Improvement in Maternal Diet Quality Across Pregnancy Is Associated With Lower Gestational Weight Gain in Women With Obesity From the Illinois Kids Development Study

Abstract

Background

It is unclear how dietary patterns are associated with gestational weight gain (GWG) because prior studies focused on individual foods and did not consider diet at multiple timepoints.

Objective

This study aimed to evaluate associations between changes in diet quality across pregnancy with GWG.

Design

This secondary data analysis included women from the Illinois Kids Development Study, a longitudinal prospective cohort.

Participants/Setting

Participants were 380 women recruited from 2013 through 2018 from obstetric clinics in Champaign-Urbana, IL, who were followed through delivery.

Main Outcome Measures

At median 13 and 35 weeks of gestation, participants completed 3-month semi-quantitative food frequency questionnaires to calculate the Healthy Eating Index 2015 and Alternative Healthy Eating Index 2010, excluding alcohol. Change in diet quality was calculated by subtracting diet quality scores at 13 weeks from those at 35 weeks. Gestational age- and prepregnancy body mass index (BMI)–specific GWG z scores were calculated using weight before pregnancy and at a median 38 weeks of gestation along with an international reference chart. Using prepregnancy BMI (calculated as kg / m²), women were classified as having underweight or healthy weight (BMI <25), overweight (BMI 25–29.9), or obesity (BMI ≥30).

Statistical Analyses Performed

Covariate-adjusted linear regression models accounting for total energy intake evaluated associations of improvement in Healthy Eating Index 2015 or Alternative Healthy Eating Index 2010 across pregnancy with GWG z scores. Differences by prepregnancy BMI were also explored.

Results

In women with obesity (who drove overall associations), each 10-point improvement in Healthy Eating Index 2015 across pregnancy was associated with −0.55 (95% CI, −0.82 to −0.27) lower GWG z scores due to decreased refined grains intake and increased seafood and plant proteins intake. A similar relationship was observed when considering the Alternative Healthy Eating Index 2010 (β = −0.48; 95% CI, −0.81 to −0.15) due to higher nuts and legumes intake.

Conclusions

Accounting for energy intake, diet quality improvement across pregnancy was associated with lower GWG z scores, particularly in women with obesity. Future studies may consider the implications of these findings for maternal and child health.

Appropriate Gestational Weight Gain (GWG) supports fetal growth and development,¹ and both excessive and insufficient GWG can result in adverse health outcomes for the birthing parent and child.² Specifically, inadequate GWG is associated with an increased risk of adverse birth outcomes, such as preterm birth, whereas excessive GWG is linked to higher risk of developing gestational diabetes, hypertension or pre-eclampsia, and delivering a large for gestational age infant.^3,4 In addition, excessive GWG has been linked to postpartum depression and postpartum weight retention for the birthing parent, which can have long-term health implications.^5,6 Therefore, to protect the health of both the child and birthing parent, it is important to identify potential modifiable risk factors associated with excessive and/or insufficient GWG.

Although increased food intake in response to the higher energy needs of pregnancy contributes in part to GWG,^7–9 diet quality–independent of energy intake–may also be a critical determinant of GWG.^10,11 Specifically, a large systematic review of randomized controlled trials (RCTs) found that low glycemic load diets (with insufficient evidence for other nutritional interventions) resulted in marked decreases in GWG,¹² indicating that diet composition during pregnancy also contributes to GWG. Beyond RCTs, which have generally intervened on limited aspects of the diet (eg, glycemic load as described above), recent observational cohort studies have evaluated associations of overall diet quality or components of whole diets with GWG using established diet quality indices that reflect dietary patterns.^13–15 Two such indices include the Healthy Eating Index (HEI) and the Alternative Healthy Eating Index (AHEI), which have been used previously for assessing the roles of dietary patterns and whole diets in pregnancy.^16–18 However, the few studies evaluating associations of these indices with GWG have been somewhat mixed, likely because they assessed diet at different timepoints in pregnancy, and typically only once.^{14,15,19–21} For example, a study of 480 Malaysian participants reported that a higher HEI score (adapted for Malaysians) in the second trimester was associated with a lower risk of inadequate GWG.²⁰ Conversely, a study of 500 US participants reported no associations of diet quality across pregnancy (assessed in the third trimester using the HEI-2015) with GWG,²¹ and a study of 457 US Hispanic and Latino participants reported no associations of diet quality prepregnancy (using the AHEI-2010) with GWG.²² Findings pertaining to which dietary components are the most important drivers of several diet quality indices have been similarly mixed. For example, some studies identified that higher intake of greens and beans (as HEI-2010 or HEI-2015 components) was associated with lower GWG,^23,24 another study reported that higher grain and cereal intakes (within the Malaysian HEI) were associated with higher GWG,²⁰ and a different study reported that lower consumption of oils (as part of the HEI-2005) was associated with excessive GWG.²⁵ Although prior research supports the role of diet quality (beyond dietary energy density) in GWG, additional studies are needed to contribute to the currently limited and mixed literature.

To provide novel insights into the role of diet quality in GWG, the 2 primary objectives of this observational study were to examine how shifts in diet quality across pregnancy (evaluated via the HEI-2015 and AHEI-2010) contribute to GWG and to consider differences in these associations by prepregnancy body mass index (BMI, calculated as kg / m²). Our hypothesis was that improvement in diet quality across pregnancy (as defined by higher mid-to-late than early pregnancy HEI-2015 and AHEI-2010 scores) would be associated with lower GWG. Because GWG recommendations are prepregnancy BMI-specific^1,26 and, as some prior studies found that prepregnancy BMI moderates associations between diet quality and GWG,^14,15,20 it was also hypothesized that findings would differ by prepregnancy BMI. As a secondary objective, to compare these findings to prior observational studies, we also explored associations of diet quality at 2 individual timepoints with GWG. Our secondary hypothesis was that associations of diet quality with GWG would differ depending on the timepoint of diet assessment.

