Paper | Published:

Predictors of weight gain in the Pound of Prevention study

International Journal of Obesity volume 24, pages 395403 (2000) | Download Citation

Subjects

Abstract

OBJECTIVES: This study examined cross-sectional and prospective relationships between macronutrient intake, behaviors intended to limit fat intake, physical activity and body weight.

DESIGN: The overall goal was to identify diet and exercise behaviors that predict and/or accompany weight gain or loss over time. Specific questions addressed included: (a) are habitual levels of diet or exercise predictive of weight change; (b) are habitual diet and exercise levels associated cross-sectionally with body weight; and (c) are changes in diet and exercise associated with changes in body weight over time?

PARTICIPANTS: Subjects were a sample of community volunteers (n=826 women, n=218 men) taking part in a weight gain prevention project over a 3-year period.

MEASURES: Body weight was measured at baseline and annually over the study period. Self-report measures of diet and exercise behavior were also measured annually.

RESULTS: Among both men and women, the most consistent results were the positive association between dietary fat intake and weight gain and an inverse association between frequency of physical activity and weight gain. Individuals who weighed more both ate more and exercised less than those who weighed less. Individuals who increased their physical activity level and decreased their food intake over time were protected from weight gain compared to those who did not. Frequency of high-intensity physical activity was particularly important for both men and women. Additionally, women who consistently engaged in higher levels of moderate physical activity gained weight at a slower rate compared to women who were less active.

CONCLUSIONS: Overall results indicated that both cross-sectionally and prospectively, the determinants of weight and weight change are multifactorial. Attention to exercise, fat intake and total energy intake all appear important for successful long term control of body weight.

Introduction

Obesity is a major public health problem in the US that affects approximately 50 million adults. Dramatic increases in secular trends over the last 10–20 y have only increased concern about this problem.1 A continuation of current trends could lead to increases in the number of people affected by obesity-related health conditions, including hypertension,2 diabetes3 and premature mortality.4

A major contributor to the high rates of obesity in US adults is weight gain with age.5,6,7 About two-thirds of obese adults become obese in adulthood. Between the ages of 20 and 50 y weight gain tends to occur at a rate of 0.45 to 0.91 kg/y and major weight gain is most likely to occur between the ages of 25 and 34 among both men and women.7

Because obesity is a problem that results from an imbalance between energy intake and energy expenditure, age-related weight gain probably involves change in patterns of eating and exercise in addition to biological processes accompanying aging (eg metabolism). However, few prospective studies of factors predictive of weight change in the adult population have been conducted and available studies have varied considerably in the breadth with which they assessed diet and exercise variables. Four studies have examined prospective associations between total energy intake and future weight gain. Two found no association 3,8 and two found a positive association among women but not men.9,10 Five prospective studies have examined dietary fat intake s a predictor of weight gain. Two found positive associations among both men and women,9,11 one found a positive association among men only,8 one found a positive association among women only,10 and one found that among women, total fat intake was positively related to recent prior weight gain, but slightly negatively associated with subsequent weight gain.3 Two studies have reported positive relationships between alcohol intake and weight gain12,13 and one has reported a negative association.14 Finally, two population studies have reported that physical activity protects against weight gain among both men and women,11,15 two have reported that physical activity is a protective factor among women only9,16 and one has reported inconclusive results.17 Interestingly, despite the fact that there is evidence that high intensity activity may confer substantial benefits for weight regulation,18 only two studies9,11 differentiated between physical activity intensity level in their results.

The present paper examines cross-sectional and prospective relationships between diet, physical activity, and body weight over time in a sample of individuals participating in a weight gain prevention study. The goal of the present analysis is to identify the diet and exercise behaviors that predict and/or accompany weight change over time. Specific questions addressed include: (a) are current levels of diet or exercise predictive of future weight gain over time; (b) are current levels of diet and exercise levels associated cross-sectionally with body weight; (c) are changes in levels of diet and exercise associated with changes in body weight over time; (d) which components of diet (eg dietary fat intake, total energy intake, etc) are most strongly associated with body weight and changes in body weight; and (e) are varying intensity levels of physical activity differentially associated with body weight and body weight change.

