1 INTRODUCTION

Obesity risk and treatment responses are widely believed to be influenced by a complex set of behavioral, psychosocial, metabolic, and environmental factors.1-5 Obesity prevalence varies greatly across different regions of the United States.6 For example, as of 2020 the prevalence of obesity in Colorado (CO) is 24.2%, while states in the southeast hovering between 35% and 40%. Specifically in Alabama (AL), the prevalence of obesity was reported at 39% in 2020.7 However, a relative paucity of research has investigated whether weight loss interventions are more or less effective in different geographical regions. The purpose of this study was to investigate whether the effectiveness of a comprehensive behavioral weight loss program differed when delivered concurrently in Birmingham, AL (high-obesity prevalence) and Denver, CO (low-obesity prevalence).

With regard to environment, both the physical or built environment and the social environment may impact weight management. The physical or built environment can impact the accessibility of healthy diets and physical activity (PA). The social environment can impact the motivation for healthy lifestyles.8, 9 Factors in the physical environment that impact weight management include access to recreational spaces and access to healthy food.10, 11 AL and CO have different food and activity environments. For example, AL has a higher availability of fast food restaurants when compared to CO, both in terms of restaurants per square mile and restaurants per person,12 which could promote poorer diet quality. Looking more closely at Denver and Birmingham's food environments and grocery stores, from 2011 to 2016 Denver saw an increase of 15.8% in numbers of grocery stores (101 to 117 stores), where Birmingham saw an increase of only 4.9% (102 to 107), which could indicate less access to these stores and possibly a dependence on ready-made foods in Birmingham.13

Engaging in outdoor physical activities such as walking and bicycling are also widely accepted as being more dangerous in AL compared to CO.14 When comparing Birmingham, AL and Denver, CO these cities have starkly different walk and bike scores. These scores were developed to measures walkability (access to walking routes to stores, schools, parks etc., safety of walking) and biking (accessibility to bike routes, road connectivity etc.). Birmingham's score is 35 for walking and 31 for biking, where Denver has a score of 61 for walking and 73 for biking.15 The average daily step count in CO has been reported as higher than in other states, including those in the southeast,16, 17 which demonstrates that individuals in CO not only have a more favorable PA environment but actually do engage in more PA when compared to other states. Climate is another non-modifiable environmental factor that could affect weight management. Merrill et al.18 concluded that areas with the highest level of PA had a more dry and moderate climate. CO's climate is drier and more moderate when compared to AL.19

The social environment could impact weight management through social norms for diet (e.g., widespread consumption of fried foods and sweetened iced tea) and exercise (cultural pressure to either avoid or engage in exercise). Psychosocial factors, such as perceived stress,20 can also influence weight status. Guite et al.21 reported that factors of the physical environment, such as greenspaces, neighborhood noise, and community facilities have a significant impact on psychological well-being. Colorado has a more favorable environment for activity, but this could suggest it has a more favorable environment for psychological outcomes, as well.

The environment can have a significant impact on obesity rates, there is still an incomplete understanding of how strong this influence is. The fact that rates of obesity vary greatly between regions in the U.S. could be due in part to differences in the food and PA environments. For example, it may be easier to lose weight in some parts of the U.S. than in others. No known previous trial has sought to compare the effectiveness of behavioral weight management programs in different environments. This study is a secondary analysis which took advantage of a clinical weight loss trial (parent study) being conducted in AL (Birmingham) and CO (Denver). The same weight loss intervention was delivered in both locations.

The purpose of this study was to determine how the state in which one lives (i.e., environment) affects the outcomes of a 16-week behavioral weight loss intervention in terms of body weight and factors associated with weight loss (PA and psychosocial factors). The hypothesis was that participants in CO would have greater weight loss, greater increase in PA, and greater improvements in psychosocial factors when compared with participants in AL after a 16-week weight loss intervention.

2 METHODS 2.1 Participants

This study is an ancillary study to a clinical trial being conducted in AL and CO for which participants with type 2 diabetes (T2D) are randomized to either a high protein (HP) or normal protein (NP) diet and asked to follow a weight management program (National Clinical Trial [NCT] 03832933). Participants were randomly assigned to either group using Statistical Analysis Software (SAS) and were stratified by sex (male or female), body mass index (BMI; <35 kg/m2 and ≥35 kg/m2), age (<50 years and ≥50 years), and time since T2D diagnosis (<3 years or ≥3 years). The HP group was asked to consume four or more servings of lean beef per week and avoid all other red meat, and the NP group was asked to avoid all red meat for the duration of the study. Participants attended weekly group classes for the 16-week intervention, which used the State of Slim (SOS) weight management program.22 This program is designed to target both diet and PA. A copy of the State of Slim (SOS) book was given to participants the first day of class along with class materials and access to the online SOS community.

