Cardiovascular disease risk factors in women with O-desmethylangolensin producing and non-producing gut microbial metabotypes: an observational study Authors: Holly Childs, Cara L Frankenfeld, Allyson Dailey, Robin Couch, Margaret Slavin Published: 2026 ABSTRACT AbstractBackgroundDaidzein, an isoflavone in soy, is metabolized to O-desmethylangolensin (ODMA) by select gut bacteria. Not all individuals have gut microbiomes capable of producing ODMA, resulting in ODMA producer and non-producer metabotypes. A limited body of research suggests that ODMA producers have lower systolic blood pressure, percent body fat and total cholesterol, compared to non-producers. More work is needed to confirm this relationship in other populations.MethodsData from the Study of Women’s Health Across the Nation (SWAN), a study of women in their middle years, was used. A subset of women who reported consumption of isoflavones in diet or supplements was included for this analysis. Urinary ODMA and daidzein of 148 participants were measured via HPLC-MS-MS to classify metabotypes: ODMA (nmol/L)/daidzein (nmol/L) = 0, non-producers (n = 23); ODMA (nmol/L)/daidzein (nmol/L) > 0 but less than 0.5, low-producers (n = 90); and ODMA (nmol/L)/daidzein (nmol/L) > 0.5, high-producers (n = 35). Multivariable linear regression models were used to compare cardiovascular disease (CVD) risk markers among groups. Models were adjusted for age, race, self-reported health, blood pressure medication, smoking, menopausal status and anthropometry.ResultsODMA non-producers had higher systolic blood pressure (adj. mean difference = 7.6 mmHg, 95% CI: 1.1, 14.1, p = 0.02) compared to high producers. Another trend but not statistically significant difference observed included higher fasting serum glucose (adj. mean difference = 6.1 mg/dl, CI: -0.9, 13.2, p = 0.09) in ODMA non-producers when compared to ODMA producers.ConclusionsIn this diverse sample of U.S. women, this study observed higher blood pressure in ODMA non-producers than ODMA producers. This aligns with prior work associating ODMA producers with a more favorable CVD risk profile compared to non-producers, and provides additional evidence that this gut microbial metabotype may be related to cardiovascular health.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12872-026-06007-0.Keywords: O-desmethylangolensin, Metabotype, Cardiovascular disease, ODMA, Isoflavone, Microbiome, Daidzein Background Cardiovascular disease (CVD) is the leading cause of death in the United States among women and men and can encompass several different heart and vascular conditions, including heart attack, stroke, heart failure, arrhythmia, and heart valve problems. Even though research has discovered important differences between men and women in terms of biological stresses and responses, only 10–35% of cardiac clinical studies are comprised of women [1]. While there are known modifiable and non-modifiable risk factors for CVD, evidence suggests that the composition of the human gut microbiome, a consortium of microbes resident in the large intestine, is associated with CVD risk markers and incidence [2]. The human gut microbiome is known for its many functions, including fermentation of food, vitamin production, immune response, and nutrient and ion absorption [3]. Additionally, experts have discovered associations between microbiome composition and several cardiometabolic diseases, including obesity, liver disease, diabetes, dyslipidemia, atherosclerosis, CVD, and neurological disease [4]. Gut microbiome composition can be affected by an individual’s genes, the environment [5], illness, use of antibiotics/drugs, and diet [6], including polyphenols in the diet [7]. Though there is significant promise in understanding the gut microbiome’s influence on CVD risks, wide variation across individuals and functional redundancy in gut microbiome ecology present challenges in identifying specific features that may be important for these associations between cardiovascular risk and the gut microbiome. Omics-based technologies (e.g. metataxanomics, metagenomics, metatranscriptomics, and metaproteomics) provide important but very complex information related to specific gene markers, gene sequencing, activity and translated ribonucleic acid (RNA) in the microbiome. In contrast, metabotyping simplifies the complexity of the millions of genes in the human gut microbiome and characterizes an individual’s gut based on a metabolite’s production or non-production (or level of production). As such, metabotyping is a means of characterizing a particular activity or function of the gut microbiome, with respect to a specific