Certain foods characterizing the alternate Mediterranean diet (aMED) are high in persistent organic pollutants (POPs), which are related to greater gestational diabetes mellitus (GDM) risk. We examined the associations of combined aMED and POP exposure with GDM.
aMED score of 1,572 pregnant women was derived from food frequency questionnaires at early pregnancy within the U.S. Fetal Growth Study and plasma concentrations of 76 POPs, including organochlorine pesticides, polybrominated diphenyl ethers, polychlorinated biphenyls (PCBs), and per- and polyfluoroalkyl substances, were measured. Associations of combined aMED score and exposure to POPs with GDM risk were examined by multivariable logistic regression models.
In 61 of 1,572 (3.88%) women with GDM, 25 of 53 included POPs had a detection rate >50%. Higher POP levels appeared to diminish potential beneficial associations of aMED score with GDM risk, with the lowest GDM risk observed among women with both high aMED score and low POP concentrations. Specifically, adjusted log-odds ratios of GDM risk comparing women with low PCB and high aMED score with those with low aMED score and high PCB concentrations was −0.74 (95% CI −1.41, −0.07). Inverse associations were also observed among women with low aMED score and high TransNo_chlor, PCB182_187, PCB196_203, PCB199, and PCB206. These associations were more pronounced among women with overweight or obesity.
Pregnant women who consumed a healthy Mediterranean diet but had a low exposure to POP concentrations had the lowest GDM risk. Future endeavors to promote a healthy diet to prevent GDM may consider concurrent POP exposure.
Introduction
Gestational diabetes mellitus (GDM) is of increasing public health and clinical concern. It affects ∼5–9% of pregnancies and has been on the rise during past decades. GDM is at the center of a diabetes-begetting-diabetes vicious cycle. Specifically, women with a history of GDM are at a substantially higher risk of developing type 2 diabetes and cardiovascular disease (CVD) later in life, and 30–70% of women with GDM will develop type 2 diabetes within 15 years following their pregnancy (1). Children born from pregnancies complicated by GDM are at elevated risk for adiposity, high blood pressure, and abnormal blood lipid levels (2). The etiology of GDM is multifaceted, involving a complex interplay of genetic, environmental, and lifestyle factors. Most nongenetic factors are modifiable. Understanding how these factors interact is crucial for preventing GDM and disrupting the diabetes-begetting-diabetes vicious cycle. Current dietary guidelines encourage the adoption of healthy dietary patterns, such as the alternate Mediterranean diet (aMED), alternate Healthy Eating Index, and Dietary Approaches to Stop Hypertension, rather than focusing on specific or individual foods or nutrients because of the inherent complexity of diets people typically consume (3). The aMED features fruits, vegetables, whole grains, nuts, and legumes, with moderate amounts of fish and poultry, while using olive oil as the main fat and limiting red meat and dairy. Greater adherence to aMED has been linked to a reduced risk of developing gestational diabetes mellitus (GDM), type 2 diabetes, and CVDs (4). However, some of its components (i.e., seafood) raise concerns about food safety as they may carry environmental chemicals (5).
Circulating concentrations of persistent organic pollutants (POPs), such as polybrominated diphenyl ethers (PBDEs) and organophosphate esters (OPEs), exhibit lipophilic properties and have the potential to accumulate in human tissues, such as the liver and adipose tissue (6). Elevated levels of these POPs are related to higher risks of cardiometabolic disorders, such as GDM and type 2 diabetes, through promoting inflammation, oxidative stress, and endocrine-disrupting effects (7).
Paradoxically, several major food components of aMED (e.g., leaf and root vegetables, deep-sea fish high in n-3 and n-6 fats, top predator fish) are major food sources for certain POPs, including organochlorine pesticides (OCPs), PBDEs, and per- and polyfluoroalkyl substances (PFASs) (8,9). Both diet and exposure to certain POPs have the capacity to elicit chronic subclinical inflammatory responses, thereby potentially increasing the risk of long-term metabolic diseases. For instance, the consumption of red or processed meat has been linked to elevated levels of C-reactive protein and glycated hemoglobin, whereas fruit intake has been shown to have an inverse association with C-reactive protein levels (10). Prenatal PFAS exposure has been found to be positively related to proinflammatory cytokines (11). Besides, POPs may have similar obesogenic effects as some food consumption (i.e., carbohydrate intake) (12). As such, the role of POPs should be considered when investigating the association of dietary patterns (i.e., aMED) with the risk of GDM, as this may inform the development of a more effective intervention strategy through dietary modifications.
