Feasibility and Effectiveness of a Worksite-Weight-Loss Program for Cancer Prevention among School-District Employees with Overweight and Obesity

The effects of Vibrant Lives, a 6-month worksite-weight-loss program, were examined in a cohort of school-district employees with overweight or obesity. The VL Basic (VLB) participants received materials and tailored text messages, the VL Plus (VLP) participants additionally received WIFI-enabled activity monitors and scales and participated in health challenges throughout the school year, and the VL Plus with Support (VLP + S) participants additionally received coaching support. The levels of program satisfaction and retention and changes in weight, physical activity (PA), and diet were compared across groups using Pearson chi-square tests, repeated-measure mixed models, and logistic regression. After the program, the VLB (n = 131), VLP (n = 87), and VLP + S (n = 88) groups had average weight losses of 2.5, 2.5, and 3.4 kg, respectively, and average increases in weekly PA of 40.4, 35.8, and 65.7 min, respectively. The VLP + S participants were more likely than the other participants to have clinically significant weight loss (≥3%; p = 0.026). Compared with the VLB participants, the VLP participants were less likely to meet the recommendations for consuming fast food (p = 0.022) and sugar-sweetened beverages (p = 0.010). The VLP and VLP + S participants reported higher program satisfaction than the VLB participants. The VL program facilitates weight loss among school-district employees with overweight and obesity by increasing their PA and healthy diet.


Introduction
Obesity is associated with several comorbidities, including cardiovascular disease and diabetes [1], and an increased risk of many types of cancer [2,3]. Physical inactivity, excessive body fat, and sedentary behavior increase the risk of cancer [4], whereas higher levels of leisure-time physical activity (PA) are associated with a lower risk of cancer, including breast, colon, and lung cancer and myeloid leukemia and myeloma [5]. Thus, among its cancer-prevention guidelines, the American Cancer Society includes recommendations on how to achieve and maintain a healthy weight, adopt a physically active lifestyle, consume

Participants
This program enrolled employees of the Pasadena Independent School District, a public-school district in southeast Texas, over the 2017-2018 school year (Year 1) and the 2018-2019 school year (Year 2). Eligible participants had a BMI greater than 25 kg/m 2 in Year 1 and a BMI greater than 27 kg/m 2 in Year 2. The analysis of data from the program evaluation was reviewed by The University of Texas MD Anderson Cancer Center Institutional Review Board and was determined to be exempt.

Intervention
The VL program is a digital worksite-weight-loss program that promotes PA and healthy eating. The VL program was adapted from the Diabetes Prevention Program [29], whose recommendations for dietary intake, PA, and weight management are consistent with the cancer-prevention guidelines of the American Cancer Society [6]. The program was provided as part of the Pasadena Vibrant Community (PVC) initiative, a place-based cancer prevention program of the University of Texas MD Anderson Cancer Center's Be Well Communities™ initiative [30]. Details of DPP in this study are described in Table 1. The initial plan for the VL program was to offer a weight-loss program to schooldistrict employees that included print materials from an adapted Diabetes Prevention Program (DPP, delivered by email), daily text messages that supported the DPP lessons (7-12 per week), an activity tracker (Fitbit Flex 2, San Francisco, CA, USA) and WIFIconnected scale (Fitbit Aria Smart Scale, San Francisco, CA, USA), and opportunities to participate in group challenges (Vibrant Lives Plus, VLP). In the second year, participants were also offered the opportunity to participate in a closed, secret Facebook group which provided nutrition, weight-loss, and exercise information and included weekly challenges, such as posting a recipe modified to be healthier, or a photograph of a new place to exercise. We also planned to randomly select approximately half of the schools to receive an additional support component; participants in these schools would receive an offer to participate in telephone coaching if they had not lost weight at the midpoint of the program (Vibrant Lives Plus with Support, VLP + S). However, the demand for the program was high; we had funding to provide the VLP/VLP + S program to 100 participants per year and 400 expressed interest in the first year alone. Therefore, we worked with the school-district personnel to select schools that would receive the full VLP program (with and without the additional support) to enroll approximately 120 participants each year, and the employees in the other schools were offered Vibrant Lives Basic (VLB), a low-cost version that included the emailed materials, text messages, and, in the second year, access to a Facebook group. In the first year, the school/location selection for VLP and VLP + S prioritized employees working in the administrative office to build leadership support for the program and schools that were implementing the Coordinated Approach to Child Health (CATCH) program [31]. Selection of the schools/locations for the second year prioritized schools/locations not selected in the first year (Table 2). Enrolled employees participated in one of the three versions of the VL program. Participants in the Vibrant Lives Basic (VLB) group received emails with the DPP materials and text messages (7-12 per week) with tips for behavior change that were coordinated with the DPP lessons.

