Factors Associated with Serum Vitamin D Metabolites and Vitamin D Metabolite Ratios in Premenopausal Women

The most representative indicator of vitamin D status in clinical practice is 25(OH)D3, but new biomarkers could improve the assessment of vitamin D status and metabolism. The objective of this study is to investigate the association of serum vitamin D metabolites and vitamin D metabolite ratios (VMRs) with potentially influential factors in premenopausal women. This is a cross-sectional study based on 1422 women, aged 39–50, recruited from a Madrid Medical Diagnostic Center. Participants answered an epidemiological and a food frequency questionnaire. Serum vitamin D metabolites were determined using an SPE–LC–MS/MS platform. The association between participant’s characteristics, vitamin D metabolites, and VMRs was quantified by multiple linear regression models. Mean 25(OH)D3 concentration was 49.2 + 18.9 nmol/L, with greater deficits among obese, nulliparous, dark-skinned women, and with less sun exposure. A lower R2 ratio (1,25(OH)2D3/25(OH)D3) and a higher R4 (24,25(OH)2D3/1,25(OH)2D3) were observed in nulliparous women, with high sun exposure, and those with low caloric intake or high consumption of calcium, vitamin D supplements, or alcohol. Nulliparous women had lower R1 (25(OH)D3/Vit D3) and R3 (24,25(OH)2D3/25(OH)D3), and older women showed lower R3 and R4. Vitamin D status modified the association of the VMRs with seasons. VMRs can be complementary indicators of vitamin D status and its endogenous metabolism, and reveal the influence of certain individual characteristics on the expression of hydroxylase enzymes.


Introduction
Vitamin D has been recently hypothesized as a potentially modifiable factor that could reduce the risk of several diseases, such as cardiovascular diseases, diabetes mellitus, multiple sclerosis [1], mental and autoimmune disorders [2], or some types of neoplasms (such as breast cancer) [3]. The US Endocrine Society considers vitamin D sufficiency when serum levels exceed 75 nmol/L [4], and the Institute of Medicine set up a cutoff of 50 nmol/L [5]. According to the last threshold, vitamin D deficiency (<50 nmol/L) has been estimated to affect around 40% of the European [6] and Spanish [7] population.
Vitamin D (calciferol) is mainly produced in the skin by the action of ultraviolet B (UVB) radiation from sunlight, which transforms 7-dehydrocholesterol into previtamin D 3 . This metabolite is considered biologically inactive until it undergoes two enzymatic hydroxylations: the first one in the liver, where previtamin D 3 is hydroxylated by the 25-hydroxylase (CYP2R1) to form 25-hydroxyvitamin D 3 (25(OH)D 3 ), and then in the kidney, where 25(OH)D 3 is converted to the biologically active hormone calcitriol or 1,25-dihydroxyvitamin D 3 (1,25(OH) 2 D 3 ). This second hydroxylation is mediated by 1αhydroxylase (CYP27B1), which is expressed mainly in the kidney, but also in extra-renal tissues such as breast cells, skin (keratinocytes), immune cells, and bone [4,8]. Vitamin D catabolism takes place in the kidney, where the 24-hydroxylase enzyme (CYP24A1) metabolizes 25(OH)D 3 to 24,25-dihydroxyvitamin D 3 (24,25(OH) 2 D 3 ), the main catabolic metabolite with some biological activity [9]. The crucial control point in vitamin D homeostasis is the renal production of 1,25(OH) 2 D 3 via 1α-hydroxylase. Calcitriol (1,25(OH) 2 D 3 ) can decrease its own production acting directly on the expression of the 1α-hydroxylase or indirectly decreasing parathyroid hormone (PTH) synthesis and, therefore, decreasing 1α-hydroxylase transcription. Rising concentrations of 1,25(OH) 2 D 3 also increase the expression of the phosphaturic factor, fibroblast growth factor 23 (FGF23), which suppresses the expression of 1α-hydroxylase in the kidney and causes up-regulation of CYP24A1 expression [9] (Figure 1). plasms (such as breast cancer) [3]. The US Endocrine Society considers vitamin D sufficiency when serum levels exceed 75 nmol/L [4], and the Institute of Medicine set up a cutoff of 50 nmol/L [5]. According to the last threshold, vitamin D deficiency (<50 nmol/L) has been estimated to affect around 40% of the European [6] and Spanish [7] population. Vitamin D (calciferol) is