However, research endeavors in both nutritional and environmental epidemiology have predominantly investigated the health effects of dietary patterns and POPs separately rather than collectively, which may partly explain some of the inconclusive or inconsistent findings (13). To our knowledge, studies on the impacts of joint exposure to dietary patterns and POPs on maternal health during pregnancy have long been overlooked, and such data are lacking. In the current study, we examined associations of aMED score and POPs jointly with GDM risk. Furthermore, we evaluated whether the associations differ by several other major risk factors of GDM, such as obesity status, in a large, racially/ethnically diverse pregnancy cohort.
Research Design and Methods
Study Population
The study population comprised 2,802 U.S. pregnant women from the Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Studies–Singletons Cohort (2009–2013). Details have been published elsewhere (14). Briefly, this pregnancy cohort encompassed women recruited from 12 U.S. clinical centers. Participants with preexisting chronic diseases, a history of pregnancy complications (e.g., hypertension, diabetes), or recognized lifestyle risk factors, including active smoking, alcohol consumption, and illicit drug use, were excluded. Written informed consent was obtained from all the enrolled individuals. The study protocol received approval from the institutional review boards of the National Institutes of Health and all participating clinical sites.
Of the 2,802 participants, measurements of POPs were available for 2,696. During the recruitment phase (gestational age 8–13 weeks), a subgroup of 1,699 participants completed a semiquantitative food frequency questionnaire (FFQ) that aimed to capture dietary habits over the preceding 3 months. The FFQ used in the study was a modified iteration of the well-validated Diet History Questionnaire-II, which has demonstrated efficacy in assessing dietary habits among nonpregnant individuals (15). Participants self-reported information on the number of times consumed per day, week, or month, along with the typical portion size of each food item. Subsequently, nutrient estimations were derived from the collected data using Diet∗Calc software (National Cancer Institute) (16). A total of 1,572 participants who had both measurements of POPs and completed FFQ data were included for the final analysis. The aMED score was derived from 11 food groups (17). Each food group was assigned a score of 1 if its consumed amount surpassed the corresponding median and 0 if it fell below. Alcohol consumption was excluded from the aMED score due to its detrimental effects on fetal development (18).
Pregnancy outcomes were identified using information extracted from medical records. GDM was defined based on a 100-g, 3-h oral glucose tolerance test at a mean (SD) gestational age of 27 (4) weeks. Carpenter-Coustan criteria were followed, which require at least two specified values from the oral glucose tolerance test procedure (fasting, 95 mg/dL; 1 h, 180 mg/dL; 2 h, 155 mg/dL; 3 h, 140 mg/dL) (19) and/or by receipt of GDM medications.
Laboratory Analysis
Chemicals were measured using maternal plasma collected at recruitment (gestational age 8–13 weeks) by the Wadsworth Center, New York State Department of Health, according to standardized procedures (20). A total of 76 POPs, including 11 OCPs, 10 PBDEs, 44 polychlorinated biphenyls (PCBs), and 11 PFASs, were measured. Details can be found in Supplementary Table 1. Chemicals with plasma concentrations below the limit of detection (LOD) were replaced with LOD / sqrt(2). The total concentrations of OCPs, PBDEs, PCBs, and PFASs were determined by summing the measured chemical concentrations within their respective chemical classes.
To account for the potential influence of lipids on the plasma concentrations of certain POPs, namely OCPs, PBDEs, and PCBs, lipid concentrations were measured using commercially available enzymatic methods (21). Total lipids were then calculated using the following equation: plasma total lipids = plasma total cholesterol × 2.27 + plasma triglycerides + 62.3 (22). The concentrations of POPs, adjusted for total lipids (except for PFASs), were reported and included in the analysis. PFASs were not adjusted for lipids due to their nonhydrophilic and nonlipophilic properties.
Covariates
Potential confounders in this study were selected a priori based on previous studies (18,23). Briefly, information on maternal characteristics and socioeconomic status was obtained at enrollment through standardized questionnaires, including maternal prepregnancy BMI (calculated by weight in kg divided by height in m2), race and ethnicity, education level, maternal income, and physical activity level. Prepregnancy BMI was highly correlated with BMI calculated by self-reported weight and height (correlation coefficient 0.99). Subscapular skinfold was measured by experienced professionals. Parity and disease history were extracted from medical records. A family history of type 2 diabetes was indicated if the participant self-reported any first-degree relative diagnosed with type 2 diabetes. Tobacco exposure was defined as either yes if plasma cotinine concentration exceeded the limit of detection or no if it did not (24). Race and ethnicity was self-reported by the participants and categorized as Hispanic, non-Hispanic White, non-Hispanic Black, and Asian or Pacific Islander. Maternal education level was classified into three groups: high school or less, some college or undergraduate, and graduate and postgraduate. Parity was categorized as primiparous or multiparous. Maternal income in the previous year was categorized into three predefined levels (U.S. <$30,000, $30,000–$74,999, and >$75,000). Alcohol consumption was excluded from the analysis due to its low prevalence among participants (n = 21 [0.75%]). The level of physical activity was assessed using the validated Pregnancy Physical Activity Questionnaire and reported as weekly MET-h/week (25). Total energy intake (in kcal/day) was derived from the FFQ administered at recruitment (26).