METHODS

Illinois Kids Development Study Recruitment and Enrollment

In this secondary data analysis, we used information collected from a subset of pregnant individuals who participated in the Illinois Kids Development Study (I-KIDS), a prospective pregnancy cohort. All participants in I-KIDS self-identified as being cis-women (ie, assigned female at birth and identifying as a woman) at the time of study enrollment and throughout pregnancy and postpartum follow-up. Therefore, all participants are referred to as pregnant women, individuals, or participants throughout the article. I-KIDS participants were recruited from 2 local obstetric clinics in Champaign-Urbana, IL, at their first prenatal care appointment. I-KIDS was designed to evaluate the impacts of prenatal chemical exposures on infant neurodevelopment. Findings related to the primary study hypotheses have been published extensively.^27–30 Details about recruitment have been described previously.^31–33 Briefly, women were eligible to participate if they were ≤15 weeks pregnant, aged 18 to 40 years, fluent in English, carrying a low-risk singleton pregnancy as determined by their physician, and able and willing to complete most study visits. The current analytic sample includes a subset of 380 pregnant women who enrolled between 2013 and 2018 and remained in the study through the birth of their infant (Figure 1). These women had available data on early and mid-to-late pregnancy diet, GWG through late pregnancy, all relevant covariates (outlined in the collection of maternal sociodemographic, lifestyle, and health information section), and plausible values for daily energy intake (500–5000 kcal/d).^34–36 This secondary data analysis had 80% power to detect a difference in our outcome (GWG z score, as described later) of ≥0.145 points, with a minimum sample size of 378 and type I error probability of .05. All women provided written informed consent to participate in I-KIDS and consented to participate in research outside the original aims of the study related to chemical exposures. The study was approved by the Institutional Review Board at the University of Illinois.

Figure 1.

Derivation of analytic samples for evaluating associations of maternal diet quality with gestational weight gain (GWG) z scores. The chart presents sample sizes for all analyses evaluating associations of maternal diet with GWG z scores using adjusted linear regression models. Extreme energy intake values for daily energy intake were <500 kcal/d and >5000 kcal/d. All women were enrolled between 2013 and 2018. ^aMissingness in diet data is primarily due to the fact that the study is still ongoing and has not submitted the next batch of food frequency questionnaires for analysis.

Collection of Maternal Sociodemographic, Lifestyle, and Health Information

At the first study visit, which took place at a median 13 weeks of gestation (minimum of 9 weeks, maximum of 15 weeks), women completed an interviewer-administered questionnaire to ascertain their sociodemographic, lifestyle, and health characteristics.³⁷ Women reported their current age, race, ethnicity, educational attainment, marital status, annual household income, parity, pregnancy intention, perceived health, and smoking status from conception through a median 13 weeks. To calculate prepregnancy BMI, women self-reported their prepregnancy weight and height. To ascertain early pregnancy stress status, women completed the Perceived Stress Scale, which is a 10-item questionnaire (scored out of 40 total points) that asks about thoughts and feelings during the past month, with higher scores indicating higher perceived stress.^38,39

Collection of Dietary Information and Calculation of Diet Quality Indices

At a median 13 weeks (minimum 9 weeks, maximum 15 weeks) and 35 weeks (minimum 33, maximum 38 weeks) of gestation, women completed semi-quantitative food frequency questionnaires (FFQs). The FFQ used in this study is a modification of the standard 2005 Block FFQ (NutritionQuest),⁴⁰ which itself has not been validated, but is an update to the Block 98 FFQ (validated in nonpregnant individuals),⁴¹ with more recent dietary data and an expanded food list. The 2005 Block FFQ was developed from the National Health and Nutrition Examination Survey 1999–2002 dietary recall data,^42,43 and consists of 110 food and beverage items. It was modified for use in pregnancy by adding a list of fish and seafood items.⁴⁴ The core nutrient values for the 2005 FFQ analysis database were developed from the US Department of Agriculture Food and Nutrient Database for Dietary Studies,⁴⁵ version 2.0. The food group values were developed through an analysis of the US Department of Agriculture Food Patterns Equivalents Database,⁴⁶ version 2005–2006. Briefly, the FFQ asked about dietary intakes during the previous 3 months, and information collected at a median 13 and 35 weeks of gestation reflects early (ie, from conception through the beginning of the second trimester) and mid-to-late (ie, from mid-second through third trimester) maternal pregnancy diets, respectively, as has been reported previously.³⁷ The FFQ provided 9 intake frequency options and 2 to 4 portion options per item. Participants were provided with visual examples depicting average portion sizes for reference on paper. Women completed a pencil-and-paper version of the survey at home and returned it to the researcher at a follow-up visit.

FFQ data were used to calculate early and mid-to-late pregnancy HEI-2015 using the simple algorithm method as defined and described by the National Cancer Institute,⁴⁷ and AHEI-2010, which has been used previously in pregnant populations.^18,48,49 The HEI-2015 is a 13-component index (scored out of 100 points) that measures adherence to the 2015–2020 Dietary Guidelines for Americans (DGA).^50,51 The index is made up of 9 adequacy (ie, total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, and fatty acids) and 4 moderation (ie, refined grains, sodium, added sugars, and saturated fats) dietary components that are each scored from 0 to 5 points or 0 to 10 points. Higher scores for the adequacy dietary components reflect higher consumption, whereas the moderation dietary components are reverse-scored so that higher scores represent lower consumption. Thus, a higher HEI-2015 total score reflects better overall diet quality and adherence to the 2015–2020 DGA.^50,51 The AHEI-2010 is an 11-component index (scored out of 110 points) based on foods and nutrients predictive of chronic disease risk and mortality.⁵² The AHEI-2010 includes 6 adequacy (ie, vegetables, fruit, whole grains, nuts and legumes, n-3 fatty acids [ie, eicosapentaenoic acid and docosahexaenoic acid], and polyunsaturated fatty acids) and 5 moderation (ie, sugar-sweetened beverages [SSBs] and fruit juice, red or processed meat, trans fat, sodium, and alcohol) dietary components. Each component is scored from 0 to 10 points. Higher scores for adequacy intake dietary components represent higher intake, whereas the moderation intake dietary components are reverse-scored so that higher scores reflect lower consumption of these foods and nutrients. Due to recommendations that pregnant women should avoid consuming alcohol during pregnancy⁵³ and because AHEI-2010 considers moderate alcohol intake as beneficial to health, the alcohol component from the total AHEI-2010 score was removed. Therefore, in this study, AHEI-2010 was scored out of 100 points, and higher AHEI-2010 total scores reflect better overall diet quality. Both HEI-2015 and AHEI-2010 were considered in our study because, in a prior study, AHEI-2010 was a more precise predictor of gestational length than HEI-2015,³⁷ highlighting the need to understand differences (if they exist) in how each index predicts other pregnancy health outcomes, including GWG.

