Intake of soy products and soy isoflavones in relation to ovarian reserve.

Participants had a median (interquartile range (IQR)) age of 35.0 (32.0–38.0) years, with a BMI of 23.4 (21.2–26.4) kg/m2, were predominantly white (83%), never smokers (74%), and had completed post-secondary education (92%) ( Table 1 ). The median soy food intake was 0.09 (0.00–0.29) servings per day, or approximately one serving every 11 days, and the median intake (IQR) isoflavones intake was 1.78 (0–6.26) mg/day with a range of 0 to 166 mg/day, comparable to that of participants in the general US population (Mean± SE: 2.26±0.12 mg/day) ( 36 ). Among participants who consumed soy food, intakes for participants in increasing categories of soy food intake were approximately one serving every 3 weeks for participants in the lowest intake category, approximately one serving every week for participants in the second intake category, approximately 2 servings every week for participants in the third intake category, and approximately 6 times a week for participants in the highest intake category. As expected, soy food and soy isoflavones intakes were positively correlated (r=0.97). Also, as expected, AFC and AMH were positively correlated (r=0.75) whereas day 3 FSH was inversely related to AFC and AMH (r=−0.38, −0.39). Compared to participants who did not eat soy, participants in the highest category of soy intake had slightly lower BMI, were less likely to be white, had higher educational attainment, were less likely to have had a previous evaluation or treatment for infertility, and were less likely to have been pregnant in the past. Intake of soy and isoflavones was higher in Asian participants than in any other race/ethnicity group ( Supplemental Table 4 ). Ovarian reserve was assessed before completion of the soy questionnaire in some participants: 171 of 667 (26%) for AFC, 32 of 239 (13%) for AMH, and 75 of 666 (11%) for FSH. We found no association between soy food or isoflavones intakes with AFC, AMH, or day 3 FSH in unadjusted analyses ( Figure 1 , Supplemental table 5 ). After adjusting for potential confounders, soy food and isoflavones intakes remained unrelated to AFC and day 3 FSH. However, participants in the highest category of soy food and isoflavones intakes had lower AMH levels in these models ( Table 2 ). This inverse association between soy intake and serum AMH persisted in sensitivity analyses where the highest intake category included only participants consuming ≥1 serving of soy/day ( Supplemental Table 6 ) and when intake was modeled using 8 categories of intake ( Supplemental Table 7 ). There was also a suggestion that soy intake was inversely related to AMH in models where intake was modeled as yes/no ( Supplemental Table 8 ). However, the inverse association between soy intake and AMH was not found when participants in the highest 5% ( Supplemental Table 9 ) or the highest 2.5% of the intake distribution ( Supplemental Table 10 ) were excluded from analysis. Furthermore, when soy food and isoflavones intakes were modeled as continuous variables, there was no evidence of an association with AMH, including no evidence of departure from linearity ( Supplemental Table 11 ). Soy food and soy isoflavones intakes remained unrelated to AFC and FSH in these sensitivity analyses ( Supplemental Table 6 – 11 ). When the three markers were dichotomized at clinically relevant cut-off values, there was a suggestion of an association between higher intake of soy and a higher prevalence of low ovarian reserve defined by AMH and FSH values but not when defined by AFC values ( Table 3 ). Additional sensitivity analyses were consistent with the primary findings. The multivariable-adjusted estimates among participants with complete dietary data were similar to those in the main analysis. Additional adjustment for dietary patterns and total energy intake attenuated the association with AMH but had little impact on the association with AFC and FSH ( Supplemental Table. 12 ). Moreover, we found no association between soy food and isoflavones intakes after AFC values greater than 20 were truncated at this value ( Supplemental Table 13 ). Results were almost identical to the ones of primary analysis in sensitivity analyses excluding participants who had an AFC ≥ 30 ( Supplemental Table 14 ) and analyses excluding participants whose assessment of ovarian reserve preceded the completion of the diet questionnaire ( Supplemental Table 15 ). The observed associations were consistent when different strategies for adjustment for race/ethnicity were considered ( Supplemental Table 16 ), and when analyses were restricted to patients in couples with a primary infertility diagnosis of was male factor ( Supplemental Table 17 ). Nevertheless, in analyses restricted to participants with available AMH, intake of soy food was also inversely related to AFC ( Supplementary Table 18 ). On the other hand, soy and isoflavone intakes were unrelated to AMH in models using inverse probability weighing to account for potential selection bias into this analysis ( Supplemental Table 19 ). Last, we evaluated whether the relation of soy and isoflavone intake with ovarian reserve was modified by participant characteristics. The association between soy isoflavone intake and AFC differed by self-reported race (Asian, Non-Asian) ( Table 4 ). In these models, soy intake was associated with higher AFC among participants who self-identified as Asian but not among all other participants. These estimates were based on sparse data, however. We found no evidence of effect modification by race/ethnicity for the relations of soy intake with AMH or FSH. There were no clear associations between soy or isoflavone intake with AFC, AMH, or FSH across strata of age, BMI, or smoking status ( Supplemental Table 20 ).

