SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Objective: To assess the relationship between dieting and subsequent weight change and whether the association varies by gender or race/ethnicity.

Research Methods and Procedures: Male (n = 4100) and female (n = 4302) participants in the National Longitudinal Study of Adolescent Health who provided information on weight and height at baseline and two follow-up assessments and were not missing information on weight control strategies or race were studied. Generalized estimating equations were used to assess whether dieting to lose or maintain weight at Wave I or II predicted BMI (kg/m2) change between adolescence and young adulthood (Wave II to III). Analyses were stratified by gender and took sampling weights and clustering into account.

Results: At Wave I, the mean age of the participants was 14.9 years. Approximately 29.3% of female participants and 9.8% of male participants reported dieting in Wave I or II. Fewer African Americans than whites (6.2% vs. 10.0% and 25.5% vs. 31.2%, p = 0.007 and p = 0.02, among males and females, respectively) reported dieting. Between Waves II and III, participants gained on average 3.3 kg/m2. Independent of BMI gain during adolescence (Waves I to II), female participants who dieted to lose or maintain weight during adolescence made larger gains in BMI during the 5 years between Waves II and III (mean additional gain, 0.39 kg/m2; 95% confidence interval, 0.08 to 0.71) than their nondieting peers. The association was not significant among the male participants. The association was largest among African-American female participants.

Discussion: The results suggest that not only is dieting to lose weight ineffective, it is actually associated with greater weight gain, particularly among female adolescents. Female African-American dieters made the largest BMI gains.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Obesity is a serious public health problem in the United States. During the past 2 decades, the prevalence of overweight has more than doubled among children and adolescents (1). According to the 1999 to 2002 National Health and Nutrition Examination Survey, approximately 31% of children and adolescents were overweight or at risk for overweight {i.e., BMI [weight (kilograms)/height (meters) squared] ≥ national 85th percentile for age and sex} (2). The prevalence of overweight is particularly high among African Americans and Hispanics. Among adolescents, approximately 37% of African Americans and 41% of Mexican Americans are at risk for overweight or overweight, compared with 28% of white adolescents (2). The high prevalence rates are of concern because obesity is a risk factor for the development of numerous chronic diseases, including hypertension (3, 4), diabetes (3, 5), and postmenopausal breast cancer (6). Preadolescents and adolescents who are overweight are likely to become overweight adults (7, 8); thus, the prevention of excessive weight and excessive weight gain during preadolescence, adolescence, and young adulthood is of paramount importance.

Despite the high prevalence of overweight and obesity, the desire to be thin is still widespread among females, particularly white and Hispanic females (9, 10, 11, 12). African-American females tend to be more tolerant of larger body sizes (11, 13); however, several studies have observed high levels of body dissatisfaction among both white and African-American females (11, 13, 14). Moreover, among young girls, dissatisfaction with body shape [as measured by the Eating Disorder Inventory (15)] was higher among white girls but increased with increasing BMI among both black and white girls (16). In addition, in a multiethnic sample of 17,000 girls in 7th to 12th grade in Minnesota, Hispanic, and African-American girls were more likely than white girls to diet frequently, use laxatives, or have vomited to control their weight (17). In addition, among 939 girls in sixth and seventh grade in California, Hispanic girls had lower body satisfaction than white girls (10). Thus, body dissatisfaction and efforts to control weight seem to be relatively common among females from a variety of racial/ethnic backgrounds. Although dieting is less common among males than females (18, 19), recent data suggest that weight concerns may be becoming more prevalent (20). Little is known about body dissatisfaction and weight control behaviors among minority and nonminority young males.

In cross-sectional studies, researchers have observed a strong association between dieting and being overweight or obese (21). In addition, in the Growing Up Today Study, a prospective cohort study of 16,000 preadolescents and adolescents, Field et al. (22) observed that independent of their weight, girls and boys who dieted gained more weight than their peers. Similar results were observed by Stice et al. (23), who followed 692 adolescent girls for 4 years, and Tanofsky-Kraff et al. (24), who followed 146 preadolescent girls and boys for an average of 4 years. Stice et al. (23) observed that dieters were more likely to become obese and Tanofsky-Kraff et al. (24) observed that dieting was predictive of greater gains in body fat. However, none of these studies had a sufficiently large sample of minority youth to assess whether the association between dieting and weight gain occurs in a variety of race/ethnic groups. To assess whether adolescents who diet to lose or maintain weight gain more weight, relative to height, than nondieters, independent of their activity, inactivity, and region of the country and to examine gender and race/ethnic differences in these associations, we analyzed data from the National Longitudinal Study of Adolescent Health (Add Health)1. Add Health is a prospective study sampled to be representative of students in grades 7 through 12 in the United States.

