Diet Diversity Is Associated with Prepregnancy Body Mass Index but Not the Pregnancy Third-Trimester Gut Microbiota Authors: Heli J Sheth, Shiyi YL Xie, Jean M Kerver, Sarah S Comstock Published: 2026 ABSTRACT AbstractBackgroundThe prevalence of women entering pregnancy with overweight or obesity is rising in the United States. The dual burden of undernutrition and overweight or obesity can lead to micronutrient deficiencies. Dietary diversity supports micronutrient adequacy and can be assessed using the minimum dietary diversity for women of reproductive age (MDD-W), a food group–based indicator using 24-h dietary recall data. Lower dietary diversity may be associated with lower gut microbiome diversity and overall community composition; however, the relationship during pregnancy is unclear.ObjectivesThis study aimed to 1) describe associations between prepregnancy body mass index (pBMI) and dietary diversity during pregnancy and 2) investigate whether the pregnancy gut microbiota diversity is associated with dietary diversity.MethodsData were collected from participants in the Michigan Archive for Research on Child Health cohort (n = 132, third trimester). Twenty-four-hour dietary recall data were used to assess dietary diversity according to the MDD-W indicator. A subset (n = 114) provided fecal samples for 16S ribosomal ribonucleic acid gene sequencing.ResultsOn the basis of pBMI, 49% of participants had normal weight, 25% had overweight, and 26% had obesity. Parity, income, and marital status differed significantly across pBMI categories. Overall, 20.5% of participants did not meet the minimum dietary diversity (MDD-W score <5). pBMI was inversely correlated with MDD-W scores, with women with obesity having lower dietary diversity than women with normal weight. Consumption of low-nutrient density foods was similar across pBMI categories. Gut microbiome α- and β-diversity were not associated with MDD-W.ConclusionsAlthough no associations between the gut microbiota and diet diversity were observed, our results confirm that women with higher pBMI may be at risk of inadequate micronutrient intake. These results emphasize the value of dietary diversity screening and intervention in care for women with higher pBMI, given the well-established role of micronutrients for maternal and fetal health during the third trimester and beyond.Keywords: pregnancy, nutrition, diet diversity, micronutrient, microbiota, gastrointestinal, diet, BMI Introduction The prevalence of women entering pregnancy with overweight or obesity has increased steadily in the United States [1,2]. Higher prepregnancy BMI (pBMI) is associated with adverse maternal and infant health outcomes, including gestational diabetes, hypertensive disorders of pregnancy, systemic inflammation, and infant complications [[3], [4], [5], [6]]. These risks underscore the need to better understand the nutritional patterns among pregnant women with higher pBMI. Women of reproductive age (WRA) are vulnerable to nutritional inadequacies due to the physiological demands of pregnancy and lactation [[7], [8], [9]]. Many women enter pregnancy undernourished, and entering pregnancy with overweight or obesity may further increase risk of micronutrient inadequacies, in part, because higher pBMI is often associated with low diet quality [7,[10], [11], [12], [13], [14], [15]]. Given that diet quality may contribute to the dual burden of undernutrition and overweight or obesity, dietary diversity is a practical indicator of micronutrient adequacy. Dietary diversity is important as it increases the likelihood of adequate micronutrient intake and consumption of nutrient-dense food groups [[16], [17], [18]]. Dietary diversity reflects the variety of food groups consumed within a given time frame and serves as a proxy for nutrient adequacy [16,19]. During pregnancy, higher dietary diversity has been associated with improved mean probability of adequacy for multiple vitamins and minerals [16,20,21]. Despite these needs, dietary diversity often declines during the third trimester, which can contribute to suboptimal nutrient intake and impact fetal growth indicators, such as weight-for-age and BMI-for-age z-scores at birth [22]. One way to measure dietary diversity is the minimum dietary diversity for women of reproductive age (MDD-W). MDD-W is a validated and dichotomous indicator of micronutrient adequacy that classifies whether a woman has consumed ≥5 of 10 defined food groups in the past 24 h [20,23,24]. Although MDD-W was developed and validated for WRA rather than specifically for pregnant women, pregnant women fall within this broader target population, and its application in pregnancy cohorts remains an area where additional evidence is needed. MDD-W has been used to assess diet quality during pregnancy and guide nutrition-related policies and programs [20,23,24]. The gut microbiota plays a key role in human health by contributing to nutrient metabolism and immune development [[25], [26], [27]]. During pregnancy, the gut microbiota undergoes characteristic shifts, partly driven by hormonal and immune changes, which may influence both adaptations during pregnancy and infant microbial colonization [28,29]. Diet is one of the most important determinants of gut microbiota composition, and greater dietary diversity is consistently associated with higher microbial richness and stability [[30], [31], [32], [33], [34]]. pBMI is linked to prenatal characteristics and variation in fecal microbial communities [[35], [36], [37]]. These findings highlight that diet during pregnancy and pBMI are major determinants of gut microbial composition during pregnancy. Previous research identified an inverse association between pBMI and dietary diversity in a separate cohort of pregnant women in Michigan [38]. Although the pregnancy diet has been associated with the gut microbiome during pregnancy [39,40], there is a lack of research specifically investigating how dietary diversity relates to gut microbiome composition and diversity in pregnant women. By applying similar methods from prior studies to a unique and larger cohort, this study evaluates whether and how previously observed patterns hold. Herein, associations between pBMI and dietary