Subjects and methods

Subject population

Data for this investigation were from baseline and three subsequent annual follow-up examinations of participants in the Pound of Prevention (POP) study. POP was a 3 y randomized trial that evaluated the efficacy of an intervention for reducing the rate of weight gain with age in a community sample of adults. Participants were recruited from a variety of sources, including direct mailing to university employee groups, newspaper advertisements, and radio public service announcements. Low-income women were recruited in person at the Special Supplemental Nutrition Program for Women, Infants and Children (WIC) clinics. Details of the study design are reported elsewhere.19 All participants were volunteers, free of major chronic diseases, and were between the ages of 20–45 y at enrollment. Participants were randomized to one of two mail-based educational programs, or to a no-contact control group and followed for a period of 3 y. Data for the present report were derived from 826 women and 218 men (93% of the total sample enrolled at baseline) who completed the baseline assessment and at least one of the three annual follow-up assessments and who were not pregnant during the 31 y study period.

Measures

Dietary intake, physical activity, and eating behaviors were measured at each of the four data collection visits as detailed below.

Dietary intake.

Block Food Frequency Questionnaire:20 The 60-item version of the Block Food Frequency Questionnaire was used to estimate usual dietary intake during the past year. For each of 60 foods, participants reported their usual serving size and frequency of consumption. Nutrients were estimated by multiplying the frequency of consumption by the amount of the nutrient in a ‘standard' portion size. Data can be analyzed to yield estimates of daily energy intake as well as macro- and micronutrients. For this report, total energy intake (kcal/day) and percentage energy from fat and from alcoholic beverages were examined.

Kristal Low-fat Eating Behavior Scale:21 The 18-item Kristal Low-fat Eating Behavior Scale assesses five theoretically based and empirically validated dimensions of eating behavior related to a low-fat diet. Common food preparation methods and patterns of food use, such as using low-fat (1%) or non-fat milk and low-fat cheeses and eating bread products without butter or margarine are assessed. Total Kristal scale scores (α=0.84) reflect frequency of engaging in behaviors related to lowering dietary fat intake, therefore, higher scores on the Kristal measure indicate the use of a greater number of low-fat eating behaviors.

Physical activity.

Physical activity was measured using a 13-items self-administered version of the Physical Activity History (PAH).22 The PAH has been used in several large epidemiologic studies and has been shown to have adequate reliability and validity.23 Participants reported the frequency with which they engaged in leisure time and occupational physical activities for 20 min or more over the last year. Activities were grouped into four categories based on the results of a principal components analysis:24 (1) high-intensity activities (ie running/jogging, biking, swimming, and exercise classes); (2) moderate-intensity activities (ie walking and home maintenance activities, such as gardening and snow shovelling); (3) group and racquet sports (ie basketball, football, soccer, tennis, squash, racquetball, and handball); and (4) occupational activity (ie lifting, carrying, digging). For these analyses, the frequency per week of each category of physical activity was used.

Body weight.

Weight was measured in light clothing without shoes on calibrated balance beam scale. Height was measured with a wall mounted ruler. Body mass index was computed as weight/height2 (kg/m2).

Smoking.

Current smoking status was self-reported at baseline and at the second annual follow-up. Because only small numbers of subjects changed smoking status over the 3 y period (2.8% quit smoking and 2.8% initiated smoking) and including change in smoking status in the analyses did not significantly change the results, only smoking status at baseline was used in the analyses presented in this paper.

Demographics.

Demographic information was self-reported at baseline and included age, sex, educational attainment, occupation, income level and marital status.

Statistical analysis

Analyses were conducted using the Statistical Analysis System.25 In order to assess the impact of study dropouts on the interpretation of results, participants who completed at least two of the four study assessment visits and thus were included in the analysis were compared with participants who completed only the baseline assessment on baseline weight, demographic and behavioral characteristics. Chi-square tests were used for categorical variables and t-tests were used for continuous variables.

