A systematic review and meta-analysis of the association between vitamin D and ovarian reserve

It is hypothesized that vitamin D deficiency could be related to ovarian reserve. This systematic review and meta-analysis was undertaken to analyze the possible association between vitamin D and ovarian reserve among adolescent and adult women. All eligible studies identified through the ISI Web of Science, PubMed, and Scopus were included up to May 2021. A random-effects meta-analysis model was implemented and a weighted mean difference (WMD) and 95% confidence interval (CI) were calculated. A total of 38 papers covering 8608 individuals were enrolled in this systematic review and meta-analysis. Antral follicle count (AFC) was significantly lower among Asians (WMD − 0.65; 95% CI − 1.28 to − 0.01; P = 0.04; I2 = 0.0%) and luteinizing hormone (LH) levels were higher in non-Asians (WMD 2.16 IU/L; 95% CI 0.20 to 4.12; P = 0.031; I2 = 9.3%) with vitamin D insufficiency/deficiency. Also, there was a negative correlation between vitamin D and LH/FSH ratio in women with normal body mass index (BMI) (Fisher’s Z: − 0.18; 95% CI − 0.37 to − 0.008; P = 0.041; I2 = 51.5%). Although there were no significant associations between serum vitamin D levels and any of the intended ovarian reserve markers, subgroup analyses have found significant findings regarding AFC, LH, and LH/FSH ratio. In order to understand the underlying mechanisms of vitamin D in female reproduction, further attempts are needed.

Vitamin D is an essential nutrient with a hormone-like activity that was initially recognized for its importance in bone health and calcium-phosphate homeostasis 1 . Though, the recent vitamin D deficiency pandemic has emphasized other functions 2 . Growing documents suggest that vitamin D deficiency upsurges the risk of various chronic disorders including obesity, type 1 diabetes mellitus, cardiovascular, infectious, and autoimmune diseases; certain types of cancer; depression, and chronic pain 2 .
More recently, a regulatory role for vitamin D has been suggested in female fertility 3,4 . In this context, previous epidemiological investigations have proposed a seasonality in female reproductive capacity which might be explained partially by seasonal variation in serum levels of vitamin D 5 . Biological activities of vitamin D are applied through the vitamin D receptors (VDR) that have been detected in the ovary especially in granulosa cells and theca cells, endometrium and placenta 6 . This diverse VDR expression proposes a potential role of vitamin D in female reproduction 7 . Though the underlying mechanism by which vitamin D may involve in reproductive physiology is poorly known, a direct link between vitamin D and ovarian steroidogenesis has been proposed. This link is derived from several in-vitro and in-vivo studies indicating that vitamin D could stimulate steroidogenesis in ovarian cells by modulating the mRNA and protein expression levels of steroidogenic enzymes including Cyp11a1, StAR, Cyp19a1, and 3β-HSD 8,9 . Reproductive potential of an individual is mainly explained by the quality and the quantity of ovarian primordial follicles that was diminished as women get older. Therefore, several markers have suggested to illustrate the ovarian reserve status. Low anti-Mullerian hormone (AMH), low antral follicle count (AFC), low luteinizing hormone (LH), high follicle-stimulating hormone (FSH), and low LH/FSH ratio may represent a diminished ovarian reserve status 10 .
Nevertheless, the findings of experimental studies are consistent enough to suggest the association between vitamin D and ovarian reserve, the evidence of human studies are commonly inconsistent, with some documents Study selection and eligibility criteria. The PICO (Population/intervention/comparison/outcome) components were as follows: P (adolescent and adult premenopausal women with vitamin D deficiency/insufficiency), I (serum levels of vitamin D), C (women with a normal level of serum vitamin D), O (ovarian reserve markers including AMH, AFC, LH, FSH, and LH/FSH ratio). The inclusion criteria were as follows: (1) original human observational studies either with case-control, cross-sectional, or longitudinal design; (2) published in the English language; (3) assessed serum levels of at least one of the ovarian reserve markers including AMH, AFC, LH, FSH, and LH/FSH ratio in association with 25(OH)D; and (4) those presented as (4.1) comparison of ovarian reserve markers (AMH, AFC, LH, FSH, and LH/FSH ratio) between women with vitamin D insufficiency/deficiency and vitamin D sufficient ones; or (4.2) correlation between 25(OH)D and ovarian reserve markers (AMH, AFC, LH, FSH, and LH/FSH ratio).
