International Journal of Obesity (2001) 25, 574±580 ß 2001 Nature Publishing Group All rights reserved 0307±0565/01 $15.00 www.nature.com/ijo
PAPER The relationship between restraint and weight and weight-related behaviors among individuals in a community weight gain prevention trial MT McGuire1*, RW Jeffery1, SA French1 and PJ Hannan1 1
University of Minnesota, Division of Epidemiology, Minneapolis, Minnesta, USA
PURPOSE: The present study evaluated the cross-section and prospective associations between the Eating Inventory's (EI) total, ¯exible and rigid dietary restraint scales and changes in weight and behaviors in a community sample of adults enrolled in a 3 y weight gain prevention study. METHODS: Subjects were participants in the Pound of Prevention (POP) study, a community-based weight gain prevention trial. RESULTS: Higher levels of baseline total, ¯exible and rigid dietary restraint were related to lower weight and more weightcontrolling behaviors at the baseline assessment. Baseline restraint measures positively predicted increases in weighing frequency over the 3 y follow-up. Increases in restraint over the follow-up period were related to decreases in weight, energy intake and television watching, and increases in self-weighing and physical activity. CONCLUSION: The EI's total, ¯exible and rigid restraint scales were not differently associated with weight and behaviors in this heterogeneous sample of adults who were attempting to lose weight. Developing methods to increase behavioral and cognitive strategies to control weight may help to prevent weight gain in clinical and community samples. International Journal of Obesity (2001) 25, 574 ± 580 Keywords: dietary restraint; weight gain prevention; diet; physical activity
Introduction
Weight gain results when energy intake is greater than energy expenditure. Genetics in¯uence components of the energy equation and cause individual variability in risk of weight gain. However, the dramatically increasing prevalence of obesity in our society in the last few decades is almost certainly due to rapidly evolving environmental factors that affect energy intake and expenditure rather than changes in biology.1 Prevention of weight gain in our current obesity-conducive environment requires the skills to adapt dietary and physical activity behaviors.2 Thus, a better understanding of the factors that are related to successful adaptation of dietary and physical activity behaviors may be
*Correspondence: MT McGuire, University of Minnesota, School of Public Health, Division of Epidemiology, 1300 South Second Street, Suite 300, Minneapolis, MN 55454, USA. E-mail:
[email protected] Received 1 June 2000; revised 10 November 2000; accepted 20 November 2000
important for efforts to address the growing health problems of obesity. Dietary restraint is a measure of behavioral and cognitive strategies that people use to control body weight. The construct of restraint has evolved over time and the behaviors that are related to restraint depend on which measure is used. The Eating Inventory (EI)3 was developed as a measure to distinguish between restrictive and disinhibited eating patterns. The EI has three scales, one of which contains 21 items that measure dietary restraint. Studies showed that the EI restraint was cross-sectionally related to lower body weight and lower dietary intake.4 ± 6 Studies also found that increases in restraint over time were related to weight loss and increases in weight-controlling behaviors.4,7 ± 11 In 1991, Westenhoeffer12 proposed that, although the EI restraint items measured behaviors used to restrict energy intake, not all of the behaviors were necessarily effective in controlling weight. In that study, more than 50 000 men and women who were participants in a computer-based weight loss program completed the German version of the EI. Results showed that the EI restraint scale contained two
Relationship between restraint and weight MT McGuire et al
subscales that measured ¯exible and rigid behavioral and cognitive strategies used to control eating. The seven-item ¯exible restraint subscale re¯ected a balanced approach to dietary intake, whereas a seven-item rigid restraint subscale re¯ected an `all-or-nothing' approach to dieting. Cross-sectionally, higher initial ¯exible restraint scores were related to lower disinhibition,12 caloric intake, and body mass index (BMI);13 higher initial ¯exible restraint also predicted better weight loss success at a 1 y follow-up.13 Higher initial rigid restraint scores were cross-sectionally related to higher disinhibition12 and BMI,13 and predicted less weight loss success in women.13 Two other cross-sectional studies that evaluated the EI's ¯exible and rigid restraint subscales failed to completely replicate Westenhoefer's12,13 rigid restraint ®ndings. One of these studies sampled 31 women, most of whom were being treated for either anorexia or bulimia, and found no relationship between rigid restraint and BMI.14 However, because the women had been in treatment for an average of 19 weeks, it is unlikely that they were engaging in many of the behaviors measured by the