diversity during pregnancy as well as associations between dietary diversity during pregnancy and gut microbiome diversity are tested. We hypothesized that higher pBMI would be negatively associated with dietary diversity during pregnancy, and that dietary diversity during pregnancy would be positively associated with third-trimester gut microbiota diversity. Methods Study participants The present study was a secondary analysis of data from participants recruited from the Michigan Archive for Research on Child Health (MARCH) prospective pregnancy cohort [41]. The MARCH cohort is part of the nationwide Environmental influences on Child Health Outcomes program [42], and includes 1130 participants recruited from 21 prenatal clinics across the state of Michigan’s Lower Peninsula [41]. Study participants were recruited at their first prenatal care visit at the sampled clinic, regardless of when it took place in their pregnancy. Women who switched providers before birth and women who left a clinic before birth were dropped from the sample. Finally, only women who were aged ≥18 y and able to answer a survey written in English were recruited into the study [41]. The analysis presented herein includes 132 pregnant women from the MARCH study in their third trimester who completed 24-h diet recalls and had complete covariate data, including self-reported height and weight data. An overlapping subset of these women provided fecal samples for microbiome analysis (n = 114) (Figure 1). This study received approval from the human subjects’ research protection program at Michigan State University. All participants provided written informed consent. Flowchart of participants included in each analysis: pBMI and MDD-W (n = 132), MDD-W and microbiota (n = 114). MDD-W, minimum dietary diversity for women of reproductive age; pBMI, prepregnancy BMI. Data collection An enrollment questionnaire was administered during each participant’s first prenatal study visit, where participants self-reported their prepregnancy height and weight, along with sociodemographic information such as age, parity, race, income, marital status, and education level. In addition, health-related behaviors and history, including smoking status and antibiotic use, were collected through this enrollment questionnaire as well as through the questionnaire administered at the time of stool sample collection. Self-administered 24-h dietary recall surveys were mailed to third-trimester participants. Participants reported all foods and beverages consumed in the 24 h before completing the survey using an open-ended question and a checklist of 27 foods, categories, beverages, and supplements, with each participant completing a single 24-h dietary recall. Specifically, participants were asked to, “Place an ‘x’ in the box next to anything you ate in the past 24 h.” The checklist included: cow milk; other milk: soy, rice, almond, coconut, goat, etc.; other dairy foods: yogurt, cheese, ice cream, pudding, etc.; other soy foods: tofu, frozen soy desserts, etc.; 100% fruit juice or 100% vegetable juice; sweet drinks: regular soda, sweet tea, sports drinks, etc.; alcohol; caffeinated beverage (coffee, tea, energy drinks, etc.); other grains: cheerios, other breakfast cereals, crackers, breads, pasta, rice, etc.; fruit or fruit purees; vegetables or vegetable purees; carrots, sweet potatoes, mangoes, apricots, bell peppers; spinach, kale, Swiss chard, romaine lettuce; French fries; meat: chicken, beef, ham, combination dinners; fish or shellfish; peanut butter, other peanut foods (Bamba), other nuts; liver or other organ meats; eggs; beans, lentils, peas; sweet foods: candy, chocolate, cookies, cakes, etc.; prenatal vitamin; other multivitamin; fish oil, DHA or EPA supplement; prebiotic supplement (i.e., inulin, etc.); probiotic supplement (i.e., culturelle, ultimate flora, etc.); calcium supplement; iron supplement; other dietary supplement, provide name:_____. Along with the survey, participants were sent fecal sample collection kits containing an OMNIgene-GUT tube (DNA Genotek Inc), detailed collection instructions, and return boxes with prepaid postage. The OMNIgene-GUT tube contained a proprietary stabilizing solution that inactivates microbial growth and preserves microbial DNA integrity at ambient temperature, allowing samples to be stored and shipped without immediate freezing. Participants were instructed to collect their fecal sample using the provided kit and to complete the dietary recall after sample collection. Both the surveys and fecal samples were returned to the laboratory by mail through the United States Postal Service. Upon arrival, fecal samples were aliquoted and stored at −80°C until further processing. pBMI classification Participants were classified according to Centers for Disease Control BMI categories: underweight, healthy (normal) weight, overweight, or obesity. Individuals with a BMI <18.5 were considered underweight, those between 18.5 and 24.9 were classified as normal (healthy) weight, between 25 and 29.9 as overweight, and ≥30 as having obesity [43]. pBMI was calculated based on self-reported height and weight before pregnancy [44,45]. Because of the small sample size, 3 participants in the underweight category were excluded from the analysis. Participant characteristics are shown in Table 1. TABLE 1Characteristics of third-trimester pregnant participants (n = 132) by BMI statusTABLE 1Normal (n = 65)Overweight (n = 33)Obese (n = 34)Prepregnancy characteristicMean (SD)Mean (SD)Mean (SD)P value Age131.0 (4.6)31.8 (6.1)30.7 (4.0)0.66 BMI122.2 (1.7)26.8 (1.7)37.0 (5.2)<0.0001 Parity10.94 (1.0)0.64 (0.9)1.47 (1.0)0.0026Prepregnancy characteristicn (%)3n (%)3n (%)3Race20.16 White63 (96.9)31 (94.0)30 (88.2) Black1 (1.5)1 (3.0)4 (11.8) Other1 (1.5)1(3.0)0 (0.0)Participant income20.0027 <$50,0009 (13.8)8 (24.2)17 (50.0) $50,000–$100,00023 (35.4)10 (30.3)9 (26.5) >$100,00031 (47.7)14 (42.4)7 (20.6) Missing2 (3.1)1 (3.0)1 (2.9)Married259 (90.8)27 (81.8)19 (55.8)0.0002College graduate (Bachelor’s)241 (63.1)21 (63.6)13 (38.2)0.06Missing0 (0.0)0 (0.0)1 (2.9)Smoking during pregnancy22 (3.1)4 (12.1)5 (14.7)0.10Missing1 (1.5)0 (0.0)0 (0.0)Antibiotic use in the past year226 (40.0)19 (57.6)19 (55.9)0.13Missing0 (0.0)0 (0.0)1 (2.9)Abbreviation: ANOVA, analysis of variance.1Continuous variables were analyzed using 1-way ANOVA.2For categorical variables presented as percentages, the χ2 test was used to assess statistical significance.3Percentages represent the proportion within each BMI category. Characteristics of third-trimester pregnant participants (n = 132) by BMI status Abbreviation: ANOVA, analysis of variance. Continuous variables were analyzed using 1-way ANOVA. For categorical variables presented as percentages, the χ2 test was used to assess statistical significance. Percentages represent the proportion within each BMI category. MDD-W scoring We assessed dietary diversity using the MDD-W indicator based on 24-h dietary recall data [24]. Foods were assigned to one of the 10 MDD-W food groups (grains/roots/plantains; pulses; nuts and seeds; dairy; meat/poultry/fish; eggs; dark green leafy vegetables; vitamin A–rich fruits and vegetables; other vegetables; and other fruits) [24]. For each participant, each food group consumed was coded as 1 (any intake) or 0 (no intake), and scores were summed (maximum score = 10). Consistent with MDD-W guidance, participants were classified as achieving minimum dietary diversity if they consumed ≥5 food groups [24]. Separately, 24-h recalls were also used to identify intake of 3 low-nutritional-density food categories (fried and salty foods, sweet foods, and sweetened beverages); participants were classified as consumers if they reported consuming ≥1 item in the category [24]. DNA extraction and gene sequencing Genomic DNA (gDNA) was extracted from fecal samples using the DNeasy PowerSoil Pro Kit (Qiagen). DNA concentration was quantified using the Quant-iT dsDNA Assay Kit, High Sensitivity (Thermo Fisher). The V4 region of the 16S rRNA gene was amplified using the Schloss laboratory primers [46], AccuPrime Pfx SuperMix (ThermoFisher Scientific), and template DNA, with all PCR reactions performed in triplicate. A no-template negative control (nuclease-free water in place of gDNA template) was included in each PCR run to monitor potential contamination. The PCR amplicons were confirmed by agarose gel electrophoresis, after which the triplicate PCR products were pooled and purified using Agencourt AMPure XP beads (Beckman Coulter) for amplicon purification. The final amplicon libraries were submitted for 16S rRNA gene sequencing at the Michigan State University Genomics Research Technology Support Facility. Sequencing was performed on an Illumina MiSeq using 250 base-pair (bp) paired-end reads with V2 chemistry. Sequencing data processing The 16S rRNA gene sequences were processed using mothur (v.1.48.2). The MiSeq standard operating procedure was followed for processing [47]. After merging forward and reverse reads, sequences containing ambiguous bases or exceeding 275 bp were removed to ensure alignment with the V4 region. The mean number of reads per sample was 53,619, with a median number of 57,171 reads, and a minimum number of 1 read and a maximum number of 153,485 reads. Samples with <1000 reads (n = 2) were excluded before rarefaction. Reads were rarefied to 15,000 reads per sample using the subsample function in the vegan package (2.7-1) in R. The SILVA reference taxonomy (Version 138) [48] was used for sequence alignment, and taxonomy was assigned based on operational taxonomic units following the phylotype approach, using version 19 of the PDS reference database [49]. Statistical analysis Data analysis was conducted using R version 4.4.1 [50], incorporating multiple packages including vegan, dplyr, and ggplot2. The study dataset, consisting of 132 participants, was cleaned by excluding women with underweight and individuals with missing pBMI values or missing covariate data. Normality was assessed using the Shapiro–Wilk test. Normally distributed variables were compared using 1-way analysis of variance (ANOVA), and nonnormally distributed variables were compared using Kruskal–Wallis tests. Categorical variables were compared using χ2 tests, including race, income, education level, marital status, smoking during pregnancy, and antibiotic use. The percentage of participants consuming each of the 10 MDD-W food groups was calculated overall and stratified by both pBMI category and MDD-W category. Spearman rank correlation was used to assess the association between pBMI and total MDD-W score. χ2 tests were used to compare the percentage of individuals achieving MDD-W scores ≥5 across pBMI categories. Post hoc pairwise χ2 tests were conducted to compare pBMI category pairs. Kruskal–Wallis tests were used to compare total MDD-W scores across pBMI categories. Pairwise Mann–Whitney U tests with Bonferroni correction were used for post hoc comparisons. χ2 tests were also used to compare the frequency of consumption of low-nutrient-density foods across pBMI categories. Taxa-level differences between MDD-W groups were assessed using negative binomial regression, and P values were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) method. Alpha Diversity, reflecting the diversity within individual samples, was evaluated using 3 indices: Chao1, Shannon, and inverse Simpson. The Shapiro–Wilk test was applied to examine the normality of each index. Group comparisons (MDD-W) for α diversity were performed using the Kruskal–Wallis test based on the designated grouping variable in this study. Beta diversity, representing differences in community composition between samples, was quantified with Sorensen and Bray–Curtis dissimilarity metrics. Principal coordinate analysis (PCoA) was used to visualize β-diversity patterns [50]. Statistical differences in β-diversity were assessed through permutational multivariate ANOVA (PERMANOVA) and permutation tests for homogeneity of dispersion. P values from microbiota diversity analyses were adjusted for multiple testing using the Benjamini–Hochberg FDR method. For α diversity, P values from Chao1 richness, Shannon diversity, and inverse Simpson diversity comparisons were adjusted. For β diversity, P values from analyses using Sorensen and Bray–Curtis dissimilarity metrics were adjusted. For multivariable models, relevant prenatal characteristics were included as covariates, including