Measures
We used a survey to collect participants' demographic data, including age, race, sex, education, and marital status. Participants reported their self-measured weights at baseline and follow-up. Participants in the VLP and VLS groups self-measured their weights on their WIFI-enabled scales.
Participants reported their PA and dietary-intake behaviors at baseline and follow-up using the survey items from the Health Information National Trends Survey [32] and Health of Houston Survey [33]. Self-reports of PA included the number of minutes of moderate-to-vigorous aerobic activity per day and the number of times in which strength exercise was performed per week. Self-reports of dietary intake included the numbers of times red meat, fast food, and sugar-sweetened beverages were consumed per week, the number of cups of fruit and vegetables that was consumed per day, and the number of days breakfast was consumed during a week.
We scored participants' responses on the reports according to whether they met the recommendations of at least 150 min of moderate-to-vigorous PA per week and at least two sessions of strength exercise per week (meet, score = 1; did not meet, score = 0); how often they consumed red meat (≤2 times/day, score = 1; ≥3 times/day, score = 0), fast food (≤2 times/week, score = 1; ≥3 times/week, score = 0), and sugar-sweetened beverages (never in the week, score = 1; at least once in the week, score = 0); their daily consumption of fruit (≥1 cup/day, score = 1; <1 cup/day, score = 0) and vegetables (≥3 cups/day, score = 1; <3 times/day, score = 0); and how often they ate breakfast (every day in the past week, score = 1; 0-6 days in the week, score = 0).
Participants in the VLP and VLP + S groups used their Fitbits to measure their physical activity. After excluding non-valid days (those with <1500 steps or <10 h of valid wear) and non-valid weeks (those with <4 days of valid wear), we summed the number of daily steps and "very active", "fairly active", and "lightly active" minutes. "Active minutes" represented the combination of the number of Fitbit-derived "fairly active" and "very active" minutes per day. Detailed information about Fitbit-data use in the VL program is provided elsewhere [34].
The program-retention rate was the percentage of participants who completed both the baseline and follow-up assessments and a program-satisfaction survey. Program satisfaction was assessed with a Likert-scale survey (1 = strongly disagree; 2 = disagree; 3 = neutral; 4 = agree; 5 = strongly agree) with the following prompts: "Being in the program has motivated me to increase my physical activity"; "Being in the program has motivated me to adopt a healthy diet"; "Being in the program has motivated me to lose weight"; "This program was effective for me"; and "I would recommend this program to a friend or family member".

Statistical Analyses
We used descriptive statistics to summarize the baseline characteristics of the participants. One-way analysis of variance and the Pearson's chi-square test were used to detect differences in continuous variables (e.g., age, BMI) and categorical variables (e.g., race, sex, education, BMI category), respectively, between the participants who completed the program, "completers", and those who did not complete the program, "non-completers", among the three groups. Changes in weight, PA, and dietary intake between baseline and follow-up among the VLB, VLP, and VLP + S groups were analyzed using multi-level repeated-measure linear mixed models and logistic-regression models; in these analyses, we clustered participants by school and controlled for the study years (e.g., Year 1 and Year 2) and race/ethnicity as covariates. Differences in program retention and satisfaction among the three groups were assessed using the Pearson's chi-square test. Fitbit-derived daily step counts and active minutes were converted to weekly average, which were reported using linear regressions for each of the 26 weeks of the program and compared between VLP and VLP + S groups using repeated-measure mixed models. All statistical analyses were performed with the STATA 15.1 statistical software program (StataCorp LP, College Station, Texas). Results were reported using an unstandardized beta (β) or odds ratio (OR) with a standard error (SE) and p-value (p < 0.05).

Results
Of the 543 participants initially enrolled in the VL program, 306 (131, 87, and 88 in the VLB, VLP, and VLP + S groups, respectively) completed the program and follow-up assessments and were included in the analyses. The percentage of program completion (program retention) by the groups is shown in Figure 1. The retention rate in the VLB group (43%) was significantly lower than that in the VLP group (73%; χ 2 = 30.85, p < 0.001) and that in the VLP + S group (73%; χ 2 = 31.51, p < 0.001).
among the three groups were assessed using the Pearson's chi-square test. Fitbit-derived daily step counts and active minutes were converted to weekly average, which were reported using linear regressions for each of the 26 weeks of the program and compared between VLP and VLP + S groups using repeated-measure mixed models. All statistical analyses were performed with the STATA 15.1 statistical software program (StataCorp LP, College Station, Texas). Results were reported using an unstandardized beta (β) or odds ratio (OR) with a standard error (SE) and p-value (p < 0.05).

