ABSTRACT
Background
Intensive lifestyle interventions (ILIs) stimulate weight loss in underserved patients with obesity, but the mediators of weight change are unknown.
Objectives
We aimed to identify the mediators of weight change during an ILI compared with usual care (UC) in underserved patients with obesity.
Methods
The PROPEL (Promoting Successful Weight Loss in Primary Care in Louisiana) trial randomly assigned 18 clinics (n = 803) to either an ILI or UC for 24 mo. The ILI group received an intensive lifestyle program; the UC group had routine care. Body weight was measured; further, eating behaviors (restraint, disinhibition), dietary intake (percentage fat intake, fruit and vegetable intake), physical activity, and weight- and health-related quality of life constructs were measured through questionnaires. Mediation analyses assessed whether questionnaire variables explained between-group variations in weight change during 2 periods: baseline to month 12 (n = 779) and month 12 to month 24 (n = 767).
Results
The ILI induced greater weight loss at month 12 compared with UC (between-group difference: −7.19 kg; 95% CI: −8.43, −6.07 kg). Improvements in disinhibition (−0.33 kg; 95% CI: −0.55, −0.10 kg), percentage fat intake (−0.25 kg; 95% CI: −0.50, −0.01 kg), physical activity (−0.26 kg; 95% CI: −0.41, −0.09 kg), and subjective fatigue (−0.28 kg; 95% CI: −0.46, −0.10 kg) at month 6 during the ILI partially explained this between-group difference. Greater weight loss occurred in the ILI at month 24, yet the ILI group gained 2.24 kg (95% CI: 1.32, 3.26 kg) compared with UC from month 12 to month 24. Change in fruit and vegetable intake (0.13 kg; 95% CI: 0.05, 0.21 kg) partially explained this response, and no variables attenuated the weight regain of the ILI group.
Conclusions
In an underserved sample, weight change induced by an ILI compared with UC was mediated by several psychological and behavioral variables. These findings could help refine weight management regimens in underserved patients with obesity.
This trial was registered at clinicaltrials.gov as NCT02561221.
Keywords: comprehensive lifestyle intervention, diet, eating attitudes, health disparities, minority groups, primary health care, weight loss, weight regain
Introduction
Obesity is a public health disease that increases the risk of type 2 diabetes, cardiovascular disease, cancer, and premature death (1, 2). Overall, obesity affects ∼40% of adults in the United States (3), and health disparities are present. Obesity is more prevalent in certain demographic groups with a low annual income (4). Moreover, compared with non-Hispanic white adults, black and Hispanic populations exhibit higher rates of obesity (5). It is thus important to identify effective weight-management methods for individuals with obesity in these populations to attain national health targets and decrease health disparities.
Usual care (UC) for weight loss and weight management within primary care typically involves behavioral counseling and therapy to improve dietary habits and physical activity, yet such regimens often yield substandard weight loss because of time constraints and a lack of training among practitioners (6). Intensive lifestyle interventions (ILIs) are recommended as alternative programs for weight loss in individuals with obesity in primary care (7). These aim to stimulate energy deficits and weight loss through reduced-calorie diets, improvements in physical activity, and behavioral therapy in an on-site and intensive (≥14 sessions in the first 6 mo) regimen delivered by trained interventionists (7, 8).
In the PROPEL (Promoting Successful Weight Loss in Primary Care in Louisiana) trial, we demonstrated that underserved patients with obesity lose more weight and improve cardiometabolic risk markers during an ILI compared with UC over 24 mo (9, 10). However, it is unclear what factors drove the increased weight loss produced by the ILI relative to UC. It is in addition not known if the factors associated with midterm (6–12 mo) weight loss during the ILI were effective at attenuating weight regain, which is common and can decrease the health benefits associated with lifestyle interventions (11). These factors could include those that have been associated with weight loss and were linked to behaviors and strategies covered in counseling sessions of the ILI, such as increased dietary restraint (i.e., the intent and ability to restrict food intake), reduced dietary disinhibition (i.e., the tendency to overeat) (12), increased intake of healthy foods with low fat (13), increased physical activity (14), and improved quality of life (15). Identifying the factors that mediated weight loss and weight-loss maintenance during the PROPEL trial is important because strategies and behaviors can be targeted and tested in future interventions, enhancing the efficacy of weight-management programs that are delivered to underserved individuals with obesity in primary care.
The aim of this exploratory investigation was to use mediation analyses to identify the mediators of weight change during an ILI compared with UC in underserved patients with obesity. We hypothesized that improvements in eating behaviors (increased dietary restraint and reduced dietary disinhibition), dietary intake (reductions in dietary fat and increases in fruit and vegetable intake), physical activity, and quality of life shown in the ILI compared with UC would mediate improved weight change during the ILI.
Methods
Patients
Primary inclusion criteria for PROPEL (NCT02561221) included an age of 20–75 y, a BMI (in kg/m2) of 30.0–50.0, and being a patient at a participating primary care clinic. Patients were excluded if they used weight-loss medication, were presently partaking in a structured weight-loss program, previously had bariatric surgery or planned to have bariatric surgery within 2 y, or had lost >10 lbs (4.5 kg) in the last 6 mo. A full list of inclusion and exclusion criteria has been previously published (9, 16), and all these criteria applied to these analyses.
Study design
The PROPEL study was a cluster-randomized trial consisting of 18 primary care clinics from 5 health systems across Louisiana. Details of the trial's design, randomization and recruitment methods, and protocol have been published (9, 16). The Pennington Biomedical Research Center Institutional Review Board approved the study. All procedures followed the ethical standards set by this Institutional Review Board, and all patients provided written informed consent. A self-report demographic questionnaire was used to obtain information about sex, race, and income.
Clinics were randomly assigned in a 1:1 allocation ratio to provide patients with an ILI or UC for 24 mo. Randomization was stratified by health system, with the random allocation method generated by a study statistician. Patients were not blinded to their group assignment because randomization occurred at the clinic level and the interventions are distinct. Efforts were nonetheless made to blind staff involved in data collection to the clinic randomization, and intervention staff were blinded to the patient's official study measures. The PROPEL trial data were collected and managed via the use of Research Electronic Data Capture (REDCap) resources hosted by the Pennington Biomedical Research Center (9, 17). The trial was conducted between April 2016 and September 2019, finishing when recruited patients who completed the trial had their month 24 assessments (9, 16).
Patients in the ILI received a pragmatic, intensive lifestyle program, which was based on previous lifestyle regimens (18–20) and consistent with the 2013 recommendations for the management of overweight and obesity set out by the American Heart Association, American College of Cardiology, and The Obesity Society (8). The ILI regimen was administered by appropriately trained health coaches embedded within primary care clinics and comprised weekly sessions in the first 6 mo (16 face-to-face and 6 delivered via telephone), followed by sessions that were held at least monthly. The objective for patients in the ILI was to lose 10% of their body weight through numerous strategies which aimed to change eating behaviors and physical activity. Strategies incorporated in the ILI included the provision of suitable prepackaged foods and meal replacements, coaching on appropriate portion sizes, and information on how to purchase and prepare healthy foods. It also included encouragement to increase physical activity to 175 min/wk in line with the physical activity goal of the Look AHEAD (Action for Health in Diabetes) trial (19). In addition to these strategies, a weight-loss calculator was used to formulate personalized energy intake targets and then display predicted weight loss to patients and health coaches (21).
Patients assigned to UC received the care routinely delivered by their clinic for the duration of the trial. They were also provided 6 newsletters that covered numerous topics such as sitting and health, goal setting, memory health, self-care, sleep hygiene, and smoking cessation. Primary care providers in the UC clinics received information at baseline and annually on the Centers for Medicare and Medicaid Services approach to behavioral therapy for obesity (22).
Measures
Body weight
Body weight was measured using a digital scale (Seca Model 876) at assessment visits conducted at baseline and at months 6, 12, 18, and 24. Patients were instructed to wear light clothes and no shoes while measurements were conducted. Anthropometric measurements were made in duplicate, although a third measurement was taken if weight differed by 0.5 kg. The mean of the 2 closest measurements was recorded.