Statistical Analysis
The distribution of participant characteristics were reported according to GDM status and analyzed using the χ2 test for categorical characteristics or ANOVA for continuous variables. All POP concentrations were subjected to natural log transformation to approximate a normal distribution and were standardized. Mann-Whitney U test was used to compare the difference in POP concentrations between participants with and without GDM. Analyses were performed on POPs with a detection rate >10% because chemicals may have toxic effects even at very low concentrations, particularly for pregnant women and developing fetuses, leaving 10 OCPs, 7 PBDEs, 27 PCBs, and 9 PFASs for the final analysis (Supplementary Table 1).
Binary logistic regression models were used to evaluate the prospective associations of combined aMED score and individual POP or POP class exposure with GDM. Individuals were categorized into four groups based on both aMED score (high vs. low, dichotomized by median) and plasma POP concentrations (high vs. low, dichotomized by median), including the low aMED score and high POP group (reference group), low aMED score and low POP group, high aMED score and high POP group, and high aMED score and low POP group. For POPs with detection rates <50%, we used the LOD/sqrt(2) as the cutoff to categorize the POP group. Logistic regression models were conducted for both lipid-corrected POPs (excluding PFASs) and POPs without correction, as some POPs themselves may disrupt lipid metabolism and additional adjustment may block the causal pathway of POP-induced GDM. Meanwhile, to further characterize the interactive effects of POPs and aMED exposure on GDM, we calculated the relative excess risk due to interaction, attributable proportion, and odds ratio (OR) of the product terms to evaluate their additive and multiplicative effects, respectively. All models were adjusted for maternal age, race and ethnicity, prepregnancy BMI, education level, maternal income, parity, physical activity, total energy intake, tobacco exposure, and family history of diabetes. ORs and 95% CIs were reported on a log scale so that the 95% CIs can be evenly distributed around the point estimate. Tolerance and variance inflation factor values were computed to assess multicollinearity among the covariates, with tolerance values <0.1 and variance inflation factor values >10 indicating the presence of multicollinearity.
Given the high correlation and structural and biological homology among certain POPs, principal component analysis (PCA) was used as an additional approach to reduce dimensionality and collinearity and prevent potentially misleading interpretations of the individual POP effects (Supplementary Fig. 1). Principal component scores were generated and categorized by their medians, used to define low and high POP exposure, and combined with the groups of dietary pattern scores to assess the associations of their combined exposure with GDM, as presented above. No rotations were used in PCA. Interactions of other major risk factors of GDM (i.e., prepregnancy BMI) with the combined exposure of POP and aMED score were examined and stratified analyses performed. Meanwhile, the subscapular skinfold measurement was used to evaluate the accuracy of prepregnancy BMI in reflecting overall body fat and obesity status. Gestational weight gain in the first trimester was additionally adjusted to control its potential confounding or mediated role in the associations of combined exposure with GDM risk. All statistical analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC) and R 4.2.1 software.
Results
Baseline Characteristics of Study Population
Compared with participants in the lowest quartile of aMED score (least healthy), those in the highest quartile (most healthy) were more likely to be older, to be Asian or Pacific Islander, and to have a normal prepregnancy BMI, higher education level, higher income, lower tobacco exposure, and higher plasma OCP and PCB concentrations (Table 1). Plasma concentrations of oxychlordane, TransNo_chlor, PCB146_161, PCB153, PCB156, PCB180, PCB170, PCB199, PCB196_203, PCB194, and PCB206 were higher among participants with GDM, whereas P_P_DDT was higher among the participants without GDM (Supplementary Table 2).