For the parent trial, 70 participants (39 CO, 31 AL; 23 male, 47 female) were recruited between May and December 2019 in the Birmingham, AL, and Denver, CO, areas using letters, Internet advertisements, and news advertisements. Participants were required to be at least 18 years old, have a BMI ≥27 kg/m2, have a T2D diagnosis within the past 6 years, be weight stable (±3 kg in the past 3 months), and be stable on all medications for the past 3 months. Exclusion criteria were: hemoglobin A1c ≥ 12%, current eating disorder (anorexia or bulimia), dependence on illicit drugs or alcohol, untreated hypothyroidism, currently using insulin or other drugs known to cause weight loss or gain, following a vegetarian or vegan diet, any illness or injury that would make it unsafe to follow a diet and/or exercise up to 70 min at a moderate intensity regularly, and women who were pregnant, lactating, trying to become pregnant, or who had been pregnant or lactating in the last six months. Criteria for diabetes diagnoses were confirmed through medical records or doctor reports, blood biomarkers were confirmed via a blood test at the screening visit, and all other criteria were confirmed by self-report means. Of the 70 participants who provided consent for the study, 51 (21 AL, 30 CO) completed the 16-week intervention for an overall retention rate of 72.9%. A CONSORT diagram for each state can be seen in Figure 1 for AL and Figure 2 for CO. All participants of the parent trial were included in this study. The study was approved by the University of Alabama at Birmingham Institutional Review Board (IRB 300002928).

Consort diagram representing participant flow in Alabama (AL). AL, Alabama; BMI, body mass index; TSH, thyroid-stimulating hormone

Consort diagram representing participant flow in Colorado (CO). A1c, Hemoglobin A1c; BMI, body mass index; CO, Colorado; TSH, thyroid-stimulating hormone

2.1.1 Diet intervention

The SOS diet plan was used for the intervention.22 Typically, the SOS plan is low in fat and high in protein and emphasizes non-starchy vegetables and whole-grain carbohydrates. This diet plan consists of three distinct phases in which participants choose from specific food options. The SOS plan also has five diet rules that are to be followed throughout each phase: (1) Eat five to six times per day. (2) Eat breakfast within 1 h of waking. (3) Do not count calories; instead, measure portions. (4) Have the right protein mix at each meal (one carbohydrate and one protein at each meal). (5) Eat a healthy fat twice a day.

Participants were given food lists as well as recommended portion sizes in ounces or cups for each food at each phase of the diet plan, consistent with the lists in the book, with the exception of the red meat recommendation. The HP and NP groups were given food lists specific to their proposed macronutrient intake. The HP group was designed to have a reduction in carbohydrates and an increase in protein intake when compared with the NP group. Dietary fat was designed to be similar between groups. Table 1 illustrates the targeted macronutrient content of each group. Additionally, the HP group was instructed to consume ≥4 servings of lean beef per week, while avoiding all other red meat, and the NP group was instructed to avoid red meat for the duration of the intervention.

TABLE 1. Approximate macronutrient % by group Nutrient HP NP Carbohydrate % 32 53 Protein % 40 21 Fat % 28 25 Note: Diet plan approximate macronutrient percentage by group. Does not reflect actual intake of participants. Abbreviations: HP, High Protein; NP, Normal Protein. 2.1.2 Anthropometrics

Body weight was measured at baseline using a DETECTO BRW1000 scale (DETECTO, Webb City, MO). Participants were weighed in a fasted state (≥8 h fast) and were asked to void and remove heavy clothing. Due to restrictions placed on clinical research activities during the COVID-19 pandemic, Week 16 visits were completed remotely with video teleconferencing, and participants used home scales to weigh themselves at that visit, following the same protocol as the baseline visit (e.g., following an ≥8-h fast, void prior to weighing, wearing light clothing).

2.1.3 Physical activity

Physical activity was measured using an ActivPAL4 accelerometer (PALTechnologies, Glasgow, Scotland).23 Participants wore an accelerometer at baseline and Week 16 for 7 days at each time point. At baseline, a staff member attached the device to each participant's right thigh, and participants were given instructions for reattaching the device, should it come off. At the Week 16 visit, participants were mailed the accelerometer and asked to place it on their right thigh during a teleconferencing call, to confirm proper placement. The device was worn all day and was used to estimate number of steps, amount of active time (light activity and moderate-to-vigorous activity), and overall activity score.