Collection and Calculation of GWG z Scores Through Late Pregnancy

Medical record abstraction for I-KIDS is ongoing, so the approach for calculating GWG in our study was based on methods described in a prior publication.³¹ Briefly, at delivery, women reported their measured weight (in lb) from their last obstetric appointment before delivery, which took place at a median of 38 weeks of gestation (minimum 32 weeks, maximum 42 weeks). Self-reported total GWG has been found to greatly agree with measured weight⁵⁴; in our study, women had a clinical weight measure approximately 2 weeks before the interview and were asked to recall their weight measured at that particular appointment, which likely improved the accuracy of self-report. To calculate GWG through late pregnancy (in kg), self-reported prepregnancy weight was subtracted from the weight measured at a median of 38 weeks of gestation. In the few cases where weight at median 38 weeks was not available (n = 21), the weight reported at the 35-week research visit at which women reported their weight from their most recent obstetric visit (occurring at median 34 weeks of gestation; minimum 32 weeks, maximum 37 weeks) was used. Given that GWG correlates with pregnancy duration,²⁶ gestational age– and prepregnancy BMI–specific GWG z scores were calculated using an international reference chart of pregnant women from Europe, Oceania, and North America.⁵⁵

Statistical Analysis

Covariate Selection and Presentation of Descriptive Statistics.

Potential confounding variables were identified using a priori consideration and prior literature,^37,56 which informed a directed acyclic graph (Figure 2) to guide the minimum sufficient adjustment set of covariates needed to ascertain the proposed relationships.^57,58 This study also investigated whether these potential confounders were separately associated with maternal diet and GWG z scores, and evaluated correlations (Pearson for continuous and polychoric for categorical variables) between all covariates to check for potential multicollinearity, but none of the included covariates were strongly correlated with each other (r ≤ 0.3). Therefore, all statistical models accounted for race and ethnicity (non-Hispanic White, others), perceived health (excellent, good, fair/poor), and smoking in the first trimester (yes, no, missing) as categorical variables with reference groups indicated in Table 1, annual household income as an ordinal variable, and prepregnancy BMI, perceived stress score, and total daily energy intake as continuous variables. These represent the following latent constructs as denoted in Figure 2: socioeconomic status, structural and cultural factors, mental health, lifestyle, general health, and energetics. Due to small sample sizes, women who self-reported being Hispanic, Black, Asian, Native Hawaiian or other Pacific Islander, American Indian or Alaska Native, multiracial, or other were classified into 1 category representing racial and ethnic minority groups. Marital status, employment, and educational attainment were not included in the final models due to strong correlations with annual household income. Maternal age was not included in models because it was not associated with GWG, and its inclusion as a covariate in models did not impact our findings (data not shown). Models evaluating “change in” diet quality accounted for mean early and mid-to-late pregnancy energy intakes, whereas those evaluating diet in early and mid-to-late pregnancy accounted for energy intake at each of those timepoints. Categorical variables were presented as n (%), and continuous variables were presented as median (25^th, 75^th percentiles).

Figure 2.

Directed acyclic graph for associations between maternal diet quality and gestational weight gain. Maternal diet quality is the exposure (green/black circle), and gestational weight gain is the outcome (blue/black circle). Red circles indicate variables associated with both the exposure and outcome that were included in the final covariate-adjusted statistical models as the minimum sufficient adjustment set. Gray circles represent latent variables.

Table 1.

Characteristics of the Illinois Kids Development Study analytic sample (n = 380)

Characteristic	Data
Age, y, median (25th, 75th percentile)	31.0 (28.0, 33.0)
Race/ethnicity,a n (%)
Non-Hispanic White [Reference]	315 (82.9)
Otherb	65 (17.1)
Educational attainment, n (%)
Some college or less	53 (14.0)
College graduate or higher	327 (86.0)
Annual household income,ac n (%)
<$60 000	102 (26.8)
≥$60 000	278 (73.2)
Marital status, n (%)
Married	351 (92.4)
Living as married	18 (4.7)
Single	11 (2.9)
Parity, n (%)
No children	195 (51.3)
1 or more children	185 (48.7)
Perceived health,a n (%)
Excellent [Reference]	143 (37.6)
Good	218 (57.4)
Fair/poor	19 (5.0)
Early pregnancy perceived stress score,ad median (25th, 75th percentile)	11.0 (7.0, 16.0)
Prepregnancy BMI,ae median (25th, 75th percentile)	24.7 (21.9, 29.1)
Prepregnancy BMI, n (%)
Underweightf (<18.5)	9 (2.4)
Healthy weight (18.5–24.9)	194 (51.0)
Have overweight (25.0–29.9)	92 (24.2)
Have obesity (≥30)	85 (22.4)
Smoking in the first trimester,a n (%)
No [Reference]	336 (88.4)
Yes	16 (4.2)
Unknown	28 (7.4)
Sex of fetus, n (%)
Female	189 (49.7)
Male	191 (50.3)
Total daily energy, kcal, intake,a median (25th, 75th percentile)
Early pregnancy	1541 (1165, 1880)
Mid-to-late pregnancy	1586 (1259, 2002)
Across pregnancyg	1588 (1248, 1906)
Gestational weight gain through late pregnancy, kg, median (25th, 75th percentile)	15.0 (10.9, 18.6)
Gestational weight gain (z score), median (25th, 75th percentile)	0.4 (−0.3, 1.1)

^aIncluded in final statistical models as covariates.