The Environment and Reproductive Health (EARTH) Study is a prospective cohort study examining how environmental and nutritional factors impact fertility ( 22 ). Female patients aged 18–45 years presenting to the Massachusetts General Hospital (MGH) Fertility Center seeking infertility evaluation or treatment were invited to participate in the study. After reviewing study procedures with staff, all participants provided written informed consent. At enrollment, participants completed a baseline questionnaire where they provided information on demographic, medical, reproductive, occupational, and lifestyle characteristics, including age, smoking status, education status, physical activity and self-identified race/ethnicity, which for the current study was conceptualized as a social construct serving as a proxy for socially-influenced behaviors associated with exposure. They also provided blood and urine specimens and underwent an anthropometric evaluation, which was used to calculate body mass index ( 22 ). Information on medical diagnoses and procedures was abstracted from electronic medical records. For this analysis, we included all participants who joined the study between April 2004 and December 2019, completed the baseline questionnaire, and had an AFC assessment as part of the diagnostic procedures. From a total of 845 participants who joined during this period, we excluded 119 patients who did not answer the baseline questionnaire, and 59 patients who had invalid AFC data (history of oophorectomy (n=5), difficulty visualizing ovaries (n=6), incomplete AFC scan data (n=6), concurrent use of leuprolide (n=42)), leaving a final sample size of 667 participants ( Supplemental Figure 1 ). Participants excluded from this analysis were less likely to be white and to have an unexplained primary infertility diagnosis, and more likely to have ovulatory dysfunction or endometriosis than those included ( Supplemental Table 1 ). We also examined the association between soy food and isoflavone intakes and serum day 3 FSH among 666 participants and serum AMH among 239 participants for whom these hormone measurements were available. AMH measurements were available in fewer participants because they were introduced to the study in 2011. Patients with AMH data had a higher proportion of previous infertility examinations but were less likely to have previously undergone infertility treatment than participants without AMH data ( Supplemental Table 2 ). After excluding participants who joined the study prior to 2011, patients with AMH data were less likely to have had previous infertility treatment and were more likely to receive a primary infertility diagnosis of diminished ovarian reserve or ovulation disorder than participants without AMH ( Supplemental Table 3 ). The study was approved by the Institutional Review Boards of MGH and the Harvard T.H. Chan School of Public Health. Soy intake was assessed at baseline using a short questionnaire focused on intakes of 15 soy-based foods; tofu, tempeh, soy sausages, soy burgers, soy packages, miso soup, soy milk, soy cheese, soy yogurt, tofu cream, soybeans, soy nuts, soy drinks, soy protein and soy bars ( 16 ). Participants were asked to report their consumption frequency of these 15 foods during the previous three months, with nine possible frequency choices ranging from ‘never or less than once per month’ to ‘twice or more times per day’ and three possible selection serving sizes: “medium” serving size (the specified serving size), small (less than specified), and large (more than specified). The isoflavones content of each food was referenced to the United States Department of Agriculture database ( 23 ). Intake of isoflavones was estimated by summing the isoflavones contribution of all food items in the questionnaire. A comprehensive dietary assessment was introduced to the study in 2007. Using data collected with this validated questionnaire ( 24 – 26 ) we identified dietary patterns in this population using principal components analysis (PCA) ( 27 ). The primary outcome of this study was AFC. AFC was measured by reproductive endocrinology and infertility physicians from the MGH Fertility Center on the third day of the menstrual cycle or progesterone-induced withdrawal bleed. We included only the first AFC measurement for each participant. Of the 667 participants, 23 participants (3.5%) had an AFC greater than 30. For these 23 participants, AFC was coded as 30 for the analysis to reduce the influence of extreme values in the results. We also evaluated AMH and FSH as secondary outcomes. FSH concentration was measured with an automated electrochemiluminescence immunoassay using the Elecsys FSH reagent kit and the Roche Elecsys immunoassay analyser (Roche Diagnostics, Indianapolis, IN, USA) from a blood sample collected on the third day of the menstrual cycle at the MGH Core Laboratory ( 28 ). Serum AMH was measured from blood samples that were collected during the follicular phase of IVF cycle using the Ansh Labs ultra-sensitive AMH/mullerian-inhibiting substance enzyme linked immunosorbent assay at Mayo Clinic Department of Laboratory Medicine and Pathology (Rochester, MN) before January 2018, and later, using the Elecsys AMH immunoassay by Roche Diagnostics (Roche Diagnostics GmbH) at the Brigham and Women’s Hospital Laboratory of Pathology in Boston( 29 ). We first categorized participants into