2.1.4 Psychological assessments

A series of questionnaires were given to participants at baseline and at Week 16 to measure factors of psychological status. The assessments used were the Cohen Perceived Stress Scale,24 the Reward Based Eating Drive (RED) scale,25 the Behavior Rating Inventory of Executive Function-Adult Version (BRIEF-A)26 assessment, and the Ryff's Psychological Well-being (RPWB) scale. All questionnaires were sent via email to the participants to be completed directly within Research Electronic Data Capture (REDCap).

The 10-item version of the Cohen Perceived Stress Scale was used to determine stress exposure in the past month at baseline and at Week 16 of the intervention. The scale measures the degree to which respondents view their lives as stressful (e.g., overwhelming, unpredictable). It uses a 5-point Likert scale of 0 (never) to 4 (very often) to assess how often the respondent has felt a certain way in the previous month. This survey has demonstrated reliability and validity in a number of populations.24, 27, 28

The RED scale is designed to assess participants' control over their eating. It uses a 5-point Likert scale, from Strongly Disagree to Strongly Agree, to assess factors that drive overeating and lack of control overeating. The scale shows high internal validity and reliability across demographic factors.25 The 13-question version developed in 2017 was used in this analysis, as it is broader and has demonstrated greater validity and reliability than the nine-question version.29

The BRIEF-A is an assessment used to determine executive function in adults created as an extension of the original BRIEF assessment developed by Gioia et al. to determine executive function in children.30 Executive function includes abilities such as autonomy, self-regulation, working memory, problem solving, and planning and organizing.26 The version used is a 34-question modified version using a 7-point Likert scale from Never a Problem to Always a Problem.

An 18-question modified version of RPWB was used to assess psychological functioning. The assessment uses a 6-point scale from Strongly Disagree to Strongly Agree to measure six domains of well-being—self-acceptance, positive relations with others, autonomy, purpose in life, personal growth, and environmental mastery.31 The RPWB has demonstrated reliability and validity in multiple demographics.32

2.1.5 Social support assessment

A series of questionnaires were given to participants at baseline and at Week 16 to evaluate their social support. The three assessments used were the Social Support and Eating Habits Survey,33 the Social Support and Exercise Survey,33 and the Important Other Climate Questionnaire (IOCQ).34 Questionnaires were sent to participants via a link in an email, which they used to access and complete the questionnaires.

The Social Support and Eating Habits survey is a 10-question assessment that uses a 6-point scale from None to Very Often and includes Does Not Apply. Participants are asked to answer each question separately for friends and family to determine support for healthy eating habits based on each group. The assessment is used to determine support for healthy eating habits over the previous 3 months. This assessment shows good internal consistency as well as test-retest reliability.33

The Social Support and Exercise Habits survey is a 13-question assessment that uses a 6-point scale from None to Very Often and includes Does Not Apply. Participants are asked to answer each question separately for friends and family to determine support for participating in PA based on each group. The assessment is used to determine support for exercise habits over the previous three months. This assessment shows good internal consistency as well as test-retest reliability.33

The IOCQ is a six-question assessment that uses a 7-point Likert scale from Not True at All to Very True. This assessment is used to determine perceived support for reaching health goals (e.g., weight loss, exercise, etc.) from important people in the participant's life. This assessment has demonstrated strong reliability and validity in other dietary intervention trials.34

2.1.6 Changes due to COVID-19

Due to the COVID-19 pandemic, certain study procedures had to be altered. First, the group classes were switched to online for both AL and CO at Weeks 4–5 and 7–8, respectively. Instead of meeting in-person, the participants were sent a link to the group video call every week for the remainder of the study. The calls were still led by trained coaches each week. Additionally, the Week 16 study visits became at-home study visits. Participants were sent a link to a video call with research staff, who conducted the visit. All the questionnaires were sent to the participant via email, just as was done for the baseline visit. The major change to the Week 16 visit was that participants were weighed using their at-home scales as opposed to using the scale for the study. Participants were still weighed following a ≥8-h fast, were asked to void, and removed any heavy clothing prior to weighing. Additionally, participants attached the accelerometers themselves following instructions and under the supervision of research staff to ensure correct placement. After the 7-day period, the accelerometer was sent back to research staff in the mail.

