Vitamin D and Reproductive Hormones Across the Menstrual Cycle

Abstract

STUDY QUESTION

How do the calciotropic hormones (25-hydroxyvitamin D, 1,25-dihydroxyvitamin D and intact parathyroid hormone (iPTH)) vary across the menstrual cycle and do cyclic patterns of reproductive hormones (estradiol, progesterone, LH, FSH) differ by vitamin D status?

SUMMARY ANSWER

Calciotropic hormones vary minimally across the menstrual cycle; however, women with 25-hydroxyvitamin D below 30 ng/ml have lower mean estradiol across the menstrual cycle.

WHAT IS KNOWN ALREADY

Prior human studies suggest that vitamin D status is associated with fecundability, but the mechanism is unknown. Exogenous estrogens and prolonged changes in endogenous estradiol (pregnancy or menopause) influence concentrations of 25-hydroxyvitamin D. In vitro, treatment with 1,25-dihydroxyvitamin D increases steroidogenesis in ovarian granulosa cells. There are little data about changes in calciotropic hormones across the menstrual cycle or cyclic patterns of reproductive hormones by categories of vitamin D status.

STUDY DESIGN, SIZE, DURATION

A prospective cohort study of 89 self-identified white women aged 18–44, across two menstrual cycles. Participants were a subset of the BioCycle Study, a community-based study conducted at the University of Buffalo, 2005–2007.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Eligible participants had self-reported regular menstrual cycles between 21 and 35 days and were not using hormonal contraception or vitamins. Early morning fasting blood samples were drawn at up to eight study visits per cycle. Visits were timed to capture information in all cycle phases. Serum samples for 89 women (N = 163 menstrual cycles) were analyzed for estradiol, progesterone, LH, FSH and 25-hydroxyvitamin D (25(OH)D). Variability in calciotropic hormones within and across menstrual cycles was assessed using intraclass correlation coefficients and non-linear mixed models. Given the relative stability of the calciotropic hormones across the menstrual cycle, non-linear mixed models were used to examine differences in the cyclic patterns of estradiol, progesterone, LH and FSH by categories of each calciotropic hormone (split at the median). These models were conducted for all ovulatory cycles (N = 142 ovulatory menstrual cycles) and were adjusted for age, BMI (measured in clinic) and self-reported physical activity.

MAIN RESULTS AND THE ROLE OF CHANCE

Median 25(OH)D concentration was 29.5 ng/ml (SD 8.4), and only 6% of women had vitamin D deficiency (<20 ng/ml). The mean concentration of 25(OH)D did not differ between the luteal and follicular phase; however, both 1,25(OH)₂D and iPTH showed small fluctuations across the menstrual cycle with the highest 1,25(OH)₂D (and lowest iPTH) in the luteal phase. Compared with women who had mean 25(OH)D ≥30 ng/ml, women with lower 25(OH)D had 13.8% lower mean estradiol (95% confidence interval: −22.0, −4.7) and 10.8% lower free estradiol (95% CI: −0.07, −0.004). Additionally, compared to women with iPTH ≤36 pg/ml, women with higher concentrations of iPTH had 12.7% lower mean estradiol (95% CI: −18.7, −6.3) and 7.3% lower progesterone (95% CI: −13.3, −0.9). No differences in the cyclic pattern of any of the reproductive hormones were observed comparing cycles with higher and lower 1,25(OH)₂D.

LIMITATIONS, REASONS FOR CAUTION

Women included in this study had self-reported ‘regular’ menstrual cycles and very few were found to have 25(OH)D deficiency. This limits our ability to examine cycle characteristics, anovulation and the effects of concentrations of the calciotropic hormones found in deficient individuals. Additionally, the results may not be generalizable to women with irregular cycles, other races, or populations with a higher prevalence of vitamin D deficiency.

WIDER IMPLICATIONS OF THE FINDINGS

These findings support current clinical practice that does not time testing for vitamin D deficiency to the menstrual cycle phase. We find that women with lower vitamin D status (lower 25(OH)D or higher iPTH) have lower mean concentrations of estradiol across the menstrual cycle. Although this study cannot identify a mechanism of action, further in vitro work or clinical trials may help elucidate the biologic mechanisms linking calciotropic and reproductive hormones.

STUDY FUNDING/COMPETING INTEREST(S)

This work was supported by the Intramural Research Programs of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health (contract numbers: HHSN275200403394C, HHSN275201100002I and Task 1 HHSN27500001) and the National Institute of Environmental Health Sciences. There are no competing interests.

Introduction

Long known as beneficial for bone health, adequate vitamin D concentrations have more recently been associated with a range of reproductive outcomes (reviewed in Irani et al., 2014; Luk et al., 2012). In particular, human studies suggest possible associations between vitamin D and embryo implantation, polycystic ovary syndrome and menstrual cycle length and regularity (Jukic et al., 2015; Jukic et al., 2016; Jukic et al., 2018; Luk et al., 2012).