^bIncludes Hispanic (n = 7), Black (n = 18), Asian (n = 22), American Indian or Alaska Native (n = 1), Native Hawaiian or other Pacific Islander (n = 0), multiracial (n = 12), and other (n = 5).

^cAnnual household income was included as an ordinal variable with 11 categories.

^dHighest possible score is 40, with higher scores indicating higher perceived stress.

^eBMI = body mass index; calculated as kg / m².

^fWomen with underweight (n = 9) were included with women with healthy weight in all BMI-specific analyses due to small sample size.

^gCalculated by taking the mean of early and mid-to-late pregnancy energy intakes.

Associations of Change in Maternal Diet Quality or Diet Quality at Individual Timepoints With GWG z Scores.

To understand whether changes in maternal diet quality or individual components across pregnancy were associated with GWG, separate linear covariate-adjusted regression models were specified to evaluate associations of changes in HEI-2015 total score, each of the 13 HEI-2015 components, AHEI-2010 total score, and each of the 10 AHEI-2010 components with GWG z scores. This resulted in 14 total linear regression models for the HEI-2015 and 11 total linear regression models for the AHEI-2010 (at each timepoint and for each BMI category). Based on prior recommendations,⁵⁹ and given the exploratory nature of this study, adjustments for multiple comparisons were not conducted. Change in maternal diet scores for the HEI-2015 and AHEI-2010 total scores and individual components were calculated as follows: mid-to-late pregnancy total score or individual component minus the corresponding early pregnancy total score or individual component. A positive change indicates improvement, whereas a negative change indicates a decline in maternal diet quality from early to mid-to-late pregnancy. Maternal diet and GWG z scores were evaluated as continuous variables, and neither diet nor GWG z scores were transformed after checking distributions and regression models for nonconstant residual variance to ensure model assumptions were met. To place the current study within the context of prior research and to understand gestational timepoint-specific associations of maternal diet with GWG z scores, maternal diet in both early (n = 396) and mid-to-late pregnancy (n = 383) were also separately evaluated using the same modeling approach described above. Women in these analyses did not differ from those in the main analytic sample (n = 380; data not shown).

Prepregnancy BMI-Specific Associations of Maternal Diet Quality With GWG z Scores.

When evaluating risk factors of GWG, it is recommended that studies include prepregnancy BMI as an effect modifier.²⁶ Therefore, in addition to evaluating overall relationships, this study also evaluated differences in associations of HEI-2015 and AHEI-2010 total scores and individual components with GWG z scores by prepregnancy BMI. To obtain the prepregnancy BMI-specific effect estimates and 95% CIs in linear regression models, a multiplicative interaction term between diet and prepregnancy BMI was included. Prepregnancy BMI was categorized into the following 3 clinical categories⁶⁰: underweight or healthy weight (BMI <25.0), overweight (BMI 25.0–29.9), and obesity (BMI ≥30.0).

Sensitivity Analyses

Because associations between improvement in maternal diet quality and GWG z scores can be influenced by maternal diet quality at baseline, a sensitivity analysis evaluating differences in the main associations by tertiles of early pregnancy diet quality was conducted. Covariate-adjusted linear regression models were specified that also included an interaction term between change in HEI-2015 or AHEI-2010 and early pregnancy diet quality (in tertiles). This sensitivity analysis was conducted in the full sample and stratified by prepregnancy BMI as discussed in the prepregnancy BMI-specific associations of maternal diet quality with GWG z scores section. An additional sensitivity analysis was conducted in which models also accounted for the HEI-2015 total score (for the individual HEI-2015 components) or AHEI-2010 total score (for the individual AHEI-2010 components) that excluded the component that was being evaluated.⁶¹ For example, for associations of the HEI-2015 total vegetables component in early pregnancy with GWG z scores, the model also accounted for the early pregnancy HEI-2015 total score minus the early pregnancy score for the total vegetables component.³⁷

Interpretation of Findings and Identifying Significant Results

The resulting β-estimates and 95% CIs from all linear regression models were scaled to represent the z score change in GWG for each 10.0-point improvement (across pregnancy) or increase (at each timepoint) in the total score (both indices), a 1.0-point improvement or increase in all AHEI-2010 components along with whole grains, dairy, fatty acids, refined grains, sodium, added sugars, and saturated fats from the HEI-2015, and a 0.5-point improvement or increase in total fruits, whole fruits, total vegetables, greens and beans, total protein foods, and seafood and plant proteins from the HEI-2015. The resulting β-estimates and 95% CIs for a z score change in GWG can be converted (considering the gestational age at late pregnancy weight and prepregnancy BMI) to represent a kilogram change.²⁶ For example, a 0.2-point z score change in weight gain through 38 weeks of gestation can be interpreted as a 0.82-kg change for women with underweight; 0.84-kg change for women with healthy weight; 1.09-kg change for women with overweight; and finally, 1.18-kg, 1.28-kg, and 1.44-kg changes for women with class I, class II, and class III obesity, respectively.³¹ All linear regression analyses were performed using SAS software, version 9.4, using the PROC GLM procedure.⁶² Associations were considered statistically significant at P ≤ .05.

