Participants and recruitment
Participants were recruited through an online marketing research group, Survey Sampling International (SSI ref. 16). In order to recruit survey participants to their participant panels in general, SSI uses pop-up windows and banner ads on various websites, including their affiliate partners' websites, social media websites, and online communities. Interested individuals are invited to be on the SSI panel of participants; they then complete a series of online multiple choice questions so that they can be characterized in terms of various demographic and other descriptive variables. SSI panel participants are then randomly assigned to be invited to complete online surveys for which they are eligible. SSI uses a confidential identification number (which they do not share with investigators) to provide respondents with an incentive (a quarterly drawing for $25,000) to complete a survey. Every respondent who completes at least one survey during a given quarter is entered into the quarterly contest. In addition, a second incentive is provided to participants aged between 18 and 23 years because of their relatively low response rate: this age group is offered 300 points (equivalent to $3) to complete the survey.
The eligibility criterion for completing the survey in the present study was only that participants were required to be over the age of 18 years, there were no selection criteria based on any weight-related variables. However, we used two different measures of fat phobia based on whether participants identified themselves overweight or not. Within the present survey, once participants had self-reported their weight status, only those who identified themselves as very underweight, slightly underweight, or a healthy weight were asked to complete the Fat Phobia Scale — short form (FPS-S ref. 17). Individuals who self-identified as slightly or very overweight completed the Weight Bias Internalization Scale(18) and were part of another study examining the impact of genetic vs. non-genetic information on Weight Bias Internalization Scale scores among self-identified overweight individuals. Thus, only those participants who self-reported themselves to be underweight or healthy weight were included in the present analyses.
For the present study, SSI randomly selected 1,477 adult males and females from its participant panel, sent them a link to our online survey, and invited them to take part. Of these, 1,475 individuals began the survey, and 1,207 (82%) completed it.
After entering the online survey, participants answered some initial questions regarding eating attitudes, height, and weight (described fully below) and were then randomly assigned to one of the written information experimental conditions (each comprising a total of 226–330 words). All participants received identical information about the risk and consequences of obesity (Table 1). One-third of them then received genetic information, while another third received non-genetic (environment) information, and the final third received gene—environment interaction information about the causes of obesity. Half of the participants also received behavioral advice about how to reduce obesity risk, but as noted above this was not the primary focus of the present analysis.
Table 1. Content of the experimental information provided to participants
In brief, the genetic information contained information about twin studies, heredity and genetic transmission of obesity relevant genes. The non-genetic (environment) information contained information about large portion sizes, easy access to nutrient poor and high calorie foods, and conditions that favor sedentary behavior. The gene—environment interaction information contained information from both the genetic and environmental information sheets and how genes interact with the environment. Finally, the behavioral advice contained information about eating healthier and exercising more frequently. Table 1 shows the full wording of all components of the written experimental information used in the study. Although there was a minor error in the text describing the benefits of physical activity (the statement “About 60 min a day may be needed to prevent weight gain” should have read “About 60 min a day may be needed to prevent weight gain after initial weight loss”) there is no reason to suspect that this would have influenced outcomes relating to the study hypotheses. Information for the content of the experimental conditions was compiled from the scientific literature (13,19,20,21), a CDC website about obesity (22), and an online news article from Newsweek about obesity and genetics (23). The experimental information was not pilot tested before use in the current study. Participants were then asked to complete the rest of the questionnaire.
Measures assessed premanipulation (i.e., before participants were randomized to the experimental conditions) included maladaptive eating attitudes, self-reported height, and self-reported weight. Measures assessed postmanipulation (i.e., after participants were randomized to the experimental conditions) included evaluation of the information they had just read, perceived weight status, obesity stigma, self-esteem, causal beliefs about obesity, family history of obesity, and demographics.
Demographic and weight-related variables. Demographic characteristics were assessed by asking participants to report their age, gender, the highest level of education completed, annual household income, ethnicity, and relationship status. BMI (kg/m2) was calculated from participant self-reports of their height and weight, and BMI categories were assigned based on CDC guidelines (22). Participants were excluded if their calculated BMI was unrealistically low (<16 kg/m2) or unrealistically high (>55 kg/m2). Participants were also asked to indicate whether they perceived themselves to be 'very underweight', 'slightly underweight', 'healthy weight', 'slightly overweight', or 'very overweight'. Reporting that they were 'slightly' or 'very' overweight was an exclusion criterion for the present analysis. Family history was assessed by asking participants whether they thought any of their first degree relatives (mother, father, siblings, and children) were overweight. The total number of first degree relatives that were overweight was calculated for each participant.