The mechanisms underlying these observed associations are unknown. Experimental models have suggested that 1,25-dihydroxyvitamin D (1,25(OH)₂D), the active metabolite of vitamin D, may act directly to influence reproductive hormone production and estrus cycle characteristics (Kinuta et al., 2000; Merhi et al., 2014; Parikh et al., 2010; Sun et al., 2010). Many reproductive tissues (Luk et al., 2012) including ovarian granulosa cells (Parikh et al., 2010) harbor the vitamin D receptor. Stimulation of human ovarian cells in vitro with 1,25(OH)₂D influences steroidogenesis (Merhi et al., 2014; Parikh et al., 2010). Mouse knock-out models lacking the vitamin D receptor exhibit hypergonadotropic hypogonadism with decreased aromatase expression, (Kinuta et al., 2000) and mice lacking the enzyme that produces 1,25(OH)₂D show reduced corpora lutea and prolonged diestrus (Sun et al., 2010).

However, research also suggests that reproductive hormones, particularly estrogen, might affect vitamin D status. Multiple hormones related to calcium metabolism (calciotropic hormones) have been observed to change during events characterized by persistent changes in the concentrations of endogenous estradiol such as pregnancy and menopause (Bouillon et al., 1981; Pop et al., 2015) as well as with the use of exogenous estrogen (contraception or hormone replacement) (Garcia-Bailo et al., 2013; Harmon et al., 2016).

While in vitro models may exclude relevant biological pathways in other tissues or cells, and pregnancy/menopause will include a multitude of metabolic changes, the menstrual cycle offers an in vivo opportunity to examine short term interactions between reproductive hormones and calciotropic hormones. Differences in the pattern of reproductive hormones in cycles with higher or lower calciotropic hormones may generate hypotheses to narrow down the relevant biological processes in these complex metabolic pathways. Furthermore, if calciotropic hormones vary with the menstrual cycle, this may have implications for the timing of clinical diagnostic tests.

The BioCycle Study provides an ideal data source for studying calciotropic hormones across the menstrual cycle. Women in this study were followed closely across two menstrual cycles and importantly were not using vitamin supplements or exogenous hormones. Using this well-defined cohort we examine within-woman correlations of three calciotropic hormones, 25-hydroxyvitamin D (25(OH)D), 1,25-dihydroxyvitamin D (1,25(OH)₂D) and intact parathyroid hormone (iPTH), and describe the changes of the three calciotropic hormones across the menstrual cycle. We then test whether higher or lower concentrations of each calciotropic hormone are associated with estradiol, progesterone, LH and FSH cycle trajectories.

Materials and Methods

Study sample

Subjects for the present study are a subset of participants in the BioCycle Study. The BioCycle Study enrolled healthy women age 18–44 years for up to two menstrual cycles. Eligible participants were community members enrolled between 2005 and 2007 from western New York. Recruitment methods included flyers in health and student health clinics, word of mouth and media advertisements. Participants had self-reported regular menstrual cycles between 21 and 35 days, were not using hormonal contraception, vitamin or mineral supplements, did not have kidney disease requiring treatment in the past year and met other inclusion and exclusion criteria as previously described (Wactawski-Wende et al., 2009). The current analysis includes a random subsample of 89 participants, including all available anovulatory cycles, resulting in 163 cycles (N = 15 women contributed samples from a single cycle, most single cycles were anovulatory); all were self-reported white race to minimize variability of vitamin D by race (Institute of Medicine Food and Drug Board, 2010).

Ethical approval

The study protocol was approved by the University at Buffalo Health Sciences Review Board, which served as the institutional review board designated by the National Institutes of Health under a reliance agreement. All participants provided written informed consent.

Data collection

Data collection and study visit alignment with menstrual cycle phase have been described in detail elsewhere (Howards et al., 2009; Wactawski-Wende et al., 2009). Briefly, early morning fasting blood samples and questionnaire data were collected at up to eight in-person study visits per cycle (days 2, 7, 12, 13, 14, 18, 22 and 27 of a standardized 28-day cycle). Serum samples were processed and stored at −80°C within 90 minutes of phlebotomy. Using menses information and fertility monitors, study visits were timed to capture information in all menstrual phases (Howards et al., 2009) (Supplementary Fig. S1). To further account for individual variation in cycle length, the data from each study visit were realigned as necessary based on measured hormones (Mumford et al., 2011). Realigned cycle day was used in all analyses.

Reproductive hormones

Estradiol, progesterone, LH and FSH were measured in serum samples at a single laboratory (Kaleida Health Laboratories, Buffalo, NY). Free estradiol was calculated with measured estradiol and concentrations of sex hormone binding globulin and albumin (Yeung et al., 2013). All samples from a given participant were analyzed in the same batch to reduce within-subject variability (Wactawski-Wende et al., 2009). Estradiol, progesterone, LH and FSH were measured using solid phase competitive chemiluminescent enzymatic immunoassay (Specialty Laboratories Inc, Valencia, CA) on the DPC Immunolite®2000 analyzer (Siemens Medical Solutions Diagnostics, Deerfield, IL). The coefficients of inter-assay variation (CV) for these assays reported by the laboratory were <14% for progesterone, <10% for estradiol and <5% for LH and FSH.

Anovulatory cycles were defined as cycles with luteal progesterone levels ≤5 ng/ml and no observed LH peak during the mid or late luteal phase visit (Lynch et al., 2014). The analytic sample had 19 anovulatory cycles identified in 17 women.