substrate/metabolite relationship [5, 8]. Thus, metabotyping can be advantageous for evaluating health outcomes associated with the functional aspect of the gut microbiome. Prior research has measured CVD associations with several gut microbial metabolites from substrates of key interest like dietary fiber and proteins. Isoflavones are a subgroup of flavonoids, a class of bioactive food components also of interest in the study of metabotypes and CVD. Daidzein, an isoflavone commonly found in soy, is metabolized to O-desmethylangolensin (ODMA) by bacteria resident in the large intestine [9]. Daidzein is a product of its precursor daidzin. Like most isoflavones, daidzin is not readily available for absorption in its current form. However, once hydrolyzed by the small intestine β-glycosidase enzyme, daidzin releases its sugar moiety, then the remaining aglycon, daidzein, becomes more bioavailable. Daidzein can then either be absorbed intact or convert to its intermediate dihydrodaidzein. Dihydrodaidzein can then be transformed to either equol or O-desmethylangolensin (ODMA). Tetrahydrodaidzein, 39-hydroxy-daidzein, 6-hydroxy-daidzein, 8-hydroxy-daidzein, 3-(4-hydroxyphenyl)-benzo-pyran-4,7-diol, and 2-dehydro-O-DMA have also been reported as other metabolites of daidzein [10]. Evidence has shown that Eubacterium ramulus [11] as well as Clostridium species [12] can be responsible for this biotransformation of daidzein to ODMA. However, ODMA production cannot be isolated to one specific bacterium because some strains of bacteria in the gut may perform many cross-connecting or over-lapping functions [13]. For these reasons, ODMA producer and non-producer metabotypes are useful in research. ODMA metabotype remains stable in most individuals over time [14], and it identifies individuals (male and female) whose guts produce different metabolites following consumption of food or supplements containing daidzein [15, 16]. Prior studies observed ODMA non-producers to have less favorable cardiometabolic disease risk profiles compared to ODMA producers. In a sample of peri/post-menopausal women who consumed at least 3 servings of soy per week, obese participants were 3 times more likely to be ODMA non-producers [17]. These results were also present in a population of different demographics. In a group of men and women age 18 to 95 years who completed a 3-day soy challenge, obese participants were 2.8 times more likely to be ODMA non-producers [18]. Body composition and lipid profiles were also observed to be less favorable for ODMA non-producers. In a sample of prehypertensive, postmenopausal women who completed a daidzein feeding challenge, ODMA non-producers had significantly higher percent body fat, systolic blood pressure (SBP) and total cholesterol compared to their ODMA producer counterparts [19]. Bone density and hormone concentrations of women in their pre- and post-menopausal years revealed further heath implications of the ODMA metabotype [20–22]. In post-menopausal women, ODMA producers had significantly greater total and leg bone mineral density [20], however these associations were not observed in women pre-menopause [21]. Post-menopausal women who were ODMA producers were observed to have greater concentrations of 2-hydroxyestrone (a protective metabolite against cancer) and 16α-hydroxyestrone (a possible genotoxic metabolite associated with breast cancer risk) when compared to ODMA non-producers [22]. The gut microbiome is involved in the metabolism of estrogens [23]. Changes in estrogens were associated with cardiometabolic disease risk [24], and ODMA metabotypes were associated with differing CVD risk profiles in past research. Therefore, further research is needed to elucidate the ODMA metabotype’s relationship to CVD risk factors in women transitioning through menopause. The purpose of the present study was to evaluate whether the associations discovered in these earlier studies (e.g. Liu et al. 2014), also apply to the population of women transitioning through menopause who participated in the Study of Women’s Health Across the Nation (SWAN). Methods Study design and participants The Study of Women’s Health Across the Nation (SWAN) (Clinical trial number: not applicable) is a large, cohort study based in the United States [25]. The goal of SWAN is to examine the health of women across their middle-aged years. Data collected included anthropometric measurements such as weight, height and waist circumference as well as blood pressure, blood samples, urine samples, dietary data, health questionnaires and assessments targeting other health concerns such as menopause symptoms, bone health, sleep and depression. Inclusion