Questionnaires
All questionnaires used in the present analysis were administered at baseline, month 6, month 12, and month 24.
The Eating Inventory (EI) is a 51-item tool that assesses dietary restraint, dietary disinhibition, and hunger (23). However, only restraint and disinhibition were assessed and thus a shortened 37-item EI was provided to PROPEL patients, with items assessing hunger removed (9). Dietary restraint is defined as the intent and ability to restrict food intake; a higher score is generally positive for weight control when disinhibition is low (24). Dietary disinhibition is defined as the tendency to overeat, and a higher value is associated with eating disorder symptoms and poor weight control (25). Greater scores for restraint and disinhibition were indicative of higher levels of the eating behavior assessed.
A customized questionnaire was administered to measure aspects of dietary intake. The questionnaire utilized scales from several sources to measure 3 outcomes: a National Cancer Institute (NCI) fat screener assessed percentage fat intake (26); a 7-item screener devised by the NCI and National 5 A Day Program examined fruit and vegetable consumption (27, 28); and 3 questions from the Brief Questionnaire to Assess Habitual Beverage Intake (BEVQ-15) assessed the frequency of alcohol intake (29).
Weight-related quality of life was measured through the 31-item Impact of Weight on Quality of Life-Lite (IWQOL) questionnaire (30, 31). This measures obesity-related aspects of quality of life, with a total quality of life score and separate scores for physical function, self-esteem, sexual life, public distress, and work or daily activities yielded. Scores are transformed to a 0–100 scale; a score of 100 represents the highest quality of life. The questionnaire asks patients to reflect on quality of life constructs because of their weight (30). Hence, in line with other analyses (15), only the total IWQOL score was utilized in the current analysis to limit the inclusion of variables that may be causally affected by weight change.
The Patient-Reported Outcomes Measurement Information System-29 (PROMIS-29) questionnaire was also administered to measure health-related quality of life (32). This 29-item questionnaire assesses health-related domains related to physical function, anxiety, fatigue, depression, sleep disturbance, ability to partake in social roles and activities, pain interference, and pain intensity. All constructs were used except for pain intensity owing to its relation with pain interference.
The International Physical Activity Questionnaire (IPAQ) short form was used to assess physical activity levels (33). The questionnaire, which asks questions related to physical activity over the previous 7 d, provides physical activity scores in median metabolic equivalent of task (MET)-minutes per week. Four constructs of physical activity were assessed in MET-minutes per week: vigorous, moderate, walking, and total. In the PROPEL trial, numerous patients had missing data for particular activity types (vigorous, moderate, and walking), meaning total MET-minutes per week scores could not be calculated for these patients per standardized scoring methods (33). Thus, in the present analysis, we only included vigorous, moderate, and walking MET-minutes per week variables.
Statistical analysis
The present article is an exploratory analysis; accordingly, the sample size acquired in the trial was studied. As summary statistics, between-group differences in change scores for questionnaire variables were determined using unadjusted independent-samples t tests. Absolute Cohen's d effect size (ES) values were also assessed for change scores (34). The magnitude of an ES value was considered trivial (<0.20), small (0.20–0.49), medium (0.50–0.79), or large (≥0.80) (34).
Our objective was to identify the mediators of weight change in the ILI compared with UC; in other words, we aimed to test the extent to which a set of variables (mediators) explained weight differences between the ILI group and the UC group. Multilevel mediation analysis was used to measure the effects conveyed by intervening variables (mediators) to the observed relation between an exposure variable and an outcome variable (35, 36). In this analysis, the mediator (change in questionnaire variables) and outcome (weight change) variables were continuous, whereas the exposure variable was binary (ILI or UC group). We built random intercept models to account for the correlation among subjects within the clinic. As part of the analysis, the total effect was estimated at the individual level; that is, the average difference in weight change (outcome variable) caused by the ILI compared with UC (exposure variable). The analysis further separated the total effect of the ILI (compared with UC) on weight change into 2 components: the indirect effects from mediators and the direct effect. The indirect effect is the effect of the ILI (compared with UC) on weight change that is driven by each proposed mediator; the direct effect is the effect of the ILI (compared with UC) on weight change that is not explained by the change in the proposed mediators.
In accord with the aims of the article, 2 conceptual models were used to guide the analyses (Figure 1). The first model aimed to determine the mediators of weight change induced by the ILI relative to UC during the first 12 mo of the trial. This was chosen to highlight mediators of midterm weight loss (6 to ≤12 mo) (37). The exposure variable was the trial group (ILI or UC), the proposed mediators were change in questionnaire variables from baseline to month 6, and the outcome variable was weight change from baseline to month 12. In the second model, the aim was to assess the mediators of weight change during the second 12 mo of the trial. This was broadly chosen to identify mediators of weight change during periods of weight-loss maintenance. This model had the same exposure variable as model 1, although change in questionnaire variables from baseline to month 12 were the proposed mediators and change in weight from month 12 to month 24 was the outcome variable. In both models, the proposed mediators preceded the outcome variable, with a time difference (6 mo in model 1 and 12 mo in model 2) between the final measurement of the proposed mediators and the outcome variable. This was to ensure temporal ordering of our exposure variable, proposed mediators, and outcome variable, limiting the confounding influence of reverse causality. In addition, we removed patients with censored weights from the mediation models. Weight measurements were censored if a patient became pregnant, developed a medical condition, or died.
FIGURE 1.
Hypothetical mediation models.
We conducted our analyses using the multilevel mediation analysis method of Yu and colleagues (36, 38), which is implemented in the mlma package in the software R. Briefly, potential mediators from our proposed mediators were informally selected if 2 conditions were satisfied. First, the proposed mediator distributed differently with or without the study's intervention (ILI compared with UC). In this regard, we used the ANOVA method to test if the mean of the variable differed between the ILI and UC. Second, the variable was significantly related to the outcome (weight change) while adjusting for all other related factors. This condition was tested through mixed-effect generalized linear models, with linear regression models used for linear outcomes or mediators. If only the second condition was satisfied, the variable was included as a covariate; yet the variable was excluded if the second condition was not satisfied (39). Further to the tests of 2 conditions, the package allows related variables to be forced into the model as mediators or covariates and it can assess joint effects of groups of mediators. Because the PROMIS-29 is used to determine overall health-related quality of life and no total score is obtained in the measure, we forced all PROMIS-29 constructs into the model as potential mediators and their joint effect was estimated. We likewise forced vigorous, moderate, and walking MET scores into the model as potential mediators and estimated the joint effect of these variables. Age, sex, race, baseline values for selected mediators, and weight (baseline weight for model 1; month 12 weight for model 2) were added as covariates. We estimated absolute total, direct, and indirect effects, as well as relative direct and indirect effects that provide the magnitude of these effects as a proportion of the total effect. For both the absolute and relative effect estimates, the SE and asymmetric 95% CIs around estimates were calculated, with inferences made using the bootstrap method. Unless noted otherwise, within the text, data are displayed as mean ± SD and 95% CI.
Results
Patient characteristics
A total of 803 patients with obesity (BMI: 37.2 ± 4.7) and a mean ± SD age of 49.4 ± 13.1 y were enrolled in the trial from 18 clinics: 452 patients from 9 clinics enrolled into the ILI and 351 patients from 9 clinics enrolled into UC (Figure 2). Details of the sample and the numbers who missed visits and withdrew are reported in the primary outcome article (9). The majority of patients were female (n = 678; 84.4%), were black (n = 540; 67.2%), and had a total household income <$40,000 (n = 515; 64.1%) (Table 1). Moreover, 247 patients (30.8%) were food insecure.
FIGURE 2.
Participant flowchart for the analyses.
TABLE 1.
