Vitamin D supplementation before in vitro fertilisation in women with polycystic ovary syndrome: multicentre, double blind, placebo controlled, randomised clinical trial.

Polycystic ovary syndrome (PCOS) is a heterogeneous clinical syndrome characterised by at least two of the following three criteria: excess of androgens, oligo-ovulation or anovulation, and the presence of polycystic ovary morphology. 1 PCOS is a common cause of infertility, affecting approximately 5-20% of reproductive aged women worldwide, with varying prevalence across different populations and diagnostic criteria. 2 3 4 5 Some studies suggest that women with PCOS undergoing in vitro fertilisation (IVF) may have increased risk of fertilisation failure, biochemical pregnancy loss, and miscarriage compared with women without PCOS. 6 Additionally, among those who achieve pregnancy, the incidence of adverse perinatal outcomes, including gestational diabetes mellitus, hypertensive disorders of pregnancy, preterm birth, and caesarean delivery, is considerably higher in women with PCOS. 7 8 9 10 The reported prevalence of vitamin D deficiency and insufficiency among women of reproductive age ranges from 30% to 40% across various large scale studies. 11 12 13 This issue is even more pronounced among women with PCOS. 14 Severe vitamin D deficiency (serum 25-hydroxyvitamin D (25-OHD) levels <10 ng/ml) is more prevalent in women with PCOS than in those without the condition (44% v 11%). 15 Women with PCOS who are scheduled for ovulation induction and have vitamin D deficiency are less likely to ovulate and have lower live birth rates. 16 17 Additionally, beyond its metabolic effects, vitamin D receptors are expressed in ovarian tissue, endometrium, and placenta, suggesting potential involvement of vitamin D in folliculogenesis, endometrial receptivity, and implantation. 18 Therefore, vitamin D supplementation has been proposed as an adjunctive treatment for women with infertility—especially PCOS—before IVF to improve fertility outcomes. Evidence on vitamin D supplementation in women with infertility is conflicting. A meta-analysis based on nine randomised controlled trials and three cohort studies suggested increased clinical pregnancy rates in women with vitamin D supplementation. 19 However, the largest included trial (630 participants) did not show improvement in clinical pregnancy rates with a single high dose bolus (600 000 IU) of vitamin D before IVF. 20 It remains unclear which metabolic or reproductive outcomes are influenced by vitamin D supplementation, let alone if the impact would be greater for women with PCOS. Existing randomised controlled trials in general infertility populations are limited by small sample sizes, short intervention durations, single centre design, and insufficient follow-up time. Therefore, we conducted a multicentre, randomised, double blind, placebo controlled trial to evaluate the effects of supplementing vitamin D (4000 IU daily) for up to about 12 weeks (up to 90 days) before IVF on live birth rates in women with PCOS.

Our study was a multicentre, randomised controlled trial originally planned at six sites across China and expanded to 24 sites to address pandemic related recruitment challenges during covid-19 lockdowns (supplementary table 1). The study was approved by the ethics committee of the Women’s Hospital of Zhejiang University (IRB-20200035-R), followed by ethics committee approval from each centre. The trial was prospectively registered with ClinicalTrials.gov ( NCT04082650 ) on 9 September 2019 and the trial protocol has been published previously. 21 The trial was started on 15 October 2020 and the follow-up of live births was completed in May 2024. Study participants were women aged 20-42 years diagnosed as having PCOS according to the Rotterdam criteria, 1 and scheduled for autologous IVF with a fresh or freeze-all embryo transfer strategy. Exclusion criteria were participants who had experienced three or more failed IVF cycles, if they were scheduled for pre-implantation genetic testing, oocyte donation IVF cycles, or with a known allergy to vitamin D. Participants receiving treatments for tuberculosis, convulsions, and epilepsy were also excluded because drugs treating these diseases may affect vitamin D or calcium metabolism. Detailed criteria for participants are summarised in the published protocol. 