Characteristics of study participants at enrollment (gestational weeks 8–13) according to aMED score, Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Study–Singletons Cohort
Characteristic . | Total population (n = 1,572) . | Quartile 1 (n = 620) . | Quartile 2 (n = 317) . | Quartile 3 (n = 294) . | Quartile 4 (n = 341) . |
---|---|---|---|---|---|
Age, years* | 28 (5.6) | 27 (5.5) | 28 (5.7) | 29 (5.3) | 30 (5.5) |
Race and ethnicity* | |||||
Hispanic | 481 (30.6) | 198 (31.9) | 109 (34.4) | 88 (29.9) | 86 (25.2) |
Non-Hispanic White | 324 (20.6) | 115 (18.6) | 63 (19.9) | 58 (19.7) | 88 (25.8) |
Non-Hispanic Black | 471 (30.0) | 227 (36.6) | 88 (27.8) | 76 (25.9) | 80 (23.5) |
Asian or Pacific Islander | 296 (18.8) | 80 (12.9) | 57 (18.0) | 72 (24.5) | 87 (25.5) |
BMI (kg/m2)* | |||||
Normal weight | 915 (58.2) | 323 (52.1) | 183 (57.7) | 182 (61.9) | 227 (66.6) |
Overweight | 429 (27.3) | 176 (28.4) | 89 (28.1) | 76 (25.9) | 88 (25.8) |
Obesity | 228 (14.5) | 121 (19.5) | 45 (14.2) | 36 (12.2) | 26 (7.6) |
Highest level of education* | |||||
High school or below | 495 (31.5) | 242 (39.0) | 94 (29.7) | 87 (29.6) | 72 (21.1) |
Some college or undergraduate | 828 (52.7) | 318 (51.3) | 173 (54.6) | 153 (52.0) | 184 (54.0) |
Graduate or postgraduate | 249 (15.8) | 60 (9.7) | 50 (15.8) | 54 (18.4) | 85 (24.9) |
Income during past year (U.S. $)* | |||||
<30,000 | 455 (28.9) | 216 (34.84) | 94 (29.65) | 69 (23.47) | 76 (22.29) |
30,000–74,999 | 387 (24.6) | 168 (27.1) | 76 (23.97) | 74 (25.17) | 69 (20.23) |
≥75,000 | 488 (31.0) | 147 (23.71) | 97 (30.6) | 99 (33.67) | 145 (42.52) |
Unknown | 242 (15.4) | 89 (14.35) | 50 (15.77) | 52 (17.69) | 51 (14.96) |
Parity | |||||
Primiparous | 718 (45.7) | 268 (43.2) | 137 (43.2) | 147 (50.0) | 166 (48.7) |
Multiparous | 854 (54.3) | 352 (56.8) | 180 (56.8) | 147 (50.0) | 175 (51.3) |
Total physical activity (MET-h/week) | 322 (166) | 311 (161) | 323 (175) | 321 (156) | 342 (173) |
Tobacco exposure* | |||||
Not exposed | 1,109 (70.6) | 393 (63.4) | 232 (73.2) | 211 (71.8) | 273 (80.1) |
Exposed | 463 (29.5) | 227 (36.6) | 85 (26.8) | 83 (28.2) | 68 (19.9) |
Family history of diabetes | |||||
Yes | 1,202 (76.5) | 456 (73.6) | 242 (76.3) | 236 (80.3) | 268 (78.6) |
No | 330 (21.0) | 144 (23.2) | 68 (21.5) | 53 (18.0) | 65 (19.1) |
Unknown | 40 (2.5) | 20 (3.2) | 7 (2.2) | 5 (1.7) | 8 (2.4) |
Total OCPs (ng/g lipid)* | 0.58 (0.72) | 0.49 (0.64) | 0.57 (0.7) | 0.62 (0.83) | 0.66 (0.82) |
Total PBDEs (ng/g lipid)* | 0.09 (0.13) | 0.1 (0.14) | 0.08 (0.11) | 0.08 (0.11) | 0.09 (0.14) |
Total PCBs (ng/g lipid)* | 0.12 (0.13) | 0.09 (0.11) | 0.12 (0.13) | 0.13 (0.17) | 0.15 (0.14) |
Total PFAS (ng/mL) | 9.33 (6.70) | 8.89 (6.68) | 9.3 (6.28) | 9.44 (7.11) | 9.72 (7.49) |
Characteristic . | Total population (n = 1,572) . | Quartile 1 (n = 620) . | Quartile 2 (n = 317) . | Quartile 3 (n = 294) . | Quartile 4 (n = 341) . |
---|---|---|---|---|---|
Age, years* | 28 (5.6) | 27 (5.5) | 28 (5.7) | 29 (5.3) | 30 (5.5) |
Race and ethnicity* | |||||
Hispanic | 481 (30.6) | 198 (31.9) | 109 (34.4) | 88 (29.9) | 86 (25.2) |
Non-Hispanic White | 324 (20.6) | 115 (18.6) | 63 (19.9) | 58 (19.7) | 88 (25.8) |
Non-Hispanic Black | 471 (30.0) | 227 (36.6) | 88 (27.8) | 76 (25.9) | 80 (23.5) |
Asian or Pacific Islander | 296 (18.8) | 80 (12.9) | 57 (18.0) | 72 (24.5) | 87 (25.5) |
BMI (kg/m2)* | |||||
Normal weight | 915 (58.2) | 323 (52.1) | 183 (57.7) | 182 (61.9) | 227 (66.6) |
Overweight | 429 (27.3) | 176 (28.4) | 89 (28.1) | 76 (25.9) | 88 (25.8) |