rigid restraint subscale. The other crosssectional study sampled a group of primarily overweight women (n 293) who were not seeking weight-loss treatment and found that higher rigid restraint was related to lower BMI.11 These con¯icting results suggest that the relationship between rigid restraint and weight may depend on the characteristics of the individuals sampled. The purpose of the present study was to assess the crosssectional and prospective associations between total, ¯exible and rigid restraint and weight and weight-related behaviors among a sample of individuals enrolled in a community weight gain prevention study. We hypothesized that the relationship between these three measures of restraint and weight would closely re¯ect those found in Westenhoefer.12,13 That is, we hypothesized that higher total and ¯exible restraint would be related to lower weight and more behaviors related to weight-control; higher rigid restraint would be related to higher weight and fewer weight-controlling behaviors.
Methods
Participants and design The subjects in the present study were participants in a 3 y community-based weight gain prevention trial, Pound of Prevention (POP). The design and methods of POP have been described in detail elsewhere.15,16 Brie¯y, the purpose of POP was to evaluate the ef®cacy of a low-intensity intervention for preventing age-related weight gain. Men and women were recruited from the Minneapolis=St Paul, Minnesota metropolitan area through newspaper advertisements, mailings to employees at the University of Minnesota, and direct telephone recruitment. Women of lower socio-economic status (SES) were recruited through Women, Infants, and Children (WIC) programs and advertisements within certain neighborhoods. Participants were
eligible for the study if they were between 20 and 45 y and not pregnant at, or 1 y prior to the beginning of the study. All participants (n 1226) were randomized into one of three treatment groups. One-half of the participants were randomized into a no-treatment control group that did not receive any treatment information during the 3 y study. Twenty-®ve percent were randomized into an educationonly group which received monthly newsletters that emphasized self-weighing, increased servings of fruits and vegetables, decreased servings of high-fat foods, and walking. The monthly newsletters were mailed to participants for the 3 y of the intervention. Participants were encouraged to send back self-addressed, stamped postcards that assessed adherence to the educational messages. The other 25% of the participants were randomized into an education plus lottery incentive group. This group received the same monthly newsletters as the education-only group but, in addition, they were entered into a lottery drawing for $100 if they returned their adherence postcard.
575
Measures Participants completed the following measures at baseline, before randomization, and once a year for 3 y. Demographics. Age (y), gender (male 1, female 2), ethnicity (white 0, other 1), and highest level of education achieved ( < college degree 0, college degree 1) were assessed during the baseline assessment. Smoking. Current smoking status (no 0, yes 1) was assessed by self-report at each yearly assessment. Restraint. The Cognitive Restraint scale of the Eating Inventory3 was used to assess restraint levels among participants. This 21-item scale measured the cognitive and behavioral strategies used to control weight and it is related to dieting status, body weight, and positive changes in weightcontrolling behaviors.4,7 ± 11 In addition to this total measure of restraint, the seven-item ¯exible restraint subscale and the seven-item rigid restraint scale were developed according to the scoring method described by Westenhoefer.13 Items on the ¯exible restraint subscale included items such as: When I have eaten my quota of calories, I am usually good about not eating any more; I deliberately take small helpings as a means of weight control; While on a diet, if I eat food that is not allowed, I consciously eat less for a period of time to make up for it. Items on the rigid subscale included items such as: I count calories as a conscious means of controlling my weight; How often are you dieting in a conscious effort to control your weight?; Do feelings of guilt about overeating help you to control your food intake? Weight. Weight and height were measured at each annual clinic visit. From this information BMI (weight (kg)= International Journal of Obesity
Relationship between restraint and weight MT McGuire et al
576