Baseline characteristics and measures of the PROPEL (Promoting Successful Weight Loss in Primary Care in Louisiana) trial cohort1
| All (n = 803) | ILI (n = 452) | UC (n = 351) | |
|---|---|---|---|
| Age, y | 49.4 ± 13.1 | 48.8 ± 12.7 | 50.1 ± 13.6 |
| Sex | |||
| Male | 125 (15.6) | 54 (11.9) | 71 (20.2) |
| Female | 678 (84.4) | 398 (88.1) | 280 (79.8) |
| Race | |||
| White | 208 (25.9) | 95 (21.0) | 113 (32.2) |
| Black | 540 (67.2) | 332 (73.5) | 208 (59.3) |
| Other | 55 (6.8) | 25 (5.5) | 30 (8.5) |
| Total annual household income, $ | |||
| <10,000 | 156 (19.4) | 86 (19.0) | 70 (19.9) |
| 10,000–19,999 | 168 (20.9) | 95 (21.0) | 73 (20.8) |
| 20,000–39,999 | 191 (23.8) | 112 (24.8) | 79 (22.5) |
| 40,000–59,999 | 117 (14.6) | 69 (15.3) | 48 (13.7) |
| >60,000 | 154 (19.2) | 83 (18.4) | 71 (20.2) |
| Missing | 17 (2.1) | 7 (1.5) | 10 (2.8) |
| Household food security status | |||
| Food insecure | 247 (30.8) | 129 (28.5) | 118 (33.6) |
| Food secure | 556 (69.2) | 323 (71.5) | 233 (66.4) |
| Weight, kg | 102.1 ± 16.7 | 101.6 ± 16.4 | 102.7 ± 17.0 |
| BMI, kg/m2 | 37.2 ± 4.7 | 37.3 ± 4.6 | 37.2 ± 4.8 |
| EI | |||
| EI, restraint | 9.6 ± 4.5 | 9.6 ± 4.5 | 9.5 ± 4.5 |
| EI, disinhibition | 6.9 ± 3.7 | 7.0 ± 3.6 | 6.7 ± 3.7 |
| Dietary intake questionnaire | |||
| NCI, percentage fat intake | 35.3 ± 6.4 | 35.9 ± 6.7 | 34.6 ± 5.9 |
| NCI, fruit and vegetable intake | 2.2 ± 1.7 | 2.2 ± 1.6 | 2.3 ± 1.8 |
| BEVQ-15, alcohol intake | 0.2 ± 0.4 | 0.2 ± 0.4 | 0.2 ± 0.4 |
| Physical activity, MET-min/wk | |||
| IPAQ, vigorous | 561.4 ± 956.1 | 504.7 ± 891.1 | 634.3 ± 1030.3 |
| IPAQ, moderate | 475.2 ± 839.4 | 435.9 ± 803.3 | 525.2 ± 881.9 |
| IPAQ, walking | 808.9 ± 1011.2 | 780.8 ± 1027.3 | 844.0 ± 991.2 |
| Weight-related quality of life | |||
| IWQOL, total score | 73.9 ± 19.0 | 72.8 ± 19.5 | 75.3 ± 18.3 |
| Health-related quality of life | |||
| PROMIS-29, sadness | 47.5 ± 8.6 | 47.0 ± 8.5 | 48.1 ± 8.7 |
| PROMIS-29, pain interference | 51.9 ± 9.6 | 51.5 ± 9.7 | 52.5 ± 9.4 |
| PROMIS-29, physical function | 48.6 ± 8.0 | 48.9 ± 7.9 | 48.1 ± 8.1 |
| PROMIS-29, social functioning | 54.8 ± 9.0 | 55.2 ± 8.9 | 54.3 ± 9.1 |
| PROMIS-29, fatigue | 50.1 ± 10.1 | 49.4 ± 9.8 | 50.9 ± 10.4 |
| PROMIS-29, anxiety | 51.9 ± 9.9 | 51.7 ± 9.7 | 52.2 ± 10.1 |
| PROMIS-29, sleep disturbance | 50.7 ± 9.4 | 50.2 ± 9.2 | 51.5 ± 9.5 |
Values are mean ± SD for continuous data and n (%) for categoric variables. BEVQ-15, Brief Questionnaire to Assess Habitual Beverage Intake; EI, Eating Inventory; ILI, intensive lifestyle intervention; IPAQ, International Physical Activity Questionnaire; IWQOL, Impact of Weight on Quality of Life-Lite; MET, metabolic equivalent of task; NCI, National Cancer Institute; PROMIS-29, Patient-Reported Outcomes Measurement Information System-29; UC, usual care.
During the trial, 24 patients had their month 12 weight censored, whereas a further 12 had their weight censored at month 24 (Figure 2). Therefore, the first mediation analysis and related summary comparisons (change scores in mediators from baseline to month 6) included 779 (439 ILI; 340 UC) patients, whereas the second mediation analysis and related summary comparisons (change scores for mediators from baseline to month 12) included 767 patients (433 ILI; 334 UC). Supplemental Tables 1 and 2 show baseline characteristics of these analytical samples.
Change scores
The analysis in the primary outcome article showed that weight loss in the ILI group was greater than in the UC group at month 12 (ILI: −7.22 kg; 95% CI: −8.25, −6.19 kg; UC: −0.99 kg; 95% CI: −2.08, 0.09 kg) and month 24 (ILI: −5.43 kg; 95% CI: −6.52, −4.34 kg; UC: −0.91 kg; 95% CI: −2.07, 0.24 kg) (9).
Unadjusted independent-sample t tests suggested that the ILI group displayed a significant and large increase in restraint compared with the UC group at month 6 and month 12 (P < 0.001; ES ≥ 1.16), whilst a 0.9-point reduction in disinhibition was shown in the ILI relative to UC at month 6 (P < 0.001; ES = 0.33) (Table 2). At month 6 and month 12, compared with the UC group, the ILI group showed a small reduction in percentage fat intake and an increase in fruit and vegetable intake (P ≤ 0.010; ES ≥ 0.20); yet both groups reported a similar change in alcohol intake at months 6 and 12 (P ≥ 0.662; ES ≤ 0.03) (Table 2). The ILI group reported an increase in all physical activity constructs at month 6 (P ≤ 0.035; ES ≥ 0.17), although only change in vigorous physical activity was greater in the ILI group than in the UC group at month 12 (Table 2).
TABLE 2.