Obesity | 228 (14.5) | 121 (19.5) | 45 (14.2) | 36 (12.2) | 26 (7.6) |
Highest level of education* | |||||
High school or below | 495 (31.5) | 242 (39.0) | 94 (29.7) | 87 (29.6) | 72 (21.1) |
Some college or undergraduate | 828 (52.7) | 318 (51.3) | 173 (54.6) | 153 (52.0) | 184 (54.0) |
Graduate or postgraduate | 249 (15.8) | 60 (9.7) | 50 (15.8) | 54 (18.4) | 85 (24.9) |
Income during past year (U.S. $)* | |||||
<30,000 | 455 (28.9) | 216 (34.84) | 94 (29.65) | 69 (23.47) | 76 (22.29) |
30,000–74,999 | 387 (24.6) | 168 (27.1) | 76 (23.97) | 74 (25.17) | 69 (20.23) |
≥75,000 | 488 (31.0) | 147 (23.71) | 97 (30.6) | 99 (33.67) | 145 (42.52) |
Unknown | 242 (15.4) | 89 (14.35) | 50 (15.77) | 52 (17.69) | 51 (14.96) |
Parity | |||||
Primiparous | 718 (45.7) | 268 (43.2) | 137 (43.2) | 147 (50.0) | 166 (48.7) |
Multiparous | 854 (54.3) | 352 (56.8) | 180 (56.8) | 147 (50.0) | 175 (51.3) |
Total physical activity (MET-h/week) | 322 (166) | 311 (161) | 323 (175) | 321 (156) | 342 (173) |
Tobacco exposure* | |||||
Not exposed | 1,109 (70.6) | 393 (63.4) | 232 (73.2) | 211 (71.8) | 273 (80.1) |
Exposed | 463 (29.5) | 227 (36.6) | 85 (26.8) | 83 (28.2) | 68 (19.9) |
Family history of diabetes | |||||
Yes | 1,202 (76.5) | 456 (73.6) | 242 (76.3) | 236 (80.3) | 268 (78.6) |
No | 330 (21.0) | 144 (23.2) | 68 (21.5) | 53 (18.0) | 65 (19.1) |
Unknown | 40 (2.5) | 20 (3.2) | 7 (2.2) | 5 (1.7) | 8 (2.4) |
Total OCPs (ng/g lipid)* | 0.58 (0.72) | 0.49 (0.64) | 0.57 (0.7) | 0.62 (0.83) | 0.66 (0.82) |
Total PBDEs (ng/g lipid)* | 0.09 (0.13) | 0.1 (0.14) | 0.08 (0.11) | 0.08 (0.11) | 0.09 (0.14) |
Total PCBs (ng/g lipid)* | 0.12 (0.13) | 0.09 (0.11) | 0.12 (0.13) | 0.13 (0.17) | 0.15 (0.14) |
Total PFAS (ng/mL) | 9.33 (6.70) | 8.89 (6.68) | 9.3 (6.28) | 9.44 (7.11) | 9.72 (7.49) |
Data are mean (SD) for continuous variables, n (%) for categorical variables, and median (IQR) for chemical concentrations.
*Statistically different across quartiles, with P < 0.05.
Associations of Combined Summed POP Concentrations and aMED Scores With GDM Risk
Higher summed POP levels appeared to dilute the potential beneficial associations of aMED with GDM risk, with GDM risk being the lowest among participants with both high aMED score and low summed POP concentrations (Fig. 1). In particular, the joint exposure of high aMED score and low summed PBDEs or PCBs was significantly associated with lower risks of GDM (logOR −0.62 [95% CI −1.09, −0.15] and −0.76 [−1.44, −0.09], respectively) compared with the reference group (low aMED score and high summed PBDEs or PCBs). The joint associations of aMED score and summed PCBs remained significant even after adjusting for total lipids (logOR −0.74 [95% CI −1.41, −0.07]). Results of the multiplicative interaction test showed a consistent inverse association with GDM risk (logOR −0.75 [95% CI −1.47, −0.03]) (Supplementary Table 2). No significant additive interaction was observed.
Associations of environmental chemicals (in classes) and aMED with GDM risk, Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Study–Singletons Cohort. All models were adjusted for maternal age, race and ethnicity, prepregnancy BMI, education level, maternal income, parity, physical activity, total energy intake, tobacco exposure, and family history of diabetes. Chemicals in model 1 were not corrected for total lipids but were corrected in model 2 (except for PFASs). All chemicals were natural log transformed and standardized to benefit the interpretation. H, high; L, low; Ref, reference.