height (m2)) was calculated. Waist circumference was also measured. Weight controlling behaviors. Current dieting status: participants were asked if they were dieting to either lose or maintain their weight (no 0; yes 1). Self weighing: participants were asked how frequently they weighed themselves each month. Dietary intake: the 60-item Block Food Frequency Questionnaire (FFQ)17 was used to assess dietary intake. Participants reported the frequency and typical serving size for common food items. Responses were used to estimate total energy intake and percentage of total energy intake from fat, carbohydrates, protein, sweets and alcohol. The 60-item FFQ was validated against food diaries18 and a longer 100-item FFQ.19 The Food Habits Questionnaire (FHQ)20 was used to assess behavioral strategies for reducing dietary fat intake. A series of questions assessed ®ve general behaviors of reducing dietary fat intake such as modifying meat preparation, avoiding fat as a seasoning, replacing fattening foods with less fattening foods, substituting high-fat foods with modi®ed fat alternatives, and replacing fattening foods with fruits and vegetables. Responses to items on this questionnaire, which range from rarely or never to usually or always, were summed to form subscale and total scale scores. Higher scores on the FHQ inversely correlate to dietary fat intake.20 Physical activity. A version of the Physical Activity History (PAH)21 was used to assess physical activity levels. Participants reported how frequently (never or less than once per month to ®ve or more times per week) they engaged in 13 activities for 20 min or more on each occasion during the previous year. The number of times answered for each activity was multiplied by a metabolic unit and then summed to obtain a total physical activity score. This version of the PAH has been shown to be negatively correlated to body weight and changes in body weight.22 As a measure of sedentary behavior, participants reported the average number of hours they watched television each day.
Statistics Baseline assessments were gathered from 1226 (998 women and 228 men) participants. Data from 1044 (826 women and 218 men) were included in the present analyses. The present analyses did not include data from the 106 women who became pregnant during the 3 y of treatment or 76 individuals who did not attend at least one follow-up assessment. The 76 individuals who had data only from the baseline assessment were compared to the 1044 individuals included in the present analyses. Baseline levels of restraint did not differ between the participants included in the analyses and those lost to follow-up. There were no gender or educational differences between the groups. Compared to the individuals lost to follow-up, the participants included in the analyses were older (35.2 6.3 vs 32.3 6.8), weighed less International Journal of Obesity
(BMI 27.0 5.7 vs 29.6 6.5), and were more likely to be white (94.3% vs 85.6%). Descriptive analyses were completed to evaluate the relationships between baseline levels of restraint and baseline demographic, weight and behavioral characteristics. Unchanging co-variates (gender and treatment group) and other co-variates (current BMI, dieting status and smoking status) are speci®ed when used. The second set of analyses assessed the relationships between baseline levels of restraint and changes in weight and behaviors within each participant over the four assessment periods. This was a within-person, prospective analysis. In order to investigate how baseline restraint (a ®xed variable) affected changes in weight and behavioral variables over time, it was allowed to interact with time. Changes in weight and weight-controlling behaviors were the dependent variables. Because this was a within-person analysis, controlling for unchanging co-variates (for example gender and treatment group) was irrelevant. However, changes in dieting status and smoking status were entered as co-variates; change in BMI was also used as a co-variate when the dependent variables were weight-controlling behaviors. The ®nal set of analyses evaluated the relationships between changes in restraint over the four assessment periods and changes in weight and behavioral variables over time. This set of analyses was also a within-person evaluation. To assess this relationship, two orthogonal restraint scores were computed: (1) an average restraint score representing the personal level was calculated as the mean restraint over the four time periods for each person, and (2) a deviation restraint score representing the within-person variance was calculated by subtracting the average restraint score from the restraint score at each time period. In this set of analyses, the deviation restraint variable was the primary independent variable and, again, changes in weight and weight-controlling variables were the dependent variables in separate analyses. Co-variates in this set of analyses included treatment group by time interaction, and changes in dieting status and smoking status; change in BMI was also used as a co-variate when weight-controlling behaviors were the dependent variables. Each set of analyses was repeated to evaluate the relationships between ¯exible and rigid restraint subscales and changes in weight and behavioral variables. Gender- and weight- (BMI < 25 kg=m2 or BMI > 25 kg=m2) speci®c analyses were also completed. These analyses did not differ from the analyses completed with the total sample so results from the total sample are presented. The proc MIXED program of the Statistical Analysis System (SAS)23 was used for all analyses. Because of multiple analyses, a signi®cance level of P < 0.01 was used.