Calciotropic hormones

Serum concentrations of total 25-hydroxyvitamin D (sum of 25-hydroxyvitamin D2 and D3) (25(OH)D), total 1,25-dihydroxyvitamin D (sum of 1,25-dihyoxyvitamin D2 and D3) (1,25(OH)₂D) and iPTH were measured at the University of Minnesota. All were assayed using enzyme immunoassays 25-OH Vitamin D ELISA (BioVendor R&D, Asheville, NC), 1,25 Dihydroxyvitamin D EIA (Immuno Diagnostics Systems, Fountain Hills, AZ) and iPTH ELISA (Abnova, Walnut, CA). Assays were conducted 9 years after original blood-draw using samples that had not been previously thawed. While concentrations of 25(OH)D are robust to variation in sampling procedure and long-term storage (Jukic et al., 2018a), less is known about 1,25(OH)₂D and iPTH. To the extent that long term storage may introduce measurement error, this error is unlikely to be differential by cycle phase or concentration of reproductive hormones. Therefore, bias is unlikely. Inter-assay CVs reported by the laboratory for in-house pooled serum controls were 10.7% for 25(OH)D, 12.2% for 1,25(OH)₂D and 16.5% for iPTH. Intra-assay CVs were between 4.1% (25(OH)D) and 6.9% (1,25(OH)₂D).

The number of blood samples assayed per cycle was determined based on the half-life of each hormone. The half-life of 25(OH)D is 2 weeks (Vieth 2011); thus, most participants had three samples assayed, a mid-follicular and mid-luteal sample from cycle 1 and a mid-follicular sample from cycle 2. 1,25(OH)₂D and iPTH have shorter half-lives and were assayed in up to six study visits in each cycle. These samples were chosen to represent the follicular (two samples), ovulatory (two samples) and luteal (two samples) phases. Final sample number assayed for each participant varied due to cycle length and available samples with sufficient volume. (Supplemental Fig. S1 shows the sampling schematic with additional details).

Covariates

Based on the literature, we examined the following covariates for associations with 25(OH)D, 1,25(OH)₂D and iPTH: ever use of hormonal birth control, time since last use of hormonal birth control, BMI (kg/m²) from measured height and weight, dietary intake of vitamin D estimated from a validated food frequency questionnaire (Wactawski-Wende et al., 2009), season of blood sample, smoking status, parity, physical activity (IPAQ—The International Physical Activity Questionnaire (Craig et al., 2003)), education and age. A measure of kidney function (estimated glomerular filtration rate, eGFR) was calculated at each visit using the Modification of Diet in Renal Disease study equation with a fasting serum creatinine (Pollack et al., 2015). All other variables were self-reported (Wactawski-Wende et al., 2009) unless noted.

Statistical Analysis

Short term, within-woman reliability of calciotropic hormones and correlation among these hormones

We calculated intraclass correlation coefficients (ICCs) and 95% confidence intervals (CIs) for each calciotropic hormone for all measures across the study period using the R package ICC (Wolak et al., 2012). We examined Pearson correlations between the calciotropic hormones in two ways: using paired measures taken on the same day (point in time) and using the mean of all measures across a given cycle. These analyses used SAS 9.4 (Cary, NC) and included both ovulatory and anovulatory cycles.

Pattern of calciotropic hormones across the menstrual cycle

Differences in 25(OH)D concentrations were explored using a mixed model accounting for multiple observations per woman, with cycle phase (follicular or luteal) or cycle (follicular measures in cycle 1 versus cycle 2) as the predictor. For 1,25(OH)₂D and iPTH, up to six measures per cycle were available allowing for more complex modeling using non-linear mixed models accounting for repeated measures across the cycle. After transforming the cycle day of measurement to an angle in a ‘circle’ of 28 days (one standardized cycle length) we fitted concentrations of 1,25(OH)₂D or iPTH to a cosinor model (Sachs et al., 2013) (only first harmonics included after examining model fit for up to 4 harmonics). The model provides estimates of the absolute difference between the predicted peak and nadir concentrations (with 95% CIs). To reflect clinical interest in possible fluctuation of calciotropic hormones across the menstrual cycle in a general population of young women, (i.e. ‘Do clinical measures require timing to specific cycle phases?’), data included all cycles (ovulatory and anovulatory) and we did not adjust for other covariates. We used SAS 9.4 (Cary, NC) for these analyses.