The exclusion criteria were as follows: (1) Experimental studies; (2) recruited pregnant, lactating, or postmenopausal women; and (3) poster abstracts, case reports, review articles, editorials, and non-original full-length articles, or those without original data or articles with no appropriate outcome measures. Two assessors independently (A.A and E.K) conducted the selection process. Any disagreement was resolved through discussion with a third reviewer (R.A). Data extraction. The following data were extracted: first author's name, year of publication, geographical location, sample size, participant characteristics including health status, age and body mass index (BMI), 25(OH) D assay method, cut-off values of vitamin D status, the season of sample collection, study design, reported ovarian reserve markers, and statistical adjustment. Quality assessment. The quality assessment of eligible studies was performed by two reviewers (A.A and E.K) individually using the Newcastle-Ottawa Scale (NOS) star system (ranged, 0-9 stars) 17 , which focuses on selection, comparability, and outcome. Studies scoring ≥ 7, 4-6, and ≤ 3 points were assumed as high, moderate, and low quality, respectively 18 .
Statistical analysis. The present study was performed to present the association between vitamin D and ovarian reserve quantitatively. Prior to the calculation of the effect size, the concentration of AMH was converted to ng/mL, LH to IU/l, and FSH to IU/l. In the current study, we calculated two types of effect sizes: (1) weighted mean difference (WMD) in AMH, AFC, LH, or FSH between vitamin D insufficiency/deficiency and sufficient vitamin D groups; and (2) Fisher's Z of the correlation between 25(OH)D and AMH, AFC, LH, FSH or LH/FSH. If a document provided the results stratified by certain variables like age, BMI, and participants' health status, it was divided into two different studies supposed to be independent of each other. In the absence of the mean and standard deviation (SD), values of median and range or median and interquartile range were converted into mean and SD based on related formulas 19 . Fisher's Z and its SE using correlation coefficients (r) and sample size (N) were calculated by the relevant formula 20 . Heterogeneity between effect size of included studies was estimated by chi-squared (χ 2 ) test and I 2 statistic [I 2 index < 40 (low heterogeneity), 40-75 (moderate heterogeneity) and > 75% (high heterogeneity)] 21 . When there was no significant heterogeneity, the effect size was calculated using a fixed-effects model. Otherwise, a random-effects model was used 22 . Subgroup analyses were done based on different characteristics of included studies, whenever possible, to check the sources of heterogeneity. Publication bias was assessed using Egger's and Begg's statistics 23 and in the presence of significant publication bias, trim & fill analysis was performed to detect any possibly missed study. The sensitivity analyses were also con-Findings from the systematic review. Four studies have examined the association between 25(OH)D and ovarian reserve markers, however, due to insufficient data, they were described qualitatively.
In the first study, Ghadimi et al. 38 , have assessed the association between vitamin D and metabolic parameters of PCOS in high-school girls (mean age: 17.85 years old) through a case-control design. The study included 104 PCOS individuals and 88 non-PCOS controls. Based on Pearson's test, no significant correlation was found between 25(OH)D levels and LH and FSH levels.
The other investigation has been conducted by Jukic et al. 33 in 2015 to explore the relationship between FSH and serum vitamin D among 527 premenopausal women (mean age: 17.85 years old). In this cross-sectional study, 25(OH)D and urinary FSH levels were inversely correlated (P = 0.003).
The other study in 2018 was conducted by Arefi et al. 36 to explore the correlation between vitamin D deficiency and ovarian reserve through a cross-sectional study of 189 Iranian infertile women (mean age: 32.21, mean BMI: 26.7 kg/m 2 ). The result of this study proposed a highly significant correlation between vitamin D and AFC (p < 0.001).
The last evidence was conducted among infertile and fertile females (18-40 years old) to investigate the correlation of vitamin D deficiency with serum AMH. The result of this cross-sectional study failed to show any significant correlation between vitamin D and AMH in either fertile or infertile women 47 .
