Psychological and weight history variables as predictors of short‐term weight and body fat mass loss

Summary Objective Identifying predictors of early weight loss may have value in predicting longer‐term success in weight loss programmes. This study examined if weight history variables (ie, weight cycling history [WCH], age of onset of obesity [AOO]), and preintervention Three‐Factor Eating Questionnaire (TFEQ) and Power of Food Scale (PFS) scores predicted weight loss (WL) and fat mass loss (FML) following a 3‐week calorie restriction intervention. Methods Thirty‐two participants (19‐60 y; body mass index [BMI] 30‐39.9 kg/m2) participated in a 3‐week calorie restriction intervention (1120 kcal/d) as part of a larger clinical trial with 28 completers included in the current analyses. Preintervention WCH, AOO, TFEQ, and PFS subscale scores were collected, and WL and FML were measured. Multiple linear regression analyses were performed to predict WL and FML for relevant covariates in this study. Results WCH, AOO, preintervention TFEQ subscale scores, and PFS subscale scores did not predict WL (all Ps > .08) or FML (Ps > .06) except, PFS‐food tasted scores significantly predicted WL (r = −0.40, P = .03). Conclusion Although these variables were not robust predictors, results for at least the PFS suggest there may be value in further exploring this measure using larger sample sizes.


O R I G I N A L A R T I C L E
as yet understudied gap in understanding of treatment response. Some common psychosocial, behavioural, and physiological factors within the context of very early treatment response may provide predictors that are informative in understanding contributors to early success or lack thereof and allow us to identify potential targets for adjusting treatment early in those at risk for nonresponse. These predictors may be different for dietary interventions utilizing typical foods eaten vs total meal replacement (TMR) programs based on the known differences of these diet types in terms of both treatment outcomes and psychological response to food cues. 5,6 Factors related to weight history such as weight cycling (repeated weight loss and regain) and age of onset of obesity have received only moderate attention in the obesity literature to date with somewhat equivocal findings. Weight cycling in individuals with obesity and those who do not yet have obesity has been associated with a higher risk of developing or exacerbating obesity and also with increased abdominal and visceral fat accumulation. [7][8][9][10] However, in terms of predicting successful weight loss, the association of weight cycling and body weight change is inconclusive. [11][12][13] There is some suggestion that history of weight cycling may act as both an impedance to weight loss success in those undergoing treatment and may also be useful as a predictor of future weight gain. 11,14,15 Early age of onset of obesity has been associated with an increased incidence of adult obesity, increased overall adiposity, as well as the number of adipose cells in the body. [16][17][18] One study suggests that children and adolescents with obesity are five times more likely to develop obesity in adult age than those who did not have obesity at an early age. 19 Some limited evidence suggests that age of onset may not influence success during weight loss treatment, 20,21 while at least one study reported that during 1 to 2 years follow up after bariatric surgery, early age of onset was associated with less weight loss. 22 Psychological influences on eating behaviour have long been of interest in relation to predicting weight loss outcomes. Numerous self-report measures have been developed in an attempt to characterize the relationship with food in people with obesity and subsequently provide quantifiable measures of potential psychological predictors.
The Three Factor Eating Questionnaire (TFEQ) 23 is a widely used and longstanding instrument of this kind. In its original 51-item version, it attempts to quantify eating behaviours along three distinct general psychological dimensions: (a) cognitive restraint, exerting cognitive control to restrict food intake; (b) disinhibition, the experience of impulsive/uncontrolled eating; and (c) susceptibility to hunger, which characterizes the likelihood of eating while in a hunger state.
The scale identifies individual patterns of eating behaviour that may lead to overconsumption of foods based on specific psychological states/dysregulation. Higher scores on the TFEQ are associated with higher body mass index (BMI). 24,25 Several studies have found that TFEQ scores can predict subsequent success with weight loss. 26,27 The Power of Food Scale (PFS) measures the psychological impact of living in food-abundant environments. Specifically, the PFS quantifies appetite for palatable foods in the context of availability, proximity, and personal experience of a food (ie, food available, food present, and food tasted). 28 The measurement of hedonically based motivations to eat in environmental context has utility in predicting long-term weight control. 28,29 Understanding how these wellvalidated psychological measures are related to very early treatment response will provide valuable insights into identifying early responders to dietary intervention.