five groups according to their soy food and isoflavone intakes. Participants who did not consume soy food served as the reference group and the remaining participants were divided into quartiles according to their intakes of soy food and isoflavones. We assessed differences in baseline characteristics across categories of soy intake using Kruskal-Wallis tests for continuous variables and Fisher’s exact test for categorical variables. We also compared the demographic data and soy isoflavone intake by race. There were five missing values (0.7%) for BMI, one missing value (0.1%) for smoking status, and two missing values (0.3%) for education level. Given the low frequency of missingness, we imputed the median value of BMI to participants missing BMI data, and the most frequent category for smoking status, and education levels to participants missing these data. We evaluated the correlations for soy food and isoflavones intakes and AFC, AMH, and FSH using the spearman correlation test. To evaluate the relation between soy food and isoflavones intakes with AFC, we fitted Poisson regression models (generalized linear models with Poisson distribution and log-link function) and presented the results as the relative difference (%) in AFC, with 95% confidence intervals, adjusting for the covariates. To account for the skewed distribution of day 3 FSH and AMH, we modeled the associations of soy food and isoflavones intakes with serum levels of these hormones using quantile regression, fitting models for the median. Regression coefficients from these models can be interpreted as differences in the median relative to the reference group (non-consumers). In addition, we dichotomized each of the three markers of ovarian reserve using previously proposed cut-off values to define low ovarian reserve for each of these markers: AFC less than 7, AMH less than 0.5 ng/ml, and FSH greater than 10 IU/L ( 30 – 35 ). We then fitted generalized linear models with Poisson distribution and log-link function where low ovarian reserve was the outcome of interest for each specific marker. Confounding was evaluated based on prior knowledge and descriptive statistics from the study population. The multivariable-adjusted models included age (continuous), race/ethnicity (white vs not), BMI (continuous), smoking status (never vs ever), education status (higher than college or not), physical activity (hours/week of moderate-to-vigorous activity exercise, continuous) and parity (nulliparous or parous). To account for the change of laboratory for AMH measurements, we additionally adjusted models for AMH with an indicator for the laboratory where measurements were performed. To test the linear trend across categories, the median value for each category was assigned and these values were used as continuous variables. We evaluated whether the associations of soy food and isoflavones intakes differed by age (< 35, ≥ 35 years), BMI (< 25, ≥ 25 kg/m2), smoking status (never, ever) race (Asian, Non-Asian) by fitting models stratified by these characteristics and using cross-product terms to evaluate their statistical significance. Last, we performed a series of sensitivity analyses to evaluate the robustness of the findings. First, to evaluate whether the findings were influenced by a priori chosen cutoff values, we performed analyses considering different categorizations: two categories (no intake and any soy intake), including in the top intake category participants with an intake ≥1 serving of soy/day (no intake, tertiles of intake among participants with soy intake >0 but <1 serving/day, and participants with intake ≥1 serving/day), and eight categories (no intake, and septiles of intake). We then tested the robustness of the results when we changed the cutoff for the extreme value of AFC from 30 to 20; in this sensitivity analysis, AFC was coded as 20 among participants whose AFC was greater than 20. We also modeled intakes as a continuous exposure and used restricted cubic splines to evaluate deviations from linearity. In addition, to minimize the impact of outliers, we fitted models excluding participants in the top 5% and the top 2.5% of intake. Next, to address the potential for residual confounding by overall dietary choices, we fitted models including terms for PCA-derived dietary pattern and total energy intake among the 576 participants who completed the FFQ. Moreover, to investigate the difference of association between soy intake and ovarian reserve by race/ethnicity, we created two additional race/ethnicity variables; the first was a categorical variable consisting of white, Asian, or other, and the second was a binary variable of Asian/non-Asian. To reduce the concern that patients may have changed their soy intake after knowing the primary infertility diagnosis or being advised by clinicians before the study enrollment, we conducted the sensitivity analysis among patients whose primary infertility diagnosis was a male factor. Next, to address potential bias introduced by participant selection into the analysis, we evaluated the relation of soy and isoflavone intakes with AFC and day 3 FSH among participants with available AMH data. In addition, we used inverse probability of weighting with stabilized weights to account for systematic differences in observed characteristics between those who had AMH data and those who did not. We first created weights by fitting