Serum 25(OH)D levels and LH. The difference of LH according to the vitamin D status was examined in six studies 28,29,48,50,52 which was not significant (WMD 0.05 IU/l; 95% CI − 0.67 to 0.76; P = 0.900) (Fig. 5). Substantial heterogeneity was not observed (I 2 = 39.7%, P = 0.141) among the included studies. Subgroup analysis of geographical areas revealed that serum LH was significantly higher among the non-Asian population with vitamin D insufficiency/deficiency compared to the control group (WMD 2.16 IU/l; 95% CI 0.20 to 4.12; P = 0.031) with no evidence of heterogeneity (I 2 = 9.3%, P = 0.294) ( Table 2). Excluding individual studies did not change the overall meta-analysis results.  (Fig. 7). Subgroup analysis was performed based on geographical areas, participants' health status, and study design, however, the overall results did not change ( Table 3). The overall meta-analysis result for AMH was not sensitive to individual studies.  (Fig. 8). There was evidence of substantial heterogeneity among the effect size of included studies (I 2 = 83.7%, P < 0.001). Subgroup analysis based on participants' health status and study design did not change the findings (Table 3). Excluding individual studies did not change the overall meta-analysis results.
The correlation between LH and 25(OH)D levels. The correlation between LH and vitamin D was not significant in the meta-analysis of seven studies 26,28,30,32,44,48 with 919 participants (Fisher's Z: − 0.09; 95% CI − 0.29 to 0.11; P = 0.372). Furthermore, evidence of significant heterogeneity was observed (I 2 = 87.2%, P < 0.001) (Fig. 9). Subgroup analysis based on participants' health status and study design did not change the overall findings (Table 3). Meta-analysis findings were not sensitive to individual studies.

Discussion
In order to identify new nutritional factors associated with women's fertility, various attempts have been conducted. However, interpreting the literature to wrap up a conclusion is a difficult process for clinicians. Therefore, a comprehensive systematic review and meta-analysis of available literature can represent the most reliable evidence. Although previous systematic review and meta-analysis examined the relationship between concentrations of vitamin D and ovarian reserve 56 , that study focused only on AMH and included only 5 articles. Therefore, it was necessary to conduct a more comprehensive systematic review and meta-analysis on this relationship. The present systematic review and meta-analysis of 36 observational studies examined the association between serum vitamin D levels and ovarian reserve markers including AMH, AFC, FSH, LH, and LH/FSH ratio in the adolescent and adult population of premenopausal women. Although, there was no significant association www.nature.com/scientificreports/ between serum vitamin D levels and any of the intended ovarian reserve markers, some of the subgroup analyses have found significant findings. AFC was significantly lower among Asians and LH was higher in the non-Asian population with vitamin D insufficiency/deficiency. Moreover, there was a negative correlation between vitamin D and LH/FSH ratio in women with normal BMI.
There are some points that should be taken into account when interpreting the results. First of all, there are substantial inter-assay differences in the performance of commercially available kits for serum vitamin D assay 57 . This notion may affect the results and play a considerable role as a heterogeneity factor. Additionally, seasonal variation in serum vitamin D should be considered when interpreting the results 58 . This notable fact has been excused in some papers 59,60 and such inconsistency among the season of sample collection could also influence our final results. Lastly, there are several factors including race, skin color, use of skin protection (sunscreen), latitude, environmental pollution, aging, cultural and lifestyle issues that all can affect the synthesis and availability of vitamin D worthy to consider when interpreting the results 4,61,62 .