Results
Demographic characteristics of the 1044 participants included in the following analyses are shown in Table 1. In
Relationship between restraint and weight MT McGuire et al
addition, this table shows average baseline and changes over the 3 y follow-up for weight and behavioral variables.
Baseline restraint and baseline characteristics Baseline total, ¯exible, and rigid restraint scores were similarly related to several other baseline variables (Table 2). Higher levels of restraint were reported from older individuals, women and non-smokers. Baseline restraint was positively related to current dieting status and negatively related to current weight, BMI and waist circumference measurement. Individuals who reported higher levels of restraint also reported more behaviors related to maintaining energy balance. For instance, higher restraint was related to lower percentage of calories from fat and sweets. Higher restraint was related to higher scores on the FHQ and its subscales.
Baseline restraint and changes in weight and behaviors There was no relationship between baseline levels of restraint and changes in weight, dietary intake or physical activity variables. However, the three baseline restraint measures were related to increases in weighing frequency. Figure 1 Table 1 Baseline demographic, weight, and behavioral variables and changes in weight and behavioral variables from baseline to 3 y followup: Pound of Prevention (n 1044) Baseline
3 y follow-up 7 baseline
Age 35.16 6.30 Ð Female (%) 79.1 Ð White (%) 88.6 Ð College (%) 87.7 Ð Maximum weight (lb) 173.47 41.87 Ð 2 27.88 6.08 Ð Maximum BMI (kg=m ) Current weight (lb) 167.12 38.44 3.86 14.26 2 27.02 5.74 0.62 2.35 Current BMI (kg=m ) Waist circumference (inches) 49.46 8.92 2.86 4.94 Currently dieting (%) 29.20 7 5.8 Currently smoking (%) 17.80 7 0.70 Restraint total score 8.11 4.38 7 0.11 3.59 Flexible restraint subscore 2.85 1.73 7 0.03 1.66 Rigid restraint subscore 3.00 1.83 7 0.05 1.57 Weighing frequency=month 4.79 9.06 7 0.33 8.59 Food Frequency Questionnaire Total calories 1755.10 1163.15 7 176.71 1091.08 Percent calories from fat 34.09 7.99 7 1.35 8.20 Percent calories from protein 16.30 3.26 0.44 3.64 Percent calories from carbohydrates 48.88 8.67 0.51 8.95 Percent calories from sweets 14.58 10.04 7 0.60 9.82 Percent calories from alcohol 2.59 4.27 0.42 3.81 Food Habits Questionnaire Total score 48.23 9.61 1.29 6.80 Modify meal preparation 11.87 3.00 0.39 2.35 Avoid fat 9.54 2.82 0.42 2.33 Replace modi®ed food for high fat 5.92 2.18 0.17 1.69 Substitute low fat for high fat 15.07 3.97 0.12 3.19 Fruit and vegetables 5.82 1.72 0.19 1.64 Physical activity score 46.77 31.65 7 2.88 30.50 Hours of television=day 2.28 2.15 7 0.38 1.78
shows the relationship between baseline total restraint and changes in weighing frequency over time for a non-smoker with a BMI of 27 and average baseline restraint scores. The ®gure shows that baseline restraint is strongly related to baseline weighing frequency. Although the relationship between baseline restraint and later reports of weighing frequency were also positive, the relationship lessened over time. The ®gures representing the relationships between ¯exible and rigid restraint and changes in weighing frequency were similar to those of total restraint but are not shown.
