Abstract
Background: Previous studies observed inverse associations of adherence to the alternate Mediterranean (aMED), Dietary Approaches to Stop Hypertension (DASH), and alternate Healthy Eating Index (aHEI) dietary patterns with risk of type 2 diabetes; however, their associations with gestational diabetes mellitus (GDM) risk are unknown.
Objective: This study aimed to assess usual prepregnancy adherence to well-known dietary patterns and GDM risk.
Design: Our study included 21,376 singleton live births reported from 15,254 participants of the Nurses’ Health Study II cohort between 1991 and 2001. Pregnancies were free of prepregnancy chronic disease or previous GDM. Prepregnancy dietary pattern adherence scores were computed based on participants’ usual intake of the patterns’ components, assessed with a validated food-frequency questionnaire. Multivariable logistic regressions with generalized estimating equations were used to estimate the RRs and 95% CIs.
Results: Incident first-time GDM was reported in 872 pregnancies. All 3 scores were inversely associated with GDM risk after adjustment for several covariables. In a comparison of the multivariable risk of GDM in participants in the fourth and first quartiles of dietary pattern adherence scores, aMED was associated with a 24% lower risk (RR: 0.76; 95% CI: 0.60, 0.95; P-trend = 0.004), DASH with a 34% lower risk (RR: 0.66; 95% CI: 0.53, 0.82; P-trend = 0.0005), and aHEI with a 46% lower risk (RR: 0.54; 95% CI: 0.43, 0.68; P-trend < 0.0001).
Conclusion: Prepregnancy adherence to healthful dietary patterns is significantly associated with a lower risk of GDM.
INTRODUCTION
Gestational diabetes mellitus (GDM)4 affects ∼7% of all US pregnancies, making it one of the most common pregnancy complications (1). Its prevalence is increasing as obesity among women of reproductive age escalates (2–4). GDM is associated with a significantly elevated risk of short-term and long-term complications for both mothers and offspring (1, 5–7). Research on early-pregnancy diet and GDM prevention is limited, with little evidence to assess its role in GDM etiology. In a recent meta-analysis of randomized trials of the effect of GDM treatment for the prevention of its subsequent morbidity, various interventions for blood glucose control, including diet, glucose monitoring, insulin use, and pharmaceuticals, did not significantly reduce the risk of some adverse perinatal and neonatal endpoints, including cesarean delivery and perinatal or neonatal death (8). Collectively, these data indicate that identifying modifiable factors for the prevention of GDM could be crucial for avoiding its associated adverse health outcomes.
Major risk factors for GDM include older age in pregnancy, a family history of diabetes, and race or ethnicity (1). Modifiable factors include excess adiposity, physical activity, and diet (9). Dietary components associated with GDM risk include macronutrients, micronutrients, and individual foods, such as refined carbohydrates, saturated and trans fats, heme iron, and processed meats (9). Whereas studying individual nutrients may lead to the understanding of important biological mechanisms, assessment of dietary patterns offers a comprehensive and complimentary approach and may be more applicable to clinical and public health interventions. Analyses of overall food patterns also account for any interactions or synergistic effects among individual foods or nutrients. If dietary patterns beneficially affect GDM risk, it would be important to disseminate such information to women of reproductive age.
Several healthful dietary pattern scores, including the alternate Mediterranean Diet (aMED), Dietary Approaches to Stop Hypertension (DASH), and alternate Healthy Eating Index (aHEI), have been inversely associated with type 2 diabetes risk among nonpregnant individuals, making them potential risk factors for GDM as well (10–12). The aMED score was adapted for the US population from a previously published pattern by Trichopoulou et al (13, 14), based on the traditional Mediterranean diet. The DASH score was developed by Fung et al (15) to measure adherence to the DASH diet, a dietary pattern originally put forth by Sacks et al (16) for the reduction of blood pressure. McCullough et al derived the aHEI score (17) from the USDA Food Guide Pyramid (18) and the 1995 Dietary Guidelines for Americans (19).
Associations of these dietary patterns with GDM risk, however, have not been investigated. The aim of this analysis was therefore to determine whether usual prepregnancy adherence to dietary patterns, including aMED, DASH, and aHEI, is associated with the risk of GDM.
SUBJECTS AND METHODS
Source population
Our research question was examined in the Nurses’ Health Study II (NHS II) longitudinal cohort. This ongoing observational prospective cohort was established in 1989, enrolling 116,671 female nurses aged 24–44 y at baseline (20). Questionnaires are distributed biennially to update lifestyle characteristics and health-related outcomes. The 1989 baseline questionnaire captured information regarding medical, diagnostic, and prescription history; date of birth; occupational status; gravidity; height and weight; and a range of other characteristics. Time-varying characteristics are updated every 2 y. Beginning in 1991 and every 4 y thereafter, participants are asked to complete a semiquantitative food-frequency questionnaire (FFQ) in addition to the main questionnaire (21). The FFQ is designed to measure average intake over the past year and has been extensively validated (22–24). Questions for various food items ask participants to report their frequency of consumption of a standard portion size, ranging from “never” to “6 or more times/d.” A validation study from a similar cohort of female participants compared foods assessed by the FFQ and multiple diet records and found a mean correlation coefficient between food items of 0.52, ranging from 0.08 for spinach to 0.90 for tea (24). This study was approved by the institutional review board of the Partners Health Care System (Boston, MA), with participants’ consent implied by the return of the questionnaires. Follow-up for each questionnaire cycle is >90% through 2001.
Study population
The first dietary questionnaire was administered in 1991 and is thus the baseline assessment for prepregnancy exposure. Investigators ceased the update of GDM occurrence in 2001; thus, follow-up is through the return of the 2001 questionnaire. NHS II participants were considered for this analysis if they reported at least one singleton live birth after the return of their 1991 questionnaire through 2001.
