The WOMAN Study is one of the few long term weight reduction studies and, similar to other studies,6
documents the difficulties of maintaining substantial weight loss over time for most of the participants. As hypothesized, most of the eating habits predicted weight change at 6 months when considered individually, and results in the intervention group mirrored the analysis including all WOMAN Study participants. Interestingly, most of these eating behaviors were independently associated with weight change in the fully adjusted model, suggesting a role for each eating behavior change in the context of short-term weight loss or control. Decreased consumption of fried foods and desserts had the largest standardized coefficients when including all WOMAN participants and within the intervention group, and these were the only significant predictors in fully adjusted analysis within the control group, highlighting these two specific targets for short-term weight change. At 48 months, frequency of eating at restaurants was consistently not associated with weight change, though the other eating behaviors continued to have associations when considered in isolation. In the fully adjusted model, only changes in meat and cheese intake were associated with weight change in the intervention group while desserts, sugar-sweetened beverages, and fruits and vegetables were independent predictors in the control group. These behavior changes could represent changes in eating behaviors that both influence weight change and are sustainable long-term.
A surprising finding was that frequency of eating at restaurants was not related to weight change outcomes at 48 months. Reported restaurant eating decreased whether subjects decreased weight or not (see ), consistent with the lack of association observed even in the single behavior analyses. Specific measurement of the frequency of dining at fast food restaurants, which has been shown to predict weight change in other studies,32, 33
was not available but may have been more strongly associated with weight change.
The associations observed at 6 months and 48 months were all consistent with a diet that decreases energy density. Lower energy density has been associated with greater weight loss or decreased weight gain in cohort studies,34, 35
secondary analyses within clinical weight loss trials,36–38
and in randomized trials.39, 40
Further, successful weight loss maintainers report a diet with lower energy density compared to normal weight or obese controls,16
indicating that decreasing energy density may be a particularly useful strategy for maintaining a weight-reduced state. At the same time, many of the behaviors that predicted weight change were specific targets of the intervention. Thus, results within the intervention group in particular may also reflect that weight loss or maintenance is more common among participants that follow dietary prescriptions more closely.
The differences in the magnitude of the associations between the intervention and control groups in this secondary analysis should be interpreted with caution. The intervention group received active treatment, especially during the first year, which resulted in a distribution of weight change in this group that was shifted toward weight loss with 85% of participants maintaining or losing weight at 48 months (). In contrast among controls, the distribution of weight change was more centered around weight maintenance, with about a third of controls achieving weight loss (29%), maintenance (37%), and weight gain (34%) at 48 months (). With this in mind, results from the intervention group may be more applicable to short- and long-term weight loss in a clinical setting; whereas, findings in the control group may be more applicable to women interested in weight loss who have not received an evidence-based intervention in a clinical setting.
Strengths of this study include a large sample size, long duration, and excellent retention in a lifestyle intervention study. Though the average weight loss at 48 months in the intervention group was <5%, a sizable subset of women had lost more than a 10 kgs at 48 months, providing variability in the outcome measure. With these strengths and the measurement of eating habits at relevant time points, this study was able to tease out which eating habits were independently related to short-term and long-term weight change in post-menopausal women.
On the other hand, several limitations deserve discussion. First, eating behaviors and physical activity were self-reported and thus vulnerable to the known biases involved with self-report, such as underreporting of undesirable habits (e.g. eating desserts). Further, seasonality may have affected certain eating habits between the baseline and 6 month assessments, e.g. fruits and vegetables, and this could explain why changes in fruits and vegetables were not independently associated with weight change at 6 months. Change in leisure-time physical activity over the first 6 months of the intervention was not included in regression models because the physical activity component of the intervention was slowly introduced during the first 6 months. However, the 48-month results reassuringly showed associations between eating behaviors and weight change to be independent of physical activity. Also, certain eating behaviors are collinear as evidenced by the attenuated associations when moving from single eating behavior models to all eating behaviors models, for example eating of fried foods and eating at restaurants. However, these eating behaviors had independent associations with weight change at 6 months and each single behavior model is reported to inform this issue. Also, information on frequency of snacking between meals, which may be another eating behavior associated with weight change, was not collected. Lastly, this secondary analysis identifies associations between changes in eating behaviors and changes in weight, but the study was not specifically designed to evaluate whether targeting changes in these behaviors would result in greater weight loss.
In summary, these results suggest that decreased consumption of desserts and sugar-sweetened beverages consistently associate with short- and long-term weight loss or maintenance, but increased fruits/vegetables in controls as well as decreased meats/cheeses in an intervention are additional factors that may help for long-term, but not necessarily short-term, weight loss or control. If the goal is to decrease the burden of obesity, the focus must be on long-term strategies because changes in eating behaviors only associated with short-term weight loss are likely ineffective and/or not sustainable. Future studies should examine whether interventions focused on changing these specific eating behaviors associated with long-term weight change could improve obesity treatment outcomes.