577
Changes in restraint and changes in weight and behaviors As shown in Table 3, increases in total, ¯exible, and rigid restraint scores were related to positive changes in weight and weight-controlling behaviors. As restraint increased over time, weight, total caloric intake, and the percentage of calories from fat and sweets decreased, and fat-reducing strategies and physical activity increased. Hours of television watching decreased as restraint increased over time.
Discussion
The present study evaluated the cross-sectional and prospective relationships between the EI's total, ¯exible and rigid restraint scores and weight and weight-controlling behaviors. This study found that the three measures of restraint were similarly associated with weight and behaviors that are related to energy balance. That is, higher total, ¯exible and rigid restraint were related to lower weight and more weightcontrolling behaviors; increases in these measures of restraint were related to weight loss and increasing weightcontrolling behaviors. Contrary to our initial hypotheses, the results from the present study did not completely replicate the ®ndings from the Westenhoefer study.12,13 In the Westenhoefer study, rigid and ¯exible restraint were related in different directions to weight and weight-controlling behaviors. There may be several reasons why the present study failed to replicate Westenhoefer's results. The ®rst reason may be due to methodological differences between the studies. In Westenhoefer's original study,12 small but signi®cant relationships were found for the original EI ¯exible and rigid restraint subscales among a sample of more than 50 000 individuals. In a follow-up study,13 ¯exible and rigid restraint results were strengthened when additional items were incorporated into the subscales, even though the sample size ( > 1800) was smaller than the original study. Because the present study used the original EI restraint subscales (the additional items were not available until after the current study's implementation) and sample size of 1044, it may have lacked the sensitivity to ®nd relationships between the restraint subscales and weight and behaviors. Another reason why the results from the present study differed from those in Westenhoefer's studies may be due to International Journal of Obesity
Relationship between restraint and weight MT McGuire et al
578 Table 2 Relationship between baseline total restraint, ¯exible restraint, and rigid restraint and baseline demographic, weight, and behavioral variables: Pound of Prevention (n 1044) Restraint
Rigid Restraint
P
b
P
0.1622 0.0254 0.0031 7 0.0001 0.0066 7 0.0135 0.0519 7 1.5241 7 0.1366 7 0.4373 0.3573
0.001 0.001 0.169 0.961 0.475 0.001 0.001 0.001 0.001 0.001 0.001
0.2831 0.0559 0.0022 7 0.0053 0.0001 7 0.0262 0.1046 7 4.8809 7 0.5265 7 1.2517 0.5278
0.012 0.001 0.703 0.369 0.995 0.001 0.001 0.001 0.001 0.001 0.003
0.3195 0.0654 0.0154 0.0018 0.0336 7 0.0296 0.1247 7 3.3801 7 0.2204 7 0.8841 0.8515
0.005 0.001 0.007 0.754 0.150 0.001 0.001 0.001 0.032 0.001 0.001
7 23.5008 7 0.4393 0.0877 0.3894 7 0.3264 0.0042
0.016 0.001 0.004 0.001 0.001 0.907
7 6.6323 7 0.6994 0.1043 0.6319 7 0.6304 0.0145
0.773 0.001 0.114 0.001 0.002 0.866
7 111.3209 7 0.7795 0.2174 0.7099 7 0.5155 0.1051
0.001 0.001 0.001 0.001 0.011 0.227
0.9806 0.1923 0.1946 0.1177 0.3738 0.1022 1.6605 7 0.0803
0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001
1.5625 0.3168 0.3027 0.1999 0.5410 0.2019 3.8349 7 0.0764
0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.069
2.4142 0.4917 0.4244 0.2767 1.0271 0.1942 2.0364 7 0.1645
0.001 0.001 0.001 0.001 0.001 0.001 0.002 0.001
b Age Gender Ethnicity Education Treatment group Currently smoking Currently dieting Current weight (lb) Current BMI (kg=m2) Waist circumference (inches) Weighing frequency=month Food Frequency Questionnaire Total calories Percent calories from fat Percent calories from protein Percent calories from carbohydrates Percent calories from sweets Percent calories from alcohol Food Habits Questionnaire Total score Modify meat preparation Avoid fat Replace modi®ed food for high fat Substitute low fat for high fat Fruit=vegetables Physical activity score Hours of television=day
Flexible restraint
b
P
Controlled for gender, treatment group, dieting status and smoking status; current BMI was controlled for in the behavioral analyses.