RESULTS

Characteristics of the Analytic Sample

Women in the analytic sample had a median age of 31 years and a median prepregnancy BMI of 24.7 (with approximately one-half [53%] classified as having underweight or healthy weight before pregnancy) (Table 1). Most women were non-Hispanic White (83%), college-educated (86%), and had annual household incomes ≥$60 000 (73%). Few women smoked (4%), and most perceived their health as excellent or good (95%) in the first trimester. Median (25^th, 75^th percentile) GWG and GWG z score through a median 38 weeks of gestation were 15.0 (10.9, 18.6) kg and 0.4 (−0.3, 1.1), respectively. Women had a median (25^th, 75^th percentile) daily energy intake of 1588 (1248, 1906) kcal across pregnancy (Table 1).

Distribution of HEI-2015 and AHEI-2010 Total and Individual Scores

The median (25^th, 75^th percentile) changes in HEI-2015 and AHEI-2010 scores from early to mid-to-late gestation were 0.5 (−4.5, 5.6) points and 0.3 (−4.0, 5.6) points, respectively, due to approximately equal numbers of women whose diet scores declined and improved across pregnancy (Figures 3 and 4, available at www.jandonline.org). Median (25^th, 75^th percentile) HEI-2015 and AHEI-2010 scores in early pregnancy were 65.0 (58.7, 71.1) points and 51.9 (44.5, 60.0) points, respectively, and in mid-to-late pregnancy were 65.3 (58.5, 72.1) points and 53.0 (45.5, 61.5) points, respectively (Table 2). Distributions of individual HEI-2015 and AHEI-2010 component scores are presented in Table 2.

Table 2.

Distributions of HEI-2015^a and AHEI-2010^b total scores and individual components in Illinois Kids Development Study participants

Diet quality index and component	Early pregnancy diet (n = 396)	Mid-to-late pregnancy diet (n = 383)	Changec in pregnancy diet (n = 380)
HEI-2015
Total score (maximum 100 points)	65.0 (58.7, 71.1)	65.3 (58.5, 72.1)	0.5 (−4.5, 5.6)
Adequacy (higher score, higher intake)
Total fruits (maximum 5 points)	5.0 (3.4, 5.0)	5.0 (3.5, 5.0)	0.0 (−0.1, 0.4)
Whole fruits (maximum 5 points)	5.0 (4.6, 5.0)	5.0 (4.8, 5.0)	0.0 (0.0, 0.0)
Total vegetables (maximum 5 points)	4.0 (3.0, 5.0)	3.8 (2.9, 5.0)	0.0 (−0.8, 0.4)
Greens and beans (maximum 5 points)	4.9 (2.8, 5.0)	4.4 (2.6, 5.0)	0.0 (−0.8, 0.2)
Whole grains (maximum 10 points)	4.0 (2.4, 6.5)	3.9 (2.5, 6.1)	—0.1 (−1.5, 1.5)
Dairy (maximum 10 points)	7.0 (5.4, 9.6)	7.3 (5.8, 9.6)	0.0 (−0.7, 1.7)
Total protein foods (maximum 5 points)	4.6 (3.6, 5.0)	4.4 (3.4, 5.0)	0.0 (−0.6, 0.3)
Seafood and plant proteins (maximum 5 points)	4.5 (2.7, 5.0)	4.7 (2.7, 5.0)	0.0 (−0.5, 0.4)
Fatty acids (maximum 10 points)	4.5 (3.0, 6.2)	4.2 (2.8, 6.2)	0.0 (−1.6, 1.3)
Moderation (higher score, lower intake)
Refined grains (maximum 10 points)	7.8 (6.1, 9.6)	8.4 (6.2, 9.9)	0.0 (−1.0, 1.7)
Sodium (maximum 10 points)	3.9 (2.4, 5.7)	4.8 (3.2, 6.3)	0.6 (−0.9, 2.4)
Added sugars (maximum 10 points)	8.7 (6.9, 9.8)	8.3 (6.7, 9.5)	—0.2 (−1.2, 0.5)
Saturated fats (maximum 10 points)	4.5 (2.1, 6.4)	4.4 (1.8, 6.3)	0.0 (−1.8, 1.4)
AHEI-2010
Total score (maximum 100 points)	51.9 (44.5, 60.0)	53.0 (45.5, 61.5)	0.3 (−4.0, 5.5)
Adequacy (higher score, higher intake)
Vegetables (maximum 10 points)	4.4 (3.0, 6.4)	4.5 (2.8, 6.6)	0.0 (−1.3, 1.4)
Fruit (maximum 10 points)	5.0 (2.7, 7.8)	5.9 (3.0, 8.5)	0.3 (−0.9, 1.9)
Whole grains (maximum 10 points)	1.9 (1.0, 3.1)	2.0 (1.2, 3.1)	0.1 (−0.8, 0.9)
Nuts and legumes (maximum 10 points)	5.2 (2.4, 10.0)	6.6 (3.0, 10.0)	0.0 (−0.8, 2.2)
EPAd + DHAe (maximum 10 points)	2.0 (1.1, 4.1)	1.9 (1.0, 3.4)	—0.1 (−0.9, 0.5)
PUFAf (maximum 10 points)	6.6 (5.6, 8.0)	6.7 (5.6, 7.7)	0.0 (−1.1, 1.0)
Moderation (higher score, lower intake)
SSBsg and fruit juice (maximum 10 points)	3.5 (0.0, 7.8)	3.3 (0.0, 7.5)	0.0 (−1.3, 1.0)
Red/processed meat (maximum 10 points)	7.5 (5.9, 8.5)	7.2 (5.6, 8.5)	—0.1 (−1.0, 0.7)
Trans fat (maximum 10 points)	8.0 (7.3, 8.5)	7.9 (7.2, 8.5)	0.0 (−0.6, 0.5)
Sodium (maximum 10 points)	6.6 (3.9, 8.7)	6.5 (3.8, 8.6)	0.0 (−1.9, 1.5)

^aHEI-2015 = Healthy Eating Index 2015.

^bAHEI-2010 = Alternative Healthy Eating Index 2010.