21 Trial coordinators at each centre explained the study protocol to potential participants before the IVF cycle. After giving informed consent, participants were randomly assigned 1:1 to receive vitamin D or placebo. Randomisation was conducted using a computer generated list, with variable block sizes of two, four, or six, stratified by study centre. This list was prepared by an independent statistician not involved in participant recruitment and data analysis, and it was not accessible to investigators or recruiting staff at any centre. The randomisation list was implemented using identical study drug containers, which were centrally labelled by two researchers who did not participate in the recruitment process. Vitamin D and placebo capsules were produced by a licensed pharmaceutical company (Sinopharm Xingsha Pharmaceuticals, Xiamen, China). Placebo capsules, composed of pharmaceutical grade gelatin, glycerin, and purified water, were identical in size, colour, appearance, and packaging to the vitamin D capsules. Participants and trial investigators were unaware of the treatment allocation, except for the two researchers who labelled the containers. Treatment allocation was not revealed until the completion of the analysis. Participants in the intervention group were treated with vitamin D 4000 IU/day (800 IU per capsule, five capsules taken once a day; each vitamin D capsule contained 720-960 IU calcium-free cholecalciferol), starting before IVF for up to about 12 weeks (up to 90 days) until the date of triggering. Participants in the placebo group were treated with five placebo capsules each day for the same duration. The IVF process is described in the published protocol. 21 In brief, pituitary downregulation was achieved using the gonadotropin releasing hormone agonist or antagonist protocol. Gonadotropin dosing was titrated based on ovarian response monitored by serial ultrasonographic follicle tracking and serum oestradiol measurements. Triggering occurred when at least two follicles had a diameter of 18 mm or greater, followed by transvaginal oocyte retrieval 36-37 hours later. Fertilisation was performed through conventional IVF or intracytoplasmic sperm injection four to six hours after retrieval. For the freeze-all strategy, embryos were frozen by vitrification on day 3 or day 5 after oocyte retrieval. Up to two of the embryos at the cleavage stage or blastocyst stage were transferred in a cycle under ultrasound guidance using a soft transfer catheter. Luteal phase support and embryo transfer strategies followed centre specific protocols. The primary outcome was live birth after the first embryo transfer. Pregnancies conceived without fertility treatment that occurred before the first embryo transfer but after randomisation were also included. Secondary outcomes were vitamin D levels on trigger day and pregnancy test day (12-14 days after embryo transfer), other fertility and pregnancy outcomes, and cumulative conception leading to live birth at six months after randomisation (supplementary table 2). We added cumulative conception leading to live birth at 12 months after randomisation as a post hoc secondary outcome because of the impact of covid-19 on delaying embryo transfer in some patients. Supplementary table 2 presents details of these outcomes and related definitions. All adverse events and serious adverse events were monitored and systematically collected at each scheduled study visit using a structured case report form. Moderate or severe ovarian hyperstimulation syndrome was prespecified as a safety outcome and was assessed for two to four weeks after triggering. Participants were provided with a structured diary to record daily drug intake and any adverse events at all study visits (see participant daily diary form in supplementary appendix). Research investigators in each centre reviewed these diaries during follow-up. Adherence was quantitatively monitored by collection of study drug containers at each clinic visit and an electronic drug diary on the patient system that recorded the doses each day. Peripheral venous blood was collected from each participant at baseline, on trigger day, and on pregnancy test day (12-14 days after embryo transfer). Blood samples were centrifuged at 3000 rpm and 4°C for 10 minutes. The supernatants were collected into two microcentrifuge tubes and were stored at −80°C until batch analysis. Serum concentrations of 25-OHD from all centres were measured using a well standardised isotope dilution liquid chromatography tandem mass spectrometry method with an automatic analyser (SCIEX Triple Quad 4500MD) using the vitamin D 200M Assay Kit (DISIGNS Diagnostics, China) in Hangzhou Dian Medical Laboratory Center Co, which held the National Center for Clinical Laboratories external quality assessment certificate. 