Figure 1 The relationship between baseline total restraint and changes in weighing frequency for a non-smoking individual with BMI of 27 and a restraint score of 8 4.
International Journal of Obesity
Relationship between restraint and weight MT McGuire et al Table 3 Relationship between changes in total, ¯exible and rigid restraint and changes in weight and behaviors from baseline to 3 y follow-up: Pound of Prevention (n 1044) Restraint b Current weight Current BMI Waist circumference Weighing frequency=month Food Frequency Questionnaire Total calories Percent calories from fat Percent calories from protein Percent calories from carbohydrates Percent calories from sweets Percent calories from alcohol Food Habits Questionnaire Total score Modify meat preparation Avoid fat Replace modi®ed food for high fat Substitute low fat for high fat Fruit and vegetables Physical activity score Hours of television=day
Flexible restraint P
b
579
Rigid Restraint P
b
P
7 0.7433 7 0.1301 7 0.1111 0.3160
0.001 0.001 0.001 0.001
7 1.3673 7 0.2602 7 0.3764 0.5572
0.001 0.001 0.001 0.001
7 1.1499 7 0.2105 7 0.2462 0.6181
0.001 0.001 0.001 0.001
7 27.9764 7 0.3255 0.1047 0.2204 7 0.3769 0.0018
0.001 0.001 0.001 0.001 0.001 0.920
7 40.1258 7 0.4835 0.1488 0.3099 7 0.6563 0.0385
0.001 0.001 0.001 0.001 0.001 0.306
7 56.8884 7 0.6108 0.2317 0.4129 7 0.5988 7 0.0382
0.001 0.001 0.001 0.001 0.001 0.361
0.6053 0.1348 0.1327 0.0647 0.2405 0.0728 1.2646 7 0.0305
0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.001
0.8484 0.2195 0.2019 0.0813 0.2949 0.1312 2.3603 7 0.0479
0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.018
1.1455 0.2251 0.2283 0.1299 0.5415 0.1361 2.0331 7 0.0460
0.001 0.001 0.001 0.001 0.001 0.001 0.001 0.038
Controlled for treatment by time interaction, changes dieting status and changes in smoking status; changes in BMI were controlled for in the behavioral analyses.
differences in sample characteristics. Previous studies have shown that about 30% of overweight individuals who seek weight loss treatment meet DSM-IV criteria for binge eating disorder (BED).24 Individuals who have BED report high levels of disinhibitory eating behaviors, more extreme dieting attitudes, and a greater history of weight losses.25 It is possible that a high proportion of Westenhoefer's sample had BED or BED symptomatology and that these individuals endorsed many rigid restraint items and failed to lose weight during treatment. Thus, in the Westenhoefer study, the individuals with these extreme behavioral strategies and attitudes toward weight control may have driven the positive relationship between rigid restraint and BMI. In the present study, only 1.6% of the sample met DSM-IV criteria for BED.26 It is unlikely that a large proportion of POP participants solely used rigid behavioral strategies to control their weight. Instead, it is likely that these individuals used a combination of strategies that are better re¯ected by the EI's total restraint scale rather than either the ¯exible or rigid restraint subscales. The results from the present study are similar to those by Williamson,11 which surveyed nearly 300 overweight women and found that total, ¯exible and rigid restraint were similarly related to BMI. The similarities may be because both studies included heterogeneous community samples and because neither sample was recruited to join a weight loss program. As previously mentioned, the majority of the individuals in these community samples were probably not engaging in extreme or rigid weight controlling behaviors. These studies suggest that the EI's rigid restraint