The analyses used to examine the three study questions employed random coefficients models appropriate for cohort data.26 Weight was the dependent variable and diet and exercise levels were the independent variables in all analyses. Random coefficient models estimate a separate intercept and linear slope for each member of the cohort. They also estimate separate components of variance for those slopes and intercepts and allow the slope and intercept for each member to covary. Standard errors for the fixed-effect predictors are constructed to reflect the appropriate sources of variation. Such models have been shown to better protect the nominal Type I error rate in the presence of heterogeneity among the member-specific slopes than the more familiar repeated-measures analysis of variance models.27

The initial random coefficients model was designed to examine whether habitual diet and/or exercise patterns are predictive of future weight gain. For this analysis, regression effects for the diet and exercise variables were decomposed into between- and within-subject domains. Specifically, for each of those variables two orthogonal scores were computed for each person: (1) an average value across the four assessment points; and (2) a deviation from that average at each assessment point. Interactions with time were computed for both the average and deviation scores for each dietary intake and physical activity variable. The coefficient for time estimates the adjusted change in weight per unit time when all variables involved in interactions with time are zero. In other words, the coefficient for time is the base slope for weight change across the study period. The time×average coefficients estimate the departure in the coefficient for time per unit difference between subjects in each average variable. The time×deviation coefficients estimate the departure in the coefficient for time per unit increase within person in each deviation variable. Under the null hypothesis of no association between levels of diet and exercise and rate of weight gain over time, these interaction coefficients have expected values of zero.

The average and deviation scores for physical activity and dietary intake variables were conceptualized as four sets of variables relevant to the prediction of weight, whether as main effects or s interactions with time. The average score for each variable is used to examine the cross-sectional relationship between that variable and body weight and the deviation score for each variable is used to examine the prospective relationship between that variable and body weight. The set of variables describing dietary intake included: total energy intake, percentage intake from fat, total Kristal score, and percentage intake from alcohol. The set of variables describing physical activity included: high-intensity physical activity, moderate-intensity physical activity, group sports and job activity. In the initial analyses, an overall significance test for each of the four sets of interaction terms was evaluated prior to examining the relative importance of the specific components that comprised each set. In order to control the Type I error rate over the four sets of variables, a significance level of 0.0125 was used in testing each set. Data were analyzed separately for men and women. Time, age, smoking status, treatment group, and the treatment by time interaction were entered as covariates.

If the overall significance test for a given interaction was not significant at P<0.0125, the interaction terms comprising that set were removed from the model and the analysis was rerun. In these reduced models, the focus was still on the overall tests of significance for the four sets of variables, but now either as main effects or interactions with time, depending on the results of the initial analyses. When an interaction set was deleted, the focus shifted to the main effects. Here, the regression coefficient for each average score estimates the difference in mean weight between persons who differ by one unit on that score. Similarly, the regression coefficient for the deviation score estimates the mean change in weight within persons associated with a one unit increase in that score. These main effects examine (a) whether habitual diet and/or exercise levels are associated cross-sectionally with body weight and (b) whether changes in diet and exercise over time are associated prospectively with changes in body weight. As for the interaction terms, the main effects were evaluated using overall tests of significance for each set of main effect terms at a significance level of 0.0125 to correct for the number of tests (n=4). If a main effect was not statistically significant, the components of that main effect are not interpreted.

Finally, the dietary intake and physical activity patterns of individuals who gained, maintained, and lost over the 3 y study period were compared in order to further clarify the relationship between diet and exercise behavior change and weight status. ‘Weight gainers' were defined as those who had gained >5 lb; ‘weight maintainers' were defined as those who had lost or gained ≤5 lb; and weight losers were defined as those who had lost >5 lb over the study period. Analyses of covariance were used to compare ‘weight gainers' and ‘weight losers' on baseline values of the dietary intake and physical activity variables. Additionally, change scores from baseline to the third annual follow-up were computed for the dietary intake and physical activity variables. Separate analyses of covariance were used to examine whether ‘weight gainers' and ‘weight losers' differed in the amount and direction of change for each dietary intake and physical activity variable. These analyses included age, smoking status, treatment group, baseline body weight and the baseline value on the respective dependent variables as covariates.

Results

A total of 1044 participants were included in these analyses and 76 were excluded. Few differences were found between these two groups. Men who were included in the analyses weighed less (P<0.038), had completed more schooling (P<0.001), and reported lower total energy intake (P<0.017) at baseline compared to dropouts. Women included in the analyses weighed less (P<0.001), were younger (P<0.003), and reported higher income levels (P<0.001) compared to dropouts.