Change scores in questionnaire variables at month 6 and month 12 during the PROPEL (Promoting Successful Weight Loss in Primary Care in Louisiana) trial1
| Baseline to month 6 | Baseline to month 12 | |||||||
|---|---|---|---|---|---|---|---|---|
| ILI (n = 439) | UC (n = 340) | P | Cohen's d | ILI (n = 433) | UC (n = 334) | P | Cohen's d | |
| EI | ||||||||
| EI, restraint | 6.3 ± 4.5 | 0.7 ± 3.6 | <0.001 | 1.37 | 5.6 ± 4.4 | 0.8 ± 3.9 | <0.001 | 1.16 |
| EI, disinhibition | −1.8 ± 3.1 | −0.9 ± 2.5 | <0.001 | 0.33 | −1.3 ± 3.1 | −1.0 ± 2.7 | 0.109 | 0.12 |
| Dietary intake questionnaire | ||||||||
| NCI, percentage fat intake | −3.7 ± 6.0 | −1.2 ± 5.5 | <0.001 | 0.43 | −2.9 ± 6.0 | −1.0 ± 5.1 | <0.001 | 0.34 |
| NCI, fruit and vegetable intake | 0.2 ± 1.6 | −0.2 ± 1.7 | 0.003 | 0.23 | 0.2 ± 1.7 | −0.1 ± 1.9 | 0.010 | 0.20 |
| BEVQ-15, alcohol intake | 0.0 ± 0.3 | 0.0 ± 0.3 | 0.662 | 0.03 | 0.0 ± 0.3 | 0.0 ± 0.3 | 0.881 | 0.01 |
| Physical activity, MET-min/wk | ||||||||
| IPAQ, vigorous | 300.5 ± 1038.2 | 43.4 ± 1172.5 | 0.003 | 0.23 | 272.8 ± 1124.7 | 74.0 ± 1173.4 | 0.030 | 0.17 |
| IPAQ, moderate | 189.0 ± 1049.8 | 17.3 ± 989.6 | 0.035 | 0.17 | 238.0 ± 1047.3 | 138.8 ± 964.0 | 0.228 | 0.10 |
| IPAQ, walking | 211.1 ± 1203.9 | 9.5 ± 1047.3 | 0.034 | 0.18 | 125.4 ± 1283.0 | −4.4 ± 1117.0 | 0.215 | 0.11 |
| Weight-related quality of life | ||||||||
| IWQOL, total score | 10.9 ± 14.0 | 3.2 ± 10.8 | <0.001 | 0.62 | 12.1 ± 14.4 | 3.4 ± 11.5 | <0.001 | 0.67 |
| Health-related quality of life | ||||||||
| PROMIS-29, sadness | −0.1 ± 7.1 | 0.7 ± 7.9 | 0.177 | 0.10 | 0.2 ± 7.7 | 0.8 ± 7.4 | 0.281 | 0.08 |
| PROMIS-29, pain interference | −1.3 ± 8.1 | 0.3 ± 8.3 | 0.008 | 0.20 | −0.9 ± 9.2 | 0.9 ± 8.3 | 0.009 | 0.20 |
| PROMIS-29, physical function | 2.3 ± 6.4 | 0.1 ± 6.7 | <0.001 | 0.34 | 1.8 ± 6.9 | −0.1 ± 6.6 | <0.001 | 0.28 |
| PROMIS-29, social functioning | 2.0 ± 7.5 | 0.1 ± 7.4 | <0.001 | 0.26 | 2.0 ± 8.2 | 0.4 ± 7.8 | 0.007 | 0.21 |
| PROMIS-29, fatigue | −3.0 ± 9.2 | −0.6 ± 8.2 | <0.001 | 0.28 | −2.2 ± 9.2 | −0.8 ± 8.9 | 0.041 | 0.16 |
| PROMIS-29, anxiety | −1.4 ± 9.2 | 0.6 ± 8.9 | 0.003 | 0.22 | −1.0 ± 8.9 | 0.0 ± 8.5 | 0.153 | 0.11 |
| PROMIS-29, sleep disturbance | −1.8 ± 8.0 | 0.3 ± 7.7 | <0.001 | 0.28 | −0.8 ± 8.7 | 0.1 ± 8.5 | 0.174 | 0.10 |
Values are mean ± SD unless indicated otherwise. Independent-sample t tests compared change scores between groups at month 6 and month 12. Absolute Cohen's d effect size values were used to compare change scores between groups at month 6 and month 12. BEVQ-15, Brief Questionnaire to Assess Habitual Beverage Intake; EI, Eating Inventory; ILI, intensive lifestyle intervention; IPAQ, International Physical Activity Questionnaire; IWQOL, Impact of Weight on Quality of Life-Lite; MET, metabolic equivalent of task; NCI, National Cancer Institute; PROMIS-29, Patient-Reported Outcomes Measurement Information System-29; UC, usual care.
There was an increase in weight-related quality of life in the ILI group relative to the UC group at months 6 and 12 (P < 0.001; ES ≥ 0.62) (Table 2). Apart from sadness (P = 0.177; ES = 0.10), all health-related quality of life constructs of the PROMIS-29 were significantly improved in the ILI at month 6 compared with UC, with trivial-to-small ESs observed (P ≤ 0.008; ES ≥ 0.20) (Table 2). At month 12, however, statistically significant improvements were only observed for pain interference, physical function, social functioning, and fatigue in the ILI group relative to the UC group (P ≤ 0.041; ES ≥ 0.16) (Table 2).
Mediation analysis
Table 3 summarizes results from the mediation analyses. In model 1 (baseline to month 12 weight change), restraint, disinhibition, percentage fat intake, and weight-related quality of life total score met the 2 criteria and were selected as potential mediators alongside the IPAQ and PROMIS-29 variables and their composite scores. Similar to the primary outcome article (9), the total effect of the ILI (compared with UC) on weight change at month 12 was −7.19 kg (95% CI: −8.43, −6.07 kg). The direct effect [i.e., effect of ILI (compared with UC) on 12-mo weight change independent of change in mediators] was −5.36 kg (95% CI: −6.90, −3.94 kg), with a relative effect estimate showing that 75% of the between-group weight change was not caused by mediators. Of the selected potential mediators, disinhibition, percentage fat intake, moderate physical activity, walking, and fatigue change from baseline to month 6 were significant mediators of the improved weight loss displayed by the ILI group compared with the UC group at month 12. Specifically, month 6 change in disinhibition, percentage fat intake, moderate physical activity, walking, and fatigue explained −0.33 kg (95% CI: −0.55, −0.10 kg), −0.25 kg (95% CI: −0.50, −0.01 kg), −0.13 kg (95% CI: −0.23, −0.03 kg), −0.11 kg (95% CI: −0.21, −0.02 kg), and −0.28 kg (95% CI: −0.46, −0.10 kg), respectively, of the 12-mo weight change caused by the ILI (compared with UC). The joint indirect effect of physical activity (composite score for vigorous physical activity, moderate physical activity, and walking) was also significant and explained −0.26 kg (95% CI: −0.41, −0.09 kg) of the 12-mo weight change caused by the ILI (compared with UC). The relative effect estimates indicated that disinhibition, percentage fat intake, physical activity (joint effect), and fatigue explained 5%, 4%, 4%, and 4%, respectively, of the improved weight change seen in the ILI group compared with the UC group at month 12. Restraint was not a statistically significant mediator (−0.70 kg; 95% CI: −1.44, 0.03 kg). Similarly, the individual and joint effects of other PROMIS-29 variables and the change in weight-related quality of life did not significantly mediate month 12 weight change induced by the ILI (compared with UC) (Table 3).
TABLE 3.