Patterns of reproductive hormones across the menstrual cycle by category of calciotropic hormones

Given the observed relative stability of 25(OH)D, 1,25(OH)₂D and iPTH across the menstrual cycle (see results below), we examined their relationships with the reproductive hormones by comparing cycles with ‘higher’ and ‘lower’ concentrations of calciotropic hormones. Given the limited number of 25(OH)D measures (mean 2.8 measures per woman), a woman’s 25(OH)D was categorized as higher or lower based on the mean of all 25(OH)D measures across both cycles, using the cut point of 30 ng/ml. This cut point approximates the median concentration in this population and corresponds to the Endocrine Society definition of ‘sufficient’ vitamin D (Institute of Medicine Food,and Drug Board, 2010). With shorter half-lives and more variability, 1,25(OH)₂D and iPTH were categorized as lower or higher based on the mean concentration of all measures in a given cycle. There are no recommendations for target serum concentrations of 1,25(OH)₂D or iPTH, and the range of observed values in this population were within clinically normal ranges (Mayo Foundation for Medical Education and,Research, 2018). Therefore, we used the median concentrations as cut points (median 1,25(OH)2D =105 pmol/l, median iPTH = 36 pg/ml). The categories suggestive of replete vitamin D status were used as the reference categories for analyses (higher 25(OH)D, higher 1,25(OH)₂D, lower iPTH).

Non-linear mixed models with four harmonic terms were used to model trajectories of estradiol, progesterone, LH, FSH and free estradiol across the menstrual cycle (Hambridge et al., 2013) using R package nlme (Pinheiro et al., 2014). Non-linear harmonic models allow for flexible modeling of each hormone and estimation of differences (with 95% CIs) in mean concentrations, amplitude of the curves and phase shifts (Yeung et al., 2013). The cycle day of each measure was scaled to the observed cycle length (first cycle day is assigned 0 and last cycle day is 1) and centered at the time of ovulation (day 0.5). The use of cycle time centered on ovulation accommodates cycles of variable follicular and luteal phase lengths and allows for appropriate cycle phase comparison of hormone concentrations. These analyses were restricted to participants with two ovulatory cycles (N = 71). Adjustment factors, chosen based on the literature and change in estimate, were age (continuous), BMI (continuous) and physical activity index (categorical: high, moderate, low). Further adjustment for parity did not change our interpretation of the data. Plots were generated at mean (age, BMI), or modal (physical activity), values of the covariates: 28 years of age, BMI of 24 kg/m² and moderate physical activity.

Results

Demographics

Participants in this analysis tended to be young (mean age, 28 years; SD, 8.5) with normal BMI (mean BMI, 24 kg/m²; SD, 3.6), educated beyond high school, not currently married, non-smokers and nulliparous (Table I). Few (7%) reported low physical activity. Median 25(OH)D concentration was 29.5 ng/ml (SD 8.4) with little variation by season of blood draw or estimated dietary calciferol intake. Few participants (N = 5 (6%)) had vitamin D deficiency (<20 ng/ml) (Institute of Medicine Food and Drug Board, 2010). No participant had a mean eGFR <60 ml/min/1.73 m² (clinical cut-point for chronic kidney disease stages 3–5). Median concentrations of 25(OH)D, 1,25(OH)₂D and iPTH were generally similar across categories of demographic characteristics (Table I), though 25(OH)D and 1,25(OH)₂D tended to decrease with increasing BMI and increase with higher physical activity.

Table I. Participant characteristics and median concentrations of 25-hydroxyvitamin D (25(OH)D), 1,25-dihydroxyvitamin D (1,25(OH)₂D) and intact parathyroid hormone (iPTH). N = 89 BioCycle participants from western New York State 2005–2007.

			25(OH)D ng/ml		1,25(OH)  2  D pmol/L		iPTH pg/ml
Characteristic		N (%)	Median	Q1, Q3	Median	Q1, Q3	Median	Q1, Q3
	Overall		29.5	(24, 35)	104.9	(90, 121)	35.3	(29, 45)
Age (years)	≤24	39 (44)	29.6	(25, 39)	109.7	(92, 120)	34.3	(29, 48)
	25 to 29	15 (17)	27.2	(21, 34)	94.1	(72, 105)	33.2	(29, 42)
	30 to 34	6 (7)	28.4	(22, 34)	116.1	(100, 130)	43.0	(33, 70)
	35 to 39	18 (20)	29.3	(25, 33)	101.8	(89, 128)	39.8	(35, 45)
	40+	11 (12)	31.2	(27, 34)	112.4	(98, 121)	29.8	(25, 38)
Body Mass Index (kg/m2)	<18.5	3 (3)	29.7	(25, 35)	109.7	(89, 143)	42.5	(36, 44)
	≥18.5 to <25	58 (65)	30.3	(26, 36)	112.0	(99, 128)	35.1	(29, 46)
	≥25 to <30	22 (25)	29.6	(21, 34)	95.8	(82, 105)	33.4	(32, 37)
	≥30	6 (7)	22.1	(20, 24)	97.8	(78, 103)	43.7	(24, 49)
Education beyond High School	Yes	82 (92)	29.2	(24, 35)	104.9	(89, 122)	35.1	(29, 46)
	No	7 (8)	31.8	(22, 34)	104.2	(92, 119)	38.1	(33, 40)
Married	Yes	31 (35)	29.5	(25, 34)	102.6	(85, 125)	36.6	(32, 43)
	No	58 (65)	29.3	(23, 35)	107.3	(92, 120)	34.6	(29, 48)
Current smoker	Yes	6 (7)	29.1	(22, 38)	106.1	(99, 135)	41.1	(34, 52)
	No	83 (93)	29.5	(25, 35)	104.9	(89, 121)	35.0	(29, 44)
Parousa	Yes	25 (28)	31.2	(25, 34)	106.3	(91, 132)	39.0	(32, 44)
	No	63 (71)	28.7	(23, 35)	104.9	(90, 120)	33.6	(29, 46)
Physical Activity	Low	6 (7)	23.6	(21, 28)	89.9	(72, 100)	36.6	(32, 46)
	Medium	30 (34)	26.8	(21, 35)	104.9	(91, 121)	32.5	(27, 49)
	High	53 (60)	32.6	(27, 36)	109.9	(91, 128)	35.8	(31, 44)
Quartile dietary intake of calciferol (μg/d)b	1st	20 (22)	27.6	(21, 34)	97.8	(88, 111)	32.6	(28, 41)
	2nd	21 (24)	33.8	(23, 40)	111.4	(100, 129)	35.3	(31, 45)
	3rd	20 (22)	30.2	(24, 35)	105.6	(93, 117)	34.6	(29, 46)
	4th	28 (31)	29.6	(27, 33)	110.5	(91, 125)	36.3	(30, 46)
Ever Use Hormonal Birth Controla	Yes	54 (61)	29.6	(23, 34)	107.7	(94, 128)	35.8	(29, 46)
	No	34 (38)	28.8	(25, 36)	103.2	(89, 117)	33.3	(29, 44)
Season of blood drawc	Winter	304 (34)	29.0	(24, 36)	108.1	(84, 130)	37.6	(28, 50)
	Spring	229 (25)	27.5	(22, 36)	100.0	(82, 119)	34.0	(26, 44)
	Summer	221 (25)	30.5	(26, 35)	110.5	(90, 135)	36.8	(28, 44)
	Fall	148 (16)	31.4	(23, 36)	103.2	(85, 129)	33.9	(26, 42)

