Target trial emulation of preconception serum vitamin D status on fertility outcomes: a couples-based approach.

Of 2,370 couples included in this analysis, 19.5% (N=463) of females were 25(OH)D deficient, with a mean (standard deviation) level of 14.9 (3.8) ng/mL. Among females with insufficient 25(OH)D (39.3%, N=931) and sufficient 25(OH)D (41.2%, N=976), the mean (standard deviation) of 25(OH)D levels were 24.8 ng/mL (2.8) and 39.3 ng/mL (10.2), respectively. Thirty percent (N=709) of males were 25(OH)D deficient with mean 25(OH)D levels of 14.6 ng/mL (3.5), whereas 40.7% (N=964) were insufficient and 29.4% (N=697) sufficient with mean levels of 24.5 ng/mL (2.8) and 37.5 ng/mL (8.2). On average, females and males with deficient 25(OH)D had higher BMIs, were more likely to self-identify as non-Hispanic White, and reported lower levels of physical activity and HEI scores than those with higher 25(OH)D concentrations ( Table 1 ). Baseline characteristics prior to multiple imputation are presented in Supplemental Table 2 . After adjustment, females with sufficient or insufficient 25(OH)D had higher likelihood of live birth compared to females with deficient 25(OH)D (insufficient vs. deficient: aRR 1.18, 95% CI 0.98–1.42; sufficient vs. deficient: aRR 1.28, 95% CI 1.05–1.56; Table 2 ). This association varied across BMI subgroups, with increasing 25(OH)D and higher likelihood of live birth among females of normal (insufficient vs. deficient: aRR 1.24, 95% CI 0.89–1.74; sufficient vs. deficient: aRR 1.39, 95% CI 1.00–1.93) or obese BMI (insufficient vs. deficient: aRR 1.30, 95% CI 0.96–1.77; sufficient vs. deficient: aRR 1.33, 95% CI 0.95–1.85), though results were imprecise. Null results were found among females of overweight BMI (insufficient vs. deficient: aRR 1.04, 95% CI 0.73–1.54; sufficient vs. deficient: aRR 1.03, 95% CI 0.70–1.51). In the overall models, there were no significant differences in the likelihood of live birth by male 25(OH)D status. Stratified by BMI, preconception 25(OH)D was positively associated with the likelihood of live birth among males with normal BMI (insufficient vs. deficient: aRR 1.60, 95% CI 1.11–2.33; sufficient vs. deficient: aRR 1.51, 95% CI 1.01–2.25) but not among males with overweight or obese BMI. There were no associations between female or male 25(OH)D and pregnancy loss overall or stratified by BMI ( Table 2 ). Similar results were demonstrated for the likelihood of live birth and pregnancy loss by female and male VDBP, free vitamin D, and bioavailable vitamin D ( Supplemental Table 3 ). Associations varied by BMI, with preconception VDBP and bioavailable vitamin D positively associated with live birth among female and male partners with normal BMI and female partners with obese BMI ( Supplemental Table 4 ). However, females with higher concentrations of VDBP had an increased risk of pregnancy loss, though confidence intervals were wide. There were no associations between calcium and live birth or pregnancy loss. Overall, the results of vitamin D biomarkers and live birth and pregnancy loss were similar to the results of 25(OH)D. Among couples, 61.2% were both insufficient/sufficient 25(OH)D (N=1450), while only 10.6% (N=252) were both deficient ( Table 3 ). Couples whom both had insufficient/sufficient 25(OH)D had increased likelihood of live birth compared to couples in whom both had deficient status (aRR 1.26, 95% CI 1.00–1.58). There were no associations between couple 25(OH)D and pregnancy loss. Of 1,773 males (74.8%) who attended the 6-month study visit, there were no differences in semen quality parameters by 25(OH)D status overall, stratified by BMI or by vitamin D biomarkers ( Table 4 , Supplemental Table 6 , Supplemental Table 7 ). Likelihood of morphology <4% normal was slightly higher with increasing vitamin D status in the continuous models, but no association was observed in the categorical models ( Table 4 ). Sensitivity analyses examining semen outcomes longitudinally did not alter results appreciably ( Supplemental Table 8 ).