pBMI, maternal age, education level, marital status, and smoking status. Alpha-diversity models were evaluated with type II ANOVA using the lm function, whereas β diversity models were analyzed with PERMANOVA via the adonis2 function. A P value of ≤ 0.05 was considered statistically significant. Multicollinearity among predictors in the multivariable models was assessed using the “check_collinearity” function from the “performance” R package [51]; variance inflation factor values <5 were considered to indicate low multicollinearity. Results Participant characteristics Participant characteristics were compared across pBMI categories: normal, overweight, and obese (Table 1). Among 132 participants, 49% were classified as having a normal BMI (n = 65), 25% as having overweight (n = 33), and 26% as having obesity (n = 34). Race, smoking during pregnancy, antibiotic use within the past year, and age did not significantly differ by pBMI status. Participant ages ranged from 20 to 52 y, with a median age of 31 y. Parity significantly differed by pBMI status, with individuals in the obese group reporting a higher number of births. Participant income was also significantly inversely associated with pBMI status. A greater proportion of participants in the normal BMI category reported annual incomes >$100,000, whereas the participants with obesity were more likely to report incomes of <$50,000. Marital status significantly differed across pBMI groups, with a higher percentage of participants with obesity being unmarried. There was a trend toward a greater proportion of participants in the normal BMI category having a bachelor’s degree or higher, compared with those in the obese category (P = 0.06). Consumption of the 10 MDD-W food groups was summarized as a percentage and described within pBMI categories, stratified by total MDD-W score (<5 compared with ≥5) (Table 2), to illustrate differences in food group consumption patterns across pBMI categories and dietary diversity groups. Overall, 20.5% of participants did not meet minimum dietary diversity. Pregnant women with MDD-W scores of <5 consumed grains; pulses, nuts, and seeds; eggs; dark green leafy vegetables; vitamin A–rich vegetables; other vegetables; and fruits less frequently compared with those with MDD-W scores of ≥5. Among the 132 pregnant women in the study, the most commonly consumed food groups were grains and dairy. Other frequently consumed groups included meat, poultry, and fish; other vegetables; and other fruits. Nuts and seeds were consumed by approximately half of the participants, whereas consumption of vitamin A–rich fruits or vegetables, eggs, and dark green leafy vegetables was lower. During the 24-h recall period, pulses were the least consumed food group, with only 25% of participants reporting intake of any pulses. Only 1 participant with low dietary diversity (MDD-W score <5) consumed pulses, and only 30.5% (n = 32) of those achieving minimum dietary diversity (MDD-W score ≥5) reported consuming pulses. TABLE 2MDD-W food group consumption [n (%)] according to BMI category and total MDD-W scoreTABLE 2Sample size, n (%)NormalOverweightObeseOverall65 (49.2)33 (25)34 (25.8)132 (100.0)Sample size758726132127105MDD-W score<5≥5<5≥5<5≥5<5≥5Grains14 (57.1)57 (98.3)5 (71.4)24 (92.3)7 (53.8)21 (100)16 (59.3)2102 (97.1)Pulses (beans, peas, and lentils)11 (14.3)19 (32.8)0 (0)9 (34.6)0 (0)4 (19)1 (3.7)232 (30.5)Nuts and seeds10 (0)36 (62.1)3 (42.9)13 (50)3 (23.1)12 (57.1)6 (22.2)261 (58.1)Dairy16 (85.7)51 (87.9)7 (100)24 (92.3)9 (69.2)21 (100)22 (81.5)96 (91.4)Meat, fish, or poultry15 (71.4)44(75.9)4 (57.1)20 (76.9)10 (76.9)21 (100)19 (70.4)85 (81)Eggs10 (0)27 (46.6)0 (0)11 (42.3)1 (7.7)9 (42.9)1 (3.7)247 (44.8)Dark green leafy vegetables10 (0)26 (44.8)0 (0)9 (34.6)2 (15.4)10 (47.6)2 (7.4)245 (42.9)Other vitamin A–rich fruits or vegetables10 (0)33 (56.9)0 (0)10 (38.5)1 (7.7)8 (38.1)1 (3.7)251 (48.6)Other vegetables12 (28.6)52 (89.7)1 (14.3)23 (88.5)5 (38.5)18 (85.7)8 (29.6)293 (88.6)Other fruits12 (28.6)51 (87.9)1 (14.3)21 (80.8)7 (53.8)16 (76.2)10 (37)288 (83.8)Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age.1Values indicate the number and percentage of participants within each BMI category who received the corresponding overall MDD-W score and consumed that particular category of food [n (%)].2Overall comparisons of the proportion of individuals consuming the specified food group for those individuals consuming <5 food groups compared with those consuming ≥5 food groups were tested using Fisher’s exact test. A superscript 2 indicates that a statistically significantly different proportion of individuals in the <5 and ≥5 MDD-W score groups consumed the specified food group. MDD-W food group consumption [n (%)] according to BMI category and total MDD-W score Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age. Values indicate the number and percentage of participants within each BMI category who received the corresponding overall MDD-W score and consumed that particular category of food [n (%)]. Overall comparisons of the proportion of individuals consuming the specified food group for those individuals consuming <5 food groups compared with those consuming ≥5 food groups were tested using Fisher’s exact test. A superscript 2 indicates that a statistically significantly different proportion of individuals in the <5 and ≥5 MDD-W score groups consumed the specified food group. MDD-W score and pBMI An inverse correlation was observed between pBMI and MDD-W scores in the third trimester of pregnancy (Figure 2A, r = –0.26, P = 0.003). The proportion of participants achieving a total MDD-W score >5 differed significantly by pBMI category (Figure 2B, P = 0.002). This proportion was highest among women with normal BMI (89.2%), followed by those with overweight (78.8%) and obesity (61.8%). Post hoc pairwise comparisons revealed a significant difference between those of normal weight and those with obesity, with a smaller proportion of women with obesity meeting the minimum dietary diversity threshold compared with that of women with normal BMI (Figure 2B). In addition, third-trimester MDD-W scores varied across pBMI categories, with a significantly lower score observed in the obese group compared with