a logistic regression model where the outcome was the presence or not of AMH data and predictors were terms for age (continuous), BMI (continuous), race/ethnicity (categorical), smoking status (binary; never smoker, ever smoker), education status (binary; higher education than college or not), physical activity time (continuous), parity (nulliparous or parous), past infertility examination (yes, no), past infertility treatment (yes, no), primary diagnosis of infertility (categorical), antral follicle count (integer), serum follicle stimulating hormone (continuous). We obtained a second set of weights by also including soy intake in the logistic model. Last, to address potential bias introduced by high AFC values or timing of assessments, we evaluated all relations after excluding participants with an AFC ≥30 and, separately, after excluding participants whose assessment of ovarian reserve preceded the completion of the baseline questionnaire. We performed the analyses using SAS v. 9.4 (SAS Institute Inc, Cary, NC, USA).

We investigated the association between intakes of soy food and soy isoflavones and three ovarian reserve markers among individuals presenting to an academic fertility center seeking infertility treatment. We found no association of soy food and isoflavones intakes with AFC or FSH. Nevertheless, high intakes of soy food and isoflavones were related to lower serum AMH levels. This relation, however, was highly sensitive to modeling specifications, was not reproduced across multiple analytic strategies, and disappeared when using inverse probability weights, suggesting that the relation between soy and AMH observed in the primary analysis may be spurious or represent a chance finding. On aggregate, our findings are not consistent with a strong positive or inverse association between soy and isoflavone intakes, which substantially overlap with the intake distribution in the general US population ( 36 ), and ovarian reserve among fertility patients. The potential reproductive effects of soy intake receive an inordinate amount of attention in lay literature targeting couples trying to conceive and couples experiencing difficulties conceiving. While a comprehensive survey of this literature is outside the scope of this article, in broad terms, lay sources often portray intakes of soy-based foods and isoflavones as potential reproductive toxicants basing these claims on structural similarity between isoflavones and sex steroids ( 2 ), in-vitro binding to estrogen receptors and estrogenic activity ( 3 ), and documented adverse reproductive effects in some mammalian species ( 4 – 9 ). These concerns notwithstanding, evidence from human studies, including evidence from randomized trials, and more recent animal models present a much different picture. Previous studies among female participants in the general population have found no association between soy intake or urinary markers of intake with time to conception ( 15 , 17 ). In addition, studies among couples undergoing infertility treatment suggest that, if anything, intake of soy or isoflavone supplements may improve infertility treatment outcomes ( 18 , 37 ). Similarly, one study reported that a higher concentration of urinary genistein was related to a lower risk of advanced (AFS Stage III-IV) endometriosis among Japanese individuals seeking infertility treatment, although this relation was modified by genetic polymorphisms in ESR2 and estimates were based on sparse data ( 38 ). Regarding ovarian reserve specifically, administration of isoflavones to (12-month) rats increased ovarian follicular reserve by increasing the number of primordial (2 to 3-fold increase) and primary (20 to 60% increase) follicles and inhibiting follicular atresia (20 to 30% decrease) ( 20 ). Although comparable data are not available in humans, a meta-analysis summarizing the findings of 11 randomized trials of soy or isoflavone supplementation on reproductive hormone levels found that soy products reduced circulating FSH and LH levels by approximately 20% among pre-menopausal participants ( 21 ). The authors of the meta-analysis remarked that the clinical significance of their findings is unclear. When our findings are examined in light of the existing literature, our interpretation is that the most lausible interpretation is that intake of soy and soy isoflavones within the observed range are unlikely to have a major deleterious effect on ovarian reserve. While we found no relation between soy intake with AFC and day 3 FSH, there was a suggestion of an inverse relation with AMH. This relation was not consistent across modeling strategies and appeared to be driven by participants in the top 2.5% of the intake distribution (≥1.6 servings/day). This pattern has three plausible interpretations: first, the observed relation is consistent with a true biological effect that is restricted to the highest intake levels in our study population; second, the association represents a chance finding; and third, the observed relation is the result of selection bias, unmeasured confounding or both. Arguing in favor of a true biological effect, the range of intake for participants in the top 2.5% of isoflavones intake in our study population (46–166mg/day), overlaps with intake levels previously related to identifiable health effects (>40mg/day) ( 39 ). Nevertheless, an inverse relation with AMH would not be