Individual singleton live births were included in the analysis if the participant did not report GDM in a previous pregnancy, a diagnosis of type 2 diabetes or cancer, or a cardiovascular disease event (myocardial infarction, stroke, coronary artery bypass graft procedure, or angina), before an otherwise eligible pregnancy. Pregnancies after GDM were not included because women with GDM in a previous pregnancy are likely to change their diet and lifestyle during the next pregnancy to prevent recurrent GDM. Pregnancies were also excluded if the participant did not return a prepregnancy FFQ, left >70 FFQ items blank, or reported unrealistic total energy intakes (<500, >3500 kcal/d).
Exposure assessment
Prepregnancy aMED, DASH, and aHEI dietary adherence scores were computed for each FFQ cycle occurring before a given pregnancy. Components included in each score are outlined in Table 1. Justification for inclusion of each component was described in detail elsewhere (14, 15, 17). The participants’ total score is a sum of the points earned across each dietary component for a given pattern, with aMED ranging from 0 to 8, DASH from 8 to 39, and aHEI from 2.5 to 87.5 possible points. A higher score indicates greater adherence. Scores were computed as the cumulative average of all prepregnancy questionnaires to reduce random within-person error and represent usual long-term intake before an index pregnancy. Missing exposure data were carried forward from the last FFQ for which data were captured.
TABLE 1.
Components of the dietary pattern adherence scores1
| aMED | DASH | aHEI | |
| Fruit (servings/d) | ↑ | ↑ | ↑ |
| Vegetables (servings/d) | ↑ | ↑ | ↑ |
| Nuts, legumes, soy (servings/d) | ↑ | ↑ | ↑ |
| Red and processed meats (servings/d) | ↓ | ↓ | |
| White:red meat ratio (servings/d) | ↑ | ||
| Fish and seafood (servings/d) | ↑ | ||
| Whole grains (servings/d) | ↑ | ↑ | |
| Cereal fiber (g/d) | ↑ | ||
| Low-fat dairy (servings/d) | ↑ | ||
| Sweetened beverages (servings/d) | ↓ | ||
| Moderate alcohol (servings/d) | ↑ | ↑ | |
| MUFA:SFA | ↑ | ||
| PUFA:SFA | ↑ | ||
| trans Fat (% of energy/d) | ↓ | ||
| Sodium (mg/d) | ↓ | ||
| Multivitamin use (servings/d) | ↑ |
aHEI, alternate Healthy Eating Index; aMED, alternate Mediterranean Diet; DASH, Dietary Approaches to Stop Hypertension; ↑, encourages greater intake; ↓, encourages no/less intake.
To compute the aMED score, participants were allotted 1 point for being above the median of servings/d for each component, with the exception of red and processed meats, which was scored 1 point for being below the median intake. One point was earned for moderate alcohol consumption, defined as 5–15 g/d (25). Median servings/d cutoffs were computed for each FFQ year from the overall NHS II population.
In a recent publication by Fung et al (26), authors pointed out that the primary sources of MUFA in a similar cohort of US women differed substantially from Mediterranean countries. Achieving a higher MUFA:SFA ratio might therefore not be representative of the healthful dietary fats intended by the Mediterranean Diet (olive oil, other plant sources) in our population (27). Although trends in olive oil consumption increased over time in their female population, it was still outweighed by beef and other meats contributing to 18–30% of MUFA intake. Prepregnancy red and processed meat intakes have been positively associated with GDM risk (28); therefore, in a secondary analysis, the MUFA:SFA ratio component was omitted from the total aMED score.
The DASH score was derived similarly to the aMED score (16). Participants received points for each dietary component based on their quintile of intake (servings/d). For example, a participant in the third quintile for usual consumption of vegetables earned 3 points. The sweetened beverages component was scored by quartiles because there was less variability in intake. Inverse scoring was used for red and processed meats, sugar-sweetened beverages, and sodium to reflect greater adherence with less intake.
Secondary analyses examined modifications to scoring the DASH diet. Results from the Optimal Macronutrient Intake Trial for Heart Health (OMNIHeart) randomized trial indicated that vegetable protein-based and vegetable MUFA-based dietary patterns were superior to a carbohydrate-based DASH diet for several cardioprotective intermediate endpoints (29). Therefore, we added components to the DASH score to reflect macronutrient composition. The OMNI-protein score included additional points for increasing quintiles of vegetable-based protein intake and for decreasing quintiles of total carbohydrate intake (% of total energy). The OMNI-fat score included additional points for increasing quintiles of vegetable-based fat intake and for decreasing quintiles of total carbohydrate intake (% of total energy).
For the aHEI pattern, points were given for intake of each component on a scale from 0 to 10, with 10 indicating adherence to the 1995 USDA recommended levels of intake, 0 for the worst intake, and intermediate scores categorized proportionately. Multivitamin use was scored based on duration, giving 2.5 points for <5 y and 7.5 points for ≥5 y of use.
Outcome assessment
The outcome of interest was incident GDM. A physician diagnosis of GDM was ascertained by self-report on each biennial questionnaire through 2001. In the case of more than one singleton birth reported within a 2-y questionnaire period, GDM status was attributed to the first singleton birth, because the questionnaire does not specify which pregnancy experienced GDM if multiple pregnancies occurred. In a validation study among a subgroup of the NHS II cohort, 94% of GDM reports were confirmed by medical records (20). Among the confirmed GDM diagnoses, physicians were most likely to use the National Diabetes Data Group criteria. From a supplemental questionnaire sent to a random sample of parous women without GDM, 83% reported a glucose screening test during pregnancy and 100% reported frequent prenatal urine screening, suggesting a high level of GDM surveillance among both cases and noncases in this cohort.
