Design and participants
The Pregnancy Eating Outcomes Study (PEAS) was a prospective observational study of women enrolled at ≤12 weeks gestation from two university-based obstetrics clinics in Chapel Hill, North Carolina from November 2014 through October 2016 and followed through one-year postpartum (39). The primary study aims were to examine the roles of reward-related eating, self-regulation, and home food availability on dietary intake and weight change during pregnancy and postpartum, including potential moderating roles of self-regulation and home food availability on the association of reward-related eating with diet and weight outcomes. Power analyses to determine sample size have been reported previously (39). Data collection was completed in June 2018. Inclusion criteria were: confirmed pregnant ≤12 weeks gestation at enrollment; uncomplicated singleton pregnancy anticipated; age ≥18 and <45 at screening; willingness to undergo study procedures and provide informed consent for her participation and assent for the baby’s participation; BMI ≥18.5 kg/m2; able to complete self-report assessments in English; access to Internet with email; plan to deliver at the UNC Women’s Hospital; and plan to remain in the geographical vicinity of the clinical site for 1 year following delivery. Exclusion criteria included: pre-existing diabetes; multiple pregnancy; participant-reported eating disorder; any medical condition contraindicating participation in the study such as chronic illnesses or use of medication that could affect diet or weight; psychosocial condition contraindicating participation in the study.
Procedures
Research staff identified potential participants through the electronic clinical appointments and medical records database. At the time of the visit, eligibility was verified and signed informed consent obtained from those electing to participate. Study visits were conducted prenatally at baseline (<12 weeks gestation), 13-18 weeks, 16-22 weeks, and 28-32 weeks gestation, and postpartum at 4-6 weeks, 6 months, and 12 months, at which time anthropometric data and biospecimens were obtained. Self-report measures were completed online within each study visit window, prompted by email reminders to participants regarding the opening and closing of each window. Participants accessed online questionnaires and the 24-hour dietary recall system through a study website developed and hosted by the study data coordinating center. Study procedures were approved by the University of North Carolina Institutional Review Board.
Measures
Dietary Intake. Participants were asked to complete a dietary recall within each study visit window using the Automated Self-Administered 24-Hour Recall (ASA24), a web-based tool for obtaining self-administered 24-hour dietary recalls developed by the National Cancer Institute and validated against the interviewer-administered automated multiple pass method (40, 41). The ASA24 prompts participants to indicate all foods consumed, including details of food preparation, brands, portion size, and additions. From this data, the program assigns food codes from the U.S. Department of Agriculture Food and Nutrient Database for Dietary Surveys and provide estimates of macronutrient, micronutrient, food categories and USDA Food Patterns Equivalents Database food groups. Participants received written instructions on use of the program, and research staff provided assistance if participants reported difficulty using the interface. Research staff at the University of North Carolina Nutrition and Obesity Research Core identified and corrected implausible entries (e.g., food items with implausible energy, fat or weight) and missing food or nutrient values and quantities. Dietary records indicating daily energy intakes of <600 kcal (36 of 1883 records, 1.9%) were excluded from analyses. Dietary records with daily energy intakes of >4500 kcal were reviewed and determined to reflect plausible intake, and thus were retained. Cutoffs for examining implausibility were based on research in non-pregnant adults indicating that energy intake cutoffs of <500 kcal and >3500 kcal produced similar estimates of associations of diet with BMI as compared with the Goldberg method and predicted total energy expenditure method (42); we increased the cutoffs for this sample to account for the increased energy requirements of pregnancy (43). Dietary intake data were used to calculate the Healthy Eating Index-2015 (HEI), an a priori indicator of diet quality that reflects conformance to the 2015 US Dietary Guidelines for Americans (44). The HEI total score ranges from 0 – 100 and is calculated by summing 13 component scores, including 9 “adequacy components” (total fruit, whole fruit, total vegetables, greens and beans, whole grains, dairy, total protein, seafood and plant proteins, fatty acids) and 4 “moderation components” (refined grains, sodium and added sugars and saturated fats), which are calculated on a per-1000 kcal or percent of kcal basis. The 9 subscales reflecting adherence to adequacy components were summed to create a HEI-adequacy score (max score = 60), and the 4 moderation subscales were summed to create a HEI-moderation score (max score = 40). Previous studies have found little change in dietary intake across trimesters (45-47); therefore, diet recalls from pregnancy and those from postpartum were pooled to calculate HEI across pregnancy (n=365) and across postpartum (n=266) using the simple HEI scoring algorithm – per person (48).
Hedonic Hunger. The Power of Food Scale (PFS) is a 15-item questionnaire that measures hedonic hunger, the appetitive response to highly-palatable food cues in the environment (49). Items querying response to the availability, presence, or taste of desirable food are rated on a 5-point Likert scale. The measure demonstrates strong internal consistency (Cronbach’s alpha = 0.91) and test-retest reliability (r = 0.77, p <0.001), and has been validated with respect to overeating (29), outcomes of weight-loss interventions (50), and brain activity in response to viewing images of food versus control (51). The PFS was completed each trimester during pregnancy and at 6 months postpartum (n=227); mean scores across pregnancy were calculated (n=377).