22 Each sample was tested three times and the average 25-OHD level was calculated as the final result. The live birth rate of women with PCOS after the first embryo transfer for each started cycle was estimated to be 38% at the Women’s Hospital of Zhejiang University. To show or refute a 10% increase in the live birth rate in the vitamin D treatment group (48%), a total of 768 participants were required (power 80% and type I error rate 5%). Considering possible protocol violations and a loss of follow-up of around 10%, we planned to enrol 860 participants (430 in each group). Statistical analysis was conducted according to a prespecified statistical analysis plan (supplementary appendix). The primary analysis was based on a modified intention-to-treat principle excluding participants after randomisation (eg, ineligible participants). Missing data for the primary outcome (live birth) and other fertility outcomes were imputed as negative irrespective of the reason why data were not recorded. We performed a complete case analysis as a sensitivity analysis. Per protocol analysis was also conducted, but these results were considered exploratory only. The per protocol population excluded participants who did not receive the intervention or placebo, or those with poor adherence (intake of study drug <85% or duration of intake <6 weeks). Descriptive statistics were used for baseline variables. We present continuous variables as means and standard deviations or medians and interquartile ranges for each group, depending on the normality of the distribution. For binary outcomes, the counts and proportions of the two groups are given. Adjusted (adjusted for study centres) and unadjusted risk ratios or mean differences and 95% confidence intervals (CIs) between the two treatment groups were calculated and the adjusted estimates used as the primary analysis. For unadjusted estimates, modified Poisson regression with a robust variance estimate was used. For adjusted estimates, a random effects model was used to account for centre effects. 23 Subgroup analysis (treatment covariate interaction) based on baseline vitamin D levels was performed for the primary outcome. Vitamin D level was considered a continuous outcome, and an interaction term was added in the regression model. Nonlinear interaction was also explored using the multivariable fractional polynomial interaction approach. 24 Models tested included linear, fractional polynomials with one power term, and fractional polynomials with two power terms, and the selection of the best fitting model was based on the smallest Akaike information criterion. 25 All analyses were performed using Stata 18.0 (StataCorp, USA). A two tailed P value <0.05 was considered to indicate statistical significance. Investigators who performed the analysis were masked to the allocation. No patients or members of the public were involved in the study design, recruitment, conduct, or interpretation of the results. The study was designed before patient and public involvement became common.

Between 15 October 2020 and 19 February 2022, 1272 women were screened, of whom 365 declined to participate and 31 were ineligible, including one who was pregnant. Finally, 876 participants were randomised and 865 were included in the primary intention-to-treat analysis, with 435 in the vitamin D group and 430 in the placebo group ( fig 1 ). Baseline characteristics were similar for each group ( table 1 ). The most common PCOS phenotypes were type D and A, accounting for about 60% and 30% of participants, respectively. Mean baseline serum 25-OHD levels were 16.5±7.2 and 16.1±6.7 ng/mL in the vitamin D and placebo groups, respectively. Adherence to the intervention was 75.2% (327/435) in the vitamin D group and 74.2% (319/430) in the placebo group. CONSORT (consolidated standards of reporting trials) flowchart. *One participant did not meet Rotterdam criteria for polycystic ovary syndrome (PCOS), one had epilepsy, and one was scheduled for preimplantation genetic testing. †One participant did not meet Rotterdam criteria for PCOS, one had abnormal liver function, one was not scheduled for in vitro fertilisation (IVF) or intracytoplasmic sperm injection (ICSI), and one was diagnosed as having tuberculosis Characteristics of participants at baseline Data are numbers (%) unless stated otherwise. ICSI=intracytoplasmic sperm injection; IQR=interquartile range; IVF=in vitro fertilisation; PCOS=polycystic ovary syndrome; SD=standard deviation. Missing data: one in vitamin D group, two in placebo group. Missing data: three in each group. Missing data: two in each group. Missing data: one in each group. Some participants had several causes of infertility and so the total percentage is >1. Missing data: 12 in vitamin D group, 11 in placebo group. Type A: hyperandrogenism and ovulatory dysfunction and polycystic ovarian morphology; type B: hyperandrogenism and ovulatory dysfunction; type C: hyperandrogenism and polycystic ovarian morphology; type D: ovulatory dysfunction and polycystic ovarian morphology. Missing data: 34 in vitamin D group, 37 in placebo group. Missing data: 28 in vitamin D group, 31 in placebo group. Based on the modified intention-to-treat analysis, 226 (52.0%) live births occurred in the vitamin D group and 216 (50.2%) in the placebo group (adjusted risk ratio 1.03, 95% CI 0.91 to 1.18; table 2 ). The per protocol analysis was consistent with the modified intention-to-treat analysis (56.9% (186/327) v 51.1% (163/319); adjusted risk ratio 1.11, 95% CI 0.97 to 1.28; supplementary table 3). Fertility outcomes (modified intention-to-treat analysis) Data are numbers (%). Adjusted for study centres. All live birth events after first transfer were included, including those occurring after six months of randomisation. Because of covid restrictions, 85 live births occurred in participants who had their first transfers after six months of randomisation, including seven live births from participants who had their first transfers after 12 months of randomisation. The median duration of vitamin D supplementation was 69 days (interquartile range 50-90) and 64 days (46-90) in the vitamin D and placebo groups, respectively. Serum 25-OHD levels were significantly higher in the vitamin D group than in the placebo group: mean levels on trigger day were 32.3±11.2 v 18.2±7.6 ng/mL (adjusted mean difference 13.6, 95% CI 10.9 to 16.3), persisting to the day of pregnancy testing (28.0±9.7 v 20.0±8.3 ng/mL; adjusted mean difference 7.3, 95% CI 4.9 to 9.6; supplementary table 4, fig 2 ). Serum 25-hydroxyvitamin D (25-OHD) levels over time. An interactive version of this graphic and downloadable data are available at https://public.flourish.studio/visualisation/27443480/ No between-group differences were observed in positive pregnancy test rates (vitamin D v placebo: 68.3% (297/435) v 68.6% (295/430); adjusted risk ratio 1.00, 95% CI 0.90 to 1.10), clinical pregnancy rates (60.7% (264/435) v 62.6% (269/430); 0.97, 0.86 to 1.10), or ongoing pregnancy rates (54.0% (235/435) v 53.5% (230/430); 1.01, 0.89 to 1.15). We found no statistically significant differences between the groups for cumulative live birth rates resulting from pregnancy within six and 12 months after randomisation (six months: 46.0% (200/435) v 46.0% (198/430); 1.00, 0.86 to 1.15; 12 months: 64.6% (281/435) v 60.5% (260/430); 1.07, 0.97 to 1.18; table 2 ). The treatment groups showed comparable ovarian response outcomes: median oocytes retrieved (vitamin D v placebo: 15 (interquartile range 11-20) v 16 (11-21)), fertilisation rate per oocyte retrieval (66.7% (50.0-81.3%) v 66.7% (50.0-80.0%)), and total gonadotropin dose (1800 IU (1400-2400) v 1800 IU (1350-2400)). The proportions of single embryo transfers and fresh embryo transfers were also similar for both groups (supplementary table 4). Pregnancy complications including miscarriage (vitamin D v placebo: 13.3% (58/435) v 15.3% (66/430); adjusted risk ratio 0.87, 95% CI 0.64 to 1.18), gestational diabetes (7.1% (31/435) v 10.7% (46/430); 0.66, 0.42 to 1.05), premature rupture of membrane (2.8% (12/435) v 4.0% (17/430); 0.70, 0.30 to 1.61), and postpartum haemorrhage (1.6% (7/435) v 0.5% (2/430); 3.47, 0.52 to 23.03) did not differ significantly between the two groups ( table 3 ). Pregnancy complications and perinatal outcomes after first transfer (modified intention-to-treat analysis) Data are numbers (%) unless stated otherwise. Risk ratios applicable for binary outcomes only. Adjusted for study centres. Including one participant with heterotopic pregnancy who had live birth. 226 participants in vitamin D group and 216 in placebo group. 188 newborns in vitamin D group and 189 in placebo group. 74 newborns in vitamin D group and 52 in placebo group. 241 newborns in vitamin D group and 222 in placebo group. 