2.2 Statistical analysis

All study data were collected and managed using REDCap electronic data capture tools hosted at the University of Alabama at Birmingham.35, 36 REDCap is a secure, web-based software platform designed to support data capture for research studies, providing (1) an intuitive interface for validated data capture; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for data integration and interoperability with external sources. All analyses were completed using SAS (version 9.4, 2002–2012 by SAS Institute Inc.). The sample size is constrained due to the study being an ancillary study of an ongoing clinical trial (NCT03832933: Dr. James O. Hill). Therefore, the statistical power calculations reflect an observable effect size given the constrained number of participants. Using an intention-to-treat approach, statistical power calculations indicate that the sample size provides 80% power to detect an effect size of 0.67 given an alpha-error probability of 0.05.

Baseline characteristics were assessed by state (AL and CO), as well as the total of the whole sample. These were assessed using descriptive statistics and are reported as the mean ± the standard deviation (SD). Differences in baseline characteristics were assessed using Student's t-tests or Chi-squared tests. Changes in weight, PA, and psychosocial factors were the primary outcomes for this study. A generalized linear model (PROC GENMOD) was used to test effects of time (baseline vs. Week 16), state (AL vs. CO) and their interaction term on changes in weight (kg), PA (number of steps, stepping time activity score), and psychosocial factors (hedonic eating, perceived stress, executive function, psychological well-being, and friend and family support). For the weight outcomes, an intention-to-treat analysis was performed and are shown in Table 4. Physical activity outcomes were analyzed using an intention-to-treat approach, and psychosocial outcomes were analyzed using a completers analysis to determine the effect within the population that completed the intervention. Models were adjusted for sex, race, and treatment group (HP vs. NP). Changes in these measures reported as LSMEANS ± SE, with α = 0.05 being used to determine statistical significance.

3 RESULTS 3.1 Baseline participant characteristics

Baseline participant characteristics are presented in Tables 2 and 3. Participants in the study were mostly female and had an overall average age of 53.86 ± 11.95 years and a BMI of 38.77 ± 6.69 kg/m2. With the exception of differences in racial and ethnic composition, participants in AL and CO were not significantly different. The sample in CO had a higher percentage of white participants when compared to participants in AL (CO: 74.36% vs. AL: 54.84%, X2 (1, N = 64) = 4.61, p = 0.03) and a higher percentage of Hispanic participants when compared to participants in AL (CO: 23.08% vs. AL: 0.00%, X2 (1, N = 70) = 8.21, p = 0.004). Income and education distribution also differed in these two states, with participants in CO having a higher income and education level when compared to participants in AL (Table 3).

TABLE 2. Baseline characteristics of participants Parameter Total Alabama Colorado Age (years) 53.86 ± 11.95 53.65 ± 12.88 54.03 ± 11.33 Body weight (kg) 108.42 ± 29.97 108.49 ± 27.34 108.37 ± 21.29 BMI (kg/m2) 38.77 ± 6.69 38.77 ± 7.43 38.77 ± 6.14 n 70 31 39 Female (n) 47 22 25 White (n) 46 17 29 Black (n) 18 12 6 Asian (n) 3 2 1 Other race (n) 3 0 3 Hispanic or Latino (n) 9 0 9 Note: Table 1 shows baseline and demographics for baseline of both states as well as the total population. Results are presented as least squares mean ± standard deviation. Data are presented as mean ± standard deviation. Abbreviations: kg, kilograms; kg/m2, kilograms per meter squared. TABLE 3. Education and income by state Parameter Alabama Colorado n (%) n (%) Education High school diploma 5 (16.12) 6 (15.38) Some college 5 (16.12) 8 (20.51) Bachelor's degree 17 (54.84) 11 (28.21) Master's degree 3 (9.68) 12 (30.77) Doctorate 1 (3.23) 1 (2.56) Did not disclose 0 (0.00) 1 (2.56) Income <$25,000 3 (9.68) 1 (2.56) $25,000–45,000 6 (19.35) 2 (5.13) $45,000–70,000 6 (19.35) 11 (28.21) $70,000–110,000 9 (29.03) 11 (28.21) $110,000+ 6 (19.35) 14 (35.90) Did not disclose 1 (3.23) 0 (0.00) Note: Table 2 shows baseline income and education data for each state. Results are reported as number of participants (n) as well as the percentage of participants within that state. 3.2 Weight loss

Both states had significant weight loss from baseline to Week 16, and there were no differences between states (Table 4). This data is also represented in Figure 3, which shows kg lost by each state (Table 4). Analyses demonstrated that 64.5% and 61.2% of participants in AL and CO, respectively, achieved ≥5% weight loss, and 41.9% and 30.8% achieved ≥10% weight loss in AL and CO, respectively.