^{a a}Missing: Parous (N = 1), Use of Birth Control (N = 1).

^{b b}Quartiles of dietary calciferol: 1^st: <=2.2 μg/day 2^nd: 2.3–3.6 μg/day 3^rd: 3.7–5.7 μg/day 4^th: >5.7 μg/day.

^{c c}Season of blood draw: Winter (Dec, Jan, Feb) Spring (March, April, May) Summer (June, July, Aug) Fall (Sept, Oct, Nov) N = 902 measures of any calciotropic hormone used as the denominator for percent for these rows.

Q1 25^th percentile, Q3 75^th percentile.

Correlation of calciotropic hormones over time and with each other

On average, each participant contributed 2.8 measures (range 1–4) of 25(OH)D; correlation across the study period was good with an ICC of 0.8 (95% CI: 0.7, 0.9). 1,25(OH)₂D and iPTH were measured on average 10.1 times per participant (range, 4–12) with ICCs of 0.4 (95% CI: 0.3, 0.5) for 1,25(OH)₂D and 0.6 (95% CI: 0.6,0.7) for iPTH.

25(OH)D was positively correlated with 1,25(OH)₂D and negatively correlated with iPTH with the strongest correlation observed between the mean value of 25(OH)D and the mean value of 1,25(OH)₂D in a given cycle (Pearson R = 0.4). 1,25(OH)₂D and iPTH showed no correlation in these data (R = 0.03) (Supplementary Table SI).

Pattern of calciotropic hormones across the menstrual cycle

Mean concentrations of 25(OH)D (30.7 ng/ml SD 8.8) from the follicular phase (N = 155 samples) did not differ from mean concentrations (30.1 ng/ml; SD, 9.0) in the luteal phase (N = 89 samples), P = 0.6 (Fig. 1). In N = 68 women with 25(OH)D measured in the follicular phase in two cycles, mean concentration of 25(OH)D in the follicular phase of cycle 1 (30.8 ng/ml; SD, 8.8) did not differ from the concentration in the follicular phase of cycle 2 (30.9 ng/ml; SD, 8.5), P = 0.8, with a median of 28.5 calendar days between the blood draws for cycle 1 and cycle 2.

Figure 1. Comparison of 25-hydroxyvitamin D concentrations in the follicular and luteal phase. Box plot represents the mean (diamond), median (horizontal line), 25th and 75th percentiles (lower and upper edges of grey box) and distribution to +/− 1.5*interquartile range (whiskers) with outliers (open circles) for measures of 25-hydroxyvitamin D from the follicular or luteal phase in N = 89 self-reported white women in western New York State, 2005–2007. Concentrations do not differ between cycle phases, P = 0.6 (mixed model).

1,25(OH)₂D varied slightly over the menstrual cycle with a nadir in the early follicular phase and a peak in the luteal phase (Fig. 2A). The difference between the peak and the nadir was 8.2 pmol/l (95% CI: 2.6, 13.9). iPTH showed a similar, but mirrored, pattern with a peak in the follicular phase and a nadir in the luteal phase (Fig. 2B). The difference between the peak and the nadir was 4.0 pg/ml (95% CI: 2.3, 5.7).

