Abstract: Background: Lifestyle interventions to reduce excess body weight associated with high cardiometabolic risk have resulted in weight loss and improved risk short-term. However, rates of recidivism are high, averaging weight regain of 50% after 12-24 months. Understanding what factors are associated with successful weight maintenance and weight regain as well as the cardiometabolic implicat... read moreions of partial or total weight regain will help formulate strategies to improve long term outcomes. Objectives: We compared published categorization criteria that differentiate maintainers and regainers using agreement statistics in the Action for Health in Diabetes (Look AHEAD) trial with replication in the Diabetes Prevention Program (DPP) (Aim 1). Next, we examined the association between weight regain at varying magnitudes and cardiometabolic risk factors in the Look AHEAD trial (Aim 2). Finally, we developed and internally validated a prediction model of weight regain using factors from physical, psychological and behavioral domains in the Look AHEAD trial (Aim 3). Methods: Publically available data from Look AHEAD (n=1791) and DPP (n=613) were used to identify participants with ≥3% initial weight loss (IWL) after lifestyle interventions. Four-year follow-up data were used for all analyses. For Aim 2, fewer Look AHEAD participants (n=1561) were included due to medication use exclusions. For Aim 1, eight previously published criteria defining body weight loss maintainers and regainers were compared with respect to concordance using agreement statistics. Criteria were assessed separately among those with 3-9% and ≥10% IWL. Next, weight regain and weight loss maintenance were defined by dichotomization with five cut points (0%, 25%, 50%, 75% and 100%) of percent of weight loss regained (weight change from years 1 to 4 as percent of weight loss during the first year). Change in cardiometabolic risk factors after IWL was compared in maintainers and regainers according to each cut point using ANCOVA models. The effect was assessed separately in those with <10% and ≥10% IWL, and in women and men. Finally, a prediction model of weight regain was developed and internally validated. Predictors from demographic, psychosocial, clinical and behavioral domains were entered into a stochastic gradient boosting prediction model. Outputs from the model were added to a stepwise logistic regression for interpretability. Results: When assessing concordance among weight loss criteria, agreement was dependent on IWL, but many criteria were in high agreement (% agreement ≥80%). The definition of successful weight loss maintenance "regaining ≤25% of IWL during maintenance" showed high agreement with most commonly used definition: "staying ≥10% below initial weight" among those with ≥10% IWL (% agreement=85% in Look AHEAD; 87% in DPP). The same ≤25% regain definition showed high agreement with the definition of staying ≥5% below initial weight among those with 3-9% IWL (% agreement=92% in Look AHEAD; 91% in DPP). When assessing cardiometabolic risk, maintainers had significant improvements from years 1 to 4 for cardiometabolic risk factors compared to regainers for all risk factors assessed. No single weight regain cut point maximized the risk difference between maintainers and regainers across risk factors and sex/ IWL subgroups. For many risk factors, increasing the cut point to allow for more regain as part of maintenance resulted in decreased cardiometabolic benefit among maintainers. On the basis of the prediction model, the top predictors of weight loss maintenance were IWL, baseline BMI, interaction between IWL and BMI, and use of meal replacements. The model performed fairly in training and test models (ROC AUC [95%CI]: 0.79 [0.76, 0.82] and 0.67 [0.64, 0.70], respectively). The logistic regression model performed similarly. Conclusion: We found percentage of weight loss regained is a favorable way to calculate weight change and derive a cut point for maintainers and regainers. Within this calculation, we found that the 25% cut point (allowing up to 25% of weight to be regained as part of maintenance) is the best criteria for sample size and second best for maximizing the risk difference between maintainers and regainers. The 0% cut point (no regain) is the best for cardiometabolic risk reduction but has sample size limitations because a small proportion of people keep off all weight lost. Finally, we identified key predictors of weight regain. The models performed fairly and should be tested in external settings. The findings move efforts forward to define successful weight maintenance and regain and can assist the development of strategies to lower rates of post-loss weight regain.
Thesis (Ph.D.)--Tufts University, 2018.
Submitted to the Dept. of Nutritional Epidemiology.
Advisor: Alice Lichtenstein.
Committee: Gordon Huggins, Jeanne McCaffery, and Paul Jacques.