criteria for the SWAN longitudinal cohort study included women 42–52 years old; African American, non-Hispanic Caucasian, Chinese, Hispanic or Japanese; no removal of the uterus and/or both ovaries; no hormone therapy affecting ovarian function; and at least 1 menstrual period in the previous 3 months. This pilot study was undertaken to evaluate feasibility of using existing stored samples and to design a larger observational study. It was designed as a cross-sectional study of SWAN data from participant visits 5 and 9. SWAN participants’ dietary data was collected via a modified 1995 version of the Block Food Frequency Questionnaire during visits 5 and 9, and the subset of women who reported consumption of isoflavones in diet or supplements was included for this present pilot study. SWAN data and urine samples were obtained from the National Institutes of Health (NIH) Aging Research Biobank [26]. Urine samples were from either the fifth or ninth visit (when dietary intake was measured). In cases where a participant provided urine samples at both visit 5 and visit 9, only visit 5 was included in this analysis because the blood sample data was also taken from visit 5, and this ensured there were no duplicates. Urine samples that were obtained from a subset of SWAN participants had an estimated dietary daidzein equivalent (DDE) of isoflavones greater than 25 mg from regular consumption of soy protein or phytoestrogen supplements (4–7 days/week). This was done to maximize the likelihood that individuals consumed daidzein (precursor to ODMA) in close proximity to urine sample collection. ODMA and daidzein concentrations were determined by analyzing urine samples via High Performance Liquid Chromatography Tandem Mass Spectrometry (HPLC-MS-MS) identification. To further maximize the likelihood that women had consumed daidzein and ODMA metabotypes would be correctly classified due to having sufficient exposure to the precursor daidzein, participants with urinary daidzein concentration less than or equal to the lower 25th percentile daidzein (0.149 nmol/L) were excluded. This yielded a final sample of n = 148 (Fig. 1), consisting of 93 samples from visit 5 and 55 samples from visit 9. Sample selection for ODMA metabotype and CVD risk analysis from the SWAN Urine sample preparation As described elsewhere [27] with minimal modification, urine samples were extracted in batches of 30. After mixing thoroughly, 100 µL of the urine was transferred to a sterile microfuge tube and 5 µL of β-glucuronidase/arylsulfatase was added. After incubation at 37 °C for 2 h in a water bath, the microfuge tube was placed on ice, and 305 µL of methanol + 7% formic acid was added. After equilibration and centrifugation at 13,000 rpm at 4 °C, samples were filtered into an autosampler vial using a 3 mL plastic syringe equipped with a 4 mm, 0.45 μm luer lock nylon syringe filter and immediately analyzed via Liquid Chromatography-Quadrupole Time-Of-Flight (LC-QToF). LC-QToF analysis Analysis was performed according to a previously reported method [27]. A total of 2 µl of the urine extracts were injected and analyzed via an Agilent 1290 Infinity UPLC equipped with a C18 reversed phased column (2.1 × 100 mm, 1.8 μm Zorbax RRH Eclipse plus C18) and coupled to an Agilent 6530 Accurate Mass QToF, running in electrospray ionization (ESI) negative mode. The mobile phase was dispensed at a flow rate of 0.25 mL/min which consisted of solvent A (13mM ammonium acetate pH 4) and solvent B (100% methanol + 0.1% acetic acid). The solvents were dispensed over the following gradient for a total run time of 15 min: 0–1 min 5% solvent B, 1–6 min 60% solvent B, 6–11 min 65% solvent B, 11–12 min 95% solvent B, 12–13 min 5% solvent B, hold for 2 min. Mass spectrometric analysis was performed by the QToF with a mass selection range of 100–1000 m/z. The capillary voltage was set to 2000 V and the fragmentor was set to 125 V. The nebulizer pressure was set to 20 psi with a sheath gas temperature of 300 °C and a flow rate of 6 L/min. The drying gas temperature was set to 350 °C with a flow rate of 8 L/min. To confirm the presence of daidzein and ODMA within the urine samples, tandem MS analysis (MS/MS) was performed using 2 µL of the urine extracts at collision energies of 10 eV, 20 eV, and 40 eV and a MS/MS fragmentation range of 50–300 m/z. Retention time (Daidzein 7.47 min; ODMA 8.29 min) and MS (Daidzein ESI -ve m/z of 253.0506; ODMA ESI -ve m/z of 257.0819) were used to find the metabolites and MS/MS was used to identify each within the samples, relative to authentic standards. QC-based drift-corrected intensities using LOESS were used to determine %CV between batches with < 30% as a cutoff (Daidzein 28.8%CV; ODMA 21.6%CV). ODMA and daidzein quantification Finally, to determine the concentration of ODMA and daidzein within the urine samples, a series of standard curves were prepared by adding the standards to the urine, and the samples were extracted and analyzed as described above. For ODMA, the concentration ranged from 0.001 µg/mL to 10 µg/mL, whereas for daidzein the concentration ranged from 0.0005 µg/mL to 10 µg/mL. The background levels of each metabolite in the urine sample were subtracted from the standard signal. ODMA metabotyping ODMA producer/non-producer groupings of metabotypes were completed based upon ratios: ODMA high-producers’ (n = 25) urinary ODMA (nmol/L)/daidzein (nmol/L) was > 0.5, ODMA non-producers’ (n = 23) urinary ODMA (nmol/L)/daidzein (nmol/L) was = 0 and ODMA low-producers’ (n = 90) urinary ODMA (nmol/L)/daidzein (nmol/L) was between 0 and 0.5. Two categories of producers (high and low) along with the one category of non-producers were used to further minimize potential misclassification due to the lack of a controlled daidzein challenge. CVD risk factors SWAN data provided CVD risk variables from participants’ visits to the study facilities at either their fifth or ninth visit. Fasting blood sample data which included all the blood lipid and CRP data were only available from the participants’ fifth visit. The blood samples were analyzed for high-density lipoprotein (HDL) in mg/dL, low-density lipoprotein (LDL) in mg/dL, C-reactive protein (CRP) in mg/L, fasting serum glucose in mg/dL, insulin in µIU/mL, cholesterol in mg/dL, and triglycerides in mg/dL. Systolic and diastolic blood pressure, in mmHg, were measured at both the fifth and ninth visits. The present study analyzed blood pressure measurements which corresponded to the same visit as the participants’ urine collection. CVD risk variables were classified into higher risk and lower risk groups based on established risk cut-offs. Variables within the following ranges were classified as elevated CVD risk: systolic blood pressure ≥120 mmHg, diastolic blood pressure (DBP) ≥80 mmHg [28], high-density lipoprotein < 50 mg/dL [29], low-density lipoprotein ≥160 mg/dL [30], C-reactive protein ≥10 mg/L [31], fasting serum glucose ≥100 mg/dL [32], insulin ≥12.94 µIU/mL [33], cholesterol ≥200 mg/dL [29], and triglycerides > 150 mg/dL [34]. Statistical analysis Variables were checked for normal distribution and ln transformed as applicable. Multivariable linear regression models were applied to compare CVD risk markers in relation to ODMA metabotype, and p < 0.05 was considered statistically significant. The regression coefficients of the ln transformed variables were exponentiated, converted to percentages according to and displayed as the adjusted expected % change among ODMA metabotype groups in Table 2. The confidence intervals of the ln-transformed variables were also converted according to the . Binomial logistic regression models were also used to calculate odds ratios for dichotomous CVD risk markers in relation to ODMA metabotype. Height (cm), weight (kg), waist circumference (cm), self-reported health status (excellent, very good, good, fair or poor), smoking status (never, former or current), age (years), race (white, black, Chinese or Japanese), blood pressure medications (whether or not taken at the time of the visit), self-reported health status and smoking status were included as adjustment variables. Analyses were conducted using R (Version 4.3.0). Table 2Comparison of CVD risk markers among ODMA metabotypes (high-producers, low-producers and non-producers)ODMAHigh-producersODMALow-producersODMANon-producers(n = 35)(n = 90)(n = 23)Average Systolic Blood Pressure, mmHg Mean (SD)108.3 (13.1)114.0 (14.8)118.6 (14.5) Adj. mean difference (95% CI)ref3.6 (-1.3, 8.5)7.6 (1.1, 14.1) p-value*ref0.150.02Average Diastolic Blood Pressure, mmHg Mean (SD)71.2 (7.9)72.4 (8.8)74.2 (10.5) Adj. mean difference (95% CI)ref0.7(-2.9, 4.3)1.9 (-2.9, 6.8) p-value*ref0.700.44Total Cholesterol, mg/dl Mean (SD)208.2 (35.4)203.1 (30.9)208.0 (40.2) Adj. mean difference (95% CI)ref-3.7 (-17.5, 10.1)3.1 (-15.1, 21.3) p-value*ref0.600.74^Triglycerides, mg/dl Mean (SD)110.1 (54.9)124.6 (63.8)140.8 (63.2) Adj. expected % change (95% CI)ref1.6% (-14.8%, 21.1%)21.7% (-3.4%, 53.5%) p-value*ref0.860.10LDL, mg/dl Mean (SD)119.7 (35.5)118.5 (26.6)122.4 (35.1) Adj. mean difference (95% CI)ref0.2 (-12.4, 12.9)4.4 (-12.3, 21.1) p-value*ref0.970.60^HDL, mg/dl Mean (SD)66.4 (19.8)59.7 (14.9)57.3 (14.3) Adj. expected % change (95% CI)ref-4.4% (-12.9%, 5.0%)-7.1% (-17.7%, 5.1%) p-value*ref0.250.24Fasting Serum Glucose, mg/dl Mean (SD)84.1 (7.7)88.5 (11.4)90.5 (25.3) Adj. mean difference (95% CI)ref3.0 (-2.4, 8.3)6.1 ( -0.9, 13.2) p-value*ref0.280.09^Insulin, uIU/ml Mean (SD)10.0 (5.4)10.5 (6.5)12.1 (6.5) Adj. expected % change (95% CI)ref-7.9% (-19.8%, 5.7%)8.4% (-9.6%, 30.0%) p-value*ref0.240.38^C-reactive protein, mg/l Mean (SD)4.0 (5.5)4.0 (6.2)7.3 (14.0) Adj. expected % change (95% CI)ref-26.4% (-51.7%. 