Covariable assessment
Age of the mother was computed from date of birth to the date of each questionnaire return. Participants reported their current weight on each biennial questionnaire. Self-reported weight was highly correlated with measured weight among a random subset of Boston-area cohort participants (r = 0.97) (30). BMI was computed as weight in kilograms divided by the square of height in meters. Total physical activity was ascertained by frequency of engaging in common recreational activities, from which metabolic equivalent of task-hours per week were derived. The questionnaire-based estimates correlated well with detailed activity diaries in a prior validation study (r = 0.56) (31). In addition, we ascertained information about the participants’ smoking status, self-reported race and ethnicity, and parental history of hypertension and diabetes. These data were also previously validated (32, 33). Total energy intake was calculated by summing up energy from all foods reported on in the FFQ.
Statistical analysis
Baseline differences between participants in quartile 1 and quartile 4 were compared for each dietary pattern by using a chi-square test (categorical variables) and univariate linear regression (continuous variables). Dietary pattern adherence scores were analyzed both continuously for a 1-SD increase and categorically as quartiles. Pairwise correlations between baseline continuous pattern scores were computed to assess the similarity of exposures. Multivariable marginal logistic regression using generalized estimating equations, specifying an exchangeable correlation structure, were analzyed to approximate the RRs and 95% CIs for associations between the dietary pattern adherence scores and GDM risk. Models for dietary score quartiles were accompanied by chi-square tests for trend across the quartiles’ median values. In addition to age- and energy-adjusted models (model 1), multivariable models (model 2, model 3) were further adjusted for a priori–selected prepregnancy covariables. These included BMI (kg/m2), physical activity (MET-h/wk), sedentary time at home (h/wk), gravidity (total pregnancies lasting ≥6 mo), smoking status, parental history of type 2 diabetes, and race or ethnicity. The DASH score model additionally adjusted for alcohol (g/d), because this is not a component of the dietary pattern and is a potential lifestyle confounder. Age, gravidity, and lifestyle covariables were updated biennially, except physical activity, which was captured every 4 y. Non–time-varying covariables, including parental history of type 2 diabetes, and race or ethnicity were captured at baseline. Categorical covariables included an indicator for missing data, if necessary.
Additional analyses were conducted to investigate effect modification by parental history of diabetes (no compared with yes) and high-risk pregnancy age group (age <35 compared with age ≥35 y). P values for heterogeneity were derived from the cross-product interaction term coefficient (continuous dietary score × binary variable) added to the main-effects multivariable model. We also performed interaction tests for the joint effect of BMI group (normal weight, BMI <25; overweight, BMI 25–29; obese, BMI ≥30) and dietary pattern scores through a score test, fitting the model with the interaction terms constrained to zero. Additionally, because BMI is a possible intermediate between diet and GDM, we estimated the proportion of the association between dietary patterns and GDM risk that is explained by prepregnancy BMI (modeled continuously), and the 95% CI and P value for significance of mediation, with an SAS macro developed by Spiegelman et al (Harvard School of Public Health; Mediate SAS; http://www.hsph.harvard.edu/faculty/donna-spiegelman/software/mediate/). This estimates mediation from the proportional reduction in the exposure regression coefficient when continuous BMI is included in the multivariable model (34).
Several sensitivity analyses were conducted to assess the robustness of our findings. This included restricting our analysis to first births from nulliparous women to reduce possible confounding by experiences from previous pregnancies. In a separate analysis for each pattern, we modeled the components simultaneously to assess a 1-point increase in total score by a given dietary aspect, holding points from all other components constant. This was performed to assess whether the contribution of any one or more individual components explained the association between the total score and GDM risk. Finally, exposure was derived from the most recent prepregnancy FFQ for comparison with the main findings. SAS command PROC GENMOD was used to compute the models, accounting for correlation within women who contributed more than one pregnancy to the analysis (SAS version 9.1; SAS Institute Inc).
RESULTS
Overall 15,254 participants met our inclusion criteria, contributing 21,376 eligible singleton births to this analysis during 10 y of follow-up. The participants’ baseline characteristics according to dietary pattern adherence score quartiles (quartile 1 compared with quartile 4) are shown in Table 2. On average, participants who were in the highest quartiles reported more physical activity, consumed more alcohol, had a higher total energy intake, and were less likely to be current smokers or ever users of oral contraceptives. Women with better adherence also had lower prepregnancy BMI. The dietary pattern adherence scores were approximately normally distributed and significantly correlated with one another (P < 0.0001), with correlation coefficients (r) between aMED and DASH of 0.71, aMED and aHEI of 0.76, and DASH and aHEI of 0.74. There were 872 cases of incident first-time GDM reported during follow-up.
TABLE 2.