TABLE 4. Weight loss outcomes State Baseline weight (n) Week 16 weight (n) % Weight loss p valuea Alabama 114.06 ± 3.66 kg (31) 101.54 ± 4.39 kg (21) 10.98 0.84 Colorado 113.00 ± 3.32 kg (39) 100.42 ± 3.67 kg (31) 11.65 Note: Table 3 shows change from baseline to Week 16 by state for body weight (kg) and change (%). Results are presented as least squares mean ± standard error. Abbreviation: kg, kilograms.

Changes in mass by state according to the Intention to Treat Analysis. AL, Alabama; CO, Colorado; kg, kilograms

3.3 Physical activity

There were no differences in any PA measures at baseline between the two states, and both states saw significant improvement in step counts, stepping time, and activity scores from baseline to Week 16. However, step counts, stepping time, and activity scores were greater in participants in AL versus CO at Week 16 (Table 5).

TABLE 5. Physical activity outcomes by state Parameter Group Baseline (n) Week 16 (n) p valuea Step count AL 6272.72 ± 581.45 (31) 10,266.97 ± 697.11 (21) CO 6320.49 ± 524.62 (38) 8428.99 ± 598.42 (30) 0.04 Step time (min) AL 83.22 ± 6.63 (31) 119.14 ± 7.91 (21) CO 81.52 ± 5.93 (38) 101.96 ± 6.82 (30) 0.09 Activity score AL 33.13 ± 0.21 (31) 34.54 ± 0.32 (21) CO 32.93 ± 0.20 (38) 33.83 ± 0.31 (30) 0.04 Note: Table 4 shows baseline and Week 16 PA levels by state. Results are presented as least squares mean ± standard error. All results are calculated using the intention-to-treat approach. Abbreviation: min, minutes. 3.4 Psychological and behavioral assessments

Executive function, perceived stress, and psychological well-being were not different between states, nor did they change from baseline to Week 16 (Table 6). Alabama participants reported significantly lower reward-based eating as measured by the RED score at both baseline and Week 16, but neither state reported a significant change in RED score over time.

TABLE 6. Psychological outcomes by state Assessment Group Baseline score (n) Week 16 score (n) p valuea Cohen perceived stress AL 18.42 ± 1.79 (20) 1 14.81 ± 1.68 (17) 0.60 CO 17.21 ± 1.33 (39) 1 13.72 ± 1.29 (28) RED scale AL 24.08 ± 2.42 (20) 18.62 ± 2.70 (17) 0.0023 CO 34.99 ± 2.12 (39) 29.11 ± 2.19 (28) BRIEF-A AL 74.24 ± 6.42 (20) 80.31 ± 6.12 (17) 0.65 CO 83.31 ± 4.67 (39) 76.79 ± 4.81 (28) RPWB AL 91.61 ± 3.10 (20) 89.92 ± 2.98 (17) 0.73 CO 84.99 ± 2.33 (39) 88.62 ± 2.41 (28) Note: Table 5 shows baseline, Week 16, and average change in psychological questionnaire scores by state. Results are presented as least squares mean ± standard error. All results are calculated using completers analysis. Abbreviations: BRIEF-A, Behavior Rating Inventory of Executive Function-Adult Version; RED, Reward-Based Eating Drive; RPWB, Ryff's Psychological Well-Being. 3.5 Social support

Baseline measures of social support did not differ between the states. At Week 16, friends diet support was significantly greater for participants in AL compared to CO. During the intervention, participants in AL saw significant improvements in family diet support, friends diet support, family exercise participation, and the IOCQ. Participants in CO saw significant improvement only in the family diet support outcome (Table 7).

TABLE 7. Social support outcomes by state Assessment Group Baseline score (n) Week 16 score (n) p valuea Family diet support AL 9.23 ± 1.22 (21) 13.64 ± 1.11 (17) 0.51 CO 11.29 ± 0.91 (39) 14.62 ± 0.93 (28) Family diet sabotage AL 11.01 ± 0.96 (21) 12.70 ± 1.04 (17) 0.20 CO 10.46 ± 0.82 (39) 11.11 ± 0.80 (28) Friends diet support AL 7.63 ± 1.04 (21) 12.59 ± 1.01 (17) 0.03 CO