mainly produced in the skin by the action of ultraviolet B (UVB) radiation from sunlight, which transforms 7-dehydrocholesterol into previtamin D3. This metabolite is considered biologically inactive until it undergoes two enzymatic hydroxylations: the first one in the liver, where previtamin D3 is hydroxylated by the 25hydroxylase (CYP2R1) to form 25-hydroxyvitamin D3 (25(OH)D3), and then in the kidney, where 25(OH)D3 is converted to the biologically active hormone calcitriol or 1,25-dihydroxyvitamin D3 (1,25(OH)2D3). This second hydroxylation is mediated by 1α-hydroxylase (CYP27B1), which is expressed mainly in the kidney, but also in extra-renal tissues such as breast cells, skin (keratinocytes), immune cells, and bone [4,8]. Vitamin D catabolism takes place in the kidney, where the 24-hydroxylase enzyme (CYP24A1) metabolizes 25(OH)D3 to 24,25-dihydroxyvitamin D3 (24,25(OH)2D3), the main catabolic metabolite with some biological activity [9]. The crucial control point in vitamin D homeostasis is the renal production of 1,25(OH)2D3 via 1α-hydroxylase. Calcitriol (1,25(OH)2D3) can decrease its own production acting directly on the expression of the 1α-hydroxylase or indirectly decreasing parathyroid hormone (PTH) synthesis and, therefore, decreasing 1α-hydroxylase transcription. Rising concentrations of 1,25(OH)2D3 also increase the expression of the phosphaturic factor, fibroblast growth factor 23 (FGF23), which suppresses the expression of 1α-hydroxylase in the kidney and causes up-regulation of CYP24A1 expression [9] ( Figure 1). by the enzyme 1α-hydroxylase (CYP27B1), predominantly in the kidneys. Vitamin D catabolism is mainly driven by the enzyme 24-hydroxylase (CYP24A1), which metabolizes 25(OH)D3 to 24,25(OH)2D3. Vitamin D homeostasis depends on 1,25(OH)2D3 concentration, which can decrease its own production by directly inhibiting the expression of 1α-hydroxylase or indirectly, by decreasing the synthesis of the parathyroid hormone (↓PTH) or increasing the expression of the phosphaturic factor, fibroblast growth factor 23 (↑FGF23). VMR: Vitamin D Metabolite Ratio.
The most abundant circulating vitamin D metabolite is 25(OH)D3. Despite not being the biologically active form, it has been the most widely used indicator of vitamin D in most epidemiological studies, due in part to the lack of selective and sensitive methods  3 . Subsequently, 25(OH)D 3 is hydroxylated to the bioactive 1,25(OH) 2 D 3 by the enzyme 1α-hydroxylase (CYP27B1), predominantly in the kidneys. Vitamin D catabolism is mainly driven by the enzyme 24-hydroxylase (CYP24A1), which metabolizes 25(OH)D 3 to 24,25(OH) 2 D 3. Vitamin D homeostasis depends on 1,25(OH) 2 D 3 concentration, which can decrease its own production by directly inhibiting the expression of 1α-hydroxylase or indirectly, by decreasing the synthesis of the parathyroid hormone (↓PTH) or increasing the expression of the phosphaturic factor, fibroblast growth factor 23 (↑FGF23). VMR: Vitamin D Metabolite Ratio.
The most abundant circulating vitamin D metabolite is 25(OH)D 3 . Despite not being the biologically active form, it has been the most widely used indicator of vitamin D in most epidemiological studies, due in part to the lack of selective and sensitive methods for the determination of dihydroxymetabolites [10]. Despite its clinical relevance, the determination of vitamin D 3 metabolites continues to be a challenge, as it provides a more complete snapshot of vitamin D 3 status due to its physical and chemical properties (hydrophobic nature, thermal and UV instability, and similar structure). In addition, several limitations hinder the utility of 25(OH)D 3 in clinical practice, such as analytical aspects and interpretation of results [9]. In response to these limitations, new candidate biomarkers have been postulated that could improve the assessment of vitamin D status and metabolism [9]. Among these emerging candidates, vitamin D metabolite ratios (VMRs) are beginning to be used in recent studies [10][11][12][13][14][15][16], since they are not affected by the concentration of vitamin D binding proteins, are good indicators of the expression of hydroxylase enzymes, and could be useful to provide a better assessment of vitamin D status [17].