Baseline (1991) characteristics by prepregnancy dietary pattern adherence score quartile1
| aMED |
DASH |
aHEI |
||||
| Q1 (n = 3297) | Q4 (n = 3661) | Q1 (n = 3018) | Q4 (n = 3927) | Q1 (n = 3496) | Q4 (n = 4385) | |
| Diet score | 1.6 ± 0.62 | 6.6 ± 0.7* | 16.9 ± 2.0 | 30.3 ± 2.1* | 25.0 ± 3.7 | 51.4 ± 6.1* |
| Age (y) | 31.4 ± 3.2 | 32.5 ± 3.3* | 31.5 ± 3.1 | 32.5 ± 3.3* | 31.5 ± 3.2 | 32.5 ± 3.3* |
| BMI (kg/m2) | 23.9 ± 4.9 | 23.1 ± 4.0* | 23.8 ± 5.0 | 23.2 ± 4.0* | 24.1 ± 5.0 | 22.7 ± 3.6* |
| BMI at age 18 y | 21.1 ± 3.2 | 20.9 ± 2.8* | 21.0 ± 3.4 | 21.0 ± 2.9 | 21.2 ± 3.3 | 20.9 ± 2.8* |
| Physical activity (MET-h/wk) | 17.7 ± 23.5 | 29.5 ± 33.6* | 16.3 ± 21.7 | 31.1 ± 35.2* | 16.0 ± 22.8 | 31.4 ± 33.8* |
| Alcohol (g/d) | 2.1 ± 4.9 | 4.0 ± 5.1* | 2.9 ± 5.4 | 3.0 ± 4.8 | 1.4 ± 5.1 | 4.7 ± 5.2* |
| Total energy (kcal/d) | 1570 ± 480 | 2130 ± 520* | 1600 ± 490 | 2090 ± 520* | 1560 ± 470 | 2090 ± 550* |
| Carbohydrate (% of energy/d) | 48 ± 7 | 53 ± 7* | 47 ± 8 | 55 ± 7* | 48 ± 7 | 54 ± 7* |
| Protein (% of energy/d) | 19 ± 4 | 19 ± 3* | 19 ± 4 | 19 ± 3* | 19 ± 3 | 19 ± 3* |
| Total fat (% of energy/d) | 33 ± 5 | 29 ± 5* | 34 ± 5 | 28 ± 5* | 34 ± 5 | 28 ± 5* |
| MUFA (% of energy/d) | 13 ± 2 | 11 ± 2* | 13 ± 2 | 10 ± 2* | 13 ± 2 | 11 ± 2* |
| SFA (% of energy/d) | 13 ± 2 | 10 ± 2* | 13 ± 2 | 10 ± 2* | 13 ± 2 | 10 ± 2* |
| Animal fat (% of energy/d) | 20 ± 4 | 15 ± 4* | 20 ± 5 | 15 ± 4* | 20 ± 4 | 15 ± 4* |
| trans Fat (g/d) | 3.2 ± 1.5 | 3.2 ± 1.5 | 3.5 ± 1.6 | 2.9 ± 1.3* | 3.5 ± 1.6 | 3.0 ± 1.4* |
| Glycemic index | 54 ± 4 | 54 ± 3* | 56 ± 3 | 53 ± 3* | 55 ± 3 | 53 ± 3* |
| Glycemic load | 100 ± 40 | 150 ± 40* | 110 ± 40 | 150 ± 40* | 100 ± 40 | 150 ± 50* |
| White [n (%)] | 3028 (92) | 3454 (94)* | 2757 (91) | 3700 (94)* | 3208 (92) | 4121 (94)* |
| Gravidity [n (%)] | ||||||
| 0 (nulliparous) | 1307 (40) | 1480 (41)* | 1271 (43) | 1535 (40)* | 1255 (37) | 1982 (46)* |
| 1 | 1034 (32) | 1040 (29) | 900 (30) | 1202 (31) | 1160 (34) | 1206 (28) |
| 2 | 634 (20) | 750 (21) | 573 (19) | 773 (20) | 723 (21) | 792 (18) |
| 3 | 195 (6) | 243 (7) | 168 (6) | 236 (6) | 218 (6) | 228 (5) |
| ≥4 | 69 (2) | 180 (4) | 58 (2) | 99 (3) | 74 (2) | 98 (2) |
| Smoking status [n (%)] | ||||||
| Never | 2337 (71) | 2567 (70)* | 1990 (66) | 2871 (73)* | 2558 (73) | 2992 (68)* |
| Former | 580 (18) | 836 (23) | 555 (18) | 848 (22) | 535 (15) | 1051 (24) |
| Current | 376 (11) | 251 (7) | 467 (16) | 200 (5) | 396 (12) | 334 (8) |
| Oral contraceptive use [n (%)] | ||||||
| Never | 493 (15) | 646 (18)* | 421 (14) | 738 (19)* | 509 (15) | 774 (18)* |
| Ever | 2803 (85) | 3014 (82) | 2595 (86) | 3188 (81) | 2986 (85) | 3608 (82) |
| Parental history of diabetes [n (%)] | 387 (12) | 382 (10) | 349 (12) | 393 (10)** | 425 (12) | 459 (10) |
Higher scores indicate greater pattern adherence. *P < 0.01, **P < 0.05. aHEI, alternate Healthy Eating Index; aMED, alternate Mediterranean Diet; DASH, Dietary Approaches to Stop Hypertension; MET-h, metabolic equivalent of task-hours; Q, quartile.
Mean ± SD (all such values).
All 3 dietary pattern adherence scores were significantly and inversely associated with a lower GDM risk. In the fully adjusted multivariable model 3, comparing participants in the fourth quartile (greatest dietary pattern adherence) with those in the first reference quartile (lowest dietary pattern adherence), the risk of GDM was 24% lower with the aMED score (RR: 0.76; 95% CI: 0.60, 0.95; P-trend = 0.004), 34% lower with the DASH score (RR: 0.66; 95% CI: 0.53, 0.82; P-trend = 0.0005), and 46% lower with the aHEI score (RR = 0.54; 95% CI: 0.43, 0.68; P-trend < 0.0001) (Table 3). Similarly, all 3 dietary patterns were significantly associated with a lower risk of GDM when modeled as a continuous variable for a 1-SD increase in adherence (aMED SD = 1.8, DASH SD = 4.8, aHEI SD = 10.4). In the fully adjusted multivariable model 3, a 1-SD increase in score was associated with a 10% lower GDM risk for the aMED pattern (RR: 0.90; 95% CI: 0.83, 0.97), 15% for the DASH pattern (RR: 0.85; 95% CI: 0.79, 0.92), and 24% for the aHEI pattern (RR: 0.76; 95% CI: 0.70, 0.82). Attenuation of the effect estimate after covariable adjustment was driven primarily by prepregnancy BMI. Prepregnancy BMI was estimated to mediate 38% (95% CI: 18, 62%; P = 0.002) of the association between aMED dietary pattern and GDM and 16% (95% CI: 10, 24%; P < 0.0001) of the association between the aHEI dietary pattern and GDM, but was not a statistically significant mediator of the DASH dietary pattern and GDM association (4%; 95% CI: 0.4, 34%; P = 0.4).
TABLE 3.