Total, direct, and indirect effects of the PROPEL (Promoting Successful Weight Loss in Primary Care in Louisiana) ILI (compared with UC) on weight change, with questionnaire variables as mediators1
| Absolute effect | Relative effect | |||
|---|---|---|---|---|
| Estimate | 95% CI | Estimate | 95% CI | |
| Model 1 (baseline to month 12 weight change)2 | ||||
| EI | ||||
| EI, restraint | −0.70 ± 0.40 | −1.44, 0.03 | 0.10 ± 0.05 | −0.01, 0.20 |
| EI, disinhibition | −0.33 ± 0.11 | −0.55, −0.10 | 0.05 ± 0.02 | 0.02, 0.07 |
| Dietary intake questionnaire | ||||
| NCI, percentage fat intake | −0.25 ± 0.12 | −0.50, −0.01 | 0.04 ± 0.02 | 0.00, 0.07 |
| Physical activity, MET-min/wk | ||||
| IPAQ, joint effect of constructs3 | −0.26 ± 0.08 | −0.41, −0.09 | 0.04 ± 0.01 | 0.01, 0.06 |
| IPAQ, vigorous | 0.02 ± 0.06 | −0.14, 0.11 | 0.00 ± 0.01 | −0.02, 0.02 |
| IPAQ, moderate | −0.13 ± 0.05 | −0.23, −0.03 | 0.02 ± 0.01 | 0.00, 0.03 |
| IPAQ, walking | −0.11 ± 0.05 | −0.21, −0.02 | 0.02 ± 0.01 | 0.00, 0.03 |
| Weight-related quality of life | ||||
| IWQOL, total score | −0.35 ± 0.21 | −0.76, 0.04 | 0.05 ± 0.03 | −0.01, 0.11 |
| Health-related quality of life | ||||
| PROMIS-29, joint effect of constructs4 | 0.06 ± 0.13 | −0.19, 0.33 | −0.01 ± 0.02 | −0.05, 0.03 |
| PROMIS-29, sadness | 0.01 ± 0.03 | −0.05, 0.06 | −0.00 ± 0.00 | −0.01, 0.01 |
| PROMIS-29, pain interference | −0.02 ± 0.06 | −0.13, 0.09 | 0.00 ± 0.01 | −0.01, 0.02 |
| PROMIS-29, physical function | 0.14 ± 0.09 | −0.03, 0.31 | −0.02 ± 0.01 | −0.04, 0.01 |
| PROMIS-29, social functioning | 0.03 ± 0.08 | −0.12, 0.20 | −0.00 ± 0.01 | −0.03, 0.02 |
| PROMIS-29, fatigue | −0.28 ± 0.09 | −0.46, −0.10 | 0.04 ± 0.01 | 0.01, 0.06 |
| PROMIS-29, anxiety | 0.12 ± 0.08 | −0.03, 0.27 | −0.02 ± 0.01 | −0.04, 0.00 |
| PROMIS-29, sleep disturbance | 0.06 ± 0.09 | −0.12, 0.25 | −0.01 ± 0.01 | −0.03, 0.02 |
| Direct effect | −5.36 ± 0.76 | −6.90, −3.94 | 0.75 ± 0.06 | 0.62, 0.87 |
| Total effect | −7.19 ± 0.60 | −8.43, −6.07 | — | — |
| Model 2 (month 12 to month 24 weight change)5 | ||||
| Dietary intake questionnaire | ||||
| NCI, fruit and vegetable intake | 0.13 ± 0.04 | 0.05, 0.21 | 0.06 ± 0.02 | 0.01, 0.11 |
| Physical activity, MET-min/wk | ||||
| IPAQ, joint effect of constructs3 | 0.03 ± 0.04 | −0.06, 0.10 | 0.01 ± 0.02 | −0.03, 0.05 |
| IPAQ, vigorous | 0.02 ± 0.04 | −0.06, 0.09 | 0.01 ± 0.02 | −0.03, 0.04 |
| IPAQ, moderate | 0.01 ± 0.02 | −0.03, 0.05 | 0.00 ± 0.01 | −0.02, 0.02 |
| IPAQ, walking | 0.00 ± 0.03 | −0.05, 0.06 | 0.00 ± 0.01 | −0.02, 0.03 |
| Health-related quality of life | ||||
| PROMIS-29, joint effect of constructs4 | 0.09 ± 0.07 | −0.06, 0.23 | 0.04 ± 0.03 | −0.03, 0.10 |
| PROMIS-29, sadness | 0.02 ± 0.02 | −0.02, 0.06 | 0.01 ± 0.01 | −0.01, 0.03 |
| PROMIS-29, pain interference | 0.06 ± 0.06 | −0.06, 0.17 | 0.02 ± 0.03 | −0.03, 0.08 |
| PROMIS-29, physical function | 0.09 ± 0.07 | −0.05, 0.23 | 0.04 ± 0.03 | −0.03, 0.11 |
| PROMIS-29, social functioning | −0.03 ± 0.05 | −0.13, 0.07 | −0.02 ± 0.03 | −0.07, 0.03 |
| PROMIS-29, fatigue | −0.01 ± 0.04 | −0.08, 0.06 | 0.00 ± 0.02 | −0.04, 0.03 |
| PROMIS-29, anxiety | −0.05 ± 0.03 | −0.10, 0.01 | −0.02 ± 0.01 | −0.05, 0.01 |
| PROMIS-29, sleep disturbance | 0.01 ± 0.02 | −0.04, 0.05 | 0.00 ± 0.01 | −0.02, 0.03 |
| Direct effect | 2.00 ± 0.49 | 1.09, 3.02 | 0.89 ± 0.04 | 0.81, 0.98 |
| Total effect | 2.24 ± 0.49 | 1.32, 3.26 | — | — |
Values and 95% CIs were calculated with the mlma package of Yu and colleagues (36, 38). Absolute effects are estimated means ± SEs, whereas relative direct and indirect effects are the corresponding direct or indirect effect divided by the total effect ± SE. EI, Eating Inventory; ILI, intensive lifestyle intervention; IPAQ, International Physical Activity Questionnaire; IWQOL, Impact of Weight on Quality of Life-Lite; MET, metabolic equivalent of task; NCI, National Cancer Institute; PROMIS-29, Patient-Reported Outcomes Measurement Information System-29; UC, usual care.
The exposure variable was group (ILI compared with UC), the proposed mediators were change in questionnaire variables from baseline to month 6, and the outcome variable was weight change from baseline to month 12. Adjusted for age, sex, race, baseline questionnaire variables for selected mediators, and baseline weight; n = 779 (439 ILI; 340 UC).
Indirect effect is a composite score of the joint effect of all constructs: vigorous MET-mins/wk, moderate MET-mins/wk, and walking MET-mins/wk.
Indirect effect is a composite score of the joint effect of all constructs: sadness, pain interference, physical function, social functioning, fatigue, anxiety, and sleep disturbance.
The exposure variable was group (ILI compared with UC), the proposed mediators were change in questionnaire variables from baseline to month 12, and the outcome variable was weight change from month 12 to month 24. Adjusted for age, sex, race, baseline questionnaire variables for selected mediators, and month 12 weight; n = 767 (433 ILI; 334 UC).
In model 2 (month 12 to month 24 weight change), only fruit and vegetable intake met the 2 criteria and was selected as a potential mediator with the IPAQ and PROMIS variables. The ILI group displayed a significant 2.24-kg (95% CI: 1.32, 3.26 kg) increase in weight from month 12 to month 24 compared with the UC group (Table 3). The direct effect in this model was 2.00 kg (95% CI: 1.09, 3.02 kg), with the relative effect estimate suggesting 89% of the increase in weight exhibited by the ILI group (compared with UC) was not explained by the selected mediators. The change in fruit and vegetable intake from baseline to month 12 was a significant mediator of the increase in weight shown by the ILI group relative to the UC group from month 12 to month 24 (0.13 kg; 95% CI: 0.05, 0.21 kg); the relative effect estimate suggested that this explained 6% of the weight gain shown by the ILI (compared with UC). None of the other indirect effects of the selected mediators were significant (Table 3), suggesting 12-mo change in these selected mediators from baseline did not explain or attenuate (i.e., inconsistent mediation) the increase in weight seen by the ILI group compared with the UC group from month 12 to month 24.
Discussion
Over 24 mo, an ILI induced weight loss relative to UC in a sample of underserved patients with obesity. These analyses showed that month 6 change in disinhibition, percentage fat intake, physical activity, and subjective fatigue partially mediated the weight change seen in the ILI group relative to the UC group at 12 mo. The ILI group lost more weight than the UC group at month 24, but weight gain of 2.24 kg was observed in the ILI compared with UC from month 12 to month 24, with fruit and vegetable intake identified as a mediator. Analyses showed that the change in questionnaire constructs explained a small amount of the between-group weight change. More specifically, each mediator explained ≤10% of the between-group weight change, and relative direct effect values indicated that ≥75% of the between-group weight change was not explained by assessed constructs. Nonetheless, although other unmeasured factors could drive between-group weight variations, these results could help improve weight regimens by highlighting critical constructs and behaviors for weight loss.
Behavioral lifestyle interventions typically offer counseling sessions that aim to improve the eating behaviors of individuals with obesity via an increase in dietary restraint and a reduction in dietary disinhibition (8). In this analysis, we observed that a decrease in dietary disinhibition was a significant mediator of 12-mo weight loss seen in the ILI group compared with the UC group. This is consonant with studies reporting that a reduction in dietary disinhibition (12) is associated with weight loss in individuals with obesity during lifestyle interventions, and it suggests that regimens provided to underserved cohorts should place particular focus on behavioral strategies linked to disinhibition. Such strategies could consist of those utilized during the PROPEL ILI behavioral sessions, including controlled eating of foods, eating habits in response to stress and negative emotion, and healthy eating during special events. In contrast to the decrease in disinhibition, the increase in dietary restraint was not a significant mediator in our analyses. This supports research showing no association between restraint and weight loss (40), although in contrast to some work (12), we may have been underpowered to detect a positive influence of dietary restraint on weight loss.