Baseline characteristics of the 1044 participants included in the analyses are shown in Table 1. Participants averaged 35 y of age and were predominantly White. The majority of men in the study were college educated, employed in professional positions, and married. Women in the sample were more diverse with regard to education, profession and marital status. To assist in the interpretation of the analyses which follow, Table 2 presents the mean changes in weight, dietary intake, and physical activity over the 3 y period of the study by gender. On average, participants gained 1.36–1.81 kg and reported decreased total energy intake, decreased fat intake, increased alcohol intake, and decreased physical activity over time. Changes in body weight, total energy intake, fat intake and alcohol intake were significantly different from zero (P<0.05) for both men and women. A statistically significant decrease in group activity and a marginally significant decrease in high-intensity activity was observed among men. Among women, statistically significant decreases in both group and high-intensity physical activity were observed. Changes in moderate physical activity and occupational activity were not significant among either men or women.

Table 1: Demographic and behavioral characteristics of study participants at baseline
Table 2: Mean (s.d.) change in weight, dietary intake, and physical activity between baseline and year 3 by gender

Table 3 presents the results of the random coefficients regression analyses for men. Overall tests for the sets of dietary and physical activity interactions with time were not statistically significant at P<0.0125, indicting that neither habitual diet or exercise patterns were predictive of rate of weight gain over time. The analysis was next rerun excluding the interaction terms. Overall tests for the set of physical activity main effects indicated that the cross-sectional association between physical activity and body weight was not statistically significant, but that the prospective association between physical activity and body weight was significant. Because the main effect for cross-sectional associations between physical activity and body weight was not statistically significant, the coefficients for the components of physical activity (ie moderate intensity, high intensity, group sports and occupational) are not interpreted. The prospective association between physical activity and body weight was primarily due to an inverse association between change in high-intensity physical activity and change in body weight. An increase of one high-intensity exercise session per week was associated with a decrease in weight of 0.54 kg. Dietary variables predicted weight in men both cross-sectionally and prospectively. Cross-sectionally, total energy intake was strongly positively associated with body weight. Prospectively, increases in percentage energy intake from fat were associated with increases in body weight and increases in Kristal Low-fat Eating Scores were associated with the decreases in body weight.

Table 3: Multivariate cross-sectional (between-person) and prospective (within-person) associations between weight and exercise and dietary intake patterns over 3 y in men

Table 4 presents the results of the random coefficients regression analyses for women. Among women, the interaction between average physical activity level and time was statistically significant, but the other three interactions were not. As a result, the analysis was rerun excluding the nonsignificant interactions with time, but retaining the average physical activity×time interaction. The overall tests for this analysis indicated that the main effects for the cross-sectional and prospective dietary intake and physical activity variables were associated with body weight in addition to the interaction. Consider first the time×average physical activity interaction. The effect for time was statistically significant, indicting that on average participants gained weight over the study period. Specifically, when the four average physical activity scores were zero, participants gained 1.27 kg/y. Examination of the individual time × average physical activity interaction terms indicated that higher habitual levels of moderate-intensity activity were associated with smaller rates of weight gain over time. Higher levels of high-intensity physical activity also appeared to have a moderately protective effect on rate of weight gain. Examination of the main effects for average physical activity indicate that apart from the effects of the time × average physical activity interactions, women who engage in more frequent high-intensity physical activity weighed less than those who engage in less frequent high-intensity activity. Consider next the main effects for the other three sets of variables. Prospectively, increases in moderate-intensity, high-intensity, and occupational physical activity were associated with decreases in body weight over time. An increase of one high-intensity exercise session per week was associated with a decrease of 0.15 kg, an increase of one moderate-intensity exercise session per week was associated with a decrease of 0.10 kg, and an increase of one vigorous activity session at work per week was associated with a decrease of 0.21 kg. Cross-sectionally in women, total energy intake and percentage energy intake from fat were both positively associated with body weight level and percentage energy intake from alcohol was inversely associated with body weight level. Prospectively, increases in total energy intake and increases in percentage energy intake from fat were associated with increases in body weight and increases in Kristal Low-fat Eating Scores were associated with decreases in body weight.

Table 4: Multivariate cross-sectional (between-person) and prospective (within-person) associations between weight and exercise and dietary intake patterns over 3 y in women

Table 5 presents the results of the analyses comparing the dietary intake and physical activity behavior change scores across the 3 y study period of ‘weight gainers', ‘weight maintainers' and ‘weight losers' for both men and women. ‘Weight gainers', ‘weight maintainers' and ‘weight losers' did not differ on baseline values of the dietary intake or physical activity variables with one exception (data not shown). Among men, ‘weight losers' were less active in group sports compared to ‘weight maintainers' and ‘weight gainers' (weight losers mean=0.7, s.e.=0.2; weight maintainers mean=1.4, s.e.=0.2; weight gainers mean=1.0, s.e.=0.1; F=3.1, P<0.05). ‘Weight losers' were also more likely to increase moderate- and high-intensity physical activity compared to ‘weight gainers'.