Results
Of the 543 participants initially enrolled in the VL program, 306 (131, 87, and 88 in the VLB, VLP, and VLP + S groups, respectively) completed the program and follow-up assessments and were included in the analyses. The percentage of program completion (program retention) by the groups is shown in Figure 1. The retention rate in the VLB group (43%) was significantly lower than that in the VLP group (73%; χ 2 = 30.85, p < 0.001) and that in the VLP + S group (73%; χ 2 = 31.51, p < 0.001). The baseline characteristics of the completers and non-completers are provided in the Table 3; no significant differences between the groups were detected. The baseline characteristics of the completers and non-completers are provided in the Table 3; no significant differences between the groups were detected. The characteristics of the 306 completers by intervention arm are provided in Table 4. Compared with the VLB and VLP + S groups, the VLP arm had a significantly higher proportion of participants who were Hispanic White or non-Hispanic White (χ 2 = 20.48, p = 0.025); no other significant differences among the groups were detected. The changes in weight, PA, and dietary intake between baseline and follow-up are shown in Figure 2. The VLB, VLP, and VLP + S groups all had significant weight losses (mean, 2.5, 2.5, and 3.4 kg, respectively; β = -2.34; SE = 0.45; p < 0.001) and increases in weekly moderate-to-vigorous PA (mean, 40.4, 35.8, and 65.7 min, respectively; β = 40.53; SE = 8.00; p < 0.001), but these variables did not differ significantly among the groups. The participants in the VLP + S were more likely to have clinically significant weight loss (≥3%) than the other participants (OR: 1.47; SE = 0.28; p = 0.045). All the groups had higher rates of moderate-to-vigorous PA (≥150 min/week of aerobic exercise; OR = 2.31, SE = 0.54, p < 0.001) and strength exercise (≥2 times/week; OR = 2.11, SE = 0.45, p < 0.001) at the end of the program, but these rates did not differ significantly among the groups. In addition, all the groups had significantly higher rates of consuming red meat two or fewer times per week (OR = 2.07; SE = 0.39; p < 0.001), consuming fast food two or fewer times per week (OR week (OR = 1.65, SE = 0.40, p = 0.037). They also had higher rates of consuming at least one cup of fruit per day (OR = 1.90; SE = 0.36; p = 0.001) and of consuming at least three cups of vegetables per day (OR = 2.00; SE = 0.44; p = 0.002). Compared with the VLB group, at the end of the program, the VLP group had lower rates of consuming fast food fewer than two times per week (OR = 0.48. SE = 0.15, p = 0.022), and not consuming sugar-sweetened beverages in the past week (OR = 0.46, SE = 0.14, p = 0.010).   The weekly average numbers of daily active minutes and steps for the VLP and VLP + S groups (n = 175) are given in Figure 3 The program-satisfaction rates of the VLP and VLP + S groups were significantly higher than that of the VLB group for each statement (Table 5).   The program-satisfaction rates of the VLP and VLP + S groups were significantly higher than that of the VLB group for each statement (Table 5).  1 Data are the number of participants (%) who responded "agree" or "strongly agree" to the prompts on the programsatisfaction survey.