Associations of environmental chemicals (in classes) and aMED with GDM risk, Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Study–Singletons Cohort. All models were adjusted for maternal age, race and ethnicity, prepregnancy BMI, education level, maternal income, parity, physical activity, total energy intake, tobacco exposure, and family history of diabetes. Chemicals in model 1 were not corrected for total lipids but were corrected in model 2 (except for PFASs). All chemicals were natural log transformed and standardized to benefit the interpretation. H, high; L, low; Ref, reference.
Associations of Combined aMED Score and Individual POP Exposure With GDM Risk
Similarly, when examining aMED score and individual POPs jointly, participants with high aMED score and low individual POP concentration appeared to have a lower risk of GDM compared with those with a low aMED score and high POP concentrations (Fig. 2). Significant associations of combined aMED score and individual POP exposures with GDM risk were mainly observed for individual PCBs. High aMED score and low concentrations of PCB182_187, PCB196_203, PCB199, PCB156, PCB180, and TransNo_chlor were significantly associated with lower risks of GDM (logOR −1.77 [95% CI −3.31, −0.23], −1.67 [−2.96, −0.37], −1.50 [−2.69, −0.32], −1.05 [−2.07, −0.04], −1.36 [−2.68, −0.04], −1.34 [−2.50,−0.18], respectively). Associations of combined aMED score and individual lipophilic POP exposures with lower risk of GDM remained significant when POPs were not corrected for total lipids, particularly for associations of combined aMED score and hexachlorobenzene, TransNo_chlor, BDE100, BDE47, PCB118_106, PCB153, PCB156, PCB170, PCB180, PCB182_187, PCB196_203, PCB199, PCB206, and PCB99 with GDM risk (Supplementary Fig. 2). Meanwhile, BDE100, PCB182_187, and aMED showed significant multiplicative interactions with GDM (logOR −0.52 [95% CI −1.04, −0.001] and −0.78 [−1.51, −0.06], respectively) (Supplementary Table 3).
Associations of individual chemicals of POPs and aMED with GDM risk, Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Study–Singletons cohort. The most unhealthy combination (low aMED score and high POP exposure) was used as the reference. All models were adjusted for maternal age, race and ethnicity, BMI, education level, maternal income, parity, physical activity, total energy intake, tobacco exposure, and family history of diabetes. All chemicals (except for PFASs) were corrected for total lipids, natural log transformed, and standardized. Orange and blue colors of error bars are used to differentiate POPs. BDE, brominated diphenyl ethers; DDD, dichlorodiphenyldichloroethane; DDE, dichlorodiphenyldichloroethylene; DDT, dichlorodiphenyltrichloroethane; H, high; HCB, hexachlorobenzene; HCH, hexachlorocyclohexane; L, low; NMeFOSAA, 2(N-methylperfluorooctanesulfonamido)acetic acid; PFDA, perfluorodecanoic acid; PFDoDA, perfluorododecanoic acid; PFHpA, perfluoroheptanoic acid; PFHxS, perfluorohexanesulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctanesulfonic acid; PFUnDA, perfluoroundecanoic acid; TransNo-chlor, trans-nonachlor.
Associations of individual chemicals of POPs and aMED with GDM risk, Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Study–Singletons cohort. The most unhealthy combination (low aMED score and high POP exposure) was used as the reference. All models were adjusted for maternal age, race and ethnicity, BMI, education level, maternal income, parity, physical activity, total energy intake, tobacco exposure, and family history of diabetes. All chemicals (except for PFASs) were corrected for total lipids, natural log transformed, and standardized. Orange and blue colors of error bars are used to differentiate POPs. BDE, brominated diphenyl ethers; DDD, dichlorodiphenyldichloroethane; DDE, dichlorodiphenyldichloroethylene; DDT, dichlorodiphenyltrichloroethane; H, high; HCB, hexachlorobenzene; HCH, hexachlorocyclohexane; L, low; NMeFOSAA, 2(N-methylperfluorooctanesulfonamido)acetic acid; PFDA, perfluorodecanoic acid; PFDoDA, perfluorododecanoic acid; PFHpA, perfluoroheptanoic acid; PFHxS, perfluorohexanesulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctanesulfonic acid; PFUnDA, perfluoroundecanoic acid; TransNo-chlor, trans-nonachlor.
Population Heterogeneity
Significant interactive effects were observed between the combined exposure of PFAS and aMED and prepregnancy BMI on GDM risk, with P-interaction = 0.046 for total PFAS and 0.005 for PFNA. In stratified analyses, the joint associations of high aMED score and low PFAS concentrations with a lower risk of GDM appeared to be more pronounced among participants with prepregnancy BMI ≥25 kg/m2 compared with those with a BMI <25 kg/m2, particularly for PFNA, perfluorooctanoic acid, and total PFASs (logOR −1.80 [95% CI −3.17, −0.43], −1.98 [−3.59, −0.36], −2.02 [−3.65, −0.39], respectively) (Fig. 3).