239 newborns in vitamin D group and 220 in placebo group. Per protocol analyses were consistent with modified intention-to-treat analyses for all secondary outcomes (supplementary tables 3, 5, and 6), except for cumulative live birth rates resulting from pregnancy within 12 months after randomisation, where a borderline higher live birth rate was found in the vitamin D group than in the placebo group (69.4% (227/327) v 61.4% (196/319); adjusted risk ratio 1.13, 95% CI 1.00 to 1.28). Severe ovarian hyperstimulation syndrome occurred in three and six participants in the vitamin D and placebo groups, respectively (adjusted risk difference −0.7%, 95% CI −2.0% to 0.6%). A 31 year old participant in the placebo group underwent therapeutic induction of labour at 19 weeks’ gestation because of fetal hydrops with nuchal lymphatic hygroma. The event was reviewed and deemed unrelated to the study intervention by an institutional review board. No other severe adverse events were reported. Supplementary table 7 summarises other adverse events. No interaction was observed between baseline vitamin D levels and treatment effect on live birth (interaction P=0.731). Fractional polynomial interaction showed that the linear term (power=1) remained the best fitting model (supplementary figure 1). Supplementary figure 2 showed a dose-response relation between vitamin D supplementation duration and serum 25-OHD levels on trigger day (upper panel) and pregnancy test day (lower panel). Complete case analysis was consistent with the main analysis (supplementary table 8).

Vitamin D supplementation did not improve live birth rates in women with PCOS undergoing IVF despite an increase in serum 25-OHD levels. No evidence was found of differences in participants with varying baseline serum 25-OHD levels. This trial’s strengths include its rigorous design (multicentre, double blind, placebo controlled), large sample size, and use of gold standard liquid chromatography tandem mass spectrometry for vitamin D central measurement. 26 However, this trial has several limitations. The observed live birth rate in the placebo group (50.2%) substantially exceeded the prespecified assumption of 38%. This rate was estimated based on data from a university affiliated clinic, where women with lower prognosis are more likely to attend. Therefore, the estimated live birth rate in the placebo group (38%) for the sample size calculation might be too conservative, given that 13/24 recruiting sites were non-university affiliated clinics after the expansion of recruiting sites. However, this may have resulted in a more representative trial population despite the higher than expected live birth rate in the control group. Our trial (n=865) was still adequately powered to detect a 10% point difference in live birth rate, even though a higher live birth rate in the control group (50.2%) was observed. While our results do not show such a large treatment effect, the 95% CI indicates that more modest effects cannot be excluded. The duration of supplementation (up to 90 days) and subsequent follow-up of pregnancies within a year may not fully capture the long term effects of vitamin D on reproductive outcomes, although the duration of vitamin D supplementation and follow-up in our study was longer than in previous trials. Although the lack of a standard IVF protocol may complicate interpretation, the key IVF parameters related to treatment protocol use between the groups are similar and the variation in the protocols potentially enhances the generalisability of our findings. Additionally, the original six month follow-up may have underestimated cumulative success rates owing to pandemic related delays in embryo transfer. Therefore, the 12 month cumulative live birth outcome—initially unplanned—was added post hoc, and it should be interpreted as exploratory. Finally, we did not collect baseline characteristics for people who were screened but declined to participate, which prevented formal evaluation of potential selection bias. Common reasons for declining participation included preference for an immediate IVF cycle without additional treatment beforehand, covid-19 lockdown and restrictions, concerns about taking a placebo, logistical challenges with follow-up visits, and other personal reasons, which might represent a slightly different population than the study population. Additionally, the mean age and body mass index of participants are similar to populations from other trials of women with PCOS in China. 27 28 29 A meta-analysis involving 2352 women with infertility showed that moderate daily dosing of vitamin D supplementation can improve the clinical pregnancy rate. 19 In women with PCOS specifically, recent evidence further suggests that vitamin D supplementation ameliorates metabolic dysfunction by reducing fasting glucose, insulin, and lipid levels, and may also enhance endometrial thickness. 