Quartiles of prepregnancy dietary pattern adherence scores and GDM risk1
| Q12 | Q2 | Q3 | Q4 | P-trend | |
| aMED | |||||
| GDM/pregnancies | 221/4601 | 321/7366 | 147/4134 | 183/5275 | |
| Model 1 | 1.0 | 0.87 (0.73, 1.03)3 | 0.66 (0.53, 0.82) | 0.61 (0.49, 0.75) | <0.0001 |
| Model 2 | 1.0 | 0.89 (0.74, 1.06) | 0.70 (0.57, 0.88) | 0.67 (0.54, 0.84) | 0.0001 |
| Model 3 | 1.0 | 0.95 (0.79, 1.14) | 0.76 (0.60, 0.95) | 0.76 (0.60, 0.95) | 0.004 |
| DASH | |||||
| GDM/pregnancies | 232/4213 | 220/5573 | 227/5806 | 193/5784 | |
| Model 1 | 1.0 | 0.69 (0.57, 0.83) | 0.66 (0.54, 0.79) | 0.52 (0.42, 0.64) | <0.0001 |
| Model 2 | 1.0 | 0.75 (0.61, 0.90) | 0.74 (0.61, 0.90) | 0.61 (0.49, 0.76) | <0.0001 |
| Model 3 | 1.0 | 0.77 (0.63, 0.93) | 0.78 (0.64, 0.95) | 0.66 (0.53, 0.82) | 0.0005 |
| aHEI | |||||
| GDM/pregnancies | 242/4661 | 252/5261 | 203/5313 | 175/6141 | |
| Model 1 | 1.0 | 0.86 (0.72, 1.04) | 0.64 (0.53, 0.79) | 0.44 (0.36, 0.54) | <0.0001 |
| Model 2 | 1.0 | 0.90 (0.74, 1.08) | 0.67 (0.55, 0.81) | 0.46 (0.37, 0.57) | <0.0001 |
| Model 3 | 1.0 | 0.96 (0.79, 1.15) | 0.75 (0.61, 0.91) | 0.54 (0.43, 0.68) | <0.0001 |
Model 1: adjusted for age (mo), total energy intake (kcal/d; quintiles). Model 2: adjusted as for model 1 plus gravidity (0, 1, 2, 3, and ≥4), smoking status (never, former, or current), physical activity (in MET-h/wk; quartiles), sedentary time (hours sitting at home/wk: 0–1, 2–5, 6–10, 11–20, or ≥21), parental history of type 2 diabetes (yes or no), and, for the DASH analysis only, alcohol (g/d: 0, 1–14, or ≥15). Model 3: adjusted as for model 2 plus prepregnancy BMI (in kg/m2; categorical <23, 24–25, 26–27, 28–30, 31–34, or ≥35). Cox proportional hazards models were used to estimate RRs and 95% CIs; chi-square test for P-trend. Higher scores indicate greater pattern adherence. aHEI, alternate Healthy Eating Index; aMED, alternate Mediterranean Diet; DASH, Dietary Approaches to Stop Hypertension; GDM, gestational diabetes mellitus; MET-h, metabolic equivalent of task-hours; Q, quartile.
Reference.
RR; 95% CI in parentheses (all such values).
Our modified aMED score that was created by removing the MUFA:SFA ratio component correlated highly with the main aMED score (r = 0.96), as did the OMNIHeart scores with the DASH score (OMNI-protein compared with DASH: r = 0.95; OMNI-fat compared with DASH: r = 0.90). As expected, results with these modified dietary pattern adherence scores were similar to their primary adherence scores. A comparison of the participants of the fourth quartile with those of the first quartile in the multivariable model 3 showed that adherence to the modified aMED pattern was associated with a 25% lower risk of GDM (RR: 0.75; 95% CI: 0.61, 0.91), OMNI-protein with a 32% lower risk (RR: 0.68; 95% CI: 0.55, 0.85), and OMNI-fat with a 26% lower risk (RR: 0.74; 95% CI: 0.60, 0.91).
Tests for heterogeneity did not show a significant effect modification by parental history of diabetes, prepregnancy BMI, or age group (data not shown). The results were similar when the analysis was restricted to first pregnancies only (n = 6350; cases = 358) and when only the most recent prepregnancy FFQ data were used (data not shown). There did not appear to be any one dietary component driving the observed associations of the overall patterns in the analyses simultaneously modeling the individual dietary pattern components (data not shown). Several components independently contributed to the inverse associations between dietary patterns and GDM risk, including fruit, nuts and soy, whole grains and cereal fiber, moderate alcohol, and decreased intakes of red and processed meats, trans fat, and sugary beverages. Despite high correlations between the diet scores, the association between the aHEI pattern and GDM risk appeared stronger than the aMED and DASH patterns. A post hoc exploratory analysis was conducted to investigate whether the strength of this association could be explained by differences in scoring methods and discriminatory abilities. Scores for each component were rederived based on deciles of intake for each dietary pattern; thus, the patterns were scaled similarly and more directly comparable. Results from this analysis suggested that some of the differences between dietary patterns could be explained by their scoring methods; however, the association for aHEI remained significant even after adjustment for aMED and DASH, suggesting that components unique to aHEI may contribute to the greater strength of the association of aHEI with GDM as compared with DASH and aMED (data not shown).
DISCUSSION
In this large prospective cohort of 21,376 pregnancies, we found that prepregnancy adherence to a variety of healthful dietary patterns is associated with a robust, significant decrease in GDM risk. The correlation between dietary pattern adherence scores was high, >0.70 for all 2-way comparisons, despite unique components to each diet. All 3 dietary patterns were significantly associated with a reduced risk of GDM, suggesting a potentially robust role of a variety of dietary factors in the development of GDM.