Associations of POPs and aMED jointly with GDM risk by prepregnancy BMI, Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Study–Singletons Cohort. The most unhealthy combination (low aMED score and high POP exposure) was used as the reference, but only results of the most healthy group (high aMED score and low POP exposure) are presented. All models were adjusted for maternal age, education level, income, parity, physical activity, total energy intake, smoking status, family history of diabetes, and race and ethnicity. All chemicals were natural log transformed and standardized. Orange and blue colors of error bars are used to differentiate POPs. NMeFOSAA, 2(N-methylperfluorooctanesulfonamido)acetic acid; PFDA, perfluorodecanoic acid; PFDoDA, perfluorododecanoic acid; PFHpA, perfluoroheptanoic acid; PFHxS, perfluorohexanesulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctanesulfonic acid; PFUnDA, perfluoroundecanoic acid.
Associations of POPs and aMED jointly with GDM risk by prepregnancy BMI, Eunice Kennedy Shriver National Institute of Child Health and Human Development Fetal Growth Study–Singletons Cohort. The most unhealthy combination (low aMED score and high POP exposure) was used as the reference, but only results of the most healthy group (high aMED score and low POP exposure) are presented. All models were adjusted for maternal age, education level, income, parity, physical activity, total energy intake, smoking status, family history of diabetes, and race and ethnicity. All chemicals were natural log transformed and standardized. Orange and blue colors of error bars are used to differentiate POPs. NMeFOSAA, 2(N-methylperfluorooctanesulfonamido)acetic acid; PFDA, perfluorodecanoic acid; PFDoDA, perfluorododecanoic acid; PFHpA, perfluoroheptanoic acid; PFHxS, perfluorohexanesulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctanesulfonic acid; PFUnDA, perfluoroundecanoic acid.
Results of the PCA showed similar trends to those from the regression models. Lower GDM risks were observed among participants with a high aMED score and low principal component scores of POPs, corresponding to a low aMED score and high POP exposure (Supplementary Figs. 3 and 4 and Supplementary Table 4). In general, the results of models using skinfold measurement as an obesity indicator and models with additional adjustment for gestational weight gain rate were consistent with the main results (Supplementary Figs. 5 and 6).
Conclusions
In this multiracial U.S. pregnancy cohort, we conducted a comprehensive evaluation that represents, to our knowledge, one of the first studies examining the interplay of aMED and multiple POPs in association with the risk of GDM. We observed that higher POPs appeared to diminish the potentially beneficial role of aMED in GDM, with the lowest risk of GDM being observed among participants with both greater adherence to the Mediterranean diet and low POP concentrations, especially PCBs. Several individual POPs with significant associations with combined aMED exposure on GDM were identified, and these associations of combined exposure were more pronounced in individuals with a prepregnancy BMI indicating overweight or obesity.
Emerging evidence has shown the beneficial effects of adherence to healthy dietary patterns, such as aMED, on diabetes, CVD, and cancer (10). Paradoxically, major food components of aMED, such as fish, may explain the observed joint association as it has been found to be a potential source of some POPs, such as PCBs, PFAS, and PBDEs (27,28), which even at low environmental doses, have been identified as an emerging risk factor for type 2 diabetes (11). Notably, PCBs were previously shown to be significantly related to fish consumption in the general U.S. population (29). Meanwhile, both PCBs and constituted food groups of aMED, such as fish, can influence the development of GDM (30,31). Other POPs, such as pesticides (e.g., dichlorodiphenyldichloroethylene), can still be detected in vegetables and fruits and were found to potentially offset the beneficial effects of vegetables and fruits on mortality, though these pesticides have been abolished for >40 years (32). Our findings also suggest that the health benefits of a dietary pattern can be strengthened if we can reduce POPs in food sources. For example, in this study, the beneficial effects of aMED on GDM appeared to be masked, in part, by the hazardous effects of some POPs. Even though sample sizes in each joint group and concentrations of POPs were relatively small, overall, we still observed significant joint associations of high aMED score and low POP concentrations, which was associated with a meaningfully reduced GDM risk. Our findings suggest that the health benefit of a high adherence to aMED can be strengthened by reducing the concentrations of potential POP exposures in diet, or the hazardous effects of some POP exposure can be attenuated or compensated by some favorable components/nutrients in healthy dietary patterns. These findings indicate a need for strong implementation of public health practice, especially for countries with relatively high POP exposure.