A core strategy recommended for weight loss in individuals with obesity is the adoption of healthy dietary patterns. This includes limiting fat and alcohol intake and incorporating fruits, vegetables, and grains into a calorie-deficit diet (8, 37). Our results suggest that a decrease in percentage fat intake was a mediator of the between-group difference in month 12 weight loss, supporting previous analyses (13, 41) and suggesting that a reduction in fat intake is a key practice that assists the development of a calorie deficit and weight loss in underserved individuals with obesity. However, consistent with previous evidence (42), change in alcohol consumption did not mediate weight loss seen in the ILI compared with UC. In addition, although it did not influence weight loss in the first year of the trial, the increase in fruit and vegetable consumption seen in the ILI relative to UC did mediate the relative weight gain from month 12 to month 24. It is possible that fruit and vegetable consumption increased energy intake during a period of relapse in the ILI group, but it should be noted that research examining the influence of fruit and vegetable intake per se on long-term weight maintenance is mixed (14, 43). Therefore, further research is needed to elucidate the role of fruit and vegetable intake during weight-management interventions in underserved individuals with obesity.
Studies show that physical activity combined with dietary modifications stimulate greater weight loss over periods of ≥12 mo than do dietary modifications alone (44). We found, in line with these findings, that increased physical activity at month 6 mediated the greater 12-mo weight loss in the ILI, particularly the increase in moderate physical activity and walking. This suggests future weight-loss regimens in similar patient populations should seek to increase physical activity to improve weight loss. It could also imply that interventions should set more ambitious activity goals that have been recommended for weight loss, such as ≥200 min/wk of walking or moderate physical activity (45). However, because physical activity did not influence between-group weight change from month 12 to month 24, future research should elucidate the long-term role of physical activity during ILIs in underserved populations, especially because physical activity may be important in preventing weight regain (46). These studies should identify methods to sustain elevations in physical activity, given there were no differences in moderate physical activity and walking between the groups at month 12.
In addition to physical activity, model 1 revealed decreased fatigue as a mediator of improved 12-mo weight change during the ILI. Speculatively, although concurrent changes in weight and fatigue may reciprocally affect each other (15), the behavioral strategies of the ILI may have decreased subjective fatigue and led to better adherence to the weight-loss regimen compared with UC. Our analysis nonetheless indicated that other health- and weight-related quality of life constructs did not drive the greater weight loss seen in the ILI group during the first 12 mo, and quality of life changes at month 12 did not influence between-group weight change during the trial from month 12 to month 24.
A strength of the analyses is that they comprise data from a cluster-randomized trial performed in patients with obesity who are underserved and understudied within clinical studies. As a result, broadly, our findings have implications for socioeconomically disadvantaged individuals who are disproportionally affected by obesity and obesity-related conditions, and who face significant barriers for treatment. Another strength is that we collected mediator and weight measurements at multiple points during the 24-mo trial, enabling us to investigate mediators of weight change during periods which, despite variations in definitions (47), can be generally considered midterm weight loss (37) and weight-loss maintenance. This means our results can be utilized to develop enhanced ILIs which target constructs that are important for long-term weight management in similar at-risk populations. A final noteworthy strength is that trained health coaches of the trial were embedded within a team in primary care. This may explain why our analyses revealed many findings that are similar to those derived from more controlled trials, and it could supply a model for weight-management regimens in primary care (9).
The current article has limitations. First, the trial consisted mostly of females, restricting our ability to generalize our results to underserved males with obesity. Second, because there were no sessions offered in the UC group, we could not incorporate number of sessions attended into our analyses. It is possible that session attendance was a significant driver of weight change, as indicated by some studies (8). Third, these analyses are exploratory, so although our analyses comprised a relatively large number of participants from an understudied population, we included several variables and may be underpowered to detect some effects. Fourth, unmeasured mediator variables and mediator–outcome confounders may be more causally linked to between-group weight differences, limiting our ability to make causal inferences. Finally, measurement errors likely explain, at least in part, why our mediators explained a small proportion of between-group weight differences (48). Indeed, we used self-report assessments of diet and physical activity, which, in contrast to objective measures (e.g., waist devices for physical activity), are prone to systematic and random errors, primarily because they rely on recall and can be influenced by demand characteristics (49). The PROPEL trial was a pragmatic trial performed in a low-literate population in primary care; hence, a large battery of sophisticated assessments was unfeasible, and we decreased the burden of some questionnaires (e.g., not administering the EI hunger subscale). Yet additional research is needed during lifestyle interventions in underserved populations to elucidate the causal drivers of weight change. Such studies could examine further potential mediators of weight change like calorie intake, consistency of eating (50), hunger, energy density (51), and sugar-sweetened beverage consumption (52). Further, where possible, they should utilize reliable and objective assessment methods, particularly for diet (e.g., emerging technologies like food photography) and physical activity (e.g., pedometers or accelerometers) (49).
In conclusion, our analyses indicated that 12-mo weight loss during an ILI compared with UC was explained by improvements in disinhibition, percentage fat intake, physical activity, and fatigue in underserved patients with obesity. These variables did not, however, attenuate the weight gain shown during the ILI compared with UC in the final 12 mo of the trial, and fruit and vegetable intake may partially explain this response. Although additional work is needed using precise assessment methods to elucidate causal drivers of weight change during ILIs, these findings highlight psychological and behavioral constructs that could be targeted to refine interventions and facilitate weight management in underserved patients with obesity.
Supplementary Material
Acknowledgements
The PROPEL Research Group includes: Pennington Biomedical Research Center: Peter T. Katzmarzyk, PhD, Robert L. Newton Jr, PhD, Corby K. Martin, PhD, John W. Apolzan, PhD, William Johnson, PhD, Kara D. Denstel, MPH, Emily F. Mire, MS, Robert K. Singletary Jr, MHS, Cheryl Lewis, MPH, Phillip Brantley, PhD, Ronald Horswell, PhD, Betty Kennedy, PhD, Dachuan Zhang, MAppStats, Stephanie Authement, RD, LDN, MS, Shiquita Brooks, RDN, LDN, Danielle S. Burrell, MEd, MCHES, Leslie Forest-Everage, MA, Angelle Graham Ullmer, RDN, LDN, MS, Laurie Murphy, RDN, LDN, Cristalyn Reynolds, MA, Kevin Sanders, MS, RDN, LDN, Stephen Bower, MS, Daishaun Gabriel, MHA, Hillary Gahagan, MPH, Tabitha K. Gray, MA, Jill Hancock, MPH, Marsha Herrera, Brittany Molinere, Georgia Morgan, MA, Brittany Neyland, Stephanie Rincones, Deanna Robertson, MA, Ekambi Shelton, MPH, Russell J. Tassin, MS, Kaili Williams; Louisiana State University Health Sciences Center at New Orleans: Benjamin F. Springgate, MD; Louisiana State University Health Sciences Center at Shreveport: Terry C. Davis, PhD, Connie L. Arnold, PhD; Ochsner Health System: Eboni Price-Haywood,MD, Carl J. Lavie, MD, Jewel Harden-Barrios, MEd; Tulane University Medical School: Vivian A. Fonseca, MD, Tina K. Thethi, MD (MEDICAL Monitor), Jonathan Gugel, MD; Xavier University: Kathleen B. Kennedy, PhD, Daniel F. Sarpong, PhD, Amina D. Massey. Study data were collected and managed with REDCap (Research Electronic Data Capture) electronic data capture tools hosted at Pennington Biomedical. REDCap is a secure, web-based application designed to support data capture for research studies, providing an intuitive interface for validated data entry, audit trails for tracking data manipulation and export procedures, automated export procedures for seamless data downloads to common statistical packages, and procedures for importing data from external sources. We thank Health One (Carmel, CA) and Nutrisystem for providing portion-controlled meals during the study.
The authors’ responsibilities were as follows—JLD, CKM, CH, and PTK: conceived the project; CKM, JWA, RLN, and PTK: conceived the study, developed the overall research plan, and performed hands-on conduct of the experiments and data collection; JLD, QY, and WC: oversaw the analysis; JLD, CKM, and PTK: wrote the paper; CKM, JWA, RLN, KDD, EFM, and PTK: oversaw the study; QY and WC: performed the statistical analysis; KDD and EFM: managed the data; PTK: had primary responsibility for the final content; and all authors: read and approved the final manuscript. All other authors report no conflicts of interest.