This was a secondary analysis using prospective data from the Folic Acid and Zinc Supplementation Trial (FAZST) and the Impact of Diet, Exercise, and Lifestyle (IDEAL) on Fertility study.[ 22 ] FAZST was a multi-center, double-blind, block-randomized, placebo-controlled trial that examined the effects of male folic acid and zinc supplementation among couples seeking infertility treatment on live birth.[ 22 , 23 ] IDEAL included more detailed longitudinal follow-up in a subset of female partners linked to FAZST.[ 24 ] The trial design, eligibility criteria, and main findings have been described.[ 22 – 24 ] Couples seeking infertility treatment at reproductive endocrinology and infertility care study centers in Utah, Iowa, Illinois, and Minnesota were recruited (2013–2017) and were eligible for enrollment. Recruitment also targeted couples planning infertility treatments at general obstetrics and gynecology practices to be more broadly inclusive of a general infertility care population seeking a range of treatment modalities.[ 22 ] Couples were ineligible if they were planning the use of donor sperm or a gestational surrogate, were pregnant at enrollment, or if the male had poorly controlled chronic diseases (e.g., cancer, diabetes, hypertension), obstructive azoospermia, or other known infertility causes. Couples were followed for up to 9 months post-enrollment and throughout any resulting pregnancy (up to 18 months). Institutional Review Boards at all study centers and the data coordinating center approved the trial. Written informed consent was obtained for all participants. A Data and Safety Monitoring Board provided external oversight. The trial was registered on ClinicalTrials.gov ( NCT01857310 ). Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of the research. Serum samples were collected at enrollment in both female and male partners and stored at −80°C until analysis. Markers of vitamin D metabolism measured for this study include 25(OH)D, VDBP, calcium, free vitamin D, and bioavailable vitamin D. Combined concentrations of 25(OH)D D2 and D3, were measured using chemiluminescent microparticle immunoassay, which is equipotent for both D2 and D3 (Abbott) with demonstrated acceptability.[ 25 ] Immunoassays have not been shown to cross-react with 3-epi-25(OH)D3.[ 26 ] The interassay laboratory coefficient of variations (CV) was <10% for lyophilized manufacturer’s controls and 17% for an in-house pooled serum control. VDBP was measured using LC-MS/MS with interassay laboratory CVs <3.4% for both partners.[ 27 ] Calcium was measured with the Roche COBAS 6000 chemistry analyzer (Roche Diagnostics, Indianapolis, IN) using the colorimetric (NM-BAPTA) method (Roche Diagnostics, Indianapolis, IN) with CV of 1.3%. Albumin was measured using the Bromcresol Purple (Roche Diagnostics, Indianapolis, IN) on the Roche COBAS 6000 chemistry analyzer (CV 2.6%). Measured values of total 25(OH)D, VDBP, and albumin were used to calculate free 25(OH)D and bioavailable 25(OH)D according to the equation by Powe et al.[ 28 ] Primary outcomes of live birth and pregnancy loss were ascertained through medical record review, including obstetric records of the prenatal care provider and hospital records for incidental visits and delivery. Pregnancy loss included all pregnancy losses before 20 weeks’ gestation, which captured early pregnancy losses, defined as hCG pregnancy losses [serum hCG >5 mIU/mL followed by a decline prior to ultrasound visualization of pregnancy], and clinically recognized pregnancy losses, defined as pregnancy seen on ultrasound followed by a loss <20 weeks’ gestation. Secondary outcomes were semen quality parameters at 6 months post-enrollment, including sperm concentration, volume, motility, morphology, total motile count, and DNA fragmentation index (DFI).[ 29 ] Male participants completed this study visit regardless of pregnancy status at the 6-month visit. Each laboratory underwent standardized training and inter-laboratory quality control testing to comply with WHO 5 criteria.[ 30 ] DFI was determined at a central laboratory via Comet assay.[ 31 , 32 ] Target trial emulation is a framework used to analyze observational data for causal inference. Causal analyses of observational data using this framework minimize common biases in observational studies and have led to effect estimates similar to those from randomized trials.[ 16 , 33 – 35 ] Additionally, a randomized trial to estimate the effect of preconception 25(OH)D on live birth and pregnancy loss among couples seeking infertility treatment would be difficult to implement given the large number of couples needed, length of follow-up time, and overall cost. We identified and summarized a target trial protocol to estimate the effect of all possible sources of combined preconception 25(OH)D on live birth and pregnancy loss ( Supplemental Table 1 ). Of note, this is not equivalent to vitamin D supplementation, rather this design takes into account all sources of vitamin D that may lead to changes in preconception 25(OH)D, including sun exposure, supplement use, and dietary intake (this is more akin to a behavioral modification trial rather than a trial trial). In the target trial, participants are randomly assigned to pre-enrollment changes to increase or lower vitamin D intake to obtain 25(OH)D concentrations at enrollment reflecting deficient, insufficient, or sufficient status. Three trials were conceptualized, only varying the unit of randomization: 1) female partner; 2) male partner; 3) couple, with follow-up from enrollment until live birth, loss to follow-up, 9-months post-randomization, or throughout any pregnancies conceived during follow-up. We considered block randomization by BMI subgroups based on the mechanism that obesity induces inflammation,[ 36 ] whereas vitamin D has an anti-inflammatory effect.