Healthful dietary patterns have consistently been associated with a reduced risk of type 2 diabetes; however, there are limited studies of prepregnancy dietary patterns and risk of GDM (35). A study by Zhang et al assessed prepregnancy dietary patterns derived by factor analysis and the risk of GDM in the same NHS II cohort (28). A prudent pattern identified by the data was positively correlated with fruit, green leafy vegetables, poultry, and fish, whereas a Western pattern was correlated with increased intakes of red meat, processed meat, refined grains, sweets, French fries, and pizza. Adherence to the prudent pattern was significantly associated with a reduced risk of GDM in a comparison of those in the highest quintile with those in the lowest quintile (RR: 0.72; 95% CI: 0.55, 0.93). Similarly, those in the lowest quintile of adherence to the Western pattern compared with the highest quintile had a significantly lower risk of GDM (RR: 0.61; 95% CI: 0.45, 0.84). Findings from this previous analysis suggested that both consumption of healthy foods and avoidance of unhealthy foods were associated with a reduced risk of GDM and are consistent with our results presented here.
Several potential mechanisms may explain the observed associations between prepregnancy dietary patterns and GDM risk, although the precise underlying molecular mechanisms are unclear. A moderate amount of the association between dietary patterns and GDM risk appears to be mediated by BMI. Sensitivity analyses in which BMI was adjusted for as a continuous variable did not substantially change the results, indicating that these findings are unlikely to be entirely explained by residual confounding of weight status. Prior evidence suggests that women who develop GDM have prepregnancy β cell dysfunction and insulin resistance, compromising their ability to adapt to the metabolic challenges presented in pregnancy (36, 37). Adherence to diets such as the aMED, DASH, and aHEI may reduce GDM risk by minimizing such susceptibilities in the time leading up to pregnancy. Common components between the dietary patterns include fruit and vegetables, minimal red and processed meats, and carbohydrate quality. Fruit and vegetables are rich in antioxidants and photochemicals, dietary fiber, and micronutrients such as magnesium and vitamin C. The combination of these might prevent metabolic deterioration by opposing free radicals and improving systemic oxidative stress (38). Both plasma ascorbic acid concentrations and dietary vitamin C intake were inversely associated with GDM risk in the prospective OMEGA cohort (39). In a previous study with the NHS II cohort, prepregnancy intake of fruit fiber was found to be inversely associated with GDM risk (40). Foods such as fruit and vegetables might also indirectly confer a benefit by replacing harmful foods in the diet. Red and processed meats are sources of saturated fat, heme iron, nitrosamines, and other constituents and have been associated with β cell damage, oxidative stress, and insulin resistance as well as incident GDM (28). Whole grains are high in insoluble fiber (low in glycemic index), which blunts absorption of glucose and subsequent insulin requirements (41), and prepregnancy intake has been associated with a reduced risk of GDM (40). The glycemic index is a measure of carbohydrate quality and is directly correlated with glucose absorption and an increase in insulin after eating a food or meal. Both have been associated with hyperglycemia and hyperinsulinemia (42) as well as incident GDM in the NHS II cohort (40).
The analysis of dietary patterns, rather than individual foods or nutrients, allows one to capture the diet as a whole, including any unknown or synergistic effects between its components. Patterns also lead to a more comprehensive clinical and public health message, as foods are not eaten in isolation but as part of meals and overall dietary habits. Assessing dietary patterns that are derived from a priori hypotheses, such as those included here, increase generalizability across populations and time. Other strengths of this analysis are that we were able to control for several prepregnancy factors and lifestyle characteristics. The largest change in the effect estimate was seen when adjusting for BMI, a well-known GDM risk factor. Adjustment for several other potential confounders led to minor changes in the relative risk; thus, it is unlikely that unmeasured or residual confounding would explain a substantial amount of the observed associations. Another strength of this analysis is the prospective assessment of prepregnancy diet. Several previous publications of prepregnancy diet and pregnancy outcomes are limited to dietary questionnaires administered after the participants became pregnant. Changes to diet in pregnancy or participants’ awareness of their outcome status might lead to exposure misclassification and/or recall bias. Capturing prepregnancy dietary data in the prepregnancy period avoids these limitations. Additionally, because our FFQs were administered every 4 years, we were able to compute an updated cumulative average of prepregnancy dietary pattern adherence scores for the portion of participants accumulating more than one prepregnancy FFQ, thus better reflecting long-term intake, and reducing random within-person error and attenuation of the effect estimate toward the null relative risk. Finally, although GDM was ascertained by questionnaire self-report, it has been previously validated and demonstrated good validity (20).
We acknowledge that there are limitations of this analysis. First, the NHS II cohort does not capture diet during pregnancy. It is likely that diet during the 2 time periods is correlated. We are therefore unable to resolve that the associations of prepregnancy diet with GDM are independent of diet during pregnancy. Future research is required to partition out the relative contributions of diet in these 2 time periods to GDM risk. Last, our study population consisted mostly of white women, thus we are unable to ascertain whether the association is similar across other race and ethnic groups that are at higher risk of GDM. However, the relative homogeneity of our population advantageously reduces unmeasured confounding.
Overall, we found strong and consistent associations between prepregnancy adherence to several dietary patterns and a lower GDM risk. Common elements of these diets include a high intake of fruit, vegetables, whole grains, and nuts and legumes and a low consumption of red and processed meats. Further research is required to know whether improving one's dietary pattern adherence during pregnancy is associated with a lower risk of GDM. These results suggest that clinical and public health efforts to encourage diets similar to the aMED, DASH, and HEI patterns for women of a reproductive age might yield benefits in the reduction of GDM risk in a future pregnancy.
Supplementary Material
Acknowledgments
The authors’ responsibilities were as follows—DKT: performed the statistical analysis; and all authors: contributed to the conception of the hypothesis and the analysis plan and to the manuscript preparation and/or review. All authors stated that they had no conflicts of interest to declare.