the normal weight group (Figure 3). Associations between MDD-W scores and prepregnancy BMI. (A) Prepregnancy BMI and MDD-W score during the third trimester of pregnancy are inversely correlated. This relationship was assessed using Spearman’s rank correlation. (B) The proportion of participants achieving a total MDD-W score ≥5 differed by BMI category (P = 0.002). A smaller percentage of women with obesity received MDD-W scores ≥5 compared with women with normal weight or overweight, with a significant difference observed between women of normal weight and those with obesity before pregnancy. Bars without a common letter differ significantly at P ≤ 0.05. Group differences were assessed using χ2 tests, followed by post hoc tests for pairwise comparisons. MDD-W, minimum dietary diversity for women of reproductive age. Third-trimester MDD-W scores differed across the 3 prepregnancy BMI categories, with a significant difference observed between the normal weight and obese groups. The violin plot illustrates the distribution and density of MDD-W scores within each BMI category, with the box plots inside displaying the median, IQR, and overall range excluding outliers. Bars without a common superscript letter differ significantly at P ≤ 0.05. Group differences were assessed using the Kruskal–Wallis test, followed by Mann–Whitney U post hoc tests for pairwise comparisons. MDD-W, minimum dietary diversity for women of reproductive age. Low-nutritional-density food consumption Consumption of foods with low nutritional density was common among participants during the third trimester of pregnancy (Figure 4). Sweet foods, defined by the MDD-W as items such as chocolates, candies, pastries, cakes, biscuits, cookies, ice cream, and popsicles, were consumed by ∼80% of the sample. Sweetened beverages, as defined by the MDD-W, included items such as soft drinks, sodas, fruit juices, chocolate drinks, energy drinks, and sweetened tea, coffee, or herbal drinks, and were consumed by ∼66% of participants. Fried and salty food consumption was less common overall and varied numerically by pBMI category. Women with obesity had the numerically highest consumption of fried foods (32%), compared with 17% of those with normal BMI and 15% of those who were overweight. However, χ2 tests indicated no statistically significant differences in the consumption of fried and salty foods, sweet foods, or sweetened beverages across pBMI categories (P > 0.05). Consumption of fried and salty foods, sweet foods, or sweetened beverages was similar regardless of the prepregnancy BMI category. Alpha diversity Alpha diversity of the gut microbiome (n = 114) was assessed using the Chao1, Shannon, and inverse Simpson indices. Regardless of the metric analyzed, α diversity of the gut microbiota was similar for participants with MDD-W ≥5 and those with MDD-W <5 (Table 3). These results indicate that gut microbial richness and evenness were comparable within participants in either group. TABLE 3Alpha diversity of the gut microbiotaTABLE 3MDD-W <52MDD-W ≥52P valueChao1 index1100.5 (60.0, 159.1)102.1 (52.5, 207.5)0.693Shannon index12.82 (1.66, 3.40)2.79 (1.09, 3.42)0.605Inverse Simpson index18.28 (2.14, 20.31)9.08 (1.65, 21.38)0.900Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age.1Alpha-diversity of the gut microbiota, measured using Chao1, Inverse Simpson, and Shannon indices, was similar for those with MDD-W ≥5 and MDD-W <5. Kruskal–Wallis tests were used to compare groups.2Median (minimum, maximum) reported. Alpha diversity of the gut microbiota Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age. Alpha-diversity of the gut microbiota, measured using Chao1, Inverse Simpson, and Shannon indices, was similar for those with MDD-W ≥5 and MDD-W <5. Kruskal–Wallis tests were used to compare groups. Median (minimum, maximum) reported. Beta diversity Beta diversity of the gut microbiome (n = 114) was evaluated using Sorensen and Bray–Curtis dissimilarity metrics and visualized using principal coordinate plots (PCoA). The composition and structure of the gut microbiota during pregnancy were similar for participants with MDD-W ≥5 and those with MDD-W <5 based on either Sorensen (Figure 5A) or Bray–Curtis (Figure 5B) distances (PERMANOVA P > 0.05). In the Bray–Curtis ordination, the ordination suggests 3 apparent clusters, primarily associated with the predominance of Bifidobacterium, Parabacteroides, or Faecalibacterium. PCoA plots of the β-diversity of the gut microbiome using (A) Sorensen (PERMANOVA = 0.6022, PERMDISP = 0.4071) and (B) Bray–Curtis (PERMANOVA = 0.3086, PERMDISP = 0.3822) dissimilarity metrics. β-diversity of the gut microbiota was similar for participants with MDD-W ≥5 and MDD-W <5 (P > 0.05). MDD-W, minimum dietary diversity for women of reproductive age; PCoA, principal coordinate analysis; PERMANOVA, permutational multivariate analysis of variance; PERMDISP, permutation tests for homogeneity of dispersion; PC, principal coordinate. Taxa-level differences by MDD-W At the taxa level, we tested differential abundance between MDD-W groups using the negative binomial test for the 19 most abundant taxa. Roseburia was the only taxon that differed significantly by MDD-W group (P < 0.05) and was decreased in the gut microbiota of participants with MDD-W ≥5 compared with those with MDD-W <5. Descriptive relative abundances (mean% ± SD) for abundant taxa are summarized in Table 4. TABLE 4Mean relative abundances (% ± SD) of common taxa overall and by MDD-W groupTABLE 4Overall1MDD-W <51MDD-W ≥51Lachnospiraceae_unclassified2.2 ± 3.61.9 ± 1.22.3 ± 4Bacteroides6.1 ± 3.46.5 ± 5.35.9 ± 2.8Oscillospiraceae_unclassified11.2 ± 9.812.2 ± 10.510.9 ± 9.7Faecalibacterium5.2 ± 44 ± 45.5 ± 4Eubacteriales_unclassified5.2 ± 4.14.8 ± 4.75.3 ± 3.9Phaeacian1.8 ± 2.81.4 ± 2.71.9 ± 2.9Bifidobacterium13.4 ± 10.59.2 ± 8.414.4 ± 10.7Blautia2.7 ± 3.42.6 ± 2.32.8 ± 3.6Bacteria_unclassified4 ± 5.26 ± 7.23.5 ± 4.4Segatella3.2 ± 3.23.7 ± 3.83.1 ± 3Parabacteroides8.9 ± 17.811.2 ± 228.3 ± 16.7Bacillota_unclassified2.6 ± 2.42.5 ± 2.12.6 ± 2.5Roseburia21.9 ± 2.52 ± 2.3a1.9 ± 2.5bStreptococcus1.4 ± 1.71.3 ± 1.31.5 ± 1.8Collinsella1.4 ± 1.91.2 ± 1.61.5 ± 1.9Paraprevotella1.3 ± 2.41.9 ± 4.31.1 ± 1.7Barnesiella1.8 ± 3.71.5 ± 3.61.9 ± 3.7Bilophila1 ± 2.81.6 ± 4.70.9 ± 2.1Prevotellaceae_unclassified1 ± 3.42 ± 4.10.8 ± 3.2Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age.1Values are mean relative abundance (%) ± SD.2Negative binomial regression identified Roseburia as the only taxon with significantly different abundance between groups (P < 0.05), as indicated with the superscript letters, whereas all other taxa were present at a similar level regardless of MDD-W. Mean relative abundances (% ± SD) of common taxa overall and by MDD-W group Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age. Values are mean relative abundance (%) ± SD. Negative binomial regression identified Roseburia as the only taxon with significantly different abundance between groups (P < 0.05), as indicated with the superscript letters, whereas all other taxa were present at a similar level regardless of MDD-W. Multivariable analysis A multivariable analysis was conducted to examine associations between prenatal characteristics and microbial diversity. The model included MDD-W score, pBMI category, age, college graduation status, marital status, and smoking status as covariates. Multivariable analysis—alpha diversity Next, we assessed whether alpha-diversity (Chao1, Shannon, and inverse Simpson) of the pregnancy gut microbiota was associated with diet diversity when analyzed using a multivariable model (Table 5). For Chao1, in the univariable model (model 1), no significant associations were observed between MDD-W category and the Chao1 index (overall P = 0.609). In the multivariable model (model 2), after adjusting for pBMI, smoking status, college graduation status, marital status, and age, no significant associations between Chao1 and MDD-W were detected (overall P = 0.185). For Shannon, in the univariable model (model 1), there was no overall significant association between MDD-W category and the Shannon index (overall P = 0.499). In the multivariable model (model 2), which adjusted for pBMI, smoking status, college graduation status, marital status, and age, the overall model was nonsignificant but indicated a trend (P = 0.103) for decreased Shannon α diversity when adjusting for MDD-W, pBMI, age, education, marital status, and smoking status. Age during pregnancy was associated with the Shannon index (P = 0.016). Other covariates were not significantly associated with the Shannon index (all P > 0.05). With the inverse Simpson index, in the univariable model (model 1), there was no overall significant association between MDD-W category and the inverse Simpson index (overall P = 0.822). In the multivariable model (model 2), which adjusted for pBMI, smoking status, college graduation status, marital status, and age, the overall model was statistically significant (P = 0.011). Age was associated with the inverse Simpson index (P = 0.001). Other covariates were not significantly associated with the inverse Simpson index (all P > 0.05). TABLE 5Multivariable linear regression modeling results for α-diversity indexes according to maternal characteristicsTABLE 5Model (overall P value)VariableRegression coefficientSEP valueChao1 indexModel 1 (P = 0.609)Intercept106.5253.204<0.0001MDD-W ≥ 5———MDD-W < 5–3.6757.1650.609Model 21 (P = 0.185)Intercept97.43619.316<0.0001MDD-W ≥ 5———MDD-W < 52.0897.4180.779Overweight BMI (vs. normal BMI)5.7247.1540.425Obese BMI (vs. normal BMI)–9.9157.7850.206Maternal age (≤31 vs. >31)–4.1666.3530.513College graduate3.6471.9120.059Marital status–6.3687.9970.428Smoking status0.2924.1740.944Shannon indexModel 1 (P = 0.499)Intercept2.7480.045<0.0001MDD-W ≥ 5———MDD-W < 5–0.0690.1010.499Model 21 (P = 0.103)Intercept2.7050.271<0.0001MDD-W ≥ 5———MDD-W < 5–0.0080.1040.938Overweight BMI (vs. normal BMI)0.09990.0980.312Obese BMI (vs. normal BMI)–0.0850.1070.431Maternal age (≤31 vs. >31)–0.2130.0870.016College graduate–0.0120.0260.657Marital status0.1030.1100.352Smoking status0.0160.0570.788InverseSimpson indexModel 1 (P = 0.822)Intercept9.8700.473<0.0001MDD-W ≥ 5———MDD-W < 5–0.2381.0570.822Model 21 (P = 0.011)Intercept9.5282.741<0.001MDD-W ≥ 5———MDD-W < 50.3541.0530.737Overweight BMI (vs. normal BMI)1.1370.9950.256Obese BMI (vs. normal BMI)–0.6761.0830.534Maternal age (≤31 vs. >31)–2.9360.8840.001College graduate–0.3900.2660.146Marital status1.7691.1120.115Smoking status0.4000.5810.492Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age.1Model 2 was adjusted for prepregnancy BMI, maternal age, education level, marital status, and smoking status. Multivariable linear regression modeling results for α-diversity indexes according to maternal characteristics Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age. Model 2 was adjusted for prepregnancy BMI, maternal age, education level, marital status, and smoking status. Multivariable analysis—beta diversity Multivariable models to control for covariates were run for both Sorensen and Bray–Curtis (Table 6). The overall Sorensen multivariable model was nonsignificant (P = 0.521). The college graduation factor was significant (P = 0.036), and the college graduation category contributed the most variance (R2 = 0.015) in the multivariable model. The overall Bray–Curtis multivariable model was nonsignificant (P = 0.308). No single factor was significant, but the marital status category contributed the most variance (R2 = 0.014) in the multivariable model. TABLE 6Beta diversity permutational multivariable analysis of variance using distance matrices according to maternal characteristicsTABLE 6Beta diversity model (overall P value)1VariableP valueR2Overall R2F-statisticSorensen (P = 0.521)MDD-W0.9280.0050.0600.934BMI0.3870.009Maternal age0.2920.010College graduate0.0360.015Marital status0.0750.013Smoking status0.9380.004Bray–Curtis (P = 0.308)MDD-W0.5940.0070.0601.117BMI0.4410.008Maternal age0.5090.008College graduate0.1680.012Marital status0.1090.014Smoking status0.3960.008Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age.1Multivariable models were adjusted for prepregnancy BMI, maternal age, education level, marital status, and smoking status. Beta diversity permutational multivariable analysis of variance using distance matrices according to maternal characteristics Abbreviation: MDD-W, minimum dietary diversity for women of reproductive