Addictive-like Eating. The modified Yale Food Addiction Scale (mYFAS), a 9-item abbreviated version (52) of the Yale Food Addiction Scale assesses the presence of eating disorder symptoms consistent with diagnostic criteria for food addiction. The measure has demonstrated psychometric properties similar to the original instrument, and greater scores were associated with higher BMI across two cohorts of women (52). The mYFAS was completed at baseline (n=344) and 6-months postpartum (n=217). Responses to each item were dichotomized to a score of 0 or 1 based on published cut-off values (52) and summed. Due to the highly skewed distribution, responses of 2 or more (12.5% of responses) were collapsed (only 2.3% of respondents scored 3 and 1.8% scored 4 or higher).
Food Reinforcement Measures. The Reinforcing Value of Food Questionnaire (RVFQ) (53) and Multiple Choice Procedure (MCP) (54) assessed the relative reinforcing value of food. The RVFQ asks participants to report the number of portions of a specified food that they would purchase for same-day intake at varying cost levels. The measure generates five indices: breakpoint (first price at which consumption was zero), intensity of demand (consumption at the lowest price), elasticity of demand (sensitivity of consumption to increase in cost; individual elasticities calculated using the modified exponential demand equation) (55), Omax (maximum expenditure), and Pmax (price at which expenditure was maximized). The measure has demonstrated validity against a laboratory task assessing food reinforcement value (53). The MCP asks participants to make a series of discrete choices between receiving an increasing amount of a monetary reward versus an alternative reinforcer. The datum of interest is the specific price at which participants begin to select the money over the reinforcer (breakpoint). The MCP has previously been validated in the assessment of reinforcement value of alcohol and cigarettes (e.g., 56), and was adapted by the investigators to assess the relative reinforcing value of food. Study participants were presented with the name and images of 18 palatable foods (e.g., cookies, donuts, ice cream, chips, nachos, French fries) and asked to rate their degree of liking of each items using a labeled hedonic scale with 10 response options ranging from “most disliked sensation imaginable” to “most liked sensation imaginable” (57). The two highest-rated foods were then used for the RVFQ and MCP, which were assessed at the first two pregnancy visits and six months postpartum (n=209 for RVFQ and 211 for MCP). For each measure, mean scores across the two pregnancy visits were calculated (n=348 for RVFQ and 350 for MCP). Due to highly skewed distributions, scores were log transformed.
Self-regulation. Two measures of self-regulation were administered. The 15-item short form of the Barratt Impulsiveness Scale (BIS-15) measures impulsivity across three dimensions – non-planning, motor impulsivity, and attentional impulsivity. The measure has demonstrated similar psychometric properties and associations with neurobehavioral traits as the original instrument (58). The Delaying Gratification Inventory (DGI) is a 35-item questionnaire measuring the tendency to forego immediate satisfaction in favor of long-term rewards across five domains – food, physical pleasure, social interaction, money and achievement (59). The subscale scores have shown good internal consistency (Cronbach’s alpha = 0.69 – 0.89) and strong test-retest reliability (r = 0.74 – 0.90). Both measures were completed at baseline (n=314 for BIS and 330 for DGI) and 6 months postpartum (n=215 for BIS and 219 for DGI). For this study, associations with the total score and the food subscale (DGI-food) were examined.
Home Food Environment. The Home Food Inventory includes a comprehensive range of foods in 15 categories and queries the presence of each food in the home (60). Participants completed the inventory at baseline (n=303) and 6 months postpartum (n=266). Consistent with the measure’s scoring protocol, a fruit and vegetable home food environment score (HFI-FV) and an obesogenic home food environment score (HFI-OBES) were calculated as counts of the number of foods in the home in each classification. The fruit and vegetable score includes 26 common fruits and 20 common vegetables. Foods classified as obesogenic include regular-fat versions of cheese, milk, yogurt, other dairy, frozen desserts, prepared desserts, savory snacks, added fats, regular-sugar beverages, processed meat, high-fat microwavable foods, candy, and access to unhealthy foods in refrigerator and kitchen.
Demographic and medical characteristics. Demographic information including household composition, marital status, education, and race/ethnicity were reported by participants at baseline. Income-to-poverty ratio was calculated from participant report of total household income and household size (61); higher values indicate greater income relative to the poverty threshold. Participant age and parity were obtained from the electronic medical record. Measured weight and height were obtained at the baseline visit and used to calculate early pregnancy BMI.
Analysis
Paired t-tests examined differences in measures of reward-related eating, self-regulation, and home food environment between pregnancy and postpartum; Pearson correlations examined associations of these variables between pregnancy and postpartum. Multiple linear regression analyses estimated associations of measures of reward-related eating, self-regulation, and home food environment with diet quality during pregnancy and postpartum, adjusting for education, income, household size, marital status, and race/ethnicity. Postpartum analyses additionally adjusted for duration of breastfeeding. Variables were standardized prior to analyses to provide standardized estimates. Multiplicative interaction terms were used to determine whether self-regulation or home food environment moderated associations reward-related eating with diet quality. Simple slopes analyses were used to interpret significant interaction terms. Based on the findings, interactions were graphed with reward-related eating as the moderator as this provided a clearer interpretation. SPSS version 21 was used for all analyses. Analyses employed complete-case analysis; p values <0.05 were interpreted as statistically significant.