Notes
Supported by Patient-Centered Outcomes Research Institute (PCORI) award OB-1402-10977 (to PTK). The statements in this article are wholly the responsibility of the authors and do not necessarily represent the views of the PCORI, its Board of Governors, or its Methodology Committee. Supported in addition by grant U54 GM104940 from the National Institute of General Medical Sciences of the National Institutes of Health (NIH), which funds the Louisiana Clinical and Translational Science Center, and by grant P30DK072476 (“Nutrition and Metabolic Health through the Lifespan”) from the Nutrition and Obesity Research Center, sponsored by the National Institute of Diabetes and Digestive and Kidney Diseases. JLD was supported by American Heart Association grant 20POST35210907. CH was supported by NIH National Research Service award T32 DK064584.
The intellectual property surrounding the mathematical code that creates the weight graph used in this study is owned by Louisiana State University/Pennington Biomedical and Montclair State University. CKM is an inventor of the intellectual property, which is included in a US and European patent application. The code has also been licensed and Louisiana State University/Pennington Biomedical, Montclair State University, and CKM have received royalties.
Supplemental Tables 1 and 2 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/ajcn/.
Abbreviations used: EI, Eating Inventory; ES, effect size; ILI, intensive lifestyle intervention; IPAQ, International Physical Activity Questionnaire; IWQOL, Impact of Weight on Quality of Life-Lite; MET, metabolic equivalent of task; NCI, National Cancer Institute; PROMIS-29, Patient-Reported Outcomes Measurement Information System-29; PROPEL, Promoting Successful Weight Loss in Primary Care in Louisiana; UC, usual care.
Contributor Information
James L Dorling, Human Nutrition, School of Medicine, Dentistry and Nursing, College of Medical, Veterinary & Life Sciences, University of Glasgow, Glasgow, United Kingdom; Pennington Biomedical Research Center, Baton Rouge, LA, USA.
Corby K Martin, Pennington Biomedical Research Center, Baton Rouge, LA, USA.
Qingzhao Yu, School of Public Health, Louisiana State University Health Sciences Center, New Orleans, LA, USA.
Wentao Cao, School of Public Health, Louisiana State University Health Sciences Center, New Orleans, LA, USA.
Christoph Höchsmann, Pennington Biomedical Research Center, Baton Rouge, LA, USA; Department of Sport and Health Sciences, Technical University of Munich, Munich, Germany.
John W Apolzan, Pennington Biomedical Research Center, Baton Rouge, LA, USA.
Robert L Newton, Jr, Pennington Biomedical Research Center, Baton Rouge, LA, USA.
Kara D Denstel, Pennington Biomedical Research Center, Baton Rouge, LA, USA.
Emily F Mire, Pennington Biomedical Research Center, Baton Rouge, LA, USA.
Peter T Katzmarzyk, Pennington Biomedical Research Center, Baton Rouge, LA, USA.
Data Availability
Complete de-identified data described in the article, code book, and analytic code are available from the corresponding author (PTK) by reasonable request pending the approval of the PROPEL publication committee.
References
- 1. Visscher TLS, Seidell JC. The public health impact of obesity. Annu Rev Public Health. 2001;22:355–75. [DOI] [PubMed] [Google Scholar]
- 2. Guh DP, Zhang W, Bansback N, Amarsi Z, Birmingham CL, Anis AH. The incidence of co-morbidities related to obesity and overweight: a systematic review and meta-analysis. BMC Public Health. 2009;9(1):88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Hales CM, Fryar CD, Carroll MD, Freedman DS, Ogden CL. Trends in obesity and severe obesity prevalence in US youth and adults by sex and age, 2007–2008 to 2015–2016. JAMA. 2018;319(16):1723–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Ogden CL, Fakhouri TH, Carroll MD, Hales CM, Fryar CD, Li Xet al. Prevalence of obesity among adults, by household income and education—United States, 2011–2014. MMWR Morb Mortal Wkly Rep. 2017;66(50):1369–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Hruby A, Hu FB. The epidemiology of obesity: a big picture. Pharmacoeconomics. 2015;33(7):673–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Carvajal R, Wadden TA, Tsai AG, Peck K, Moran CH. Managing obesity in primary care practice: a narrative review. Ann N Y Acad Sci. 2013;1281(1):191–206. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Jensen MD, Ryan DH. New obesity guidelines: promise and potential. JAMA. 2014;311(1):23–4. [DOI] [PubMed] [Google Scholar]
- 8. Jensen MD, Ryan DH, Apovian CM, Ard JD, Comuzzie AG, Donato KAet al. 2013 AHA/ACC/TOS guideline for the management of overweight and obesity in adults: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and The Obesity Society. J Am Coll Cardiol. 2014;63(25):2985–3023. [DOI] [PubMed] [Google Scholar]
- 9. Katzmarzyk PT, Martin CK, Newton RL, Apolzan JW, Arnold CL, Davis TCet al. Weight loss in underserved patients—a cluster-randomized trial. N Engl J Med. 2020;383(10):909–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Höchsmann C, Dorling JL, Martin CK, Newton RL, Apolzan JW, Myers CAet al. Effects of a two-year primary care lifestyle intervention on cardiometabolic risk factors—a cluster-randomized trial. Circulation. 2021;143(12):1202–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Aronne LJ, Hall KD, Jakicic JM, Leibel RL, Lowe MR, Rosenbaum Met al. Describing the weight-reduced state: physiology, behavior, and interventions. Obesity (Silver Spring). 2021;29(Suppl 1):S9–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Batra P, Das SK, Salinardi T, Robinson L, Saltzman E, Scott Tet al. Eating behaviors as predictors of weight loss in a 6 month weight loss intervention. Obesity (Silver Spring). 2013;21(11):2256–63. [DOI] [PubMed] [Google Scholar]
- 13. Zheng Y, Sereika SM, Danford CA, Imes CC, Goode RW, Mancino Jet al. Trajectories of weight change and predictors over 18-month weight loss treatment. J Nurs Scholarsh. 2017;49(2):177–84. [DOI] [PubMed] [Google Scholar]
- 14. Coughlin JW, Gullion CM, Brantley PJ, Stevens VJ, Bauck A, Champagne CMet al. Behavioral mediators of treatment effects in the Weight Loss Maintenance Trial. Ann Behav Med. 2013;46(3):369–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Palmeira AL, Markland DA, Silva MN, Branco TL, Martins SC, Minderico CSet al. Reciprocal effects among changes in weight, body image, and other psychological factors during behavioral obesity treatment: a mediation analysis. Int J Behav Nutr Phys Act. 2009;6(1):9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Katzmarzyk PT, Martin CK, Newton RL, Apolzan JW, Arnold CL, Davis TCet al. Promoting Successful Weight Loss in Primary Care in Louisiana (PROPEL): rationale, design and baseline characteristics. Contemp Clin Trials. 2018;67:1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture (REDCap)—a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform. 2009;42(2):377–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Diabetes Prevention Program (DPP) Research Group . The Diabetes Prevention Program (DPP): description of lifestyle intervention. Diabetes Care. 2002;25(12):2165–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.; The Look AHEAD Research Group, Wadden TA, Smith West D, Delahanty L, Jakicic J, Rejeski JThe Look AHEAD Research Group et al. The Look AHEAD study: a description of the lifestyle intervention and the evidence supporting it. Obesity (Silver Spring). 2006;14(5):737–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Rickman AD, Williamson DA, Martin CK, Gilhooly CH, Stein RI, Bales CWet al. The CALERIE study: design and methods of an innovative 25% caloric restriction intervention. Contemp Clin Trials. 2011;32(6):874–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Thomas DM, Martin CK, Heymsfield S, Redman LM, Schoeller DA, Levine JA. A simple model predicting individual weight change in humans. J Biol Dyn. 2011;5(6):579–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Jacques L, Jensen TS, Schafer J, McClain S, Chin J. Decision memo for intensive behavioral therapy for obesity (CAG-00423 N). [Internet]. Baltimore, MD: Centers for Medicare & Medicaid Services; 2011; [cited 26 March, 2019]. Available from: https://www.cms.gov/medicare-coverage-database/details/nca-decision-memo.aspx?&NcaName=Intensive Behavioral Therapy for Obesity&bc=ACAAAAAAIAAA&NCAId=253&. [Google Scholar]