11.9%)-12.7% (-49.8%, 51.9%) p-value*ref0.150.63*adjusted for age, race, self-reported health status, smoking status, menopause status, blood pressure medications, height, weight, and waist circumference^ln transformed variables to mitigate skewness (note: the adjusted expected % change of the ln transformed variables have been exponentiated from the regression coefficients and converted to percentages using Comparison of CVD risk markers among ODMA metabotypes (high-producers, low-producers and non-producers) *adjusted for age, race, self-reported health status, smoking status, menopause status, blood pressure medications, height, weight, and waist circumference ^ln transformed variables to mitigate skewness (note: the adjusted expected % change of the ln transformed variables have been exponentiated from the regression coefficients and converted to percentages using Results Participant demographics Overall, the mean age of the participants in this sample was 46 years, and racial subgroups included white, Chinese, Japanese or black participants (Table 1). A total of 15% of the participants were ODMA non-producers. Distribution of self-reported health status, smoking status, menopausal status and blood pressure medication were relatively consistent across the ODMA metabotype groups. Table 1SWAN data sub-analysis - sample descriptives (women aged 42–52)ODMAHigh-producersODMALow-producersODMANon-producers(n = 35)(n = 90)(n = 23)Age in years, baseline Mean (SD)46.4 (2.6)46.3 (2.6)46.1 (2.3)Race, baseline Black9 (25.7%)12 (13.3%)7 (30.4%) Chinese or Japanese4 (11.4%)25 (27.8%)5 (21.7%) White22 (62.9%)53 (58.9%)11 (47.8%)Self-reported health status, baseline excellent11 (31.4%)22 (24.4%)8 (34.8%) very good11 (31.4%)33 (36.7%)6 (26.1%) good or fair or poor13 (37.1%)35 (38.9%)9 (39.1%)Smoking status, baseline never24 (68.6%)58 (64.4%)18 (78.3%) former or current11 (31.4%)32 (35.6%)5 (21.7%)Menopausal status, visit 5 post14 (40.0%)39 (43.3%)10 (43.5%) peri/pre/unknown21 (60.0%)51 (56.7%)13 (56.5%)Taking blood pressure medication when blood pressure measured no29 (82.9%)71 (78.9%)20 (87.0%) yes6 (17.1%)19 (21.1%)3 (13.0%)Height, cm, visit 5 Mean (SD)164.1 (5.0)162.7 (8.1)163.2 (6.5)Weight, kg, visit 5 Mean (SD)70.9 (15.5)73.4 (18.8)75.6 (19.0)Waist circumference, cm, visit 5 Mean (SD)83.1 (14.4)87.2 (15.8)87.7 (15.1) SWAN data sub-analysis - sample descriptives (women aged 42–52) CVD risk associations After adjustment for age, race, self-reported health status, smoking status, menopausal status, blood pressure medication, height, weight and waist circumference, ODMA non-producers’ mean systolic blood pressure was significantly higher than ODMA high-producers (Table 2). Another trend but not statistically significant difference observed included higher fasting serum glucose (adj. mean difference = 6.1 mg/dl, CI: -0.9, 13.2, p = 0.09) in ODMA non-producers when compared to ODMA producers. (Table 2). No statistically significant differences in average diastolic blood pressure, total cholesterol, triglycerides, LDL, HDL, insulin, or CRP were observed across ODMA metabotype groups. In a dichotomous analysis, no statistically significant associations were observed between ODMA metabotype and the CVD risk variables, each classified into elevated risk and lower risk groups (Table 3). With the exception of CRP and insulin, direction of non-significant associations was associated with adverse CVD risk factors in the ODMA non-producers compared to high producers. Table 3Odds ratios for CVD risk markers among metabotypes (reference = ODMA high-producers)ODMA low-producer (n = 90)ODMA non-producer (n = 23)Odds ratio95% CIp-valueOdds ratio95% CIp-valueElevated SBP* (≥ 120 mmHg)1.850.63, 5.960.282.550.64, 10.730.19Elevated DBP* (≥ 80 mmHg)1.130.35, 4.180.842.160.52, 9.610.29Low HDL* (< 50 mg/dl)0.950.32, 3.030.931.830.47, 7.270.38Elevated LDL* (≥ 160 mg/dl)0.750.19, 3.280.681.420.23, 8.220.69Elevated CRP* (≥ 10 mg/l)0.890.13, 6.620.900.360.03, 3.630.38Elevated Fasting Serum Glucose* (≥ 100 mg/dl)5.370.69, 119.920.173.940.12, 141.380.40Elevated Insulin* (≥ 12.94 uIU/ml)0.670.18, 2.670.560.910.15, 5.230.92Elevated Cholesterol* (≥ 200 mg/dl)0.750.31, 1.760.511.420.45, 4.620.56Elevated Triglycerides* (> 150 mg/dl)0.460.14, 