^cChange in pregnancy diet was calculated by subtracting early pregnancy scores from mid-to-late pregnancy scores.

^dEPA = eicosapentaenoic acid.

^eDHA = docosahexaenoic acid.

^fPUFA = polyunsaturated fatty acids as a percentage of total kilocalories.

^gSSB = sugar-sweetened beverage.

Associations of Improvement in Diet Quality Across Pregnancy and Timepoint-Specific Diet Qualities With GWG z Scores

Improvement in HEI-2015 total score across pregnancy was associated with lower GWG z scores, such that each 10-point improvement in diet quality from early to mid-to-late pregnancy was associated with −0.15 (95% CI, −0.29 to −0.01) points lower GWG z scores (Figure 5A, Table 3, available at www.jandonline.org). In contrast, the AHEI-2010 total score across pregnancy was not associated with GWG z scores (Figure 5B, Table 3, available at www.jandonline.org). However, individually, each 1-point increase in SSBs and fruit juice (AHEI-2010) score across pregnancy (reflecting decreased intake from early to mid-to-late gestation) was significantly associated with −0.03 (95% CI, −0.06 to 0.00) points lower GWG z scores (Figure 5B, Table 3, available at www.jandonline.org).

Figure 5.

Overall and prepregnancy body mass index (BMI)-specific associations of change in (A) HEI-2015 and (B) AHEI-2010 from early to mid-to-late gestation with gestational weight gain (GWG) z scores. Data are presented as the change in GWG z score for each 10-point improvement in the total score (both indices) or 1.0 point (all AHEI-2010 and most HEI-2015 components) or 0.5-point (some HEI-2015 components) improvement of individual component scores from early to mid-to-late pregnancy. For HEI-2015, higher component scores indicate higher intake, except for the following reverse-coded components (where a higher score indicates lower intake): refined grains, sodium, added sugars, and saturated fats. Similarly, for AHEI-2010, higher component scores indicate higher intake, except for the following reverse-coded components (where a higher score indicates lower intake): sugar-sweetened beverages and fruit juice, red/processed meat, trans fat, and sodium. Change in diet quality was calculated as follows: mid-to-late pregnancy total score or individual component minus early pregnancy total score or individual component, respectively. Linear regression models accounted for race/ethnicity, smoking in the first trimester, annual household income, prepregnancy body mass index, average total daily energy intake across pregnancy, perceived health, and early pregnancy perceived stress. The numeric data for these findings are found in Table 3 (available at www.jandonline.org). ^aHEI-2015 = Healthy Eating Index-2015. ^bAHEI-2010 = Alternative Healthy Eating Index-2010. ^cEPA = eicosapentaenoic acid. ^dDHA = docosahexaenoic acid. ^ePUFA = polyunsaturated fatty acids as a percentage of total kilocalories. ^fSSB = sugar-sweetened beverage. n = 380. *P ≤ .05. **P < .01. ***P < .001.

Individually, neither early nor mid-to-late-pregnancy HEI-2015 total scores were associated with GWG z scores (Table 4, available at www.jandonline.org). However, each 1-point increase in early-pregnancy consumption of whole grains (HEI-2015) was significantly associated with 0.04 (95% CI, 0.00 to 0.08) points higher GWG z scores (Table 4, available at www.jandonline.org). Similarly, neither early nor mid-to-late-pregnancy AHEI-2010 total scores were associated with GWG z scores (Table 5, available at www.jandonline.org), but each 1-point increase in the intake of SSBs and fruit juice (reflecting lower intake) in mid-to-late pregnancy was significantly associated with −0.03 (95% CI, −0.06 to 0.00) points lower GWG z scores (Table 5, available at www.jandonline.org).

Prepregnancy BMI-Specific Associations of Maternal Diet Quality With GWG z Scores

When differences in associations by prepregnancy BMI were considered, improvements in HEI-2015 and AHEI-2010 total scores across pregnancy were not associated with GWG z scores in women with underweight or healthy weight, and in women with overweight. However, in women with underweight or healthy weight, each 1-point increase in red or processed meat (AHEI-2010; reflecting decreased intake from early to mid-to-late pregnancy) was associated with 0.07 (95% CI, 0.00 to 0.14) points higher GWG z scores, and in women with overweight, each 1-point increase in SSBs and fruit juice (AHEI-2010; reflecting decreased intake from early- to mid-to-late pregnancy) was associated with −0.07 (95% CI, −0.14 to −0.01) points lower GWG z scores (Figure 5, Table 3, available at www.jandonline.org). In women with obesity, each 10-point improvement in HEI-2015 and AHEI-2010 total scores across pregnancy was associated with −0.55 (95% CI, −0.82 to −0.27) points and −0.48 (95% CI, −0.81 to −0.15) points lower GWG z scores, respectively. When assessing individual dietary components that drove these associations in women with obesity, increased intakes from early- to mid-to-late- pregnancy of seafood and plant proteins (HEI-2015; β = −0.12; 95% CI, −0.22 to −0.01) and nuts and legumes (AHEI-2010; β = −0.09; 95% CI, −0.16 to −0.01), but decreased intake of refined grains (HEI-2015; β = −0.11; 95% CI, −0.21 to −0.01), were significantly associated with lower GWG z scores (Figure 5, Table 3, available at www.jandonline.org).