- 23. Stunkard AJ, Messick S. The three-factor eating questionnaire to measure dietary restraint, disinhibition and hunger. J Psychosom Res. 1985;29(1):71–83. [DOI] [PubMed] [Google Scholar]
- 24. Stewart TM, Martin CK, Williamson DA. The complicated relationship between dieting, dietary restraint, caloric restriction, and eating disorders: is a shift in public health messaging warranted?. Int J Environ Res Public Health. 2022;19(1):491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Bryant EJ, King NA, Blundell JE. Disinhibition: its effects on appetite and weight regulation. Obes Rev. 2008;9(5):409–19. [DOI] [PubMed] [Google Scholar]
- 26. Thompson FE, Midthune D, Subar AF, Kipnis V, Kahle LL, Schatzkin A. Development and evaluation of a short instrument to estimate usual dietary intake of percentage energy from fat. J Am Diet Assoc. 2007;107(5):760–7. [DOI] [PubMed] [Google Scholar]
- 27. Havas S, Heimendinger J, Damron D, Nicklas TA, Cowan A, Beresford SAAet al. 5 a day for better health—nine community research projects to increase fruit and vegetable consumption. Public Health Rep. 1995;110(1):68–79. [PMC free article] [PubMed] [Google Scholar]
- 28. Thompson FE, Kipnis V, Subar AF, Krebs-Smith SM, Kahle LL, Midthune Det al. Evaluation of 2 brief instruments and a food-frequency questionnaire to estimate daily number of servings of fruit and vegetables. Am J Clin Nutr. 2000;71(6):1503–10. [DOI] [PubMed] [Google Scholar]
- 29. Hedrick VE, Savla J, Comber DL, Flack KD, Estabrooks PA, Nsiah-Kumi PAet al. Development of a brief questionnaire to assess habitual beverage intake (BEVQ-15): sugar-sweetened beverages and total beverage energy intake. J Acad Nutr Diet. 2012;112(6):840–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Kolotkin RL, Crosby RD. Psychometric evaluation of the impact of weight on quality of life-lite questionnaire (IWQOL-Lite) in a community sample. Qual Life Res. 2002;11(2):157–71. [DOI] [PubMed] [Google Scholar]
- 31. Kolotkin RL, Crosby RD, Kosloski KD, Williams GR. Development of a brief measure to assess quality of life in obesity. Obes Res. 2001;9(2):102–11. [DOI] [PubMed] [Google Scholar]
- 32. Cella D, Riley W, Stone A, Rothrock N, Reeve B, Yount Set al. The Patient-Reported Outcomes Measurement Information System (PROMIS) developed and tested its first wave of adult self-reported health outcome item banks: 2005–2008. J Clin Epidemiol. 2010;63(11):1179–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Craig CL, Marshall AL, Sjöström M, Bauman AE, Booth ML, Ainsworth BEet al. International Physical Activity Questionnaire: 12-country reliability and validity. Med Sci Sports Exerc. 2003;35(8):1381–95. [DOI] [PubMed] [Google Scholar]
- 34. Cohen J. Statistical power analysis for the behavioral sciences. Hillsdale, NJ: Lawrence Erlbaum Associates; 1988. [Google Scholar]
- 35. MacKinnon DP, Fairchild AJ, Fritz MS. Mediation analysis. Annu Rev Psychol. 2007;58:593–614. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Yu Q, Li B. Third-variable effect analysis with multilevel additive models. PLoS One. 2020;15(10):e0241072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Chopra S, Malhotra A, Ranjan P, Vikram NK, Sarkar S, Siddhu Aet al. Predictors of successful weight loss outcomes amongst individuals with obesity undergoing lifestyle interventions: a systematic review. Obes Rev. 2021;22(3):e13148. [DOI] [PubMed] [Google Scholar]
- 38. Yu Q, Yu M, Zou J, Wu X, Gomez SL, Li B. Multilevel mediation analysis on time-to-event outcomes: exploring racial/ethnic disparities in breast cancer survival in California. Res Methods Med Health Sci. 2021;2(4):157–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Yu Q, Li B. mma: an R package for mediation analysis with multiple mediators. J Open Res Softw. 2017;5(1):11. [Google Scholar]
- 40. Dorling JL, Bhapkar M, Das SK, Racette SB, Apolzan JW, Fearnbach SNet al. Change in self-efficacy, eating behaviors and food cravings during two years of calorie restriction in humans without obesity. Appetite. 2019;143:104397. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Delahanty LM, Meigs JB, Hayden D, Williamson DA, Nathan DM. Psychological and behavioral correlates of baseline BMI in the Diabetes Prevention Program (DPP). Diabetes Care. 2002;25(11):1992–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Morgan PJ, Hollis JL, Young MD, Collins CE, Teixeira PJ. Workday sitting time and marital status: novel pretreatment predictors of weight loss in overweight and obese men. Am J Mens Health. 2018;12(5):1431–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Kaiser KA, Brown AW, Brown MMB, Shikany JM, Mattes RD, Allison DB. Increased fruit and vegetable intake has no discernible effect on weight loss: a systematic review and meta-analysis. Am J Clin Nutr. 2014;100(2):567–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Donnelly JE, Herrmann SD, Lambourne K, Szabo AN, Honas JJ, Washburn RA. Does increased exercise or physical activity alter ad-libitum daily energy intake or macronutrient composition in healthy adults? A systematic review. PLoS One. 2014;9(1):e83498. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Donnelly JE, Blair SN, Jakicic JM, Manore MM, Rankin JW, Smith BKet al. American College of Sports Medicine Position Stand. Appropriate physical activity intervention strategies for weight loss and prevention of weight regain for adults. Med Sci Sports Exerc. 2009;41(2):459–71. [DOI] [PubMed] [Google Scholar]
- 46. Jakicic JM, Rogers RJ, Davis KK, Collins KA. Role of physical activity and exercise in treating patients with overweight and obesity. Clin Chem. 2018;64(1):99–107. [DOI] [PubMed] [Google Scholar]
- 47. Berger SE, Huggins GS, McCaffery JM, Lichtenstein AH. Comparison among criteria to define successful weight-loss maintainers and regainers in the Action for Health in Diabetes (Look AHEAD) and Diabetes Prevention Program trials. Am J Clin Nutr. 2017;106(6):1337–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48. Pieters R. Meaningful mediation analysis: plausible causal inference and informative communication. J Consum Res. 2017;44(3):692–716. [Google Scholar]
- 49. Dhurandhar NV, Schoeller D, Brown AW, Heymsfield SB, Thomas D, Sørensen TIAet al. Energy balance measurement: when something is not better than nothing. Int J Obes. 2015;39(7):1109–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Wing R, Phelan S. Long-term weight loss maintenance. Am J Clin Nutr. 2005;82(1):222S–5S. [DOI] [PubMed] [Google Scholar]
- 51. Ledikwe JH, Blanck HM, Khan LK, Serdula MK, Seymour JD, Tohill BCet al. Dietary energy density is associated with energy intake and weight status in US adults. Am J Clin Nutr. 2006;83(6):1362–8. [DOI] [PubMed] [Google Scholar]
- 52. Chen L, Appel LJ, Loria C, Lin P-H, Champagne CM, Elmer PJet al. Reduction in consumption of sugar-sweetened beverages is associated with weight loss: the PREMIER trial. Am J Clin Nutr. 2009;89(5):1299–306. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Complete de-identified data described in the article, code book, and analytic code are available from the corresponding author (PTK) by reasonable request pending the approval of the PROPEL publication committee.