This study sought to evaluate potentially influential factors in serum levels of vitamin D 3 , 25(OH)D 3 , 1,25(OH) 2 D 3 , 24,25(OH) 2 D 3 , and the four VMRs directly connected by a substrate/product relationship (25(OH)D 3 /VitD 3 , 1,25(OH) 2 D 3 /25(OH)D 3 , 24,25(OH) 2 D 3 / 25(OH)D 3 , and 24,25(OH) 2 D 3 /1,25(OH) 2 D 3 ) in middle-aged women close to menopause, a period with higher risk of developing vitamin D deficiency [18]. The knowledge of the vitamin D status and its metabolism in this group of women, as well as the sociodemographic factors and lifestyles that are associated, is of great interest to prevent or mitigate bone loss and other conditions related to both menopause and vitamin D deficiency.

Study Population
Between June 2013 and May 2015, 1466 premenopausal women, aged 39 to 50, who worked at the Madrid City Council, were invited to participate in the DDM-Madrid study, aimed to assess the effect of vitamin D on mammographic density. These women were recruited in the Madrid Medical Diagnostic Center (Madrid Salud), where they attended to undergo their routine gynecological check-up. Participants were excluded if they were postmenopausal (at least 1 year without menstruation); were pregnant or breastfeeding; had breast cancer; or had undergone a mastectomy, breast reconstruction, or breast augmentation.

Recruitment and Data Collection
Women were invited to participate in the study by phone, when the selection criteria were verified. Overall participation rate was 88%. The day that each participating woman had her medical examination scheduled, the interviewers administered a standardized epidemiological questionnaire, drew a blood sample, and took anthropometric measurements (height, weight, and waist and hip circumference). The questionnaire collected sociodemographic variables, information on childhood and youth, personal and family medical history, gynecological and obstetric history, work history, skin type and sunbathing habits, sleep habits, tobacco and alcohol consumption, and physical activity. Participants also completed a validated [19] 117-item semi-quantitative food frequency questionnaire that included eating habits during the previous 12 months. Blood samples were centrifuged, aliquoted, and stored at −80 • C in the Carlos III Institute of Health Biobank. The DDM-Madrid study was conducted in accordance with the Declaration of Helsinki guidelines. All participants signed an informed consent, and the protocol was approved by the Ethics and Animal Welfare Committee of the Carlos III Institute of Health. Further details regarding the study design have been previously published [20,21].

Biochemical Analyses
The determination of vitamin D metabolites was carried out in the Metabolomics Unit of the University of Córdoba using an automatic solid-phase extraction unit on-line connected to a liquid chromatograph-tandem mass spectrometer arrangement (SPE-LC-MS/MS). This method was validated by a standard reference material, applying the Vitamin D Standardization Program (VDSP) protocols [22], and according to external quality assurance scheme (DEQAS) [23]. Briefly, 200 µL of filtered serum spiked with deuterated standards of the analytes was introduced for cleanup-chromatographic separation as required tandem mass spectrometry detection. Calibration curves for quantification were obtained using the ratio between the chromatographic peak area of each analyte and that of the corresponding deuterated standard. More information on sample preparation and LC-MS/MS analysis can be found in the article by Mena-Bravo et al. [24], and in Appendix A.

Statistical Methods
After excluding 27 women whose serum vitamin D levels could not be measured, and 17 women with lack of information in key covariates, the final sample size included 1422 participants.