1.550.211.610.40, 6.620.50*adjusted for age, race, self-reported health status, smoking status, menopause status, blood pressure medications, height, change in weight from 1st to 2nd visit, and change in waist circumference from 1st to 2nd visit Odds ratios for CVD risk markers among metabotypes (reference = ODMA high-producers) *adjusted for age, race, self-reported health status, smoking status, menopause status, blood pressure medications, height, change in weight from 1st to 2nd visit, and change in waist circumference from 1st to 2nd visit Discussion The purpose of this study was to evaluate whether the CVD risk associations observed in earlier ODMA metabotype studies (e.g. Liu et al. 2014, Guo et al. 2010), were also observed in a middle-aged study population of women transitioning through menopause. The present study evaluated 148 female SWAN participants, age 42–52 years, who were white, African American, Chinese, Hispanic or Japanese, who visited the study sites located in the United States. Although metabotyping was completed without a soy challenge, the present study’s sample containing 15% ODMA non-producers falls within this expected range for both the total sample as well as the racial sub-groups, which is consistent with prior observations’ estimate that 10–47% of the population are ODMA non-producers [19, 21]. Also consistent with previous literature, ODMA non-producers had a shorter average height, heavier average weight, and larger waist circumference than the ODMA high-producers. Although the present pilot study was underpowered, the results suggest less favorable CVD risk profiles for ODMA non-producers when compared to ODMA high-producers. ODMA non-producers had significantly higher systolic blood pressure compared to ODMA high-producers. This observation is similar to results of Liu et al.; in a sample of 595 post-menopausal women in Hong Kong, ODMA non-producers had significantly higher systolic blood pressure (p = 0.02). Additionally, in this sample of women undergoing a menopausal transition, the average mean difference of 7.6 mmHg systolic blood pressure between groups may also be clinically significant, as research has shown that a 5–10 mmHg decrease in systolic blood pressure can have a 10 to 20% decrease in CVD events, respectively [35]. Blood collection was performed at the fifth visit, and 55 of the 148 urine samples were collected at the ninth visit. In order to explore the potential for reverse causality, a sensitivity analysis was conducted, excluding the ninth visit data. As referenced in Additional file 1, the systolic blood pressure point estimates both with the ninth visit data and without the ninth visit data remained comparable. Additionally, given concerns that blood pressure medication may confound the SBP results even after adjustment, a sensitivity analysis was conducted, excluding the 28 individuals who were taking blood pressure medication. As referenced in Additional file 1, the SBP point estimate including the 28 individuals taking blood pressure medication remained comparable to the SBP point estimate when these 28 individuals were excluded. Gut microbiome composition and cardiovascular disease risk factors are interrelated, as certain cardiovascular disease risk factors have been known to alter metabolism and gut microbial composition. For example, evidence suggests that hypertensive individuals tend to have gut microbiomes of lower diversity and higher levels of Clostridiales and Bacterodiales species, so there can be an alternative interpretation of these results [36]. Liu et al. (2014) observed a significantly higher total cholesterol in ODMA non-producers compared to producers; however, we did not observe a difference in total cholesterol between metabotypes in this SWAN sample. Additionally, Guo et al. (2020) observed, in a study of 202 individuals aged 20–69 who consumed isoflavone supplements, no statistically significant difference in blood pressure, glucose, total cholesterol, triglycerides, LDL and HDL between ODMA producers and non-producers. These inconsistent results among these three studies may be attributed to heterogeneity in sample characteristics, study design, as well as ODMA metabotype classification methods. To compare, the sample characteristics differed among these studies. The Liu et al. study examined a total of 595 post-menopausal pre-hypertensive women, age 48–70 years, who had been recruited for a randomized controlled trial designed to examine the effect of soy on blood pressure and other CVD risk in Hong Kong. The Guo et al. study evaluated a total of 102 male and 100 female participants aged 20–69 years in Guangzhou, China. The