[ 37 ] Considering these individual relationships, we hypothesize that treatment effect heterogeneity in preconception vitamin D on fertility outcomes may exist. We acknowledge the limitations of an intervention in this setting is not well-defined as there are multiple potential interventions on preconception 25(OH)D; therefore, future investigators would need to choose a specific intervention by which to alter 25(OH)D. To emulate the target trial, we applied the same eligibility criteria, start and end of follow-up, and statistical analysis from the target trial protocol. Randomization of pre-enrollment 25(OH)D was emulated via outcome regression adjustment for variables identified from directed acyclic graphs (DAG) ( Supplemental Figure 1 ). Individual 25(OH)D status was categorized as deficient (<20 ng/mL), insufficient (20 to <30 ng/mL), or sufficient (≥30 ng/mL) based on cut points from the Institute of Medicine and Endocrine Society.[ 38 ] Tercile categories for other biomarkers were used, given the lack of standardized cut points. Couple 25(OH)D status was categorized as both deficient, male deficient and female not deficient (including insufficient/sufficient), female deficient and male not deficient, and both not deficient; we also considered an alternate categorical grouping combining not sufficient compared to sufficient. Female and male partner characteristics were compared by preconception 25(OH)D status. Baseline characteristics obtained upon enrollment included age, BMI (calculated from measured height and weight at enrollment), self-identified race and ethnicity, education, employment, marital status, annual household income, health insurance, fertility treatment insurance, time trying to conceive prior to enrollment, physical activity,[ 39 ] parity, infertility diagnosis, and diet quality.[ 39 ] [ 40 , 41 ] Diet quality was assessed using the Healthy Eating Index (HEI) calculated from a food frequency questionnaire.[ 42 ] Multiple imputation accounted for missing exposure, outcome, and covariate data. Mean 25(OH)D concentrations were compared by primary outcomes and frequency of outcomes by 25(OH)D status for both partners individually and jointly. Log-binomial regression models were used to estimate relative risk (RR) and 95% confidence intervals (CIs) for live birth and pregnancy loss for each marker of vitamin D metabolism (25(OH)D, VDBP, calcium, free vitamin D, bioavailable vitamin D). Adjusted RR (aRR) accounted for confounders guided by DAGs ( Supplemental Figure 1 ), specifically parity, age, physical activity, nutritional status, body mass index, race, time trying to conceive prior to trial enrollment, season of enrollment, education level, and treatment assignment. Though serum was collected at the time of randomization and should not be associated with FAZST treatment assignment, we included treatment assignment in the models to account for any potential influence of supplementation in the male partners despite the null findings of the trial overall.[ 23 ] Stabilized inverse probability weights were used to control for potential selection bias in pregnancy loss analyses, as the analysis was restricted to couples with hCG-detected pregnancies.[ 43 ] Weights included factors associated with the probability of pregnancy, including baseline age, BMI, race and ethnicity, physical activity, parity, season of randomization, treatment assignment, and 25(OH)D concentrations. Given that 25(OH)D status cut points defined by the Institute of Medicine are designed for non-pregnant persons with outcomes of bone health in mind, we modeled 25(OH)D as a nonlinear spline to allow for greater flexibility and to test if the cutoffs were appropriate ( Supplemental Figure 2 ). We examined preconception BMI (underweight [present in <5% of cohort] and normal, overweight, obese) as an effect measure modifier using stratified models on the multiplicative scale. Semen quality parameters at 6-months post-enrollment were evaluated as binary outcomes using clinically-defined low semen parameters based on WHO cut points, except DFI, which does not have a standardized cut point and was examined as continuous variables.[ 29 ] Log-binomial regression models were used to estimate RR and 95% CIs for binary outcomes for each marker of vitamin D metabolism. Linear regression models were used for DFI. Models included stabilized IPW to account for potential selection bias due to missing the 6-month visit calculated using baseline characteristics and pregnancy status at that visit. Effect modification by BMI was examined. As a sensitivity analysis, longitudinal semen quality outcomes from baseline, 2-, 4-, and 6-months post-enrollment were examined using generalized estimating equations with robust standard errors. All analyses were performed using SAS version 9.4 (Cary, NC).