subscale may not give additional or different information about weight controlling behaviors than the EI's total restraint scale. It is interesting that no study that evaluated EI's ¯exible and rigid restraint has replicated Westenhoeffer's study. Although the studies' methodology, sampling methods and treatment characteristics may explain the differences in results, cultural differences may also contribute to result differences. In addition, all of the studies mentioned thus far used the seven-item ¯exible and rigid restraint subscales. It will be interesting to see how the revised rigid and ¯exible restraint subscales are related to weight and weight-controlling behaviors in samples that include individuals other than those in weight loss intervention studies. Only a few studies evaluated how initial EI restraint predicts future changes in weight and behaviors4,8 ± 10,27 and only one4 found any relationship. In that study, lower initial restraint scores predicted better weight losses at the end of treatment. Because behavioral weight loss treatments speci®cally focus on improving behavioral strategies that are re¯ected in the EI's restraint scale, participants with the lowest restraint levels may have bene®tted the most from treatment. The results from that study are somewhat contrary to those in the present POP analyses. In the present study, higher baseline restraint predicted increases in selfweighing over the 3 y follow-up. Self-weighing was speci®cally recommended as a way to prevent weight gain over time. Increases in self-weighing were also the strongest correlates of weight change over the 3 y of the treatment.16 Thus, the relationship we observed between baseline International Journal of Obesity
Relationship between restraint and weight MT McGuire et al
580
restraint and changes in self-weighing may be a re¯ection of participants' initial motivation and, subsequently, the participants' adherence to treatment recommendations. This interpretation would be consistent with observations in other studies of restraint predicting treatment adherence.28 This large, community-based weight gain prevention intervention found that, contrary to clinical samples,12,13 rigid restraint was inversely related to weight and weightrelated behaviors. This study suggests that in a sample with very low rates of binge eating and other extreme eating behaviors,26 the EI's total restraint score is as good a re¯ection of weight and behaviors as the ¯exible and rigid restraint subscales. The present study also showed that initial levels of restraint predicted changes in self-weighing, suggesting that initial restraint may be a marker of motivation for treatment adherence. Weight loss maintenance and weight gain prevention have important public health implications especially as our nation becomes more overweight and obese. Continuing to understand effective and ineffective methods used for weight control may improve clinicaland population-based efforts to reduce the prevalence of obesity.
References
1 Hill JO, Peters JC. Evnironmental contributions to the obesity epidemic. Science 1998; 280: 1371 ± 1374. 2 Weinsier RL. Genes and obesity: is there reason to change our behaviors? Ann Intern Med 1999; 130: 938 ± 939. 3 Stunkard AJ, Messick S. The three-factor eating questionnaire to measure dietary restraint, disinhibition and hunger. J Psychosom Res 1985; 29: 71 ± 83. 4 Foster GD, Wadden RM, Swain RM, Stunkard AJ, Platte P, Vogt RA. The eating inventory in obese women: clinical correlates and relationship to weight loss. Int J Obes Relat Metab Disord 1999; 22: 778 ± 785. 5 French SA, Jeffery RW, Wing RR. Food intake and physical activity: a comparison of three measures of dieting. Addict Behav 1994; 19: 401 ± 409. 6 Laporte DJ, Stunkard AJ. Predicting attrition and adherence to a very low calorie diet: a prospective investigation of the eating inventory. Int J Obes 1990; 14: 197 ± 206. 7 Bjorvell H, Rossner S, Stunkard A. Obesity, weight loss, and dietary restraint. Int J Eat Disord 1986; 5: 727 ± 734. 8 Clark MM, Marcus BH, Pera V, Niaura RS. Changes in eating inventory scores following obesity treatment. Int J Eat Disord 1994; 15: 401 ± 405. 9 Karlsson J, Hallgren P, Kral J, Lindroos AK, Sjostrom L, Sullivan M. Predictors and effects of long-term dieting on mental well-being and weight loss in obese women. Appetite 1994; 23: 15 ± 26. 10 Pekkarinen T, Takala I, Mustajoki P. Two year maintenance of weight loss after a VLCD and behavioural therapy for obesity: correlation to the scores of questionnaires measuring eating behaviour. Int J Obes Relat Metab Disord 1996; 20: 332 ± 337.