Attenuation of the health benefit of aMED on GDM risk by POP exposure, as observed in this study, can be explained, at least partly, by the interactive mechanisms underlying the oxidative stress induced by POP exposure and the antioxidant and anti-inflammatory compounds provided by a healthy dietary pattern (33). For example, consuming a Mediterranean diet rich in minimally processed plant foods, fish, and monounsaturated fats has been associated with a reduced risk of metabolic diseases through several important adaptations, including lipid-lowering effects and protection against oxidative stress, inflammation, platelet aggregation, and gut microbiota–mediated metabolites (34). These pathways through which healthy dietary patterns exert beneficial effects largely overlap with those through which POPs induce toxic metabolic effects, thereby providing an explainable basis for their potential interactions (35).
This study revealed that associations of dietary patterns and POPs (i.e., PFAS) jointly with GDM risk was particularly significant in participants who had overweight or obesity , potentially due to the interaction of obesogenic effects and low-grade inflammation (36). POPs have a complicated interrelationship with lipid metabolism as they can induce chronic inflammation of adipose tissue and release proinflammatory cytokines; in turn, obesity increases the release of free fatty acids into the circulation and promotes fat deposits in ectopic sites, which also can affect POP concentrations in peripheral circulation and adipose tissue (37). Thus, the more pronounced inverse association observed among participants with obesity are biologically plausible. Our findings indicated that higher adherence to a healthy dietary pattern and low POP concentrations during pregnancy may have the most pronounced benefits among women with high prepregnancy BMI.
Strengths of the study included the prospective cohort design, racially and ethnically diverse population, and the comprehensive POP profile. aMED and >70 POPs of different classes were covered and examined, which made this study unique and comprehensive as the interplay of combined exposure to dietary patterns and POPs with GDM had not yet been investigated. Pregnant women were investigated due to their high vulnerability to unhealthy dietary patterns and toxic POPs (38). However, several potential limitations merit consideration. First, even though detailed covariates related to the exposures of interest were adjusted and 76 of the most abundant POPs were examined, unmeasured confounders cannot be ruled out. Second, the plasma concentrations of the various POPs in our study population were relatively low compared with the general U.S. population at the same period and with concentrations observed in Asian countries. This may be due to the prohibition of POP production and to the degradation of these POPs over time (39,40). Thus, the effect estimation may be underestimated due to the relatively lower POP concentrations. Third, the generalizability of our findings needs to be further explored as the Mediterranean dietary pattern considered in this study may not fully capture the dietary habits of populations from different regions or countries.
In summary, in this prospective cohort study encompassing a diverse, multiracial population in the U.S, we found that overall, the lowest GDM risk was observed among individuals who consumed a healthy diet (i.e., high adherence to aMED) and were exposed to low concentrations of POPs, especially PCBs. These combined associations were more pronounced in pregnant women who had overweight or obesity before pregnancy. Findings from the current study convey important messages to pregnant women and their clinicians, which is to encourage the adoption of healthy dietary patterns while mitigating potential exposures to foodborne POPs that can further maximize the health benefits of a Mediterranean diet. Collective efforts from both government and society, along with coordinated global regulatory actions, are crucial for reducing exposure to hazardous chemicals. For pregnant women, it is crucial to minimize exposure to chemicals from food and other sources. This may be achieved by avoiding clothing with flame retardants, thoroughly rinsing and peeling fruits and vegetables, and avoiding the consumption of fatty tissue, organs, and skin of fish. More preventive tips for different chemicals can be found through the Program on Reproductive Health and the Environment (41). Besides, confirmation of these findings through other studies encompassing diverse populations with larger sample sizes and different habitual diets would strengthen the robustness of our results.
Clinical trial reg. no. NCT00912132, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.27091498.
Article Information
Acknowledgments. C.Z. is an editor of Diabetes Care but was not involved in any of the decisions regarding review of the manuscript or its acceptance.
Funding. This research was supported by American Recovery and Reinvestment Act funding and Eunice Kennedy Shriver National Institute of Child Health and Human Development intramural funding (grants HHSN275200800013C, HHSN275200800002I, HHSN27500006, HHSN275200800003IC, HHSN275200800014C, HHSN275200800012C, HHSN275200800028C, HHSN275201000009C, and HHSN275201000001Z).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. G.Y. participated in the data analysis and interpretation and wrote the initial draft of the manuscript. W.W.P., J.Y., C.G., J.G., and Z.C. helped to edit the manuscript. C.Z. conceived and designed the study, obtained funding, and supervised the study. All authors reviewed, edited, and approved the final version of the manuscript. G.Y. and C.Z. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of data and the accuracy of data analysis.
Prior Presentation. Parts of this study were presented in oral form at the 84th Scientific Sessions of the American Diabetes Association, Orlando, FL, 21–24 June 2024.
Handling Editors. The journal editor responsible for overseeing the review of the manuscript was Matthew C. Riddle.