Table 5: Mean (s.e.) change in dietary intake and physical activity between baseline and year 3 by weight change statusa

Among men, ‘weight losers' reported a greater increase in Kristal Low-fat Eating Scores compared to ‘weight maintainers' and ‘weight gainers' ‘weight maintainers' reported a greater increase in Kristal scores compared to weight gainers. There was also a nonsignificant trend for male ‘weight losers' to report a greater reduction in percentage energy intake from fat compared to ‘weight gainers' and ‘weight maintainers'. ‘Weight losers' reported increases in high-intensity physical activity whereas ‘weight maintainers' and ‘weight gainers' reported decreases in high-intensity physical activity. However, ‘weight losers' reported greater decreases in group sports compared to ‘weight maintainers' and ‘weight losers'. Among women, ‘weight losers' and ‘weight maintainers' reported greater reduction in total energy intake and greater increases in Kristal lowfat eating scores than ‘weight gainers'. ‘Weight losers' and ‘weight maintainers' also reported smaller decreases in moderate-intensity physical activity compared to ‘weight gainers'. Finally, ‘weight losers' reported increases in high-intensity physical activity whereas, on average, ‘weight maintainers' and ‘weight gainers' reported decreases in high-intensity physical activity.

Discussion

This study examined cross-sectional and prospective relationships between body weight, macronutrient intake, behaviors intended to limit fat intake, and physical activity in a sample of community volunteers taking part in a weight gain prevention project. Over 3 y of observation, the average weight gain of the study group was between 1.36 and 1.81 kg. Over the same time period, study participants reported reducing energy intake, reducing fat intake, reducing physical activity and increasing alcohol intake.

The overall results indicate that both cross-sectionally and prospectively, the determinants of weight and weight change are multifactorial. The most consistent findings with respect to the question of why people gain weight with age were the positive association between dietary fat intake and weight gain and an inverse association between frequency of physical activity and weight gain. A consistent relationship between total energy intake and body weight in women was also observed. The results are generally consistent with previous longitudinal studies of determinants of weight gain with age.3,8,9,10,11,15,16 However, it is believed that this study is an improvement over prior studies in (1) including wide range of diet and exercise variables in both men and women; (2) having weight and behavior measures at 4 separate time points; (3) including all available data points in the analysis; and (4) controlling for potential confounders (eg total energy intake) not available in some earlier studies.

Clearly evident in these results is that the range of weight changes observed over the 3 y in this study were large. The top quartile of men gained 8.16 kg over 3 y, while the lowest quartile lost 4.54 kg. Among women, the comparable gain was 9.53 kg and the comparable loss was 5.90 kg. Examination of data comparing individuals who lost, maintained and gained weight over the study period also indicates that weight regulation over time is multifactorial. Specifically, individuals who lost weight over the 3 y study period reported substantial changes in both energy intake and energy expenditure in the direction predicted by the energy balance equation relative to those who gained large amounts of weight. Successful weight maintenance also requires attention to both diet and physical activity. Frequency of high-intensity physical activity appears to be particularly important for both men and women, with moderate-intensity activity significant for women only.

In this sample, individuals who weighed more both ate more and exercised less than those who weighed less. Individuals who increased their physical activity level and decreased their food intake over time were protected from weight gain compared to those who did not. Additionally, women who consistently engaged in higher levels of moderate physical activity gained weight at a slower rate compared to women who were less active. Observed relationships were for the most part similar in men and women with the exception of the inverse association between alcohol intake and weight in women. The much larger sample size for women, however, provided better precision in estimation and thus lower P values.