consistent with findings from randomized trials regarding the effect of soy on FSH ( 21 ) nor with experimental findings in rodents ( 7 , 8 ). Moreover, a true deleterious effect is also not internally consistent with our findings for AFC and FSH nor with the suggestive effect modification by self-reported race/ethnicity. These inconsistencies argue strongly against interpreting this association as reflecting a true deleterious effect of soy intake on ovarian reserve as reflected in AMH levels. The lack of consistency of this finding across a variety of analytic strategies along with the consistent lack of association with AFC and FSH suggest instead that the observed relation with AMH could represent a chance finding. In general, AMH levels are strongly correlated with AFC ( 40 ), as was indeed the case in our study. Although AFC, AMH, and FSH can serve as markers of ovarian reserve, they represent different, but related, biological constructs. AFC is a better predictor of poor ovarian response during ART treatment ( 41 ), AMH is mostly representative of the post-primordial to preantral follicle pool ( 42 ), while FSH reflects the last two weeks of follicular maturation when follicles become gonadotropin sensitive. Nevertheless, we could not identify a biologically plausible mechanism in which isoflavones would have an impact on AMH without also having an impact on FSH and particularly on AFC, supporting an interpretation of these results as a chance finding. A third possibility is that the inverse relation of soy and isoflavones with AMH may be the result of selection bias, unmeasured confounding, or both. While unmeasured confounding can never be completely excluded, arguing against a strong unmeasured confounding, sensitivity analyses adjusting for key covariates only available in subsets of the study population, such as dietary patterns, yielded similar results to those observed in the primary analysis. Moreover, unmeasured confounders are likely to influence the associations with all markers of ovarian reserve rather than with AMH alone. This leaves selection bias as a likely explanation for our findings for AMH. It is important to keep in mind that ovarian reserve assessments, including AMH, were obtained from measurements conducted for clinical purposes. While AFC and day 3 FSH were routine assessments for all patients throughout the duration of the study, AMH measurements were not only introduced later but were also initially offered only to participants whose treating physician suspected diminished ovarian reserve. This practice is reflected in our study data where participants with available AMH measurements were more likely to receive a primary infertility diagnosis of diminished ovarian reserve. This suggests that the association with AMH could be explained by selection bias introduced when clinicians selected whether the participants needed an AMH evaluation or not. In other words, the probability of entering the analytical dataset (having AMH data) is related to the outcome (diminished ovarian reserve). At the same time, AMH testing unblinded participants to their ovarian reserve status making it more likely that they made behavioral changes, like their diet, in response to this knowledge, thereby making the distribution of exposure (soy intake) related to the probability of having AMH data available. There are other scenarios in which soy intake could be related to the probability of inclusion into the AMH analysis. For example, soy intake could change the relative distribution of conditions associated with infertility among couples presenting to a fertility center, as suggested by the findings of the Japanese study reporting an association between urinary genistein and endometriosis among fertility patients ( 38 ). The suspicion of selection bias as an explanation for the association between soy intake and AMH levels is further supported by our observation of an inverse association with AFC in the subset of participants for whom AMH was available, which was not present in any other analysis, and the lack of association between soy intake and AMH in analyses employing inverse probability weighing to account for systematic differences between patients with and without AMH measurements. These additional findings further support an interpretation of the association between soy intake and AMH as explained by selection bias. Interestingly, we found suggestive evidence that soy intake is related to higher ovarian reserve among patients of self-identified Asian descent. Asian participants had higher soy intakes than others in the study population, in agreement with nationally representative data ( 36 ) and were more likely to have intakes previously related to measurable biologic effects in humans. Nevertheless, this association was also observed at lower intakes suggesting that differences in intake levels may not entirely explain this difference. Others have previously reported that individuals of Asian descent are more efficient in metabolizing isoflavones in their gut than western individuals ( 43 ). Although this finding offers a plausible biological explanation for these findings, it is very important to keep in mind that this relation is based on very sparse data and therefore should be interpreted with extreme caution. It is important to consider the findings in the context of this study’s limitations and