Among couples seeking infertility treatment, preconception vitamin D for both partners with normal BMI or females with obese BMI was associated with an increased likelihood of live birth. Ensuring sufficient levels of preconception vitamin D may provide an opportunity to improve fertility.

The majority of preconception markers of vitamin D metabolism examined, particularly 25(OH)D, VDBP, and bioavailable vitamin D status, were positively associated with the likelihood of live birth among couples with normal BMI and females with obese BMI in a cohort of couples seeking infertility treatment. No associations were found with pregnancy loss or semen quality parameters, suggesting that effects of vitamin D on fertility are not mediated by semen quality. These findings underscore the importance of preconception vitamin D, including bioavailability and consideration of BMI for both partners, among couples seeking infertility treatment. Our results are consistent with prior studies finding a positive association between female vitamin D and live birth, though effect modification by BMI has not previously been examined.[ 5 , 7 – 9 , 13 ] Importantly, we found that concentrations of vitamin D metabolites in the female partners were associated with live birth among those with normal weight or obesity; although, this was not observed with calcium. Of note, calcium is very tightly controlled, and therefore, there may not have been enough variation in calcium levels to detect a difference in outcomes by calcium tercile groups. Obesity and weight gain is associated with chronic inflammation, which is associated with negative fertility outcomes, and vitamin D has well-known anti-inflammatory properties.[ 37 , 44 ] Individuals with normal BMI may have lower chronic inflammation levels than those with classified overweight or obese BMI status, which may be because those of normal BMI have lower levels of systemic inflammation, allowing vitamin D to function more effectively in improving fertility outcomes. These results may suggest that future interventions targeting couples of normal BMI may be particularly beneficial, as increasing vitamin D status may increase likelihood of live birth. Further, our study adds to the literature by including couples seeking various infertility treatments and primiparous females. We also considered various biomarkers of vitamin D in addition to 25(OH)D, which highlight that VDBP, free vitamin D, and bioavailable vitamin D are also relevant for reproductive health. 25(OH)D is considered a reliable, standard clinical measure of circulating vitamin D, and the consistency of results across multiple biomarkers confirms the positive 25(OH)D findings of the primary analysis. Limited epidemiological data exist on the impact of male vitamin D on fertility, despite the presence of vitamin D receptors and metabolizing enzymes in the testis and prostate.[ 2 ] However, reported associations between vitamin D status and semen quality are inconsistent.[ 11 , 12 ] In a randomized controlled trial involving 330 males with diagnosed infertility, preconception vitamin D supplementation did not significantly improve live birth rates compared to placebo.[ 12 ] Interestingly, we found that vitamin D status in male partners was only associated with live birth among those with normal BMI, and no associations were found with semen quality parameters. However, the confidence interval for the association with live birth was wide, suggesting that these results should be interpreted cautiously. The results of the previous study may be more relevant to males with impaired semen quality than the broad population seeking infertility care. Our findings align with existing epidemiological literature suggesting that any effect of preconception vitamin D on fertility outcomes is not mediated by semen quality.[ 11 , 12 ] However, mechanisms by which preconception vitamin D impacts fertility outcomes not mediated through semen quality remain to be explored, though may be through effects on hormonal regulation, immune modulation, or epigenetic modifications. No differences in risk of pregnancy loss were found based on vitamin D status, aligning with prior studies. One study found a modestly protective effect of preconception vitamin D on pregnancy loss,[ 5 , 45 ] but there may have been unmeasured confounding due to the omission of preconception nutrition.[ 46 ] Interestingly, one prospective study found that deficient preconception vitamin D status were associated with reduced fecundability among women without infertility history, yet this analysis did not include nutritional status of the participants.