An interesting aspect of the present findings was the observation that a measure of volitional behaviors intended to reduce fat intake, the Kristal Low-fat Eating Behavior Scale, predicted weight change independent of concurrent measures of fat intake and total energy intake based on a food frequency questionnaire. Although the low-fat eating behavior measure and the measure of percentage fat intake derived from the food frequency measure are moderately correlated with each other, they complement one another and measure related, but unique, aspects of dietary intake. For example, the Block Food Frequency Questionnaire is known to underestimate calorie and fat intake and also does not include detailed information on fat-modified foods that some people may be using to manage their weight. Examination of the items on the Kristal Low-fat Eating Behavior Scale indicate that this instrument measures small changes that people make in their diets to reduce overall fat intake, such as typical food preparation methods and consistent use of reduced-fat products. This suggests that dietary assessments that include assessment of typical food preparation methods and patterns of food use may provide additional valuable information related to important health outcomes.

A second noteworthy finding in the study was the strength of the association found between weight change and high-intensity physical activity relative to lower intensity activities. It is generally thought that energy expenditure in physical activity rather than activity intensity is what determines its effects on body weight.28 However, there are a number of advantages that high-intensity activity may hold for weight management. In their recent review of the literature, Hunter et al18 suggest that high-intensity exercise is associated with reduced efficiency and, in turn, greater energy expenditure and that high-intensity activity has stronger and more enduring effects on resting energy expenditure. They also suggest that high intensity improves fitness which has spillover effects on the ease with which low-intensity activity can be performed. Moreover, it has been demonstrated that higher levels of exercise intensity are associated with decreased levels of adiposity29 and greater stimulation of the potential of skeletal muscle to utilize lipids30 and a reduced postexercise compensation in energy intake.31 Additional questions deserving of further study include: do people who engage in high-intensity physical activity achieve beneficial levels of energy expenditure with less time commitment? Is high-intensity activity more predictive of weight change because it is more accurately measured? Finally, is higher intensity physical activity a marker of higher levels of overall activity that is not measured?

The fact that there was a general trend downward in reported energy intake (about 200 kcal/day) and exercise (about 0.5 exercise sessions per week) over the 3 y study period deserves comment. Although downward trends in both dietary intake and exercise are to be expected with age, the amount of change observed here in just 3 y does not seem plausible for adults in this age range and is not consistent with the observation of substantial mean weight gain. In our view the most likely explanation of the findings is the tendency of people to be less attentive to lengthy questionnaires after repeated administration and thus to underreport behaviors. This repetition fatigue phenomenon has been well documented in dietary assessment32,33 and is probably true for physical activity as well. In any event, it is not believed that such method drift would bias our results since, if anything, subject inattention would introduce error into measurements that could weaken rather than strengthen observed covariations.

The primary public health and clinical implication of the present research is that the energy balance problems causing weight gain with age involve multiple aspects of both diet and exercise. Attention to exercise, fat intake and total energy intake all appear important for successful long-term control of body weight. Although the weight changes associated with changes in diet and exercise behaviors are small in an absolute sense, these findings are of significant public health importance when viewed from a population perspective. A challenge to public health professionals is to develop educational and environmental interventions which support these healthy diet and exercise patterns.

References

  1. 1.

    , , , . Increasing prevalence of overweight among US adults: the National Health and Nutrition Examination Surveys, 1960 to 1991 JAMA 1994 272: 205–211.

  2. 2.

    , , , . Obesity and hypertension: epidemiological and clinical issues Eur Heart J 1987 8: 57–70.

  3. 3.

    , , , , , , . Weight as a risk factor for clinical diabetes in women Am J Epidemiol 1990 132:: 501–513.

  4. 4.

    , . Variations in mortality by weight among 750,000 men and women J Chronic Dis 1979 32: 563–576.

  5. 5.

    , . Trends in fatness and the origins of obesity Pediatrics 1976 57: 443–456.

  6. 6.

    , . Natural history of obesity in 6,946 women between 50 and 59 years of age Am J Public Health 1980 70: 385–388.

  7. 7.

    , , , . The 10-year incidence of overweight and major weight gain in US adults Arch Intern Med 1990 150: 665–672.

  8. 8.

    , , , . Proportion of energy intake from fat and subsequent weight change in the NHANESI Epidemiologic Follow-up Study Am J Clin Nutr 1995 61: 11–17.

  9. 9.

    , , , . A longitudinal analysis of the impact of dietary intake and physical activity on weight change in adults Am J Clin Nutr 1992 55: 818–822.

  10. 10.

    , , , , . Determinants of weight gain and overweight and adult Finns Eur J Clin Nutr 1991 45: 419–430.