Research Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Sample

In 1994 to 1995, a sample of 20,745 students from 80 high schools and 52 middle schools in the U.S., which were selected with unequal probability of selection, completed an in-home questionnaire (Wave I). The Add Health study design incorporated systematic sampling methods and implicit stratification to ensure that the sample was representative of U.S. schools with respect to region of country, urbanicity, school size, school type, and ethnicity. Approximately 1 year later, 14,738 school-aged adolescents (including drop-outs) completed another in-home questionnaire (Wave II). In April 2001 through August 2002, 15,197 of the original Wave I participants completed an in-home interview (Wave III).

Measures

Age, physical activity, inactivity, weight change efforts, weight, and height were assessed on each of the three waves of data collection. Race and region of residence (South, Northeast, Midwest, and West) were assessed at baseline (Wave I). Participants who were younger than 14 years of age at Wave I were classified as young adolescents, and those who were 14 years of age or older at Wave I were classified as older adolescents.

Physical Activity

Physical activity was assessed with three questions at Waves I and II: “During the past week, how many times did you play an active sport, such as baseball, softball, basketball, soccer, swimming, or football?”; “During the past week, how many times did you go roller-blading, roller-skating, skate-boarding, or bicycling?”; and “During the past week, how many times did you do exercise, such as jogging, walking, karate, jumping rope, gymnastics, or dancing?” A summary score of times per week that the participant engaged in physical activity at Waves I and II was computed from these three questions.

Inactivity

Physical inactivity was assessed with the three questions at Waves I and II: “How many hours per week do you watch television?”; “How many hours per week do you watch videos?”; and “How many hours per week do you play video or computer games?” Because the categories are not mutually exclusive, in the analyses we restricted inactivity to hours per week of television viewing.

Weight Control Behaviors

At Waves I and II, participants were asked, “Are you trying to lose weight, gain weight, stay the same weight, or not trying to do anything about your weight?” Participants who responded that they were trying to lose or maintain weight were then asked, “During the past seven days, which of the following things did you do to lose weight or to keep from gaining weight?” The behaviors were dieted, exercised, self-induced vomiting, took diet pills, took laxatives, or other. Two sets of dieting variables were created. Indicators for the number of years that a participant dieted (0, 1, or 2 years) were created. In addition, participants who reported dieting in either the Wave I or II assessments were classified as dieters. The dichotomization into dieters (at Wave I or II) vs. no dieting was used as the primary predictor of BMI change. In secondary analysis, we assessed whether unhealthy weight control behaviors (vomiting, using laxatives, or using diet pills) were predictive of BMI change between Waves II and III. Many people used more than one weight control measure. Exercise and diet to lose or maintain weight were the most common combination. Among the male participants, there were too few who only dieted and did not exercise to control weight to assess the independent associations of dieting and exercise alone with subsequent weight gain. Therefore, exercise to lose or maintain weight was not examined as an independent predictor of change in BMI.

Weight, Height, and BMI

At Wave I, weight and height were self-reported, whereas at Waves II and III, in addition to self-report, weights and heights were measured. We calculated BMI [weight (kilograms)/height (meters)2] using self-reported weight and height information at Wave I and measured weights and heights at Waves II and III. Because the correlation between self-reported and measured weight (r = 0.96) and BMI (r = 0.94) at Wave II was very strong, we felt confident that the self-reported information was a good proxy for the true measured values when they were not available. If measured weights and heights were not available at Wave II or III, self-reported information was used to estimate the prevalence of overweight and obesity based on the International Obesity Task Force (IOTF) cut-offs (25), which are age- and gender-specific and provide comparability in assessing overweight and obesity from adolescence to adulthood. The IOTF standards were used instead of the Centers for Disease Control and Prevention standards (26) because they were developed so that the gender and age-specific cut-off points map to a BMI > 25 or 30 kg/m2 for overweight and obesity, respectively, at age 18 (25). Adolescents who had a BMI between the 85th and 94th IOTF percentiles for age and gender were classified as overweight, and those ≥95th percentile were classified as obese.