| Design overview
Participants who met all eligibility criteria were enrolled in the original study. Participants attended visit 1 for baseline (preintervention) measurement: (a) self-report assessment measures (BMTR Lab Weight History form (included age of onset, weight cycling history)), (b) mea- where they completed the preintervention self-report measures related to the relationship with food (ie, TFEQ and PFS) followed by the fMRI scan (note fMRI data were not included in the current study) after which they were randomly assigned to either TMR with Optifast 800 shakes or TD groups. Participants attended two check-in visits between assessment visits. After completion of the 3-week dietary intervention, (visit 5) body weight, body fat mass was measured.

| Dietary intervention
Immediately following the initial fMRI scanning conducted in visit 2, all the participants were randomly assigned to either TMR or TD group.

| TMR group
Every participant assigned to the TMR group was instructed in the appropriate use of the meal replacement product. Optifast 800 TMR shakes (Nestlé HealthCare Nutrition Inc, New Jersey) were provided in a quantity that would last for a 1-week period with the prescribed energy intake of 1120 kcal/d. Every participant was instructed to abstain from all items of food for this 3-week intervention period, except the meal replacement product and water/noncaloric beverages. Participants were requested to return to the BMTR lab after completion of each week consecutively for 2 weeks to receive next week's product and to be weighed, have blood pressure measured, and speak with an investigator to review any issues related to diet adherence.

| TD group
Participants assigned to the TD group were instructed to use portion control to maintain their daily calorie intake at 1120 kcal/d but not make significant changes in typical foods eaten. Participants were directed to return to the BMTR lab at every 1-week interval over the following 2 weeks to measure their weight and blood pressure and to review any issues related to diet adherence.

| Self-report measures
The BMTR Demographic, Health, and Weight History Form is a selfreport measure used to collect basic demographic, medical, and psychosocial information. This study used the following data from this form: (a) history of weight cycling; participants were asked "How many times have you lost 10 lbs. or more when you weren't sick and then All the scales are scored individually according to published guidelines, and these scales are summed to produce the total score. From recently available data, the test-retest reliability was adequate (0.71) with good internal consistency (Cronbach α.91). 28

| Statistical analysis
Data were examined using the R statistical software package (version 3.4.3). 31 All continuous variables (eg, TFEQ score, PFS score, weight, and fat mass) were examined for normality and outliers and for compliance with the assumptions of linear regression. Descriptive statistics of all the measured variables were calculated and reported as mean and standard deviation (M ± SD). Chi-square tests of homogeneity were used to interrogate categorical variables for potential differences by sex or diet grouping.
Multiple linear regression analyses were performed to predict postintervention vs preintervention changes in body weight and fat mass using preintervention TFEQ subscales, preintervention PFS score, age of onset of obesity and weight cycling as predictors (separately) adjusting for age, sex, diet group, and preintervention body weight, fat mass, and BMI. Stepwise model selection was used to report the best model as chosen by Akaike information criterion (AIC), where a lower AIC value denotes a better model.
Univariate linear models were used in order to assess various bivariate associations between continuous and categorical measurements. Additionally, simple linear regression models were used in order to estimate point estimates and confidence intervals for various bivariate correlations among the continuous measurements. Nominal significance using an alpha threshold of.05 is used throughout.

| Ethics
This study was approved by the Texas Tech University, Human Subjects, Institutional Review Board (TTU IRB #505380). The original study is registered in Clinicaltrials.gov (NCT02637271). All procedures were carried out in accordance with the Declaration of Helsinki, 2000. 32 Upon meeting the eligibility criteria, a written informed consent was obtained from all participants.