Descriptive characteristics of participants were summarized as absolute values and percentages. Geometric means (GM), and the 25th, 50th, and 75th percentiles of Vit D 3 , 25(OH)D 3 , 1,25(OH) 2 D 3 , and 24,25(OH) 2 D 3 , according to women characteristics were also described. Comparisons were also made using the Wald test, with linear regression models adjusted for the weekly sun exposure score, vitamin D intake, and season. The weekly sun exposure score was calculated, taking into account the daily time in sun and the skin area exposed, according to the study by Hanwell et al. [25]. GM of the following VMRs were also calculated: R1: 25 Since the distribution of metabolite and VMR concentrations were positively skewed, the values were log-transformed to improve normality. To assess their association with women characteristics, we estimated geometric mean ratios (GMR) and 95% confidence intervals through multiple linear regression models, adjusted for weekly sun exposure score, vitamin D intake and season, and for those variables that were associated with each metabolite's concentration (p < 0.10) in the above-described Wald test analysis. For VMRs, models were adjusted for the same 3 mentioned variables plus those variables that, in this last analysis, showed to be relevant for any of the two metabolites of each ratio (p < 0.05). Differences in the associations of VMR according to vitamin D status (deficiency: 25(OH)D 3 < 50 nmol/L and non-deficiency: 25(OH)D 3 > 50 nmol/L) were also explored. Possible effect modifications were tested using the likelihood ratio test. Finally, to take into account the problem of multiple comparisons or multiple testing, p-values were also suitably adjusted by controlling the expected proportion of false positives, as proposed by Benjamini and Hochberg [26]. All analyses were performed using STATA/MP 14.0 software.

Results
The mean age of the participants was 44 years. As can be seen in Table 1, 23% of the women were overweight, and almost 10% were obese. Most were university graduates (61%). The percentage of nulliparous, non-smoking, abstemious, and sedentary women was 24%, 39%, 20%, and 42%, respectively. Hypercholesterolemia was reported in 13% of the women, and 10% were in treatment with corticosteroids. Most of the participants had a type IV skin phototype. The mean (+standard deviation) consumption of calories and calcium was 1976 + 681 Kcal/day and 1129 + 491 mg/day, respectively. Sun exposure, according to the weekly sun exposure score, was low in 47% of women, and vitamin D intake was lower than 5 µg/day in 72%. Most of the samples were obtained in spring (33%) and fall (29%).  The mean 25(OH)D 3 concentration was 49.2 + 18.9 nmol/L. More than half of the participants (59%) had vitamin D deficiency (25(OH)D 3 < 50 nmol/L). Serum levels were significantly higher in women with adequate body mass index (BMI), with one or two children, with higher sun exposure, in the most physically active women, in those taking vitamin D supplements, and in samples collected during the summer months. Both native vitamin D and 24,25(OH) 2 D 3 showed the same pattern as 25(OH)D 3 regarding BMI, physical activity, sun exposure, and season. 24,25(OH) 2 D 3 levels were also lower in nulliparous and in current smokers, were inversely associated with age, and positively associated with calcium and vitamin D intake. Vitamin D 3 levels were also higher in corticosteroid users. Finally, 1,25(OH) 2 D 3 levels were higher in nulliparous women with higher calorie intake and in samples obtained in winter (Table 1). Table 2 shows the association between the concentrations of vitamin D metabolites and women's characteristics. Obese women had lower levels of Vit D 3 , 25(OH)D 3 and 24,25(OH) 2 D 3 , while physically active women had higher concentrations of these metabolites. Parous women, as well as those taking vitamin D supplements, had higher concentrations of 25(OH)D 3 and 24,25(OH) 2 D 3 . Sun exposure was positively associated with Vit D 3 , 25(OH)D 3 , and 24,25(OH) 2 D 3 levels. Concentrations of these three metabolites were also higher in samples obtained in summer, and lower in the samples collected in winter. Women using corticosteroids had higher Vit D 3 concentrations (GMR = 1.09; 95%CI = 1.01-1.17). Current smokers presented lower levels of 24,25(OH) 2 D 3 (GMR = 0.93; 95%CI = 0.87-0.99). Phototype V-VI was associated with decreased 25(OH)D 3 concentrations (GMR = 0.90; 95%CI = 0.83-0.99). Finally, participants with higher calcium intake and lower calorie consumption had lower levels of 1,25(OH) 2 D 3 . Table 2. Association between vitamin D metabolite concentrations and characteristics of women.