Children who reported a BMI < 12 kg/m2 (which was considered unlikely due to being off the BMI charts) or were above the 99.7th percentile for age and gender in terms of BMI, height change, or BMI change were considered to be data errors and, therefore, set to missing and not used in the analysis. The same outlier detection rule has been used for longitudinal analyses in the Growing Up Today Study (27). To explore the impact of errors in height, children whose height measurements were more than 1 inch less than a previous report were examined. If the same lower height was reported in Waves II and III, the Wave I height was corrected to be the lower value and the participant was retained in the analysis; however, if lower height was reported on only Wave II or III, the information was set to missing, and the participant was excluded from the analysis. After excluding outliers and data errors, two BMI change variables were computed: BMI change between Waves I and II (adolescent weight change) and BMI change between Waves II and III (weight change from adolescence to young adulthood). BMI change between Waves II and III was the outcome in the main analyses. The development of obesity by Wave III was the outcome in some secondary analyses.

Sample for Analysis

Participants who did not complete all three waves of data collection (n = 8096); were missing information on weight or height at any wave (98 male and 295 female participants); were outliers on weight or height (three male and 21 female participants), BMI (two male and one female participants), and height change or BMI change (45 male and 48 female participants) at any of the three waves of data collection; had their height decrease by more than 1 inch (743 male and 1068 female participants); or were missing information on race (one male and one female participant) were excluded from the analysis, thus leaving 4101 male and 4302 female participants who were 11 to 20 years of age in 1995 to 1996 (Wave I) for analysis.

Analysis

All analyses were stratified by gender and conducted with SUDAAN 9.0 (Research Triangle Institute, Research Triangle Park, NC) or SAS 8.2 software (SAS Institute Inc., Cary, NC). To account for the stratified sampling strategy and clustering, sample weights were used in all analyses. Generalized estimating equations (GEEs) were used to assess predictors of change in BMI from adolescence to young adulthood (Wave II to III). All models controlled for age at Wave I, mean activity in adolescence (i.e., mean of Waves I and II measures), mean inactivity (i.e., television viewing) in adolescence, BMI at Wave I, change in BMI between Waves I and II, race, region of the country at Wave I, and length of follow-up. Dieting to lose or maintain weight was modeled with dichotomous variables, participants that indicated in Wave I or II that they had dieted during the past week to lose or maintain weight were classified as dieters. Adolescents who did not report dieting at either time-point were the reference group to which dieters were compared. All p values are two-sided, with p < 0.05 considered statistically significant.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

At Wave I, the mean age of the sample was 14.9 years, and the mean BMI of the participants was 22.2 kg/m2 for the male participants and 22.0 kg/m2 for the female participants (Table 1). Approximately 19.9% of the participants were overweight, and an additional 8.2% were obese. Between Waves I and II, which was approximately 1 year, participants increased their BMI by a mean of 0.6 kg/m2, and the prevalence of obesity increased to 10.3%, whereas the prevalence of overweight decreased slightly. The increase in the prevalence of obesity was greater for female than male participants (Figure 1), and more female participants were trying to lose weight (54.5% vs. 26.5%, respectively).

Table 1.  Demographic and descriptive statistics* of participants in the National Longitudinal Study of Adolescent Health (Waves I and II)
 Females (n = 4302)Males (n = 4100)
  • *

    Weighted for national representation.

Age in years [mean (SD)]14.9 (0.1)15.0 (0.1)
BMI at Wave I [mean (SD)]22.0 (0.1)22.2 (0.1)
Change in BMI from Wave I to II [mean (SD)]0.7 (0.1)0.6 (0.1)
Overweight or obese at Wave I  
 Overweight17.9%21.7%
 Obese7.2%9.1%
Tried to lose weight  
 1 Year (Wave I or Wave II)21.4%12.8%
 2 Years (Wave I and Wave II)33.1%13.7%
Tried to maintain weight  
 1 Year (Wave I or Wave II)27.7%31.1%
 2 Years (Wave I and Wave II)22.7%20.8%
Dieted to lose or maintain weight  
 1 Year (Wave I or Wave II)21.9%8.1%
 2 Years (Wave I and Wave II)8.4%1.7%
Race/ethnicity  
 White69.6%66.6%
 Hispanic10.1%11.8%
 African-American14.0%14.1%
 Asian3.6%4.2%
 Other2.6%3.3%
Region of residence (Wave I)  
 Midwest33.0%30.3%
 West14.0%14.6%
 South35.4%38.3%
 Northeast17.7%16.8%
image

Figure 1. Prevalence of overweight and obesity at each wave among 8402 adolescents in the Longitudinal Study of Adolescent Health.