age. Multivariable models were adjusted for prepregnancy BMI, maternal age, education level, marital status, and smoking status. Discussion In this study, a significant inverse association was observed between MDD-W status during the third trimester and pBMI, with women with obesity significantly less likely to meet minimum dietary diversity compared with those with normal BMI. Nearly 40% of women with obesity failed to achieve the MDD-W threshold, compared with only 11% of those with normal BMI. This finding indicates that women entering pregnancy with higher pBMI are at increased risk of inadequate micronutrient intake. Despite differences in dietary diversity, no statistically significant association was observed between pBMI category and whether participants reported consuming any low-nutrient-density foods during the 24-h dietary recall period. In addition, we found no significant associations between diet diversity (MDD-W) and α diversity of the pregnancy gut microbiota across 3 indices (Chao1, Shannon, and inverse Simpson) in either univariable or multivariable models. Age was significantly associated with both Shannon and Inverse Simpson indices. Although the overall β-diversity models were nonsignificant, college graduation was associated with, and explained the most variance in Sorensen distance, and marital status explained the most variance in Bray–Curtis distance. This study extends our prior findings [38] by replicating the inverse association between pBMI and dietary diversity in a larger Michigan pregnancy cohort and incorporating pregnancy gut microbiome diversity analyses. Although the proportion of participants not meeting minimum dietary diversity was lower in the current cohort than in the prior cohort, both studies showed that women with obesity had the lowest dietary diversity, whereas women with normal BMI were most likely to meet the MDD-W threshold. This pattern is consistent with findings from other pregnancy cohorts in the United States and Canada, where higher pBMI and socioeconomic factors, including lower education, lower income, parity, and single status, have been associated with lower diet quality [52,53]. Together, these findings suggest that the association between higher pBMI and lower dietary diversity observed in this cohort reflects broader patterns linking maternal weight status, socioeconomic factors, and diet quality during pregnancy. Building on our group’s prior work [35] that examined pBMI and pregnancy gut microbiota diversity within the MARCH cohort, the current analysis included a larger sample size and incorporated dietary diversity into the multivariable framework. The pBMI distribution in the current cohort was similar to that reported in our prior work, and both studies consistently identified maternal age as associated with Shannon and Inverse Simpson indices, supporting its relationship with α diversity during pregnancy. In contrast, our prior work [35] found trends in associations between pBMI and α diversity when diet was not considered, whereas the current study found no association between MDD-W and α-diversity indices after adjustment for pBMI. However, the directionality of associations between obese BMI and α-diversity metrics was the same as in our prior analysis. For β diversity, our prior work [35] reported significant overall Sorensen and Bray–Curtis models, whereas the overall models in the present study, inclusive of pBMI and MDD-W, were nonsignificant. These findings add to the broader pregnancy diet–microbiome literature suggesting that associations between maternal diet and gut microbiome diversity are variable, as well as indicating that much larger sample sizes will be needed to disentangle the effects of pBMI and diet. Prior studies have reported associations of overall diet quality and Mediterranean diet adherence with gut microbiome diversity during pregnancy [54,55], whereas other studies found limited differences in gut microbiome composition by healthy compared with less healthy dietary patterns among pregnant women with excess body fat [56]. Additional studies have reported associations between specific dietary components, body composition, and gut microbiome composition or function during pregnancy [57,58]. Including both pBMI and MDD-W in the same multivariable framework may have partitioned shared variance between related markers of prenatal nutritional status, thereby reducing the apparent independent association of each predictor with microbiome diversity. Together, these comparisons suggest that associations among diet, pBMI, and the pregnancy gut microbiome may vary by study population, dietary assessment method, microbiome outcome, and model structure. Age may be associated with measures of gut microbiota diversity for a variety of reasons. Age during pregnancy often covaries with diet quality, which is also relevant to the gut microbiota [59]. Older mothers tend to report better diet quality and higher dietary diversity, whereas younger mothers are less likely to achieve MDD-W [60]. Age also tracks with higher educational attainment and marriage stability, both of which are more prevalent among older mothers [59,61]. In our models, maternal age at birth was associated with lower α diversity, which was opposite to the hypothesized direction based on its association with better diet quality, higher education, and marital stability. Thus, age may be acting as a proxy for other unmeasured factors relevant to gut microbiota diversity rather than representing a direct effect of age itself. These findings should be interpreted cautiously because the sample size limited our ability to evaluate potential interaction effects, including interactions between maternal age and pBMI. Strengths of this study include a relatively large, well-characterized pregnancy cohort, standardized MDD-W scoring, and paired microbiome-diet analysis. Multivariable analysis accounted for sociodemographic and behavioral covariates for microbiome analyses. Limitations include the cross-sectional study design, which limits causal interpretation, and the use of 16S rRNA gene sequencing rather than whole-genome shotgun sequencing, which limits taxonomic resolution and constrains inference regarding microbial