Figure 2. Trajectory of 1,25-dihydroxyvitamin D (Panel A) and intact parathyroid hormone (Panel B) across the menstrual cycle. Modelled trajectory of 1,25-dihydroxyvitamin D (1,25(OH)₂D) (panel A) and intact parathyroid hormones (iPTH) (panel B) across the menstrual cycle from N = 89 self-reported white women from western New York State, 2005–2007. 1,25(OH)₂D and iPTH were measured up to six times per cycle. Linear mixed-model regression accommodating repeated measures fit with an unadjusted cosinor model after transforming the cycle day of the study visit to an angle. Black solid line is the fitted mean with the grey representing the 95% confidence interval. Y-axes include the 10th to the 90th percentile of the full observed distribution for each hormone.

Associations between calciotropic hormones and reproductive hormones

Compared with women who had higher mean 25(OH)D (≥30 ng/ml), participants with lower 25(OH)D concentrations (below 30 ng/ml) had 13.8% lower mean estradiol (95% CI: −22.0, −4.7) across the cycle. The estradiol peak was also 0.05% lower (95% CI: −0.07, −0.004) and occurred 0.2 days later in the cycle (95% CI: 0.002, 0.4) (Table II, Fig. 3A). The comparison was similar for free estradiol with a 10.8% lower mean concentration (95% CI: −18.0, −2.9). There was an isolated phase shift for progesterone (0.2 days later in the cycle (95% CI: 0.03, 0.3)) among participants with lower mean 25(OH)D (Table II, Fig. 3C). We found no differences by 25(OH)D status in the concentration of FSH or LH (Table II).

Table II. Parameters describing trajectory of reproductive hormones across ovulatory cycles comparing higher and lower categories of 25-hydroxyvitamin D (25(OH)D), intact parathyroid hormone (iPTH) and 1,25-dihydroxyvitamin D (1,25(OH)₂D), for 71 women with ovulatory cycles.

		Estimate  a  95% Confidence Interval
		25(OH)D  b  < 30 ng/ml	iPTH  c  > 36 pg/ml	1,25(OH)  2  D  c  < 105 pmol/L
	Referent	≥ 30 ng/ml	≤ 36 ng/ml	≥ 105 pmol/L
Hormone	Parameter
Estradiol	Change in Mean (%)	−13.8 (−22.0, −4.7)	−12.7 (−18.7, −6.3)	0.8 (−6.2, 8.2)
	Change in Amplitude (%)	−0.05 (−0.07, −0.004)	−0.02 (−0.05, 0.04)	0.01 (−0.03, 0.09)
	Phase Shift (day)	0.2 (0.002, 0.4)	0.2 (0.08, 0.4)	−0.1 (−0.3, 0.1)
Free estradiol	Change in Mean (%)	−10.8 (−18.0, −2.9)	−8.5 (−14.6, −2.0)	0.5 (−6.3, 7.8)
	Change in Amplitude (%)	−0.07 (−0.1, −0.01)	−0.02 (−0.06, 0.06)	0.01 (−0.04, 0.1)
	Phase Shift (day)	0.2 (−0.007, 0.4)	0.2 (0.07, 0.4)	−0.03 (−0.2, 0.2)
FSH	Change in Mean (%)	3.1 (−7.8, 15.3)	0.6 (−7.4, 9.3)	0.5 (−7.4, 9.1)
	Change in Amplitude (%)	0.01 (−0.01, 0.04)	0.004 (−0.01, 0.04)	−0.02 (−0.03, 0.009)
	Phase Shift (day)	−0.02 (−0.2, 0.2)	0.2 (−0.03, 0.5)	−0.2 (−0.5, 0.01)
LH	Change in Mean (%)	4.3 (−6.9, 16.8)	−3.6 (−11.7, 5.1)	−0.3 (−8.7, 8.8)
	Change in Amplitude (%)	−0.01 (−0.03, 0.02)	−0.02 (−0.03, 0.002)	0.03 (−0.002, 0.08)
	Phase Shift (day)	0.2 (−0.08, 0.4)	0.2 (−0.01, 0.4)	−0.3 (−0.5, −0.02)
Progesterone	Change in Mean (%)	3.6 (−4.0, 11.8)	−7.3 (−13.3, −0.9)	2.1 (−5.0, 9.7)
	Change in Amplitude (%)	−0.09 (−0.1, −0.03)	0.01 (−0.04, 0.09)	0.01 (−0.04, 0.1)
	Phase Shift (day)	0.2 (0.03, 0.3)	−0.1 (−0.3, −0.02)	−0.1 (−0.3, 0.01)

^{a a}Estimate from non-linear mixed model adjusted for body mass index, age and physical activity.

^{b b}25(OH)D is the mean of all measures for a given woman.

^{c c}iPTH and 1,25(OH)₂D are categorized using cycle specific means for each woman.

Bold estimates p < 0.05

Figure 3. Fitted concentration of total estradiol (upper panels) and progesterone (lower panels) by category of 25-hydroxyvitamin D (25(OH)D; panel A, C) and intact parathyroid hormone (iPTH, panel B, D) across an ovulatory cycle. Results of a non-linear mixed model comparing participants above or below the median concentration of 25(OH)D or iPTH, adjusted for age, BMI and physical activity. Trajectories plotted for mean or modal values of covariates. A total of N = 71 self-identified white women from upper New York State with two observed ovulatory cycles. Panels A, C solid line: participants with mean of all 25(OH)D measures above 30 ng/ml (referent); dashed line: participants with mean below 30 ng/ml. Panels B, D solid line: participants with a cycle specific mean iPTH below 36 pg/ml (referent); dashed line: participants with mean above 36 pg/ml. Participants with lower vitamin D status (dashed lines in upper panels (A, B) representing lower 25(OH)D or higher iPTH) had approximately 13% lower mean estradiol. Participants with higher iPTH (dashed line in panel D) had approximately 7% lower progesterone.