Discussion
The VL program resulted in weight loss as well as positive changes in PA and diet in an ethnically diverse employee population, regardless of the version of the intervention they received. These findings are in line with those of other studies of internet-based worksite interventions [16,[35][36][37], and build on them, as this study is one of the few to focus on a school setting in a racially and ethnically diverse community, in which 71% of the population are Hispanic and 18.2% live below the poverty line (compared to 12.8% nationally) [38].
The present study evaluated a DPP behavior-change intervention adapted to be remotely delivered using digital strategies to encourage weight loss and healthier behaviors among school-district employees with overweight and obesity. The DPP was translated into practice in the workplace and showed its feasibility and effectiveness for weight loss. Two previous studies demonstrated that adaptations and modifications of DPP improved feasibility and effectiveness: minimizing program intensity and costs and varying delivery formats, such as small face-to-face groups, online, or telephone coaching improved retention rates [39,40]. However, the results of the adaptations are mixed: face-to-face group-based intervention showed higher retention but no group differences in weight loss [39], while telephone coaching showed more weight loss than the use of a face-to-face small group [40]. In our study, the digital delivery of DPP content resulted in weight loss, increases in PA, healthier diets, and high retention. Weight loss in the VL interventions resulted in greater or comparable weight loss (−3.4 to −2.5 kg) to these two studies (Ing et  Internet-based worksite PA interventions have shown higher attrition rates and lower participation rates [21,23]. However, Tate et al. showed that, compared with participants who were randomized to a weight-loss program alone, participants who additionally received counseling through emails, messages, virtual discussion boards, and other feedback had more weight loss and higher participation rates [41]. In the present study, participants with access to more interactive program components did not differ significantly in weight loss or changes in physical activity from those who completed the VLB program. However, the retention rate in the VLB group was dramatically different, with more than half of the participants (57%) dropping out before the end of the program. Other studies have shown that self-monitoring strategies that employ wearable fitness-tracking devices may help participants achieve their goals and motivate them to lose weight and become physically active [42,43]. In accordance with their higher retention rates, the VLP and VLP + S groups also had higher satisfaction rates than the VLB group did. Thus, the interactive components of the program, such as the activity monitor, connected scale, and challenges based on the data these devices provided, appeared to increase program retention and engagement.
At the end of the VL program, the participants in the VLP + S group had greater weight loss and a greater increase in the number of minutes of moderate-to-vigorous PA than the participants in the VLB or VLP groups, but these differences were not significant. In addition, compared with the VLP group, the VLB group had higher percentages of participants who consumed fast food two or fewer times per week and who had not drunk a sugar-sweetened beverage in the past week. These unexpected results may have been due to the high level of attrition in the VLB group, whose retention rate (43%) was significantly lower than those of the VLP and VLP + S groups (both 73%). Some of the outcomes of the VLB group may have been better than those of the VLP group because the 43% of VLB participants who completed the program may have been more motivated than the other subsets of participants, who received the two other versions of the program.
During the program, the participants in the VLP and VLP + S groups logged more than 8000 steps per day on average, which is in line with a reasonable threshold of free-living PA (7000-8000 steps/day) and with current PA guidelines [44]. The Fitbit-derived numbers of daily steps and active minutes varied weekly; in particular, the average numbers of steps and active minutes were lower from weeks 9 to 11 and from weeks 21 to 22. These periods corresponded to a school holiday at the end of December (Winter Break) and another in mid-March (Spring Break), respectively. Previous studies of worksite interventions that used pedometers also showed that employees' step counts decreased during the winter [45], which may have been due to employees taking fewer steps outside work during the holidays [46]. Given these patterns of PA, future participants in the VL and similar programs should be encouraged to remain physically active and engage in extra activity during holidays such as Spring Break and Winter Break. The VLP and VLP + S groups participated in a "no gain" challenge over the holidays, in which they received prizes if they did not gain weight between Thanksgiving and New Year's, but in future, efforts including physical-activity maintenance may help to boost this effect. Some limitations should be considered when interpreting the results of our study. First, the participants' self-reported PA did not capture increases in light PA or reductions in sedentary behavior, which also burns calories and can aid in weight loss. Second, the measurement of diet was limited by the need to reduce the assessment burden in this worksite-delivered program. The dichotomous variables of dietary intake, which indicated meeting healthy-eating recommendations, did not enable a direct association between diet changes and weight loss among the participants because the items do not capture all intake and do not allow the calculation of energy intake. Third, because all the participants were school-district employees from a specific geographic region (southeast Houston, in Texas) and most were women, it would be difficult to generalize our findings to other populations. Finally, because the study lacked a control arm, we cannot be certain that the effects we observed were caused by the intervention or by other programs or secular trends. However, given that adults in the US typically gain approximately one pound per year [46,47], it seems unlikely that the weight loss would have occurred without intervention. Nevertheless, our study shows that, given the relatively high rates of program retention and satisfaction, particularly in the VLP and VLP + S versions, which had more interactive components, the VL program was generally well received. In addition, its findings provide new evidence to help investigators and practitioners consider the use of effective and feasible worksite weight-loss and behavior-change-intervention programs.

Conclusions
The VL program is a promising approach to facilitating weight loss among schooldistrict employees with overweight and obesity by increasing their PA and healthy eating behaviors. This scalable, light-touch intervention may help employees improve their behaviors, achieve their weight-loss goals, and, consequently reduce their cancer risk. While all the versions of the program had similar effects on weight loss and physical activity among those who completed it, the inclusion of more interactive components (e.g., activity monitors, scales, and challenges) may increase retention.