When considering timepoint-specific associations, HEI-2015 and AHEI-2010 total scores in both early and mid-to-late pregnancy were not associated with GWG z scores in women with underweight or healthy weight or overweight. However, in women with overweight, each 1-point increase in early-pregnancy whole grains (HEI-2015), whole grains (AHEI-2010), and nuts and legumes (AHEI-2010) was associated with 0.09 (95% CI, 0.02 to 0.16) points, 0.13 (95% CI, 0.01 to 0.25) points, and 0.08 (95% CI, 0.02 to 0.15) points higher GWG z scores, respectively (Tables 4 and 5, available at www.jandonline.org). In addition, in women with overweight, each 0.5-point increase in mid-to-late pregnancy total fruits (HEI-2015: β = 0.10; 95% CI, 0.01 to 0.19) and seafood and plant proteins (HEI-2015: β = 0.09; 95% CI, 0.01 to 0.16), and each 1-point increase in mid-to-late pregnancy whole grains (AHEI-2010: β = 0.15; 95% CI, 0.01 to 0.30) and nuts and legumes (AHEI-2010: β = 0.11; 95% CI, 0.05 to 0.17) was associated with higher GWG z scores, but each 1-point increase in mid-to-late pregnancy SSBs and fruit juice (AHEI-2010: reflecting lower intake; β = −0.06; 95% CI, −0.11 to 0.00) was associated with lower GWG z scores. In women with obesity, early-pregnancy HEI-2015 and AHEI-2010 total scores were not associated with GWG z scores, however, each 1-point increase in early-pregnancy sodium (HEI-2015; reflecting lower intake) was associated with 0.10 (95% CI, 0.02 to 0.19) points higher GWG z scores and each 1-point increase in early-pregnancy polyunsaturated fatty acids (AHEI-2010) was associated with −0.13 (95% CI, −0.26 to −0.01) points lower GWG z scores (Tables 4 and 5, available at www.jandonline.org). In women with obesity, mid-to-late pregnancy HEI-2015 total score was associated with −0.27 (95% CI, −0.50 to −0.04) points lower GWG z scores, with lower consumption of added sugars being the primary significant driver of this association (β = −0.09; 95% CI, −0.16 to −0.01) (Table 4, available at www.jandonline.org). Although the AHEI-2010 total score in mid-to-late pregnancy was not associated with GWG z scores in women with obesity, lower consumption of SSBs and fruit juice was significantly associated with lower GWG z scores (β = −0.06; 95% CI, −0.12 to 0.00) (Table 5, available at www.jandonline.org).

Sensitivity Analyses

As presented in Table 6 (available at www.jandonline.org) findings from sensitivity analyses, in which associations of improvement in diet quality with GWG z scores were stratified by early pregnancy diet quality, were generally consistent with the results from the main analyses. Associations of improvement in HEI-2015 with GWG z scores in women with obesity appeared to be stronger in women with healthier baseline diets compared with those with less healthy baseline diets, but this analysis was likely underpowered, especially when evaluating BMI-specific relationships. Associations of individual components with GWG z scores were largely unchanged after additional adjustments for diet quality, subtracting the dietary component of interest, in models evaluating improvement in diet and diet at individual timepoints (Tables 7–9, available at www.jandonline.org).

DISCUSSION

Overall Summary of Findings

In our study, in women with obesity, improvements in diet quality across pregnancy were associated with lower GWG z scores. This was independent of energy intake. These findings were primarily driven by increased consumption of seafood and plant proteins and nuts and legumes, as well as a decrease in the consumption of refined grains across pregnancy. Given the homogeneous nature of this study sample, additional research is warranted to understand whether improvement in diet quality across pregnancy supports healthy GWG, especially in women with obesity.

Improved Diet Quality Across Pregnancy Was Associated With Lower GWG in Women With Obesity

To our knowledge, no prior studies examined shifts in diet quality across pregnancy as measured by the HEI-2015 or AHEI-2010. We reported that improvement in diet quality from early to mid-to-late pregnancy, measured by HEI-2015, was associated with lower GWG in the full sample, which was driven by women with obesity, whereas improvement in AHEI-2010 was associated with lower GWG only in women with obesity. Interestingly, dietary contributors to lower GWG among women with obesity differed depending on whether diet was assessed in mid-to-late pregnancy or calculated as an improvement in diet quality from early to mid-to-late pregnancy. Specifically, when diet was assessed in mid-to-late pregnancy, lower intake of added sugars appeared to be the primary contributor to lower GWG. In contrast, the association between improvement in diet quality across pregnancy and lower GWG was driven primarily by increased intakes of higher adequacy components, including seafood and plant proteins (HEI-2015) and nuts and legumes (AHEI-2010). These findings highlight notable changes in diet that women make across pregnancy. Future studies are needed to build on these findings that increasing consumption of nutrient-dense foods (rather than simply eliminating or restricting certain foods) could limit GWG in women with obesity. Overall, prior RCTs have generally focused on low glycemic load diets as an approach for decreasing GWG.¹² Our study results point to additional considerations for future studies evaluating the roles of diet in lowering GWG, including increased consumption of foods containing macro- and micronutrients critical for supporting fetal development and other pregnancy outcomes.

Timepoint-Specific Associations of Diet Quality With GWG Differed by Prepregnancy BMI

To compare our study with prior research, relationships of diet quality at individual timepoints in pregnancy with GWG were also considered. In our study, diet quality in early or mid-to-late pregnancy was not associated with GWG, which is generally consistent with prior studies.^{10,13,24,63–65} In contrast to our study’s findings, in a recent study in California–Pregnancy Environment and Lifestyle Study (PETALS)–first-trimester diet quality (HEI-2010) was associated with a higher risk of excessive GWG during the second and third trimesters.¹⁴ However, unlike our study, PETALS was composed of a more diverse sample of pregnant individuals, classified total GWG using the Institute of Medicine categories for adequate GWG,¹ and used an earlier version of the HEI (dichotomized into “good” and “bad” diet quality at the highest and lowest quartiles, respectively), which may also explain differences in findings.