Vit D 3 25(OH)D 3 1,25(OH) 2 D 3 24,25(OH) 2 D 3
Characteristics   GM: geometric mean; MET: metabolic equivalent. a Geometric mean ratio adjusted for body mass index, physical activity, use of corticosteroids, weekly sun exposure score, vitamin D intake, and season. b Geometric mean ratio adjusted for body mass index, parity, physical activity, weekly sun exposure score, vitamin D intake, and season. c Geometric mean ratio adjusted for parity, energy intake, weekly sun exposure score, vitamin D intake, and season. d Geometric mean ratio adjusted for age, body mass index, parity, tobacco, physical activity, weekly sun exposure score, calcium intake, vitamin D intake, and season. e Using the geometric mean as the reference.  (Table 3).   Table 3. Cont. GM: geometric mean; MET: metabolic equivalent. a Geometric mean ratio adjusted for body mass index, parity, physical activity, use of corticosteroids, phototype, weekly sun exposure score, vitamin D intake, and season. b Geometric mean ratio adjusted for body mass index, parity, phototype, weekly sun exposure score, energy intake, calcium intake, vitamin D intake, and season. c Geometric mean ratio adjusted for age, body mass index, parity, tobacco, physical activity, phototype, weekly sun exposure score, vitamin D intake, and season. d Geometric mean ratio adjusted for age, parity, tobacco, physical activity, phototype, weekly sun exposure score, energy intake, calcium intake, vitamin D intake, and season. e Using the geometric mean as the reference. f Taking into account daily time in sun and skin exposure according to Hanwell Tables 4 and 5 show the association of VMR with women's characteristics in participants with deficient (25(OH)D 3 < 50 nmol/L) and non-deficient (25(OH)D 3 > 50 nmol/L) serum vitamin D levels. For most of the studied associations, no differences were observed between these two groups. However, among participants with vitamin D deficiency, those who were taking corticosteroids had lower values of the R1 ratio than those who did not take corticosteroids, while no statistically significant differences were observed in participants with non-deficient levels of vitamin D (P-het = 0.027). The association of hypercholesterolemia treated with statins with VMR (decreasing the R2 values and increasing the R4 values) was only observed among women with non-deficient serum vitamin D levels. Finally, vitamin D status modified the association of the first three ratios with the season of the year, while R1 was only associated in women with sufficient levels of vitamin D (P-het < 0.001), R2 was altered only in women with deficient levels of this vitamin (P-het = 0.020), and the high R3 value in summer was only observed among participants with non-deficient vitamin D concentrations.  GM: geometric mean; MET: metabolic equivalent. a Geometric mean ratio adjusted for body mass index, parity, physical activity, use of corticosteroids, phototype, weekly sun exposure score, vitamin D intake, and season. b p-value for heterogeneity. c Geometric mean ratio adjusted for body mass index, parity, phototype, weekly sun exposure score, energy intake, calcium intake, vitamin D intake, and season. d Using the geometric mean as the reference. e Taking into account daily time in sun and skin exposure according to Hanwell

Discussion
To our knowledge, this is the first study providing information on the association of serum VMRs with several sociodemographic and lifestyle-related characteristics in premenopausal women. Our results show a notable vitamin D deficiency in the participating women, as well as the influence of certain factors (such as age, parity, and several lifestyles) on the vitamin D serum levels, its metabolites, and VMR.
Vitamin D deficiency (<50 nmol/L of 25(OH)D 3 ) is a global problem [4] that affects around 40% of the European population [6,27], and the Southern European countries [7]. In Spain, despite abundant sunshine, it has been estimated that 40% of the Spanish adult population have serum concentrations of 25(OH)D 3 below 50 nmol/L, and 18% below 25 nmol/L. These figures are 35% and 27% when we refer exclusively to the elderly population and postmenopausal women [7]. In our study, more than half (59%) of the participants had deficient levels of vitamin D, and only 9% had optimal levels (>75 nmol/L). Nulliparous women, and those with obesity or with darker skin, presented lower levels of 25(OH)D 3 , while women with greater sun exposure, those who took vitamin D supplements, were physically active, drank more alcohol, and those whose samples were collected in summer had higher concentrations. Regarding BMI, our results are in line with other Spanish [28] and international studies [29,30], in which