Footnotes
Abbreviations used: aHEI, alternate Healthy Eating Index; aMED, alternate Mediterranean Diet; DASH, Dietary Approaches to Stop Hypertension; FFQ, food-frequency questionnaire, GDM; gestational diabetes mellitus, NHS II; Nurses’ Health Study II; OMNIHeart, Optimal Macronutrient Intake Trial for Heart Health
REFERENCES
- 1.ADA Gestational diabetes mellitus. Diabetes Care 2004;27(suppl 1):S88–90 [DOI] [PubMed] [Google Scholar]
- 2.Albrecht SS, Kuklina EV, Bansil P, Jamieson DJ, Whiteman MK, Kourtis AP, Posner SF, Callaghan WM. Diabetes trends among delivery hospitalizations in the United States, 1994-2004. Diabetes Care 2010;33(4):768–73 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Dabelea D, Snell-Bergeon JK, Hartsfield CL, Bischoff KJ, Hamman RF, McDuffie RS. Increasing prevalence of gestational diabetes mellitus (GDM) over time and by birth cohort: Kaiser Permanente of Colorado GDM Screening Program. Diabetes Care 2005;28:579–84 [DOI] [PubMed] [Google Scholar]
- 4.Ferrara A, Kahn HS, Quesenberry CP, Riley C, Hedderson MM. An increase in the incidence of gestational diabetes mellitus: Northern California, 1991-2000. Obstet Gynecol 2004;103:526–33 [DOI] [PubMed] [Google Scholar]
- 5.Langer O, Mazze R. The relationship between large-for-gestational-age infants and glycemic control in women with gestational diabetes. Am J Obstet Gynecol 1988;159:1478–83 [DOI] [PubMed] [Google Scholar]
- 6.McMahon MJ, Ananth CV, Liston RM. Gestational diabetes mellitus. Risk factors, obstetric complications and infant outcomes. J Reprod Med 1998;43:372–8 [PubMed] [Google Scholar]
- 7.Bellamy L, Casas JP, Hingorani AD, Williams D. Type 2 diabetes mellitus after gestational diabetes: a systematic review and meta-analysis. Lancet 2009;373:1773–9 [DOI] [PubMed] [Google Scholar]
- 8.Horvath K, Koch K, Jeitler K, Matyas E, Bender R, Bastian H, Lange S, Siebenhofer A. Effects of treatment in women with gestational diabetes mellitus: systematic review and meta-analysis. BMJ 2010;340:c1395. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zhang C, Ning Y. doi: 10.3945/ajcn.110.001032. Effect of dietary and lifestyle factors on the risk of gestational diabetes: review of epidemiologic evidence. Am J Clin Nutr 2011;94(suppl):1975S–9S. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Liese AD, Nichols M, Sun X, D'Agostino RB, Jr, Haffner SM. Adherence to the DASH Diet is inversely associated with incidence of type 2 diabetes: the insulin resistance atherosclerosis study. Diabetes Care 2009;32:1434–6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Fung TT, McCullough M, van Dam RM, Hu FB. A prospective study of overall diet quality and risk of type 2 diabetes in women. Diabetes Care 2007;30:1753–7 [DOI] [PubMed] [Google Scholar]
- 12.Salas-Salvadó J, Bullo M, Babio N, Martinez-Gonzalez MA, Ibarrola-Jurado N, Basora J, Estruch R, Covas MI, Corella D, Aros F, et al. Reduction in the Incidence of type 2 diabetes with the Mediterranean Diet: results of the PREDIMED-Reus nutrition intervention randomized trial. Diabetes Care 2011;34:14–9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Trichopoulou A, Costacou T, Bamia C, Trichopoulos D. Adherence to a Mediterranean diet and survival in a Greek population. N Engl J Med 2003;348:2599–608 [DOI] [PubMed] [Google Scholar]
- 14.Fung TT, McCullough ML, Newby PK, Manson JE, Meigs JB, Rifai N, Willett WC, Hu FB. Diet-quality scores and plasma concentrations of markers of inflammation and endothelial dysfunction. Am J Clin Nutr 2005;82:163–73 [DOI] [PubMed] [Google Scholar]
- 15.Fung TT, Chiuve SE, McCullough ML, Rexrode KM, Logroscino G, Hu FB. Adherence to a DASH-style diet and risk of coronary heart disease and stroke in women. Arch Intern Med 2008;168:713–20 [DOI] [PubMed] [Google Scholar]
- 16.Sacks FM, Obarzanek E, Windhauser MM, Svetkey LP, Vollmer WM, McCullough M, Karanja N, Lin PH, Steele P, Proschan MA, et al. Rationale and design of the Dietary Approaches to Stop Hypertension trial (DASH). A multicenter controlled-feeding study of dietary patterns to lower blood pressure. Ann Epidemiol 1995;5:108–18 [DOI] [PubMed] [Google Scholar]
- 17.McCullough ML, Feskanich D, Stampfer MJ, Giovannucci EL, Rimm EB, Hu FB, Spiegelman D, Hunter DJ, Colditz GA, Willett WC. Diet quality and major chronic disease risk in men and women: moving toward improved dietary guidance. Am J Clin Nutr 2002;76:1261–71 [DOI] [PubMed] [Google Scholar]
- 18.USDA. The Food Guide Pyramid Washington, DC: US Government Printing Office, 1992 [Google Scholar]
- 19.Kennedy E, Meyers L, Layden W. The 1995 Dietary Guidelines for Americans: an overview. J Am Diet Assoc 1996;96:234–7 [DOI] [PubMed] [Google Scholar]
- 20.Solomon CG, Willett WC, Carey VJ, Rich-Edwards J, Hunter DJ, Colditz GA, Stampfer MJ, Speizer FE, Spiegelman D, Manson JE. A prospective study of pregravid determinants of gestational diabetes mellitus. JAMA 1997;278:1078–83 [PubMed] [Google Scholar]