[ 8 ] This suggests that vitamin D may increase the probability of conceiving without affecting pregnancy loss, consistent with our findings. Our study adds to the existing literature by examining preconception vitamin D among a broad cohort of couples seeking infertility therapy. While more than 20% used IVF, the majority used lower technology treatments, and thus, these results are generalizable to couples seeking multiple types of infertility treatment. Additionally, we used data from an RCT with high follow-up rates and reliable measurement of exposure and outcome.[ 23 ] Further, we analyzed multiple biomarkers of vitamin D and considered individual partner and joint couple models. Our conclusions are strengthened by the finding that the relationship between vitamin D and live birth was mainly consistent across biomarkers of vitamin D, and similar to findings on 25(OH)D. This study has several limitations inherent to the study design and primary goals. First, preconception vitamin D was not randomized, thus we attempted to emulate randomization using adjustment for a wide range of possible confounders. Although we could adjust for factors not included in prior studies, such as diet and physical activity, the possibility of unmeasured confounding (e.g., genetics) remains. Second, different interventions could be used to modify preconception vitamin D status, and the specific type of intervention that would have the desired impact on live birth remains uncertain; further study is needed to explore the effectiveness of various interventions. A randomized controlled trial of oral vitamin D supplementation for females using IVF with insufficient or deficient vitamin D status showed no difference in clinical pregnancy rates; however, the extent to which these findings are generalizable to couples using a broader range of infertility treatments is unknown.[ 47 ] Additionally, the study population was primarily non-Hispanic White, with high educational attainment and health insurance, which limits the generalizability of the findings. Finally, we did not have data on vitamin D supplementation among participants in the study. Therefore, the findings from our study cannot be directly translated into guidelines given that vitamin D levels are a result of heterogeneous sources.

The importance of vitamin D in reproduction is evidenced by its involvement in key processes of conception including spermatogenesis, folliculogenesis, and implantation.[ 1 , 2 ] Vitamin D receptors are present in central and peripheral reproductive organs, such as the hypothalamus, and male and female reproductive tracts.[ 1 , 2 ] However, 18% of US adults do not meet the recommended status for vitamin D.[ 3 ] Confirming the importance and understanding the role of vitamin D for reproductive outcomes are vital, as low-cost interventions like supplementation and diet to increase vitamin D might enhance reproductive health.[ 4 ] Previous studies have suggested a positive association between preconception 25-hydroxyvitamin D (25(OH)D) and live birth and a negative association with pregnancy loss among females without infertility history.[ 5 – 8 ] Among those seeking infertility treatment, females with sufficient concentrations of 25(OH)D may have higher rates of clinical pregnancy or live birth with in vitro fertilization (IVF).[ 7 , 9 , 10 ] Most studies have not considered the role of vitamin D in male partners. There is evidence suggesting preconception 25(OH)D may impact semen quality, although findings have been inconsistent.[ 11 , 12 ] Only one small study of 132 couples examined 25(OH)D in both female and male partners, finding a positive association between 25(OH)D and clinical pregnancy and live birth.[ 13 ] Most studies focus on couples undergoing IVF and may lack generalizability to couples seeking more common, less invasive treatments. Additionally, these studies often concentrate solely on 25(OH)D, without considering other markers of vitamin D metabolism, such as vitamin D binding protein (VDBP) or calcium which may be especially relevant given the potential influence of vitamin D bioavailability in influencing reproductive outcomes.[ 14 , 15 ] This study aims to emulate a target trial using observational data to assess the associations between preconception vitamin D (including serum 25(OH)D concentrations and other markers of vitamin D metabolism) and live birth and pregnancy loss in a large cohort of couples seeking infertility treatment.[ 16 ] We also investigate the potential effect modification by body mass index (BMI), as BMI is associated both with serum vitamin D and with outcomes of infertility treatment.[ 17 , 18 ] [ 19 ] BMI has been associated with chronic inflammation and vitamin D has known anti-inflammatory properties. We hypothesize that vitamin D may have different effects by BMI, and other analyses of preconception exposures have observed treatment effect heterogeneity by preconception BMI.[ 5 , 20 , 21 ] Furthermore, we examine associations between vitamin D and semen quality parameters among male partners to explore potential mechanisms.