30 31 32 33 34 These findings collectively support a biologically plausible hypothesis that vitamin D could improve pregnancy outcomes in women with PCOS by targeting systemic metabolic disturbances and local endometrial receptivity. Although observational and mechanistic data are compelling, this large randomised controlled trial with a long duration of vitamin supplementation reveals no significant improvement in live birth rates despite achieving an increase in vitamin D levels. The per protocol analysis produced a slightly higher point estimate than the intention-to-treat analysis, but both sets of results were statistically non-significant with overlapping confidence intervals, indicating consistent findings across analyses. Results from per protocol analyses should be interpreted cautiously owing to potential selection bias, but we could not rule out a small benefit of vitamin D supplementation on live birth rate. No treatment interaction was observed for vitamin D baseline levels, suggesting that correcting deficiency alone may not suffice to improve pregnancy outcomes in PCOS related infertility. Several factors may explain why vitamin D supplementation did not result in higher live birth rates in this trial. It remains unclear whether vitamin D deficiency is a prognostic marker associated with infertility in women with PCOS or whether it plays a causal role. Although observational studies have consistently shown that vitamin D deficiency commonly coexists with PCOS and predicts reproductive success in these women, 16 17 this association does not imply causation. If vitamin D deficiency does not contribute causally to infertility in this population, then supplementation is unlikely to improve reproductive outcomes. Additionally, genetic factors related to vitamin D metabolism and receptor activity could affect individual responses to vitamin D supplementation. 35 Finally, some earlier evidence suggests that 25-OHD levels of at least 38 ng/mL were beneficial for live birth rate after IVF, 17 whereas only around 25% of participants in the vitamin D group reached this level in our trial. Therefore, our trial may be considered underpowered to detect benefits in this subgroup. We should also acknowledge that owing to covid restrictions during the trial recruiting period, participants may have been exposed to less sunlight, resulting in lower 25-OHD levels compared with the general population. Clinicians should consider these factors when advising patients and weigh the benefits and drawbacks of vitamin intake in the context of individual patient needs and preferences. Although our trial shows that vitamin D supplementation does not improve IVF success rates in PCOS, there are still uncertainties to be addressed in future research. Firstly, the dose-response relations between serum vitamin D increases and fertility outcomes remain unclear. Secondly, the decline in 25-OHD levels between trigger day and pregnancy testing may reflect cessation of supplementation or physiological changes. We do not know whether extending vitamin D supplementation through embryo transfer or even early pregnancy, or specifically intervening in women with severe deficiency (<10 ng/mL), would improve reproductive outcomes. We should also acknowledge that high dose vitamin D supplementation from the implantation period to early pregnancy may impose ethical challenges in some settings because of the lack of evidence on safety. Large scale randomised controlled trials with longer treatment periods are warranted to show the effect of vitamin D supplementation on live birth rate among women with severe vitamin D deficiency. Finally, while our study provides robust evidence for Chinese women with PCOS, further multicentre randomised controlled trials in diverse ethnic populations and settings are needed to confirm our findings. In conclusion, while vitamin D supplementation of 4000 IU/day for up to 90 days increases serum 25-OHD levels after the intervention, this does not translate to higher live birth rates in women with PCOS undergoing IVF. However, we could not rule out a modest benefit of vitamin D supplementation on live birth rates. Observational studies consistently show associations between vitamin D deficiency and poorer reproductive outcomes in women with polycystic ovary syndrome, and suggest a role of vitamin D supplementation in improving metabolic parameters (eg, insulin resistance) Small, underpowered trials suggested potential beneficial effects of vitamin D supplementation on in vitro fertilisation (IVF) Existing studies have limitations, including small sample size, single centre design, short durations of interventions, and lack of masking This large, multicentre, randomised clinical trial shows that vitamin D supplementation (4000 IU/day for up to 90 days) increases serum 25-OHD levels, but does not improve live birth rates in women with polycystic ovary syndrome undergoing IVF These findings challenge the routine use of vitamin D supplementation as an adjunctive treatment in women with polycystic ovary syndrome undergoing IVF