INTRODUCTION

Long-term weight loss maintainers (WLM) consume a low-energy dense diet that is micronutrient rich.1 Successful WLM also report high dietary consistency, tending to eat the same foods during the week as on weekends and during the holidays vs. non-holiday times.2 Moreover, high levels of cognitive restraint, defined as conscious control overeating, have been extensively reported among WLM.3 Although the content, consistency, and ability to restrict eating have been studied among WLM, other behavioural and attitudinal factors that may influence food choice have received less attention.

Prior research in general populations has shown that beliefs about food, health, weight, food familiarity and perceived sensory properties, current mood, ethical concerns, and food price may shape food choices.4-7 Moreover, an ability to focus on the future8 and resist eating in response to tempting food cues9-11 have been identified as promising strategies for weight management. However, research to date generally has been limited by a restricted range of measures, lack of measurement of weight status, and a narrow range of populations, such as consumers,7 employees,9 undergraduate students8 or convenience6 samples. No known research has comprehensively surveyed the factors that relate to the food choices among people with varying weight statuses, including among long-term WLM. The scientific report for the 2020 Dietary Guidelines Advisory Committee12 recognised the need for more research to better understand not only on what people choose to eat, but also the social, economic, and environmental factors that shape dietary patterns. A comprehensive understanding of the diverse factors that may influence eating decisions is critical for informing the development of effective interventions that aim to modify unhealthy dietary patterns and promote long-term successful weight control.

The present study aimed to examine the factors that distinguish food choices among long-term WLM in a widely used a commercial weight management program (WW International, Inc.) compared to weight stable individuals with obesity. The study hypothesised that WLM would report food choices that were more motivated by health and weight-related factors and less motivated by responses to palatable food cues than weight stable individuals with obesity. WLM were also hypothesised to score higher on future orientation then weight stable individuals with obesity.

METHODS Design

The WW Success Registry (WWSR) is an observational study of individuals who have lost weight in the WW International, Inc. program and were successful at long-term (≥1 year) maintenance of substantial (≥9.1 kg) weight loss.1, 13, 14 In this cross-sectional study, long-term WLM following the weight management program (WW) are compared with weight stable individuals with obesity (“controls”) to distinguish the factors associated with successful maintenance of weight loss.

Participants and eligibility

Procedures were approved by the Institutional Review Board, and all participants provided informed consent electronically via Research Electronic Data Capture (REDCap).

Weight loss maintainers

Prospective WLM were recruited through an email sent by WW to members who had reported a loss of ≥9.1 kg >1 year ago when following WW. Interested individuals were referred to the study website hosted by the university for online screening, consent, and enrollment. Eligibility was based on self-reported weight, height, weight change, and duration of weight loss. To be eligible for enrollment, individuals were aged ≥ 18 years and had maintained a >9.1 kg loss from WW entry for ≥1 year. The criterion 9.1 kg was selected to approximate a clinically significant 10% weight loss,15 assuming a starting weight of 90 kg among people entering WW and other weight loss programs.16 Use of an absolute weight loss value was also intended to simplify messaging for recruitment and eligibility screening and has been used successfully in the National Weight Control Registry.17

Weight stable individuals with obesity

Weight stable individuals with obesity were recruited through local and national advertising channels, including Facebook, ResearchMatch.org, Amazon, Mechanical Turk, and via the Academic Center for Health Research registry. Interested individuals were referred to the study website hosted by the university for online screening, consent, and enrollment. Eligibility was based on self-report and included age ≥ 18 years, with a body mass index (BMI) > 30 kg m–2 and reported weight stability (within 2.3 kg) for at ≥ 5 years prior to enrollment.18 Control participants were not currently enrolled in WW. Control participants were provided 1 month of the WW online program (WW Digital) free of charge after completion of the survey.