International Journal of Obesity
11 Williamson DA, Lawson OJ, Brooks ER, Wozniak PJ, Ryan DH, Bray GA, Duchmann EG. Association of body mass with dietary restraint and disinhibition. Appetite 1995; 25: 31 ± 41. 12 Westenhoefer J. Dietary restraint and disinhibition: is restraint a homogeneous construct? Appetite 1991; 16: 45 ± 55. 13 Westenhoefer J, Stunkard AJ, Pudel V. Validation of the ¯exible and rigid control dimensions of dietary restraint. Int J Eat Disord 1999; 26: 53 ± 64. 14 Shearin EN, Russ MJ, Hull JW, Clarkin JF, Smith GP. Construct validity of the three-factor eating questionnaire: ¯exible and rigid control subscales. Int J Eat Disord 1994; 16: 187 ± 198. 15 Jeffery RW, French SA. Preventing weight gain in adults: design methods and one year results from the pound of prevention study. Int J Obes Relat Metab Disord 1997; 21: 457 ± 464. 16 Jeffery RW, French SA. Preventing weight gain in adults: the pound of prevention study. Am J Public Health 1999; 89: 747 ± 751. 17 Block G, Subar AF. Estimates of nutrient intake from a food frequency questionnaire: the 1987 National Health Interview Survey. J Am Diet Assoc. 1992; 92: 969 ± 977. 18 Block G, Hartman AM, Naughton D. A reduced dietary questionnaire: development and validation. Epidemiology 1990; 1: 58 ± 64. 19 Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 1986; 124: 453 ± 469. 20 Kristal AR, Shattuck AL, Henry HJ. Patterns of dietary behavior associated with selecting diets low in fat: reliability and validity of a behavioral approach to dietray assessment. J Am Diet Assoc 1990; 90: 214 ± 220. 21 Jacobs DR, Hahn LP, Haskell WL, Pirie P, Sidney S. Validity and reliability of short physical activity history: CARDIA and the Minnesota heart health program. J Cardiopulmon Rehab 1989; 9: 448 ± 459. 22 French SA, Jeffery RW, Forster JL, McGovern PG, Kelder SH, Baxter JE. Predictors of weight change over two years among a population of working adults: the healthy worker project. Int J Obes Relat Metab Disord 1994; 18: 145 ± 154. 23 SAS Institute. SAS language and procedures: usage version 6.12. SAS Institute: Cary, NC, 1996. 24 Spitzer RL, Devlin M, Walsh BT, Hasin D, Wing R, Marcus M, Stunkard A, Wadden T, Yanovski S, Agras S, Mitchell J, Nonas C. Binge eating disorder: a multisite ®eld trial of the diagnostic criteria. Int J Eat Disord 1992; 11: 191 ± 203. 25 Spitzer RL, Yanovski S, Wadden T, Wing R, Marcus MD, Stunkard A, Devlin M, Mitchell J, Hasin D, Horne RL. Binge eating disorder: its further validation in a multisite study. Int J Eating Disord 1993; 13: 137 ± 153. 26 French SA, Jeffery RW, Sherwood NE, Neumark-Sztainer D. Prevalence and correlates of binge eating in a nonclinical sample of women enrolled in a weight gain prevention program. Int J Obes Relat Metab Disord 1999; 23: 576 ± 585. 27 McGuire MT, Wing RR, Klem ML, Lang W, Hill JO. What predicts weight regain in a group of successful weight losers? J Consult Clin Psychol 1999; 67: 177 ± 185. 28 Stunkard AJ, Wadden TA. Restrained eating and human obesity. Nutr Rev 1990; 48: 78 ± 86.