Evaluation of the information. Participants were asked to rate the information in the experimental conditions regarding whether the information they had just read was (i): “easy to read”, (ii): “relevant to me”, (iii3): “useful to me”, and (iv) “provided me with new information”. The five response options for each statement were “strongly agree”, “agree”, “neither agree nor disagree”, “disagree”, and “strongly disagree”.
Psychological variables. Self-esteem was assessed using a single-item measure (24), “I have high self-esteem” with respondent options ranging from one (not very true of me) to five (very true of me). This measure has shown strong convergent validity for men and women, for different ethnic groups and for both college students and community members, with the Rosenberg Self-Esteem Scale (25).
To assess maladaptive eating attitudes, we used items from the Eating Disorder Examination Questionnaire (26). This questionnaire examines four subscales: restraint, weight concern, shape concern, and eating concern. To our knowledge, the Eating Disorder Examination Questionnaire has not previously been used in a general population sample, and we excluded some of the questions that seemed inappropriate to ask in a survey of a nonclinical population (e.g., the item “Over the past 28 days, how many times have you made yourself sick (vomit) as a means of controlling your shape or weight?”). This led to the exclusion of roughly half of the questions in the restraint, weight concern, and shape concern subscales, and all but one of the eating concern subscale, leaving questions 1, 3, 4, 8, 12, 20, 22, 23, 25, 26, 27, and 28. Items were rejected based on discussion and consensus between the study investigators. Because half of the questions from the restraint, weight concern, and shape concern subscales were included in the survey, a mean score for these subscales could be calculated. According to Fairburn et al. (27), “if ratings are only available on some items, a score may nevertheless be obtained by dividing the resulting total by the number of rated items so long as more than half the items have been rated” (pages 4 and 5). Alpha reliability scores were 0.86 for the restraint subscale, 0.90 for the shape concern subscale, and 0.85 for the weight concern subscale. Eating concern was assessed using question 20 on the Eating Disorder Examination Questionnaire, this item assesses guilt about eating because of its effect on shape and weight.
Causal beliefs about obesity were assessed using 10 items adapted from the revised Illness Perception Questionnaire (IPQ-R ref. 28). Participants were asked on a 5-point Likert scale how much they agreed or disagreed that each of the following causes obesity: “stress or worry”, “a germ or virus”, “diet or eating habits”, “chance or bad luck”, “hereditary—it runs in families”, “overwork”, “ageing”, “the environment”, “lack of exercise”, and “a person's genes”. The five response options for each item were strongly agree, agree, neither agree nor disagree, disagree, and strongly disagree.
Obesity stigma. The FPS-S was used to assess obesity stigma among survey respondents who did not perceive themselves to be overweight (17). The FPS-S is a semantic differential scale containing 14 items that assess attitudes about obese people (lazy—industrious, no will power—has will power, attractive—unattractive, good self-control—poor self-control, fast—slow, having endurance—having no endurance, active—inactive, weak—strong, self-indulgent—self-sacrificing, dislikes food—likes food, shapeless—shapely, undereats—overeats, insecure—secure, and low self-esteem—high self-esteem). In each word pair, the adjectives are placed at opposite ends of a scale that ranges from 1 to 5. Scores below 3 indicate positive attitudes toward obese people (i.e., no obesity stigma), and scores above 3 indicate negative attitudes toward obese people (i.e., obesity stigma). The alpha reliability in the present study was 0.93. We chose this scale over other possibilities (e.g., the universal measure of bias—fat version, ref. 29) in part because it is relatively straightforward for participants to complete, comprising pairs of words rather than whole sentences.
Demographic and weight-related characteristics (gender, age, education, income, ethnicity, BMI, family history, relationship status, and perceived weight status) of the sample were described using frequencies, means, and s.d., and compared between the experimental groups using ANOVAs to check for potential confounders. The effects of the experimental manipulations on obesity stigma were examined using ANOVA. In order to do this, two new categorical variables were created representing (i) obesity cause information type (genetic/environmental/gene—environment interaction) and (ii) behavioral advice (yes/no), and then entered simultaneously into an ANOVA. The main effects of each of these as well as the interaction between the two were examined. We also examined associations between obesity stigma, demographic, and weight-related characteristics and psychological variables (self-esteem, maladaptive eating attitudes, and obesity causal beliefs): unadjusted ANOVAs and χ2 were used for continuous and categorical independent variables, respectively. We then used ANCOVA to examine the associations between obesity stigma and any variables that were significantly associated with obesity stigma in the prior unadjusted models. P values of less than 0.05 were considered significant in all statistical analyses. All statistical analyses were performed using IBM SPSS statistics 19 (Chicago, IL).