Download figure to PowerPoint

Among male participants, the prevalence of trying to lose weight was significantly higher among Hispanics (36.5%) than whites (25.2%, p = 0.0003). Among female participants, there was a suggestion that Hispanics (58.0%) were more likely than whites (55.2%) to be trying to lose weight, but the difference was not significant. The prevalence was lowest among the African Americans (23.2% of male and 51.1% female participants), but the difference with whites was significant only among the female participants (p = 0.02). However, in terms of dieting, African Americans were significantly less likely to have dieted at Wave I or II than whites (Figure 2). In contrast, the prevalence of dieting was similar among Hispanics and whites.

image

Figure 2. Racial/ethnic group differences in the prevalence of dieting to lose or maintain weight at Wave I or II among 8402 adolescents in the Longitudinal Study of Adolescent Health. * Compared to whites of the same gender, p < 0.05.

Download figure to PowerPoint

Between Waves II and III, a period of approximately 5 years, the mean BMI of the participants increased by approximately 3.3 kg/m2, and there was a large increase in the prevalence of overweight and obesity. At Wave III, the prevalence of overweight and obesity were 21.3% and 21.9%, respectively, among female participants and 30.5% and 19.7%, respectively, among male participants. BMI at Wave I was positively associated with gain in BMI between Waves II and III. For each 1 kg/m2 difference in BMI at Wave I, young adults gained ∼1.1 additional BMI units during the follow-up (Waves II to III).

Dieting and Weight Change

Female dieters gained more weight than their nondieting peers. Female participants who dieted for 1 year (β= 0.33 kg/m2) gained less weight than those who dieted for 2 years (β = 0.42 kg/m2) during adolescence (i.e., Waves I and II), but there were too few male participants who dieted for 2 years to conduct meaningful analyses; thus, dieting was modeled as dieting for at least 1 year. Independent of their Wave I BMI and change in BMI between Waves I and II, female dieters significantly gained more weight during the follow-up [β = 0.39 kg/m2 additional gain; 95% confidence interval (CI), 0.08 to 0.71] than their nondieting peers (Table 2). The magnitude of the association with dieting was stronger among female participants who were older adolescents (β = 0.49) vs. younger adolescents (β = 0.18) at Wave I. Among male participants, dieting was not significantly associated with change in BMI, overall or within age group (older adolescents vs. younger adolescents); however, the point estimates overall and among the adolescents and young adults were similar to those observed among the female participants.

Table 2.  Predictors of change in BMI over 5 years of follow-up (Waves II to III)* among adolescents and young adults in the National Longitudinal Study of Adolescent Health
 FemalesMales
  • *

    β Values and 95% CIs from multivariate gender-specific GEE models that account for the sampling strategy by using weights and include all the variables in the table and activity level in adolescence, mean hours per week television viewing in adolescence, region of the country at Wave I, and race.

Age−0.18 (−0.28 to −0.08)−0.20 (−0.29 to −0.11)
BMI at Wave I1.15 (1.10 to 1.19)1.05 (1.00 to 1.10)
Change in BMI from Wave I to II0.91 (0.84 to 0.99)0.66 (0.58 to 0.73)
Dieting in adolescence  
 NeverReferentReferent
 ≥1 Year0.39 (0.08 to 0.71)0.46 (−0.20 to 1.12)

Among female participants, the association of dieting to subsequent weight gain varied by race/ethnicity. Among both whites and African Americans, dieters gained significantly more BMI units than their nondieting peers (β = 0.39 and 1.09 kg/m2, respectively; Table 3). There was a suggestion that African-American dieters gained more than white dieters, but the interaction was not significant (β = 0.64, p = 0.1). Dieting was not associated with weight change among Hispanic female participants. Among the male participants, dieting was not a significant predictor of subsequent change in BMI in any race/ethnic group (data not shown).

Table 3.  Racial/ethnic group-specific estimates of the association of dieting to change in BMI over 5 years of follow-up (Wave II to III)* among female adolescents and young adults in the National Longitudinal Study of Adolescent Health
 WhitesAfrican AmericansHispanics
  • *

    β Values and 95% CIs from multivariate gender and racial/ethnic-specific GEE models that account for the sampling strategy by using weights and include all the variables in the table and age, region of the country at Wave I, activity level in adolescence, and mean hours per week television viewing in adolescence.