| RESULTS
Fifteen participants in the TMR group and thirteen participants in the TD group completed the larger study (see Figure 1: CONSORT diagram) and were included in the analyses. The two groups did not differ by age, sex, weight, fat mass, height, and BMI. The preintervention characteristics of this study are presented in Table 1, and weight history data with frequency are presented in Table 2.
After randomly assigning participants into TMR and TD groups, history of weight cycling and age of onset of obesity showed no significant difference between sex (male and female) and diet group (TMR and TD) ( Table 3). Also, there was no difference between male and female distribution within the diet groups (Table 4). Cross-tabulations were analysed using chi-square tests of independence. The 3-week calorie-restricted diet resulted in a significant reduction in body weight and body fat mass (Tables 5 and 6).

| Association of weight history variables and BWL and FML
Data from history of weight cycling and from age of onset of obesity were analysed to determine correlations with BWL and FML (Tables 7   and 8). The analyses showed that age of onset of obesity and history of weight cycling were not associated with BWL and FML.

| Association of TFEQ and PFS subscales with BWL and FML
Linear modelling showed that the correlations between preintervention TFEQ subscales and PFS subscales with body weight and fat mass loss were not significant, except PFS-Food Tasted score was negatively correlated with BWL (P = .03) (Tables 9 and 10).

| DISCUSSION
This study considered whether weight history and psychological relationship with food variables were associated with BWL and FML in a sample of adults with class I and II obesity undergoing a 3-week isocaloric TMR-and TD-based weight loss intervention. History of weight cycling and age of onset were not significantly associated with initial,   Overall, the hypotheses that TFEQ and PFS subscales would negatively predict BWL and FML were not supported; this was somewhat unexpected in the context of the literature.. 27,35,36 There are two    fat mass, and BMI in the exploratory analyses. All of these variables may significantly contribute BWL and FML, so their inclusion helps untangle their potential influences on the outcomes of primary interest. Finally, the research design, data analysis approach, and statistical reporting improves substantially on several deficiencies in the existing literature, which will provide a strong foundation from which to construct future studies examining this novel approach to the topics.
This study has some limitations. The lack of significant associations with simple linear modelling between predictor variables and BWL and FML versus a significant association between predictor variables and BWL and FML with multiple linear modelling suggest that this study may be lacking enough power (due to relatively low sample size or narrow ranges of covariate values). For example, using a narrow BMI range may have limited the range of scores on all the predictor variables since each of the variables is independently associated with BMI. Thus, having a high BMI score will more likely lead to reporting a high score on this survey measurement subscales, causing less overall variability in the total scores of the TFEQ and PFS instruments. Moreover, while this study observed a significant loss of weight and fat mass during the 3-week intervention, these losses are only over a short time frame and may solely be specific to a population with high BMI. Thus, a larger sample size with wider BMI and other covariates ranges, as well as a longer intervention period, is prudent.

| CONCLUSION
In conclusion, weight history variables (ie, weight cycling history and age of onset of overweight or obesity) did not predict BWL and FML following a 3-week calorie restriction intervention. Scores of TFEQ subscales (ie, dietary restraint, disinhibition, and susceptibility to hunger) did not predict BWL and FML. Scores of PFS (ie, food available, food present, and food tasted) were not a robust predictor of BWL and FML. The results as a whole suggest that being in the TMR group has inverse effects on TFEQ subscales when predicting BWL and on PFS scores when predicting FML. A possible explanation may be that the TMR diet is thought to act through a narrowing of the food stimulus environment. In this context, this result makes intuitive sense: A person on TMR is not being exposed to the food environment; they are not shopping for food or cooking, and they often report consciously trying to avoid food commercials, restaurants, or other situations heavy with food cues. Thus, while they may be more susceptible in the food environment, the motivation towards ingesting food is negated by the dearth of cues. Furthermore, this was not seen in the TD group where the exposure in the food environment remains relatively unchanged, lending further support for this potential explanation of these findings.