functional potential. Another limitation is the reliance on a single self-administered 24-h recall without portion size measurement, meaning minimal intakes of any single food category were possible. The lack of portion size information limits our ability to capture detailed nutritional intake, which could, in turn, explain the absence of observed associations with the microbiome. In addition, vegetarian and vegan dietary status was not collected, limiting our ability to determine whether specific dietary patterns may have influenced food group consumption. Pregnancy-related food avoidance was not assessed and may have influenced reported intake during the 24-h recall period. Medical conditions that may influence dietary intake were not included as covariates in the present analyses and may have affected reported food group consumption. Recruitment occurred at the first prenatal care visit, which could range widely across gestation. In addition, self-reported prepregnancy height and weight may introduce recall or reporting bias [62]. However, previous studies showed that self-reported prepregnancy height and weight demonstrate strong agreement with measured values and are considered valid for estimating pBMI, especially when early pregnancy measurements are unavailable [44,45]. As with any MDD-W-based analysis, actual micronutrient adequacy cannot be confirmed without measurement of nutrient status biomarkers. The primarily white study population (≥88% in all pBMI groups) may limit generalizability to more diverse populations. The results of this study reinforce the need for targeted nutrition guidance for women entering pregnancy with higher pBMI to encourage proper dietary diversity, with an emphasis on expanding intake of micronutrient-rich food groups such as pulses, leafy greens, and vitamin A–rich produce. Future research should track dietary diversity from preconception through postpartum and pair MDD-W scoring with quantitative analysis as well as assessments of nutritional status using biomarkers. In addition, studies should test interventions that focus on making foods from underconsumed food groups more accessible and palatable while monitoring participant adherence, ultimately aiming to improve dietary diversity. In conclusion, in this study, higher pBMI was associated with lower dietary diversity, whereas no association was observed between dietary diversity and the third-trimester pregnancy gut microbiome. Future studies should prioritize improving dietary diversity during pregnancy, particularly among women with higher pBMI, with an emphasis on increasing intake of less-consumed MDD-W food groups, including grains, pulses, nuts and seeds, eggs, dark green leafy vegetables, and vitamin A–rich fruits or vegetables. This study accounted for only recent intake and did not observe an association with gut microbiota characteristics. However, future studies accounting for overall dietary patterns may detect effects of diet diversity on gut microbiota characteristics. Declaration of generative AI and AI-assisted technologies in the writing process DOI,No AI or AI-assisted technologies were used in the writing process. Author contributions Conceptualization: SSC; Methodology: SYLX, SSC; Software: HJS, SYLX; Validation: HJS, SYLX; Formal Analysis: HJS, SYLX; Investigation: HJS, SYLX, JMK, SSC; Resources: JMK, SSC; Data Curation, HJS, SYLX; Writing - Original Draft: HJS, SYLX; Writing - review and editing: HJS, SYLX, JMK, SSC; Visualization: HJS, SYLX; Supervision: SSC; Funding acquisition: JMK, SSC. Funding This research was funded in part by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant number R01 DK135054 as well as the Office of the Director at the National Institutes of Health under award numbers UG3OD023285 and UH3OD023285, the Michigan Health Endowment Fund under award numbers G-1608-140432 and R-1605-140007, and from Michigan State University Center for Research in Autism, Intellectual, and Neurodevelopmental Disabilities (C-RAIND). This research was also supported in part by a CHARM (Child Health Advances from Research with Mothers) small grant funded by the Vice Presidents for Research of Michigan State University, University of Michigan, and Wayne State University. HJS has been supported by undergraduate research grants from the MSU Department of Food Science and Human Nutrition and the College of Agriculture and Natural Resources. SYLX has been supported by graduate fellowships from the MSU Department of Food Science and Human Nutrition. Conflict of interest SSC is an Editorial Board Member for the Journal of Nutrition. All other authors report no conflicts of interest. Acknowledgments This research could not have been carried out without the collaboration of Hurley Medical Center and Hurley Residency Clinic in Flint, MI, United States; Hutzel Medical Center; DMC Center for Obstetrics and Gynecology; and University Health Center in Detroit, MI, United States; Munson Hospital and Grand Traverse Women’s Clinic in Traverse City, MI, United States; University of Michigan Hospital; Von Voigtlander Women’s Center; West Ann Arbor Health Center; and Obstetrics and Gynecology at Briarwood Center in Ann Arbor, MI, United States; St. John’s Providence Park; Metro Partners in Women’s Health; and Women’s Health Consultants in Novi, MI, United States; St. Joseph Mercy Hospital; IHA Canton Obstetrics and Gynecology; and IHA Domino Farms in Ann Arbor, MI, United States; Sinai Grace Hospital, Sinai Grace OB-GYN Women’s Health Centers; Northwest Women’s Care; and DMC Northwest Women’s Care in Detroit, MI, United States; Beaumont Dearborn Hospital and Oakwood OB-GYN Associates in Dearborn, MI, United States; Covenant Hospital, Central Michigan University Health in Saginaw, MI, United States; SHMG OB/GYN; Grand Rapids Women’s Health, and Spectrum Health in Grand Rapids, MI, United States; Blue Water OB/GYN; NorthPointe OB/GYN; McLaren Port Huron Hospital in Port Huron, MI, United States; the MSU Faculty Clinic; the MSU Residency Clinic; and Sparrow Hospital, in Lansing, MI, United States. Declaration of Generative AI and AI-Assisted Technologies in the Writing Proceess No AI or AI-assisted technologies were used in the writing process. References