Measures

All measures were administered online via REDCap immediately after consent. All participants were asked standard demographic information (age, education level, income, marital status) and details about weight history (age of onset of overweight, maximum lifetime weight), as well as current weight and height. The validity of self-reported weight history has been established previously.3 Also, self-reported weights have been shown to correlate strongly with measured weights.19

The Food Choice Questionnaire (FCQ)20 was used to measure diverse factors that influence food decisions, including subscales for health, mood, convenience, sensory appeal, natural content, price, weight control, familiarity, and ethical concern. Participants were asked to rate the importance of diverse determinants of food choice by responding to the prompt, “It is important to me that the food I eat on a typical day …”. Examples of items were “is low in calories” (weight control), “is cheap” (price), “is packaged in an environmentally friendly way” (ethical concern), and “cheers me up” (mood). Scores were on a scale where 1 = not at all important and 4 = very important. Scores on subscales are added and ranked to indicate relative importance of factors in shaping food choices.20 The FCQ has been shown to have acceptable reliability (>0.70) and the internal consistency coefficients on its subscales range from 0.72 to 0.86.20

The 12-item Consideration of Future Consequences Scale (CFC)20, 21 was used to measure the extent to which people consider potential distant outcomes of their current behaviours (e.g., “I am willing to sacrifice my immediate happiness or well-being in order to achieve future outcomes.”). Each statement is rated on a scale from 1 (“extremely uncharacteristic”) to 5 (“extremely characteristic”). The scale scores range from 12 to 60, and a high CFC score indicates greater importance being placed on the future consequences of a behaviour, whereas a lower CFC score indicates greater importance being placed on the more immediate consequences of behaviour. Cronbach's α values for the CFC range from 0.80 to 0.86.22, 23 Eating in the Absence of Hunger (EAH) was measured using the EAH-C24 scale. This scale is composed of 14-items that assess three dimensions related to stimuli that generate beginning or continuing to eat food in the absence of hunger. “Continuing EAH” is defined as continuing to eat immediately after being satiated at mealtime, and “beginning EAH” is defined as beginning to eat when not hungry several hours after being satiated.24 Within these, the scale includes three motivators of eating in the absence of hunger: Negative affect (feeling sad or depressed, angry or frustrated, anxious or nervous); external eating (e.g., food looks, tastes or smells good and/or being in the presence of others who are eating); and fatigue/boredom. EAH-C was originally developed and validated for children and adolescents24 but modified for college students and found to have high internal consistency across subscales (0.83–0.92).25 In a subset of participants (n = 1162 [30.5%] WLM and 139 [26.8%] controls), the Diet History Questionnaire (DHQ-II) from the National Institutes of Health was used to measure self-reported calorie and macronutrient intake.26 The DHQ-II was included as an exploratory and optional measure in the WWSR.1

Statistical analysis

Independent t-tests and a chi-squared analyses were used to compare socio-demographic characteristics of WLM vs. controls and completers vs. non-completers. Subsequent general linear models compared WLM and controls on scores of the FCQ measure (i.e., health, convenience, mood, sensory appeal, weight control, price, natural content, familiarity, and ethical concern), the CFC, and the EAH (beginning and continuing domains for external eating, negative affect, fatigue/boredom) and adjusted for a priori covariates of age, race (white vs. non-white), employment (employed vs. not), education (≥college education vs. < college), income (<$25,000, $25,000–75,000, > 75,000/year), maximum lifetime weight, sex assigned at birth (male vs. female), and marital status (married vs. not).

Discriminant function analysis was used to determine the variables that most discriminated WLM from controls among the set of variables (subscale scores only) that were found to differ between the two groups in the initial general linear model analyses. The resulting standardised canonical coefficients represent the measure of association between the discriminant function (based on the linear combination of variables) and each predictor variable and indicate the relative importance each variable in distinguishing the two groups (similar to a beta weight in a multiple regression). Within each group (i.e., WLM and controls), BMI and dietary intake (in a subset) were examined in relation to the FCQ, CFC, and EAH questionnaire scores, adjusting for the same covariates. To guard against type 1 error due to multiple analyses, statistical significance was set to p < 0.01 and small effect sizes (ηp2 < 0.03) were considered as not significant. SPSS, version 25.0.0 (IBM Corp.) was used for all of the analyses.