strengths. First, measurement error of intake is always a concern in studies based on food questionnaires. In addition, we assessed soy intake in the last three months investigating only the short-term impact of soy consumption, which may not be the relevant time frame for ovarian reserve. Although studying longer-term soy intake assessment could provide further insights, similar questionnaires have found that self-reported intakes are correlated with biomarkers of intake ( 44 , 45 ), and we have identified differences in semen quality parameters ( 10 ) and live birth rates ( 18 ) using this questionnaire in the same study. Second, causal inference is impaired due to the cross-sectional design of the study. Third, all participants were seeking infertility care at an academic fertility center and therefore findings may not be generalizable to females in the general population. Finally, as is the case with any observational study, we cannot exclude the possibility of residual or unmeasured confounders. Nevertheless, this is more often a concern in studies identifying an association than in studies failing to identify one like ours. Moreover, we included in our models a number of known and suspected potential confounders of the hypothesized association making this concern less likely. The study has multiple strengths. First, we were able to assess ovarian reserve using three different markers in the same participants: AFC, AMH, and day 3 FSH. Second, we obtained information on a wide variety of lifestyle factors, which allowed for extensive statistical control for potential confounders. Last, the observed range of soy intake was comparable to that of the general population of the United States ( 46 ), increasing the generalizability of our findings. In summary, we found no association of soy intake with AFC and day 3 FSH among participants with soy intakes comparable to those observed in the general US population. Although we found an inverse relation between soy intake and AMH, this association was highly sensitive to modeling assumptions and disappeared when employing inverse probability weighted models, suggesting a spurious relation or a chance finding. When interpreted collectively and in light of existing literature, our results are not consistent with a strong deleterious or beneficial relation between intake of soy foods within intake levels typically observed in Western populations and ovarian reserve. Given the internal inconsistencies of the findings, however, additional evaluation of this question in larger and better designed studies may be warranted.

Isoflavones, including daidzein, genistein, and glycitein, are non-steroid phytochemicals primarily found in soy and to a lesser extent in other legumes ( 1 ). Isoflavones share structural and functional similarities with endogenous estrogens ( 2 ) and are therefore often referred to as phytoestrogens. Phytoestrogens have a weak interaction with estrogen receptors α and β ( 3 ). Previous work in animals has shown that exposure to phytoestrogens can have major adverse impacts on reproduction. Severe breeding problems and other reproductive anomalies were first described in the 1940s in sheep feeding on phytoestrogen-rich pastures ( 4 ). Fertility problems were then linked to the use of phytoestrogen-rich feed in captive cheetahs ( 5 ), and experimentally in rodents ( 6 – 9 ). These findings have raised public awareness and concern about soy-based foods as potential reproductive hazards. Findings of studies in humans on the potential role of soy food and phytoestrogens on reproduction, however, have not matched the level of concern raised by the animal literature. Although some studies suggest a modest negative impact on semen quality ( 10 , 11 ), these relations have not been consistent across studies ( 10 , 12 – 14 ) and do not appear to impact fertility, ( 15 , 16 ) even in populations where associations with semen quality are simultaneously identified ( 16 ). Moreover, studies evaluating either women’s isoflavones intake or biomarkers of intake in relation to fertility have found either no relation ( 15 , 17 ) or a benefit ( 18 ) to fertility. Regarding ovarian reserve specifically, Medigovic and colleagues investigated the effects of genistein and daidzein on ovarian function in 12 months old female rats, which are considered as middle-aged based on average timing of menopause in rats (15–18 months) ( 19 ), revealing that rats exposed to genistein and daidzein had more healthy primordial and primary follicles and less atretic follicles ( 20 ). In addition, a meta-analysis summarizing 11 randomized trials of soy or isoflavone supplementation revealed that short-term soy or isoflavone supplementation reduced FSH by approximately 20% ( 21 ). Nevertheless, to our knowledge there are no human studies that have evaluated the relation between soy or isoflavone intake with markers of ovarian reserve other than FSH. To address this knowledge gap, we examined the association of intakes of soy food and isoflavones with three markers of ovarian reserve: antral follicle count (AFC), serum levels of anti-Müllerian hormone (AMH), and day three follicle stimulating hormone (FSH), among individuals presenting for infertility evaluation and treatment at an academic fertility center. We hypothesized that participants with higher intake of soy would have higher ovarian reserve measured by AFC, AMH and FSH.