obesity was significantly associated with lower 25(OH)D 3 levels. Circulating vitamin D concentrations are partially determined by genetic factors, and play an important role in the process of adipogenesis and inflammation status in adipocytes and adipose tissue [31]. Due to its fat solubility, vitamin D is retained by the body fat mass, resulting in lower availability of vitamin D for metabolic function in obese people [31,32]. Regarding parity, although a recent study has shown no association [33], Andersen et al. observed that the prevalence of vitamin D insufficiency was less frequent in nulliparous women [34]. The lower levels detected in our nulliparous participants could be due to lifestyles that imply less sun exposure or greater protection from the sun, different eating habits (egg and dairy products consumption was significantly lower in nulliparous participants), or the involvement of endogenous factors (such as the influence of hormones on vitamin D metabolism). Several observational studies have shown that vitamin D deficiency is a risk marker for reduced female fertility and various adverse pregnancy outcomes [35,36]. Leisure-time physical activity appears to be an effective manner of maintaining adequate vitamin D concentrations [37]. Such association has often been attributed to confounding factors, but recent studies indicate that exercise may have a direct and causal effect on vitamin D status, possibly through the mobilization of adipose-derived vitamin D and/or 25(OH)D 3 [38], or through an increase in muscle use producing the release of 25OHD from its interior [39]. The association between alcohol consumption and vitamin D serum levels remains controversial, although recent studies, with large sample sizes, showed positive associations [40]. Consistent with our findings, other factors related to sun exposure, such as short time spent in the sun, low amount of skin surface exposed, samples collected in winter/early spring, and increased skin pigmentation have been associated with higher risk of 25(OH)D 3 deficiency in the literature [41]. Finally, and as expected, the intake of vitamin D supplements increased serum levels of 25(OH)D 3 . However, the intake of these supplements is very infrequent, both among the women of our study and in Spain in general [42]. Only 19% of our participants took the 5 µg/day of vitamin D recommended by the Spanish Federation of Societies of Nutrition, Food, and Dietetics (FESNAD) in 2010 [43], and only 0.4% took the 15 µg/day recommended in 2019 by the European Food Safety Authority (EFSA) for the adult population [44].
Although 25(OH)D 3 is still recommended as the marker of choice by current guidelines from scientific organizations, growing evidence indicates significant limitations that hamper the utility of this analyte in clinical practice, including analytical aspects and interpretation of results [9]. VMRs are promising emerging biomarkers that may provide additional information in assessing vitamin D status [9,45]. The first ratio (25(OH)D 3 /Vit D 3 ), represents the activity of 25-hydroxylase enzyme in the liver, which is the main enzyme responsible for the conversion of vitamin D 3 to the main circulating form of this vitamin, the 25(OH)D 3 . The values of this ratio were similar to those described in the study by Mena-Bravo et al. [10]. This ratio was higher in women taking supplements and among participants with sufficient 25(OH)D 3 levels whose samples were collected in fall or winter, and lower among nulliparous women and corticosteroid users with deficient vitamin D levels.
The second ratio (1,25(OH) 2 D 3 /25(OH)D 3 ) represents the 1α-hydroxylase activity, an enzyme encoded by the CYP27B1 gene in the kidney, where 25(OH)D 3 is converted to the active 1,25(OH) 2 D 3 . We found a higher ratio in obese women (mainly in those with deficient serum vitamin D levels), in nulliparous women, in those with more caloric diets, and in women with deficient vitamin D concentrations whose samples were collected in winter. On the contrary, this ratio was lower in women with higher consumption of alcohol, calcium, vitamin D supplements, and statin users; in women with greater sun exposure; and in samples collected in spring and summer (both results only detected in women with vitamin D deficiency). The values of this ratio in our participants are slightly higher than those described by Mena-Bravo et al. [10] (average ± SD: 0.0029 ± 0.002) and, although we have not found studies reporting characteristics associated with this ratio, there is evidence that high dietary calcium intake reduces 1α-hydroxylase activity (reflected in a lower R2 ratio), while low calcium intake down-regulates 24-hydroxylase expression [9,46].