RESULTS Participants

Of the 8047 WLM and controls, 4325 completed the FCQ, which was situated in the second half of a lengthy questionnaire. Comparing participants who completed (n = 4325) vs. those who did not complete (n = 3722) the questionnaire, completers were older (53.6 [12.9] vs. 51.4 [12.8] years; p = 0.0001), more likely to be white (93.8% vs. 68.4%; p = 0.0001), less likely to be Hispanic (3.6% vs. 5.9%; p = 0.0001), and less likely to be employed (65.2% vs. 73.4%; p = 0.0001). Also, a greater proportion of controls than WLM completed the questionnaire (61.6% vs. 52.9%; p = 0.0001). Among participants, WLM and controls differed on several demographic factors (Table 1). Weight loss maintainers were more likely than controls to be older (54.5 vs. 46.7 years; p = 0.0001), female (91.8% vs. 78.6%; p = 0.0001), white (95.1% vs. 83.6%; p = 0.0001), married (74.7% vs. 51.1%; p = 0.0001), with an annual family income exceeding $75,000 (65.4% vs. 29.1%; p = 0.0001), and with at least a college education (89.5% vs. 84.6%; p = 0.002). Subsequent analyses statistically adjusted for these variables.

Table 1. Characteristics of weight loss maintainers following a commercial weight management program (WW International, Inc.) vs. weight stable individuals with obesity (controls) WLM Controls p value n = 3806a SD n = 519a SD Age (years), mean 54.5 12.6 46.7 13.1 0.0001 Female (%) 91.8 78.6 0.0001 Currently in WW (%) 90.3 0 0.0001 Lifetime maximum weight, mean (kg) 105.8 23.0 121.4 27.9 0.0001 Weight at start of WW (kg), mean 101.7 21.3 Not applicable Not applicable Current weight (kg), mean 76.5 16.5 111.1 23.2 0.0001 Weight loss since WW start (kg), mean 25.2 12.7 Not applicable Not applicable Duration 9.1 kg loss from WW start weight (years), mean 3.2 3.2 Not applicable Not applicable Weight lost from maximum weight (kg), mean 29.3 15.4 10.3 14.3 0.0001 Current BMI (kg m–2), mean 27.6 5.4 39.6 7.8 0.0001 BMI categories 0.0001 Obese (%) 22.3 100% Overweight (%) 44.3 0 Normal weight (%) 33.4 0 Underweight (%) 0.0 0 Income (total in family per year) 0.0001 <$25,000 (%) 4.5 21.7 $25,000–75,000 (%) 30.1 49.2 ≥$75,000 (%) 65.4 29.1 Race/ethnicity 0.0001 White (%) 95.1 83.6 Black (%) 2.3 11.9 0.0001 Hispanic (%) 3.3 6.0 0.003 Employed (%) 63.8 75.2 0.0001 College education or more (%) 89.5 84.6 0.002 Married (%) 74.7 51.1 0.0001 Abbreviation: WLM, weight loss maintainers. Motivators of food choices

In both groups, the top three factors rated as most important in food choices were health, convenience, and mood, and the lowest ranking factors were ethical concern, familiarity, and natural content (Table 2). Although health was the strongest reported motivation for food choice in both groups, WLM scored significantly higher than controls in the extent to which health influenced their food decisions (18.9 vs. 16.3; ηp2 = 0.052; p = 0.0001) (Table 2). WLM also scored significantly higher than controls in reports of making food choices based on beliefs that the food aided in weight control (9.9 vs. 7.3; ηp2 = 0.159; p = 0.0001). WLM reported lower scores than controls in making food decisions based on price (8.4 vs. 9.1; ηp2 = 0.10; p = 0.0001). Both groups scored similarly on the extent to which convenience, mood, sensory appeal, natural content, familiarity, and ethical concerns shaped food choices. Examining future orientation, WLM reported greater consideration of future consequences (44.3 [95% confidence interval = 44.0–44.5] vs. 38.4 [37.8–39.1]; ηp2 = 0.060; p = 0.0001). Also, WLM reported less eating in the absence of hunger during a meal in response to external cues (7.1 [7.0–7.2] vs. 7.5 [7.2–7.7]; ηp2 = 0.058; p = 0.0001). WLM relative to controls reported consuming a smaller proportion of daily calories from fat (0.32 [0.32–0.33] vs. 0.38 [0.36–0.39]; ηp2 = 0.05; p = 0.001) and a higher proportion of daily calories from protein (0.18 [0.18–0.19] vs. 0.16 [0.16–0.17]; ηp2 = 0.03; p = 0.0001) (Table 2). No meaningful differences were observed in scores for eating in the absence of hunger at the initiation of a meal or in response to negative affect or fatigue/boredom (Table 2).