present study included women in various stages of menopause transition, including pre-, peri- and post-menopause. Previous literature also revealed associations between ODMA metabotype and other health outcomes such as hormone concentrations [22] and bone mineral density [20, 21] in women, but each stage was examined independently (i.e. exclusively pre-menopause or exclusively post-menopause). This study builds on prior observations, examining a sample of women throughout all stages of menopause transition. The menopause transition is associated with fluctuating and decreased estrogen levels, and evidence suggests an association between gut microbiome composition and estrogen levels. Findings demonstrate higher estrogen levels associated with higher gut microbiome diversity [37] as well as higher abundance of Bacteroidetes and lower abundance of Firmicutes and Ruminococcaceae [38]. Some women undergo hormone replacement therapy (HRT) as a means to combat menopausal symptoms, and preliminary evidence suggests it can impact gut microbiome composition, improving gut dysbiosis. However, most of these studies were conducted in mice [39]. More human subjects research is needed to characterize HRT’s impact on gut microbiome composition and any potential relationship with the ODMA metabotype. Study design may have also contributed to the inconsistent results among these three studies. While the present study did not incorporate a soy or daidzein challenge, both Liu et al. and Guo et al. controlled for daidzein consumption by either a 7-day, 60 mg daidzein challenge (Liu) or a 3-day isoflavone supplement (Guo). While a soy challenge is considered the gold standard for ODMA metabotyping, it was not available for this pilot study, and this may have further contributed to misclassification in the ODMA metabotype. Prior research has observed that secondary analyses of populations who are low-soy consuming may be prone to false-negative non-producer classification, due to an inability to detect ODMA at such low levels [18]. In terms of power to detect statistically significant differences, the Liu et al. study had an advantage of being larger (n = 597) while the present study as well as the Guo et al. study acknowledged one of the limitations of the study was the lack of power to detect significantly different CVD risk markers between groups due to its relatively small sample size. Contrary to Liu et al., Reverri et al. assessed cardiometabolic risk factors across ODMA metabotypes after 4 weeks soy food consumption and concluded there was no significant difference in serum lipids or glucose between ODMA producers and non-producers [40], and when comparing glycemic control, observed no significant changes in results with the addition of ODMA metabotype as a covariate [41]. Differences between study populations and metabotyping procedures make it difficult to draw direct comparisons between results. Reverri et al. required all participants (n = 17) to be hypertensive with at least one other cardiometabolic risk factor such as elevated fasting glucose, triglycerides, waist circumference or reduced HDL. Examining a population already at-risk changes the risk profile and precludes the delineation between populations at risk and not at risk, as was done in the present study. A secondary goal of this pilot study was to evaluate feasibility of identifying ODMA producers and non-producers using stored samples from people consuming daidzein but without a daidzein challenge. The present SWAN study used a product-precursor ratio to classify ODMA metabotype, where ODMA high-producers had a urinary ODMA (nmol/L)/daidzein (nmol/L) ratio > 0.5. Given the lack of consensus for an optimal ODMA concentration cut-off value for metabotyping, thresholds for ODMA low and high producers for the present study were defined by visually inspecting histogram distributions among individuals with detectable ODMA and separating them based on distinct low and high concentration clusters. Guo et al. also used ratios to define ODMA metabotypes, where ODMA producers were based upon an ODMA: daidzen ratio > 0.018. These ODMA metabotyping differences are consistent with the characteristics of each study’s sample, the type of biospecimen analyzed as well as the study design. Urine, for example, provided a clear distinction between ODMA producer and non-producer metabotypes at the absolute ratio of 0.018, but no distinct demarcation was seen for serum [42]. Since both the Liu and Guo study designs included a daidzein challenge, the lower ODMA (nmol/L)/daidzein (nmol/L) cutoff for producer classification was reasonable, with confidence that the ODMA concentration