BMI at Wave I1.14 (1.09 to 1.20)1.11 (1.01 to 1.21)1.12 (1.01 to 1.24)
Change in BMI from Wave I to II0.89 (0.79 to 1.00)0.70 (0.63 to 0.96)0.80 (0.60 to 1.00)
Dieting in adolescence   
 NeverReferentReferentReferent
 ≥1 Year0.37 (0.00 to 0.74)1.06 (0.08 to 2.04)0.17 (−0.79 to 1.14)

In secondary analyses, we assessed the association of weight control efforts to subsequently becoming obese. The results were similar to those when BMI change was the outcome. We observed that female dieters were 36% more likely (odds ratio, 1.32; 95% CI, 1.04 to 1.68) than nondieters to become obese (i.e., BMI ≥ 30 kg/m2) during the follow-up, whereas among males participants, there was no relationship between dieting and becoming obese (odds ratio, 1.12; 95% CI, 0.72 to 1.76). In other secondary analyses we assessed whether the use of unhealthy weight control methods (using vomiting, laxatives, or diet pills) was related to change in BMI. Only 31 boys had used unhealthy weight control behaviors, so the analysis was restricted to female participants. Approximately 3.1% (n = 160) of female participants had used unhealthy weight control behaviors. Among those female participants, 114 also dieted to lose or maintain weight. The point estimate for the association between using unhealthy weight control behaviors and subsequent BMI gain (β = 0.55) was similar to the association seen for dieting and weight gain, but it was not significant (p = 0.07). The association may have been at least partially due to confounding by dieting status because dieting alone and dieting or using unhealthy weight control methods (i.e., using laxatives, vomiting, or diet pills) had similar associations with subsequent BMI change.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We observed that in a representative sample of adolescents and young adults in the United States, female dieters made larger gains in BMI than nondieters. Although dieting was not significantly associated with BMI change among the male participants, the point estimates were similar among male and female participants. The associations with dieting were independent of the young adult's baseline weight and weight change in adolescence. The findings among the female participants are consistent with those of Stice (23), Field (22), and Tanofsky-Kraff (24), who observed in their respective prospective studies that dieters gained more weight than nondieters. Stice et al. (23) observed that among 692 adolescent girls who were followed for 4 years, dieters gained more weight than nondieters. In the Growing Up Today Study, Field et al. (22) observed that dieting was associated with weight gain over 3 years among the 8203 girls and 6769 boys who were preadolescents and adolescents. Moreover, in a cohort of 146 children who were 6 to 12 years old at baseline and who were followed for a mean of 4 years, Tanofsky-Kraff (24) observed that both binge eating and dieting independently predicted greater increases in body fat. None of these studies had a sufficiently large sample of minority youth to examine whether the association between dieting and weight or body fat change varied by race. Our results suggest that among female dieters, African-American dieters gain at least as much additional weight as white dieters. However, dieting was not associated with weight gain among the Hispanic females. The gender and race/ethnic group differences may reflect differences in how dieting is conceptualized among males and females and among white, African-American, and Hispanic adolescents.

One criticism of the earlier studies is that they have not taken into account that young people may be dieting because they are overweight; thus, it is unclear whether it is their weight gain trajectory that led them to diet or the dieting that predicted greater weight gain. In the present analysis, we adjusted for baseline BMI and weight change during the period that dieting was assessed. We observed that both baseline BMI and BMI change were significant predictors of later BMI change, but independent of these two confounders, female dieters gained more weight than nondieters. It is unclear whether the lack of a similar finding among the boys was a result of insufficient statistical power or a true lack of association. In the Growing Up Today Study, the analysis included almost twice as many male participants (22) as we had in this analysis and included a broader measure of dieting (frequency of dieting in the past year vs. dieting in the past week) that may have resulted in less misclassification. Other large-scale prospective studies are needed to better understand the association between weight control behaviors and subsequent weight change among young males.