Table 2. Food choice motivations in weight loss maintainers and weight stable individuals with obesity (controls) WLM Controls n = 3327a n = 507a Mean 95% CI Mean 95% CI Group effectb Food choice questionnaire Health 18.9 18.8, 19.0 16.3 16.0, 16.6 ηp2 = 0.052; p = 0.0001** Ranking 1 1 Convenience 15.2 15.1, 15.3 15.4 15.1, 15.7 ηp2 = 0.001; p = 0.127 Ranking 2 2 Mood 13.4 13.2, 13.5 14.0 13.6, 14.4 ηp2 = 0.002; p = 0.002 Ranking 3 3 Sensory appeal 12.2 12.1, 12.3 12.0 11.8, 12.3 ηp2 = 0.0001; p = 0.204 Ranking 4 4 Weight control 9.9 9.8, 9.9 7.3 7.1, 7.5 ηp2 = 0.159; p = 0.0001** Ranking 5 6 Price 8.4 8.3, 8.5 9.1 8.9, 9.3 ηp2 = 0.10; p = 0.0001** Ranking 6 5 Natural content 8.2 8.1, 8.2 7.3 7.0, 7.5 ηp2 = 0.013; p = 0.0001 Ranking 7 7 Familiarity 6.8 6.7, 6.8 7.1 6.9, 7.2 ηp2 = 0.003; p = 0.002 Ranking 8 8 Ethical concern 5.8 5.7, 5.9 5.9 5.7, 6.1 ηp2 = 0.0001; p = 0.350 Ranking 9 9 Consideration of Future consequences, Total score (possible range from 12 to 60) 44.3 44.0, 44.5 38.4 37.8, 39.1 ηp2 = 0.060; p = 0.0001** Eating in absence of hunger 32.4 32.1, 32.7 32.8 31.8, 33.6 ηp2 = 0.0001; p = 0.523 Beginning to eat while not hungry External eating (20 maximum) 7.5 7.4, 7.5 7.8 7.5, 8.0 ηp2 = 0.002; p = 0.018 Negative affect (15 maximum) 10.9 10.8, 11.0 11.3 11.0, 11.6 ηp2 = 0.001; p = 0.026 Fatigue/boredom (10 maximum) 3.8 3.7, 3.8 3.8 3.6, 3.9 ηp2 = 0.0001 p = 0.906 Continuing to eat after satiated 33.8 33.5, 34.1 33.9 33.1, 34.8 ηp2 = 0.0001; p = 0.750 External eating (15 maximum) 7.1 7.0, 7.2 7.4 7.2, 7.7 ηp2 = 0.058; p = 0.001** Negative affect (15 maximum) 8.3 8.2, 8.4 8.4 8.2, 8.7 ηp2 = 0.0001; p = 0.388 Fatigue/boredom (15 maximum) 6.1 6.0, 6.2 5.9 5.7, 6.2 ηp2 = 0.0001; p = 0.296 Dietary intakec Daily calorie intake 1499 1467, 1531 1618 1524, 1711 ηp2 = 0.005; p = 0.02 Calories from fat (%) 0.32 0.32, 0.33 0.38 0.36, 0.39 ηp2 = 0.05; p = 0.001** Calories from carbohydrate (%) 0.50 0.50, 0.51 0.46 0.44, 0.48 ηp2 = 0.02; p = 0.0001 Calories from protein (%) 0.18 0.18, 0.19 0.16 0.16, 0.17 ηp2 = 0.03; p = 0.0001** Abbreviations: CI, confidence interval; WLM, weight loss maintainers; ηp2, partial eta square. Multiple discriminant analysis

Multiple discriminant analysis was conducted to determine the factors that most strongly discriminated WLM from controls. Sta