A major purpose of the present paper was to determine whether behaviors that are associated with BMI in general population samples are similarly associated with BMI in overweight and obese persons seeking treatment. The answer to this question appears to be yes. In this sample, over two-thirds of participants were clinically obese; about one-third were severely obese. The results of the present investigation show that observations from general population samples that frequency of consumption of specific high-fat foods (particularly meat products) are positively associated with body weight [8
] and, to a lesser extent, that intake of fruits and vegetables are inversely associated with body weight [10
], are generalizable to much higher body weights as well. Likewise, positive changes in dietary intake over the course of the study (i.e., decreases in intake of high-fat foods; increases in intake of fruits, vegetables, or high-fiber foods) were associated with favorable changes in BMI. In other words, obese and very obese treatment may have more difficulty balancing their eating and activity behaviors, but the relationships between specific behaviors and body weight are on the same continuum as the rest of the population.
Results in this population relating to dieting history and behaviors are also strongly similar to those seen in weight-maintaining populations or those with lower BMI [24
]. Individuals with higher BMI diet more and participate in formal diet programs more often than those with lower BMI. It is likely that those who have engaged in subsequent attempts to reduce BMI do so because they have not been successful in earlier attempts; it is possible either that they have not engaged in weight loss behaviors at the level or duration required to sustain weight loss [27
], or that the pattern of repeated weight losses and gains promotes larger weight gains over time [28
]. New findings from this research extend the observations about greater global dieting concerns to possession of diet-related books. Those with higher BMI own more books about weight loss, nutrition, and diet cooking. Of some interest is the observation that those who owned subscriptions to magazines about exercise had lower BMI than those who did not, although the percentage of individuals with exercise magazines was small (< 10%).
Interestingly, higher baseline consumption of several high-fat foods (e.g., hot dogs, pastries, French fries, ice cream), as well as total fat for men, was associated with lower 24-month BMI. This pattern was most notable for men, who, relative to women, tend to consume a higher-fat diet in general and who may not value healthy eating habits as strongly [29
]. In every case, however, reductions in consumption of these specific foods over time, and overall reductions in fat intake, were associated with better BMI outcomes for both men and women. It is likely that these high-fat foods are readily identified targets for intake modification among high consumers who engage in a weight loss program, thus leading to an association of baseline consumption with BMI reduction over time. Whereas effect size indicators suggest that changes in composite indices of fat or fiber intake were stronger predictors of BMI change that individual foods, changes in intake of some individual foods predicted BMI change as well as composite measures (e.g., French fries or fruit for men; margarine, vegetables, or cereal for women). It is likely that these foods, which may be identified easily as sources of fat or dietary fiber, also became obvious targets for reduced or increased consumption within the context of a dietary change plan, as advocated in the weight loss program in which study participants were enrolled.
It should be noted that in the Weigh-to-Be trial, participation in weight loss activities was high even in the usual care group, such that the three groups (usual care, mail intervention, and phone intervention) did not differ in terms of weight change at the 2-year follow-up [15
]. Thus, regardless of treatment group assignment in this trial, those individuals who consumed more of these foods at baseline may have had more options for overall fat reduction at their disposal during the weight loss intervention, thus assisting their weight loss efforts.
This study was not without limitations. Dietary intake was assessed by screening instruments selected for the main study that may be limited in their ability to capture complete, well-differentiated dietary information. Attrition from the study over the two years was relatively high; however, the factors on which study completers differed from dropouts (i.e., age, education, BMI) were included in the statistical models, and adjustment for attrition by imputation of missing values using the last observation carried forward [30
] did not change the findings presented here. Lastly, BMI is not a direct measure of adiposity; as such, BMI may misclassify overweight or obese status depending on age, gender, ethnicity, athleticism, or redistribution of proportions of fat mass versus lean mass by resistance training [31