- 21.Willett W. Nutritional epidemiology. 2nd ed. New York, NY: Oxford University Press, 1998 [Google Scholar]
- 22.Willett WC, Reynolds RD, Cottrell-Hoehner S, Sampson L, Browne ML. Validation of a semi-quantitative food frequency questionnaire: comparison with a 1-year diet record. J Am Diet Assoc 1987;87:43–7 [PubMed] [Google Scholar]
- 23.Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, Hennekens CH, Speizer FE. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122:51–65 [DOI] [PubMed] [Google Scholar]
- 24.Salvini S, Hunter DJ, Sampson L, Stampfer MJ, Colditz GA, Rosner B, Willett WC. Food-based validation of a dietary questionnaire: the effects of week-to-week variation in food consumption. Int J Epidemiol 1989;18:858–67 [DOI] [PubMed] [Google Scholar]
- 25.Fung TT, Hu FB, McCullough ML, Newby PK, Willett WC, Holmes MD. Diet quality is associated with the risk of estrogen receptor-negative breast cancer in postmenopausal women. J Nutr 2006;136:466–72 [DOI] [PubMed] [Google Scholar]
- 26.Fung TT, Rexrode KM, Mantzoros CS, Manson JE, Willett WC, Hu FB. Mediterranean diet and incidence of and mortality from coronary heart disease and stroke in women. Circulation 2009;119:1093–100 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Willett WC, Sacks F, Trichopoulou A, Drescher G, Ferro-Luzzi A, Helsing E, Trichopoulos D. Mediterranean diet pyramid: a cultural model for healthy eating. Am J Clin Nutr 1995;61:1402S–6S [DOI] [PubMed] [Google Scholar]
- 28.Zhang C, Schulze MB, Solomon CG, Hu FB. A prospective study of dietary patterns, meat intake and the risk of gestational diabetes mellitus. Diabetologia 2006;49:2604–13 [DOI] [PubMed] [Google Scholar]
- 29.Appel LJ, Sacks FM, Carey VJ, Obarzanek E, Swain JF, Miller ER 3rd, Conlin PR, Erlinger TP, Rosner BA, Laranjo NM, et al. Effects of protein, monounsaturated fat, and carbohydrate intake on blood pressure and serum lipids: results of the OmniHeart randomized trial. JAMA 2005;294:2455–64 [DOI] [PubMed] [Google Scholar]
- 30.Rimm EB, Stampfer MJ, Colditz GA, Chute CG, Litin LB, Willett WC. Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990;1:466–73 [DOI] [PubMed] [Google Scholar]
- 31.Wolf AM, Hunter DJ, Colditz GA, Manson JE, Stampfer MJ, Corsano KA, Rosner B, Kriska A, Willett WC. Reproducibility and validity of a self-administered physical activity questionnaire. Int J Epidemiol 1994;23:991–9 [DOI] [PubMed] [Google Scholar]
- 32.Troy LM, Michels KB, Hunter DJ, Spiegelman D, Manson JE, Colditz GA, Stampfer MJ, Willett WC. Self-reported birthweight and history of having been breastfed among younger women: an assessment of validity. Int J Epidemiol 1996;25:122–7 [DOI] [PubMed] [Google Scholar]
- 33.Troy LM, Hunter DJ, Manson JE, Colditz GA, Stampfer MJ, Willett WC. The validity of recalled weight among younger women. Int J Obes Relat Metab Disord 1995;19:570–2 [PubMed] [Google Scholar]
- 34.Lin DY, Fleming TR, De Gruttola V. Estimating the proportion of treatment effect explained by a surrogate marker. Stat Med 1997;16:1515–27 [DOI] [PubMed] [Google Scholar]
- 35.Esposito K, Kastorini CM, Panagiotakos DB, Giugliano D. Prevention of type 2 diabetes by dietary patterns: a systematic review of prospective studies and meta-analysis. Metab Syndr Relat Disord 2010;8:471–6 [DOI] [PubMed] [Google Scholar]
- 36.Metzger BE, Buchanan TA, Coustan DR, de Leiva A, Dunger DB, Hadden DR, Hod M, Kitzmiller JL, Kjos SL, Oats JN, et al. Summary and recommendations of the Fifth International Workshop-Conference on Gestational Diabetes Mellitus. Diabetes Care 2007;30(suppl 2):S251–60 [DOI] [PubMed] [Google Scholar]
- 37.Buchanan TA. Pancreatic B-cell defects in gestational diabetes: implications for the pathogenesis and prevention of type 2 diabetes. J Clin Endocrinol Metab 2001;86:989–93 [DOI] [PubMed] [Google Scholar]
- 38.Hamer M, Chida Y. Intake of fruit, vegetables, and antioxidants and risk of type 2 diabetes: systematic review and meta-analysis. J Hypertens 2007;25:2361–9 [DOI] [PubMed] [Google Scholar]
- 39.Zhang C, Williams MA, Sorensen TK, King IB, Kestin MM, Thompson ML, Leisenring WM, Dashow EE, Luthy DA. Maternal plasma ascorbic acid (vitamin C) and risk of gestational diabetes mellitus. Epidemiology 2004;15:597–604 [DOI] [PubMed] [Google Scholar]
- 40.Zhang C, Liu S, Solomon CG, Hu FB. Dietary fiber intake, dietary glycemic load, and the risk for gestational diabetes mellitus. Diabetes Care 2006;29:2223–30 [DOI] [PubMed] [Google Scholar]
- 41.de Munter JS, Hu FB, Spiegelman D, Franz M, van Dam RM. Whole grain, bran, and germ intake and risk of type 2 diabetes: a prospective cohort study and systematic review. PLoS Med 2007;4:e261. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Brand-Miller JC, Thomas M, Swan V, Ahmad ZI, Petocz P, Colagiuri S. Physiological validation of the concept of glycemic load in lean young adults. J Nutr 2003;133:2728–32 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.