Comparisons of estradiol in cycles with iPTH above the median versus those below the median were comparable to cycles with lower 25(OH)D: a 12.7% lower mean estradiol (95% CI: −18.7, −6.3) with a small decrease in amplitude and a slight shift in phase (Table II, Fig. 3B). We also saw this pattern for free estradiol. Cycles with iPTH above the median also demonstrated a 7.3% lower progesterone (95% CI: −13.3, −0.9) (Table II, Fig. 3D), but no differences in LH or FSH. We observed no differences for patterns of any reproductive hormones for cycles with 1,25(OH)₂D below versus above the median (Table II).

Discussion

In this population of healthy women, concentrations of 25(OH)D did not vary over the menstrual cycle, while 1,25(OH)₂D and iPTH showed small magnitude fluctuations that mirrored each other with the highest 1,25(OH)₂D, and lowest iPTH, in the luteal phase. Women with 25(OH)D concentrations below 30 ng/ml exhibited lower mean estradiol in ovulatory cycles.

These results should be considered in the context of a population where almost half had a sufficient vitamin D concentration (>30 ng/ml) and only 6% were deficient (<20 ng/ml). Concentrations of 25(OH)D were stable across three measures over two menstrual cycles (usually representing approximately 2 months) with an ICC of 0.8. The high ICC is comparable to those observed over longer timespans (Major,et al., 2013). No difference in 25(OH)D concentrations between the follicular and luteal phase of the menstrual cycle was evident, suggesting that 25(OH)D does not respond to short-term cyclic changes in reproductive hormones. This finding is consistent with the half-life of 2 weeks (Vieth, 2011) and with other studies examining concentrations of 25(OH)D across natural cycles (Buchanan et al., 1986a); Franasiak,Franasiak; Tjellesen et al., 1983) or following ovarian stimulation for fertility treatment (Potashnik et al., 1992). Measurement of 25(OH)D in clinical settings need not be timed to menstrual cycle phase.

As compared to 25(OH)D, however, 1,25(OH)₂D and iPTH have much shorter half-lives and we observed more intra-individual variability and small fluctuations across the menstrual cycle. We observed higher iPTH in the follicular phase and higher 1,25(OH)₂D in the luteal phase in keeping with some (Buchanan et al., 1986; Buchanan et al., 1986; Gray et al., 1982; Tjellesen et al., 1983) but not all (Baran et al., 1980; Muse et al., 1986) prior studies in normally cycling women. The observed inverse relationship between concentrations of 1,25(OH)₂D and iPTH is consistent with known endocrine feedback systems (higher concentrations of 1,25(OH)₂D suppress iPTH). While concentrations of 1,25(OH)₂D and iPTH may be influenced by biological changes during the menstrual cycle, the magnitude was small relative to the overall distribution of the measures (Fig. 2) and unlikely to be clinically significant.

When examining the possibility that higher or lower concentrations of calciotropic hormones could influence cyclic trajectories of the reproductive hormones, we found that mean levels of estradiol and free estradiol were lower among women with concentrations of 25(OH)D below 30 ng/ml or concentrations of iPTH above 36 pg/ml (Fig. 3). The decrease in mean estradiol concentration appeared to be most marked during the ovulatory and mid-luteal peaks. The lack of changes in the patterns of LH or FSH supports the hypothesis that the mechanism of action does not involve the hypothalamus/pituitary.

Concentrations of iPTH tend to increase with lower concentrations of 25(OH)D leading to the possibility that the same women who are categorized as lower 25(OH)D are also categorized as higher iPTH. Although 25(OH)D and iPTH were weakly negatively correlated (Supplementary Table SI) in this cohort, women with 25(OH)D below 30 ng/ml were equally split between higher and lower categories of iPTH, suggesting that the findings observed with lower 25(OH)D are not primarily due to concurrently higher iPTH, and vice versa.

Despite the differences seen between women with higher concentrations of iPTH or lower concentrations of 25(OH)D, we found no differences in patterns of reproductive hormones between women with lower compared to higher 1,25(OH)₂D, the active vitamin D metabolite. Ovaries express 1 α-hydroxylase (Hewison et al., 2011) an enzyme that converts 25(OH)D to 1,25(OH)₂D. If 25(OH)D is acting at the ovary to influence estradiol concentrations, peripheral serum concentrations of 1,25(OH)₂D may not capture the relevant tissue specific (ovarian) 1,25(OH)₂D concentrations.

