7).

Findings from this large-scale, population-based cohort suggested that independent of the genetic predisposition of CKD and ACR, there was an inverse relationship of baseline serum 25OHD with new-onset CKD in participants with diabetes, but not in participants without diabetes.

The Cardiovascular Health Study (CHS) reported that a lower 25OHD concentration was associated with a higher risk of rapid eGFR loss in 1,705 older adults (7). Another cohort result from the Third National Health and Nutrition Examination Survey (8) found a marginally significant association between 25OHD deficiency and a higher risk of ESKD. Of note, both of the studies (7,8) included some patients with CKD at baseline, and neither study was able to draw accurate conclusions about the association of 25OHD concentrations with new-onset kidney end points in a population without CKD. Therefore, to date, the prospective association between 25OHD concentrations and new-onset CKD remains uncertain. Thus, the present analysis used the largest sample to date of United Kingdom adults over recent decades, with a duration >10 years, used serum 25OHD as both a continuous and a categorical variable in the evaluation of the longitudinal relationship between baseline serum 25OHD and risk of new-onset CKD, and has provided some new insights in this field.

Our study showed that among participants with diabetes, a significantly lower risk of new-onset CKD was found in participants with serum 25OHD ≥50 nmol/L compared with those with serum 25OHD <25 nmol/L. This might be plausible in mechanisms. First, activated vitamin D, 1,25-dihydroxyvitamin D₃ (1,25[OH]2D₃) could regulate the RAAS by suppressing renin biosynthesis (12,27). Second, in rat models, treatment with 1,25(OH)2D₃ counteracted effects on glomerular podocyte injury (28). Third, the renoprotective effects from vitamin D receptor could be reflected by modulating endothelial function (29), inhibiting mesangial cell proliferation, reduction in the expression transforming growth factor-β1 protein (30), and existing anti-inflammatory effects against the NLRP3/interleukin-1β axis in mouse renal tubular epithelial cells (31). These mechanisms may possibly explain the protective effect of higher vitamin D status on the kidney in participants with diabetes.

However, there was no significant association between serum 25OHD and new-onset CKD in participants without diabetes. The possible explanations may include: First, a series of different types of studies have reported that lower serum 25OHD was associated with insulin resistance and increased risk of diabetes (10,32,33). Accordingly, both previous studies (7,8) and our present study have shown that participants with diabetes had significantly lower concentrations of 25OHD and were therefore more likely to develop kidney impairment caused by low vitamin D, thereby benefiting from higher concentrations of vitamin D.

Second, patients with diabetes are characterized by more activation of the RAAS, which may lead to renal hypoxia and maladaptive changes in blood flow, metabolism, and vasculitis of nephrology (13,34). Therefore, RAAS inhibition, the main beneficial effect of vitamin D on the kidneys, may be more pronounced in patients with diabetes (12,27,35).

Third, the 25OHD concentrations in the UK Biobank participants limited the current study to examine the possible increased CKD risk associated with very low concentration of serum 25OHD in participants without diabetes. In fact, our study really showed a weaker inverse association between serum 25OHD and HbA_1c levels in those with diabetes. Overall, more studies are needed to confirm our findings and further explore the underlying mechanisms.

In addition, our results showed that there was no statistically significant interaction between baseline serum 25OHD and genetic susceptibility to CKD or ACR on the risk of new-onset CKD in patients with diabetes. Therefore, the relationship of serum 25OHD with new-onset CKD might be independent of genetic risks, suggesting that optimal serum 25OHD concentration is important for primary prevention of CKD regardless of genetic risks of CKD.

The possible limitations still need to be mentioned. First, although a broad number of covariates were included in the adjustments, unknown confounding from unmeasured factors could not be completely excluded.

Second, causality could not be determined by the nature of the observational cohort design, and intervention studies are needed to further assess and confirm the present findings.

Third, serum 25OHD was only assessed at baseline in the UK Biobank, and repeat measurements of 25OHD data would have allowed a more accurate assessment of its effect on kidney disease. However, a previous population-based study in Norwegian for the tracking of serum 25OHD levels during 14 years showed that most subjects with low serum 25OHD concentrations were unlikely to have substantial improvements in their 25OHD concentrations over time, and therefore provided some supports for using a single measurement of circulation 25OHD in epidemiologic studies to predict future health outcome (36).

Fourth, since the outcomes were captured from hospital inpatient records, some undiagnosed cases may have been missed. However, there was no evidence that serum 25OHD concentration could significantly modify the distribution of missed cases. Future research with more frequent assessments of eGFR and ACR are needed to confirm out findings.

Fifth, information on betel consumption was not recorded in the UK Biobank; therefore, we cannot evaluate the modifying effect of betel consumption on the findings.

Finally, the participants in the UK Biobank study were predominantly of European descent. Although there was no significant interaction between ethnicity (White, Asian, Black) and serum 25OHD on new-onset CKD among patients with diabetes, due to the relatively small sample size of Asian and Black participants, the generalizations of the results to other races and populations remains to be determined. Overall, due to these limitations, our findings should be further confirmed in more studies.

In conclusion, our analysis among 348,243 adults from the UK Biobank revealed that there was an inverse association between serum 25OHD and new-onset CKD in participants with diabetes. However, there was no significant association between serum 25OHD and new-onset CKD in participants without diabetes. Our findings, if further confirmed, underscore the importance of maintaining adequate vitamin D concentrations for primary prevention of CKD in those with diabetes.

Acknowledgments. The authors thank the UK Biobank participants. This research used data from the UK Biobank Resource under Application Number 73201.

Funding. The study was supported by the National Key R&D Program of China (2022YFC2009600) and the National Natural Science Foundation of China (81973133, 81730019).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. C.Z., P.H., and Z.Y. performed the data management and statistical analyses. C.Z., Z.Y., and X.Q. conducted the research. C.Z., J.N., and X.Q. designed the research. C.Z. and X.Q. wrote the draft. C.Z., P.H., Z.Y., Yu.Z., Ya.Z., S.Y., Q.W., M.L., J.N., and X.Q. revised and approved the final manuscript. X.Q. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.