There are some limitations to the current study. First, the assessment of dieting was asked only about the last week. It is, therefore, likely that some sporadic dieters were misclassified; thus, it is likely that our estimates underestimate the true association. Moreover, no information on the composition of the self-selected diet was collected in Add Health; thus, it is unclear what was considered dieting to these young people. Second, there was no measure of binge eating at Wave I or II; thus, we were unable to assess whether dieting was associated with binge eating and whether binge eating was an independent predictor of BMI change or a confounder of the dieting effect on BMI change. In the Growing Up Today Study, binge eating was strongly associated with dieting and independently predicted weight gain among the male participants (22), and Tanofsky-Kraff found that dieting and binge eating were both independent predictors of increases in body fatness (24). Thus, it is likely that a measure of binge eating might have been useful for better understanding the role of dieting in weight change. One other limitation is that at Wave I, only self-reported weight and height information was collected. Because people tend to underestimate their weight and overestimate their height (28, 29, 30), there is likely some misclassification of weight status at Wave I, and BMI change between Waves I and II is likely slightly biased. However, the correlation between self-reported and measured weight (r = 0.96) and BMI (r = 0.94) at Wave II, which was collected 1 year after Wave I, was very strong; thus, the self-reported information appeared to be a good proxy for the true measured values when they were not available. Moreover, in a validation study conducted with the Add Health population, Goodman et al. (31) observed that less than 4% of youth were misclassified using self-reported weight and height information. Moreover, in another validation study using Add Health data, we have found that self-reported weight was slightly lower than measured weight at Waves II and III, but weight change based on self-reported weights underestimated true weight change by only 2.2 (female participants) to 2.7 (male participants) pounds. Although overweight and obese female participants underreported their weight more than their leaner peers, they were consistent in their underreporting. Consequently, the discrepancy between weight change based on serial self-reports vs. measured weights was significantly smaller among the obese female participants vs. healthy-weight female participants (0.5-pound overestimation vs. 2.5-pound underestimation, p < 0.001). Among the male participants, the same pattern was evident. Being African American or Hispanic, physical activity level, hours per week watching television, and weight change efforts were not related to the discrepancy between weight change based on self-reported vs. measured weights. Thus, the use of self-reported weight and height at Wave I is unlikely to have introduced substantial bias because Wave I data were collected only 1 year earlier than Wave II, the data used in the validation analyses. An additional limitation is the lack of control for maturation and the possibility of misclassification of BMI change due to biological changes of puberty. Despite these limitations, there are many strengths of the current study. First, the sample was large, racially and ethnically diverse, and representative of youth in the United States. Second, measured weights and heights were used to estimate BMI at Waves II and III. Third, this is the first prospective investigation of dieting and BMI change that accounts for prior weight status and weight change. Finally, as has been previously shown, there is a tremendous amount of weight gain and a high incidence of becoming obese by Wave III in the Add Health study (32); thus, it is a good sample to study predictors of weight change and the development of obesity.

Although it may seem counterintuitive that self-selected dieting is associated with greater weight gain as opposed to weight loss, it is important to consider that the concept of dieting implies a relatively short-term change in dietary intake vs. a more permanent lifestyle change. Because drastic changes in dietary intake are rarely sustainable, it is not surprising that self-described dieting does not protect against weight gain. It should be noted that self-described dieting reflects a variety of approaches to changing dietary intake (33, 34, 35, 36); thus, self-selected diets may be quite unlike those assigned in clinical trials that evaluate the effectiveness of a variety of dietary approaches to weight loss. Moreover, in population-based observational studies, dieting is associated with binge eating (22, 37), which suggests that some young adults may alternate between periods of restrictive eating and periods of overeating. It is possible that the overeating episodes negate any positive effect of the periods of reduced intake. For other youth, the diets may be of insufficient duration to have a beneficial impact on weight. Taken together, the data suggest that many people are unhappy with their weight and willing to make at least short-term changes in dietary intake, but either they are self-selecting ineffective diets and/or are unable to make long-term beneficial changes in dietary intake. It remains unclear why self-selected dieting predicts greater weight gain, but although researchers continue to study the mechanisms, it would be prudent to encourage young adults to adopt a modest and, therefore, sustainable weight control strategy that includes physical activity and does not require severe restriction of total calories or components of the diet, such as percentage of calories from fat (38, 39). Although in the short term a restrictive diet may be beneficial for weight loss (40), in the long term, our data suggest that dieting to control weight is not only ineffective, it may actually promote weight gain, at least among young women and obese youth.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The analysis was supported by the National Institutes of Health (Grant DK-065085) and by the Boston Obesity Nutrition Research Center (Grant DK 46200). This research uses data from Add Health, a program project designed by J. Richard Udry, Peter S. Bearman, and Kathleen Mullan Harris, and funded by the National Institute of Child Health and Human Development (Grant P01-HD31921), with cooperative funding from 17 other agencies. We acknowledge Ronald R. Rindfuss and Barbara Entwisle for assistance in the original design. Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W. Franklin Street, Chapel Hill, NC 27516-2524 (E-mail: addhealth@unc.edu). We thank Elizabeth Goodman for comments and suggestions on the manuscript and Nan Laird for input on approaches to analyzing the data.

Footnotes
  • 1

    Nonstandard abbreviations: Add Health, National Longitudinal Study of Adolescent Health; IOTF, International Obesity Task Force; GEE, generalized estimating equation; CI, confidence interval.

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References