Our finding of depressed estradiol in women with lower 25(OHD) concentration, without a change in the LH or FSH trajectories, suggests that calciotropic hormones may affect the ovary directly. The presence of the vitamin D receptor in ovarian granulosa cells provides a mechanism through which vitamin D could act directly to modulate estrogen production (Parikh et al., 2010). Human in vitro models have been conducted using 1,25(OH)₂D and not 25(OH)D; however, the in vitro models generally support the hypothesis that calciotropic hormones can act in the ovary to modulate the production of ovarian hormones. Addition of 1,25(OH)₂D to cultured human ovarian cells resulted in increased concentrations of progesterone (Merhi et al., 2014; Parikh et al., 2010) and estradiol (Parikh et al., 2010). In human ovarian tissue, reported changes with 1,25(OH)₂D treatment include increased aromatase activity (Parikh et al., 2010) and decreased anti-Mullerian hormone (AMH) receptor and FSH receptor expression—changes that suggest influence on folliculogenesis and steroid production (Merhi et al., 2014). However, the role of 1,25(OH)₂D on aromatase and AMH action appears to be tissue-specific (Lundqvist,et al., 2011), and the human in vitro studies used tissue collected following ovarian stimulation in women seeking treatment for infertility. The human in vivo literature examining the effect of vitamin D status on the concentration of reproductive hormones in non-pregnant reproductive age women is restricted to cross-sectional measures of calciotropic hormones and reproductive hormones (Chang et al., 2014; Knight et al., 2010). Although we observed an association between vitamin D status and the trajectory of estradiol, a causal direction cannot be inferred from our study. We cannot rule out the possibility that just as exogenous estrogen treatment seems to increase 25(OH)D concentrations (Garcia-Bailo et al., 2013; Harmon et al., 2016), longer-term exposure to higher endogenous estradiol levels may also increase 25(OH)D concentrations possibly through hepatic pathways where most vitamin D is converted to 25(OH)D. Further in vitro and in vivo studies are needed to elucidate mechanisms.

Biospecimens used in this analysis had been at −80°C for up to 9 years. Although concentrations of 25(OH)D are stable with long-term storage (Jukic et al., 2018a) less is reported about the impacts of long term storage on the concentrations of 1,25(OH)₂D and iPTH. Possible degradation of these metabolites over time may have occurred resulting in measurement error, although the ranking of individuals (higher versus lower) should remain valid. Additional work on the possible effects of storage time and conditions for 1,25(OH)₂D and iPTH would provide useful information on the potential for measurement error.

We compared vitamin D and hormonal trajectories during two menstrual cycles, but longer-term effects of these interactions may be reflected in overall menstrual cycle characteristics including cycle length and regularity (Mumford et al., 2012). Given that an eligibility criterion for the study was self-reported cycle length of 21 to 35 days, we did not have sufficient variability to examine associations of calciotropic hormones with these cycle characteristics. However, we see some suggestive evidence of a lengthening of the follicular phase among women with lower 25(OH)D. Among the few (N = 22) ovulatory cycles with a follicular length longer than 17 days, approximately two-thirds were from women with 25(OH)D concentrations <30 ng/ml. The pattern of lower estradiol and progesterone has also been reported in ovulatory cycles in women with sporadic anovulation (Hambridge,et al., 2013). Anovulation and delayed ovulation may result in long or irregular cycles. Low vitamin D status has been linked to long menstrual cycles, (Jukic et al., 2016) irregular cycles (Jukic et al., 2015) and delayed ovulation (Jukic et al., 2018b).

Along with limited cycle length variability, our total study sample was small. Few participants had deficient 25(OH)D concentrations and we included only Caucasian women. Participants with lower vitamin D status in this population were still mostly within a normal range; therefore, we were unable to test association with concentrations characteristic of vitamin D deficiency. Therefore, our results may not be generalizable to more unselected populations. The study protocol did, however, yield frequent well-timed specimens across the menstrual cycle and thereby allowed for modeling of reproductive hormones in a non-clinical population that likely represents usual menstrual cycle patterns.

Conclusion

In conclusion, we found that concentrations of 25(OH)D do not vary across the menstrual cycle and that women with lower concentrations of 25(OH)D, or higher concentrations of iPTH, have lower mean concentrations of estradiol. The first finding confirms current clinical practice that does not time clinical measures of 25(OH)D to the menstrual cycle. Although this study was not designed to identify a mechanism of action, the findings provide further support to observed associations between vitamin D status and reproductive function.

Further in vitro work or clinical intervention trials are needed to elucidate the mechanism(s) of action. Similar studies in larger populations with more vitamin D deficiency and greater variability in menstrual cycle characteristics are needed to expand the generalizability of these findings.

Author’s roles

K. Kissell, K. Kim, L. Sjaarda, N.J. Perkins, E.F. Schisterman, S.L. Mumford: study design and data acquisition. Q.E. Harmon, A.M.Z. Jukic, D.M. Umbach, D.D. Baird: data analysis. Q.E. Harmon wrote the first draft of the manuscript. All authors contributed to critical discussion, revisions to the manuscript and approval of the final submission.

Funding

Intramural Research Programs of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health (contract numbers: HHSN275200403394C, HHSN275201100002I and Task 1 HHSN27500001); National Institute of Environmental Health Sciences.

Conflict of interest

The authors report no conflicts of interest.