  11. 11.

    , , et al.Predictors of weight change over two years among a population of working adults: the Healthy Worker Project Int J Obes Res 1992 18: 145–154.

  12. 12.

    , . Drinking and its relation to smoking, blood pressure, blood lipids, and uric acid Arch Intern Med 1983 143: 1366–1374.

  13. 13.

    , . Alcohol consumption and its relationship to smoking, weight, blood pressure, and blood lipids: the Albany Study Arch Intern Med 1986 146: 262–265.

  14. 14.

    , , , , . A prospective study of alcohol intake and change in body weight among US adults Am J Epidemiol 1994 140: 912–920.

  15. 15.

    , , , , . Effects of life-style on body mass index change Epidemiology 1994 5: 599–603.

  16. 16.

    , , , . Can physical activity mitigate the effects of aging in middle-aged women? Circulation 1992 85: 1265–1270.

  17. 17.

    , , , , , . Recreational physical activity and ten-year weight change in a US national cohort Int J Obes 1993 17: 279–286.

  18. 18.

    , , , . A role for high intensity exercise on energy balance and weight control Int J Obes 1998 22: 489–493.

  19. 19.

    , . Preventing weight gain in adults: the Pound of Prevention Study Am J Public Health 1999 89: 747–751.

  20. 20.

    , , , , , . A data-based approach to diet questionnaire design and testing Am J Epidemiol 1986 124: 453–469.

  21. 21.

    , , , , , , . Long-term maintenance of a low-fat diet: durability of fat-related dietary habits in the Women's Health Trial J am Diet Assoc 1992 92: 553–559.

  22. 22.

    , , , , . Validity and reliability of short physical activity history: CARDIA and the Minnesota Heart Health Program J Cardiopulmonary Rehabil 1989 9: 448–459.

  23. 23.

    , , , . A simultaneous evaluation of 10 commonly used physical activity questionnaires Med Sci Sports Exercise 1993 25: 81–91.

  24. 24.

    , , , . Socioeconomic status differences in health behaviors related to obesity: the Healthy Worker Project Int J Obes 1991 15: 689–696.

  25. 25.

    . SAS language and procedures: Usage version 6.12. Cary, NC: SAS Institute, 1996.

  26. 26.

    , . Hierarchical Linear Models: Applications and Data Analysis Methods. Newbury Park, CA: Sage, 1993.

  27. 27.

    , , , , . Analysis of data from group-randomized trials with repeat observations on the same groups Stat Med 1998 17: 1581–1600.

  28. 28.

    , . Physical activity and long-term maintenance of weight loss Obes Res 1994 2: 587–599.

  29. 29.

    , , , , , , . Effect of intensity of physical activity on body fatness and fat distribution Am J Clin Nutr 1990 51: 152–157.

  30. 30.

    , , . Impact of exercise intensity on body fatness and skeletal muscle metabolism Metabolism 1994 43: 814–818.

  31. 31.

    , , , . Acute effects of exercise on energy intake and feeding behaviour Br J Nutr 1997 77: 511–521.

  32. 32.

    , , , , , , , , , , . Monitoring dietary change in a low-fat diet intervention study: advantages of using 24-hour dietary recalls versus food records J Am Diet Assoc 1996 96: 574–579.

  33. 33.

    , , , . Hypertension prevention trial: do 24-hour food records capture usual eating behavior in a dietary change study? Am J Clin Nutr 1990 51: 253–257.

Download references

Acknowledgements

This research was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant DK45361, with additional funding from the Centers for Disease Control and Prevention.

Author information

Affiliations

  1. Division of Epidemiology, University of Minnesota, School of Public Health, Minneapolis, MN 55454-1015, USA

    • NE Sherwood
    • , RW Jeffery
    • , SA French
    •  & PJ Hannan
  2. Department of Psychology, University of Memphis, Memphis, Tennessee, USA

    • DM Murray

Authors

  1. Search for NE Sherwood in:

  2. Search for RW Jeffery in:

  3. Search for SA French in:

  4. Search for PJ Hannan in:

  5. Search for DM Murray in:

Corresponding author

Correspondence to NE Sherwood.

About this article

Publication history

Received

Revised

Accepted

Published

DOI

https://doi.org/10.1038/sj.ijo.0801169

Further reading