The third ratio (24,25(OH) 2 D 3 /25(OH)D 3 ) is mediated by the CYP24A1 gene that encodes the enzyme 24-hydroxylase, which catalyzes the conversion of 25(OH)D 3 into 24,25(OH) 2 D 3 . When sufficient amounts of biologically active vitamin D are available, CYP24A1 is up-regulated and more 24,25(OH) 2 D 3 is formed [9]. This ratio may be of potential use as an indicator of vitamin D deficiency and as a predictor of the change in 25(OH)D 3 after vitamin D supplementation. It may also help explain some of the interindividual differences in the response of serum 25(OH)D 3 to the same administered dose of vitamin D [9,13,[47][48][49]. In some studies, low levels of this ratio seem to be related to the increasing all-cause mortality in patients with chronic kidney disease and risk of hip fracture in older adults [16,50]. However, in our study, we found no differences in this ratio between the participants that were taking vitamin D supplements and those who did not (regardless their vitamin D status), in line with what was observed in previous studies [14,45], and contrary to what was observed in Tang's study [11]. Older women, nulliparous women, and those whose samples were collected in spring had a lower R3 ratio, although the association of this ratio with season varied as a function of vitamin D levels. Regarding R3 mean values, two previous studies have described figures that are in line with those obtained in our study [10,11].
Finally, the fourth ratio (24,25(OH) 2 D 3 /1,25(OH) 2 D 3 ) could also be a good indicator of vitamin D status. Tang et al., observed an inverse correlation between the 1,25(OH) 2 D 3 /24,25(OH) 2 D 3 ratio and the 25(OH)D 3 levels, so that when vitamin D levels were insufficient, the production of 1,25(OH) 2 D 3 was favored to the detriment of its conversion to 24,25(OH) 2 D 3 [11]. This phenomenon is also compatible with our results, since the GM of R4 was lower in women with vitamin D deficiency than in those with non-deficient levels. The R4 ratio was higher in women who consumed calcium and vitamin D supplements, in participants with high sun exposure, in those who were physically active and in samples collected in summer and fall. On the contrary, the oldest women, nulliparous women, those whose samples were collected in winter or spring, and the participants with non-deficient levels of vitamin D who consumed many calories had a lower R4 ratio.
The cross-sectional design of this study limits the possibility of establishing a temporal relationship between the exposures and vitamin D metabolite levels. This sample includes only premenopausal women recruited from a single center, so the results cannot be extrapolated to the general population. In addition, only a single blood sample was collected at the beginning of the study, so the participants' usual vitamin D status may not have been adequately reflected. On the other hand, given that we have used a novel approach to provide a more complete picture of vitamin D3 metabolism, the results of this study should be considered as hypothesis-generating and should be viewed with caution. Precisely, due to its hypothesis-generating approach and the exploratory nature of the study, corrections for multiple testing were not applied in the main analyses [51], although results adjusted using the Benjamini and Hochberg method [26] are reported in the Supplementary Material (Tables S1 and S2). Finally, even though we have included the main variables described in the literature as associated with vitamin D levels in our models, the possibility of residual confounding cannot be ruled out.
The greatest strength of our study is its novelty. In a relatively large sample of participants, we were able to quantify vitamin D metabolites, the ratios between them, and the factors that contribute to explain metabolic variations. In addition, SPE-LC-MS/MS is a sensitive automated method for the analysis of serum vitamin D and metabolites that provides reliable and robust results. This method was validated by a standard reference material and according to DEQAS [23].

Conclusions
In general, vitamin D metabolite profile in nulliparous women and older women was compatible with lower activity of the enzyme 24-hydroxylase, which catabolizes 25(OH) 2 D 3 to 24,25(OH) 2 D 3 . Furthermore, nulliparous women and those who consumed more calories showed an increase in calcitriol levels to the detriment of the concentrations of the other two metabolites. The opposite was observed among the participants with greater consumption of calcium, alcohol, or with greater sun exposure. Finally, the association of VMR with seasons was different depending on vitamin D status. These results highlight the added value of VMR as complementary indicators of vitamin D status and its endogenous metabolism, being considered better predictors of vitamin D treatment response and clinically important outcomes. The results also reveal the potential contribution of certain factors in the greater or lesser expression/activity of hydroxylase enzymes.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/nu13113747/s1, Table S1: Association between vitamin D metabolite concentrations and characteristics of women. Analysis adjusted for multiple testing. Table S2: Association between vitamin D metabolite ratios and characteristics of women. Analysis adjusted for multiple testing. Funding: This study was funded by the Spanish Ministry of Health (EC11-273) and by the Carlos III Institute of Health (PI15CIII/0029). The article presents independent research. The views expressed are those of the authors and not necessarily those of the Carlos III Institute of Health.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics and Animal Welfare Committee of the Carlos III Institute of Health.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
Data Availability Statement: Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.