can be attributed to the soy/daidzein consumed and metabolized by the gut bacteria as opposed to ODMA derived from milk products [43]. Our study has some limitations. Participants included in this analysis reported regular consumption of daidzein 4–7 days per week, and this may or may not have been temporally aligned with urine sample collection for ODMA analysis. To mitigate bias resulting from this potential temporal misalignment, this analysis excluded participants whose urinary daidzein concentration was less than or equal to the lower 25th percentile of daidzein in this study population. While this additional exclusion may have decreased sample size, limiting power to detect statistically significant associations, this was done in order to decrease bias and limit the probability of ODMA metabotype misclassification and increase the quality of the data. Despite the decrease in sample size due to these analytical decisions, CVD risk marker observations were similar in magnitude and direction to the Liu et al. (2014) study [19], suggesting consistency in observations of some unfavorable CVD risk markers associated with the ODMA non-producer metabotype. Adequately powered future studies may be able to confirm whether the present study’s systolic blood pressure and glucose trends represent true associations or random noise. Conclusions In this diverse sample of U.S. women, this study observed similar results to prior work where ODMA non-producers had significantly higher systolic blood pressure compared to producers. Those studies also found other factors indicating a more unfavorable CVD risk profile for ODMA non-producers, which were not observed in this study. Future studies which use a standardized challenge are preferrable, but in the absence of that, this study shows that a larger, secondary observational study could also be a viable approach to better understand these CVD associations and how the ODMA metabotype affects women’s health when the study participants consume sufficiently high daidzein in the study. Supplementary Information Supplementary Material 1. Acknowledgements We would like to acknowledge the contribution of the Study of Women’s Health Across the Nation (SWAN) Investigators and the National Institute on Aging (NIA) AgingResearchBiobank (https://agingresearchbiobank.nia.nih.gov) where the SWAN collection of biospecimens and data are maintained. Abbreviations O-desmethylangolensin Study of Women’s Health Across the Nation Cardiovascular disease High-density lipoprotein Low-density lipoprotein C-reactive protein Systolic blood pressure Diastolic blood pressure High performance liquid chromatography tandem mass spectrometry Ribonucleic acid National Institutes of Health Dietary daidzein equivalent Liquid chromatography-quadrupole time-of-flight Electrospray Ionization Authors’ contributions HC and CF analyzed and interpreted the data regarding ODMA metabotype and CVD risk factors and were major contributors in writing the manuscript. MS was a major contributor in writing the manuscript. RC and AD performed MS-MS identification for equol, ODMA and daidzein. All authors read and approved the final manuscript. Funding George Mason University, College of Public Health pilot funding. Data availability The SWAN public use datasets analyzed during the current study are available in the ICPSR repository, [https://www.icpsr.umich.edu/web/NACDA/series/253](https:/www.icpsr.umich.edu/web/NACDA/series/253) . Other datasets used and/or analyzed during the current study are available from the corresponding author on request. Declarations Ethics approval and consent to participate At the time of the original SWAN study, participants provided written consent to participate. We employed a secondary analysis using de-identified publicly available SWAN data and did not include any personal identifiable information of participants. This study protocol conforms to the provisions of the Declaration of Helsinki. The present study’s secondary analysis was approved by the University of Maryland College Park (UMCP) Institutional Review Board (IRB), College Park, MD (IRBNet ID 2039359-1). The IRB determined the present project did not meet the definition of human subject research under the purview of the IRB according to federal regulations. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The SWAN public use datasets analyzed during the current study are available in the ICPSR repository, [https://www.icpsr.umich.edu/web/NACDA/series/253](https:/www.icpsr.umich.edu/web/NACDA/series/253) . Other datasets used and/or analyzed during the current study are available from the corresponding author on request.