Despite not observing any significant associations of diet quality with GWG in the overall sample, associations of mid-to-late pregnancy diet quality with GWG differed by maternal prepregnancy BMI. Specifically, in women with obesity, higher mid-to-late pregnancy diet quality (as evaluated by the HEI-2015) was associated with lower GWG. A small number of other studies have reported BMI-specific associations of diet quality with GWG, but findings have been mixed in terms of which BMI category is most responsive to diet. In contrast to our study’s findings, in a multi-ethnic sample of US individuals, better diet quality in the first trimester, as measured by the HEI-2010, was associated with excessive GWG in individuals without overweight or obesity.¹⁴ In individuals from Malaysia, higher Malaysian HEI total scores in the second and third trimesters of pregnancy were associated with reduced risk of inadequate GWG in underweight or healthy weight individuals, but were associated with an increased risk of excessive GWG in individuals with overweight and obesity.²⁰ Findings from our study suggest that in individuals with obesity, higher mid-to-late pregnancy diet quality, independent of energy density, may support lower GWG, but additional studies are needed to corroborate and understand the implications of these findings.

Possible Biological Underpinnings

In our study, all models evaluating associations of diet quality or individual components with GWG accounted for usual total daily energy intake, indicating that the observed shifts in GWG were driven by factors other than the diets’ inherent energy density. For example, in our study, whole-grain consumption was positively associated with GWG in women with overweight in both early and mid-to-late pregnancy. Prior studies have found that whole-grain consumption is associated with GWG, but without consideration of prepregnancy BMI.^20,35 Similarly, prior RCTs in pregnant individuals have found that higher fiber consumption supports insulin control and lowers lipid levels⁶⁶; whether this insulin-mediated mechanism explains findings in our study needs to be further explored. Furthermore, in our study, in women with obesity, lower consumption of added sugars in mid-to-late pregnancy contributed to the association between better diet quality and lower GWG. Similarly, in women with overweight and obesity, lower consumption of SSBs and fruit juice in mid-to-late pregnancy was associated with lower GWG. Interestingly, the literature surrounding sugar consumption and GWG is mixed. In Japanese individuals, dietary patterns that were higher in sugar consumption were associated with higher average GWG compared with dietary patterns that were lower in sugar.⁶⁷ In a study of individuals from Michigan that evaluated several individual diet components (rather than overall diet quality), higher added sugar intake was associated with a lower likelihood of excessive GWG among individuals with healthy weight.⁵⁶ Conversely, in a small study of individuals from Saginaw, MI, added sugars intake (from the HEI-2015) was not associated with GWG.²⁴ Substantially more studies are needed to understand the role of diet in the underlying physiology of GWG, especially in individuals with obesity, which would support the development of evidence-based dietary guidelines.

Strengths and Limitations

Our study has several limitations. First, the HEI-2015 and AHEI-2010 were developed to evaluate alignment to the 2015–2020 DGA (HEI-2015)^50,51 and chronic disease risk and mortality (AHEI-2010)⁵² in the general population. Second, a possible limitation is that women recalled their food intake over the previous 3 months via FFQs instead of other commonly used methods, like 24-hour recalls. Although optimal dietary intake data collection includes the use of both FFQs and 24-hour dietary recalls,^68,69 we used an FFQ in our study instead of the 24-hour recall method to balance staffing limitations and participant burden with data completeness, as has been outlined previously.⁷⁰ Despite the fact that this is the only diet measurement method used in our study, FFQs are most useful for assessing habitual dietary intake and were therefore selected in our study to reflect diet across many months of pregnancy.⁷¹ Third, we did not collect data on physical activity, which may be an important determinant of GWG. However, recent systematic reviews have highlighted a lack of strong consensus on the role of physical activity in GWG (particularly compared with diet),^12,72 so additional observational and experimental studies may be needed that consider both factors and their causal relationships with GWG. In addition, a directed acyclic graph was used to consider several other lifestyle factors that may align with physical activity, including energy intake. Fourth, most I-KIDS participants are non-Hispanic White women of relatively high socioeconomic status. GWG and diet quality in women from our study are either in line with (GWG) or slightly higher than (diet quality) those in the general population,^73–75 which may allow for generalizability of these findings to other pregnant populations with similar socioeconomic status profiles. Finally, because so few women (n = 9) had underweight in our study, they were combined with women who had healthy weight in pre-pregnancy BMI-specific analyses. Therefore, future studies are needed to understand the associations between diet quality and GWG in women with underweight.

Our study also has noteworthy strengths. First, our study contributes to the limited literature applying multiple diet-quality indices to evaluate health outcomes in pregnant individuals.^18,49 Considering both indices (ie, HEI-2015 and AHEI-2010) is an important strength, given that each index models potentially different dietary exposures. Second, our study is one of the few to evaluate diet quality at multiple timepoints in pregnancy, which is critical as nutrient needs change across pregnancy.⁷⁶ Importantly, we also considered shifts in pregnancy diet quality across pregnancy to provide more realistic information about dietary changes women make in pregnancy. The role of maternal baseline diet quality in the main associations was also considered, although findings should be interpreted with caution, given the small sample sizes, especially in BMI-specific analyses. However, these preliminary results suggest that baseline diet quality may be important, which will need to be followed up in future larger studies. Finally, we also evaluated differences in associations between diet quality and GWG by maternal prepregnancy BMI, and specifically considered women with overweight separately from those with obesity.²⁶

CONCLUSIONS

In a sample of primarily non-Hispanic White US pregnant women from the Midwest with relatively high socioeconomic status, improvement in diet quality from early to mid-to-late pregnancy was associated with lower GWG in women with obesity. In addition, higher overall diet quality in mid-to-late pregnancy was associated with lower GWG in women with obesity. Given the importance of GWG for the health of the birthing parent and child, additional studies should build on these findings and consider potential short- and long-term health implications of these observations.

Supplementary Material

Figures 3 and 4 and Tables 3, 4, 5, 6, 7, 8, and 9 are available at www.jandonline.org

RESEARCH SNAPSHOT.

Research Question:

Do shifts in diet quality across pregnancy contribute to gestational weight gain independent of energy intake?

Key Findings:

In this longitudinal cohort of primarily non-Hispanic White, midwestern US women, in women with obesity, improvement in diet quality from early to late pregnancy was associated with lower gestational weight gain through late pregnancy. Importantly, these findings were independent of energy intake.