AMH is a glycoprotein hormone related to inhibin and activin and belongs to the family of transforming growth factor β (TGF-β), which has substantial functions in ovarian folliculogenesis 63 . AMH decreases follicle sensitivity to FSH 64 . Thus, there is absolute evidence that AMH is involved in the initiation of growth in follicles and FSH sensitivity 65 . The mechanism by which vitamin D may affect AMH and FSH is unclear. Similar to human studies, the findings of experimental studies regarding this association are also inconclusive [66][67][68] . Vitamin D may influence ovarian steroidogenesis, development of the follicles, and ovarian reserve 7 . AFC is a main prognosticator of the ovarian reserve and response to hormonal and follicle stimulation 69 . Furthermore, the related mechanism of vitamin D was regulated through VDR 70,71 , Thus, the ovarian reserve markers levels might be affected by the VDR polymorphism 72 . Interestingly, Szafarowska et al. reported that there is an association between polymorphisms of the VDR gene and AMH; however, they have not found any correlation between AMH levels and vitamin D concentrations in PCOS women 73 . On the other hand, some studies have represented that vitamin D deficiency and also single nucleotide polymorphism (SNP) of VDR did not affect dysmenorrhea, Table 2. Subgroup analysis of the differences in ovarian reserve markers between women with vitamin D insufficiency/deficiency and controls. AFC antral follicle count, AMH anti-Mullerian hormone, FSH follicle stimulating hormone, LH luteinizing hormone, BMI body mass index. 1 Calculated by Random-effects model as weighted mean difference.      74 . Several studies have suggested that reduced vitamin D concentrations in PCOS and obese women may be associated with infertility 75,76 . A possible mechanism regarding the recent association might be decreased insulin sensitivity due to vitamin D deficiency 77 . Considering this hypothesis, insulin could elevate androgen biosynthesis and reduce sex hormone-binding globulin (SHBG) which resulted in hyperandrogenism. An excess amount of androgens is converted to estrogen. High estrogen concentration promotes the secretion of LH and represses FSH of the anterior pituitary 78 . Based on our findings, a negative correlation was observed regarding LH/FSH ratio with vitamin D suggesting that vitamin D status may contribute to hormonal dysregulation, even in women with normal BMI. According to the current meta-analysis, the issue that vitamin D levels are associated with ovarian reserve markers is still a controversial subject. Evidence is still unreliable as the randomized controlled trials are scarce, and the findings of available evidence are extremely heterogeneous. On the other hand, the overall result of vitamin D and AFC showed a marginally significant association, whereas, the exclusion of Drakopoulos et al. and Fabris et al. studies from the analysis revealed a significant association between vitamin D and AFC. Based on the results of these studies, the change in results can be interpreted by Drakopoulos et al. 's study that was the only study to show that AFC was higher in women with vitamin D deficiency compared to those without vitamin D deficiency. In addition, Fabris et al. 's study demonstrated the least difference between the two groups with and without vitamin D deficiency in relation to AFC. As a result, the exclusion of these studies was able to make a significant result overall.
One of the substantial limitations in our study is the lack of evaluation of ethnicity in relation to vitamin D status, given that the vast majority of patients included in studies were Caucasian. Also, different methods of measuring vitamin D and the health status of the participants can be considered as other factors. Nevertheless, other parameters in the present meta-analysis were not sensitive to individual studies. The present meta-analysis had other limitations. There was significant heterogeneity in our study that may have affected the results and diminished the generalizability of the findings. The probable sources of heterogeneity might be differences in age, BMI, study design, vitamin D and ovarian reserves markers assay methods and kits, the season of sample collection, geographical variation, and the quality of the studies. Furthermore, the observational design of the included studies precludes us to examine the causality. Another limitation that may influence the findings is regarding vitamin D binding protein concentrations and VDR's polymorphisms that were not measured by the included studies. Moreover, seasonal variation of vitamin D was not taken into account in some of the included studies. In addition, the selected subgroup for the current study was not pre-specified that might be a source of bias and a limitation of the present systematic review and meta-analysis. Furthermore, the ovarian reserve mostly   www.nature.com/scientificreports/ refers to the number and quality of dormant primordial follicles that cannot be explained completely by serum levels of AMH, AFC, LH, FSH, and LH/FSH ratio. On the other hand, these biochemical markers were selected as surrogate variables to illustrate the ovarian reserve. Another point to consider is the included studies did not use the same cut-off values for determining the patients' vitamin D status. Moreover, there are potential confounders in the association between vitamin D and ovarian reserve including age, BMI, dietary intake, smoking, physical activity, and etc. Since most of the included studies did not comprehensively adjust for these confounders, this issue can influence our findings and should be considered as possible limitation.

Conclusion
Based on what was discussed, although, there was no significant association between serum vitamin D levels and any of the intended ovarian reserve markers, some subgroup analyses have found significant findings. AFC was significantly lower among Asians and LH was higher in non-Asian population with vitamin D insufficiency/ deficiency. Moreover, there was a negative correlation between vitamin D and LH/FSH ratio in women with normal BMI. In order to understand the underlying mechanisms of vitamin D in female reproduction, further attempts are needed.