Promotion of Physical Activity During Weight Loss Maintenance: A Randomized Controlled Trial

Abstract

Objective:

Behavioral weight loss (BWL) programs are not sufficiently effective at promoting high levels of moderate-to-vigorous physical activity (MVPA), despite the clear health benefits of exercise and the possibility that high levels of MVPA may improve long-term weight loss. This three-arm RCT tested the hypotheses that 1) BWL interventions with an intensive focus on exercise would result in higher amounts of MVPA and greater long-term weight loss, compared to standard BWL, and 2) among interventions with an intensive focus on exercise, outcomes would be superior when skills for exercise promotion were taught from an acceptance-based theoretical framework (which fosters willingness to accept discomfort in the service of valued behaviors), versus a traditional behavioral approach.

Methods:

Three hundred and twenty adults with overweight/obesity received group-based BWL for induction of weight loss (Months 1–6) and were randomized to receive one of three interventions for weight loss maintenance (Months 7–18): continued standard behavioral treatment (BT), behavioral treatment with an emphasis on exercise (BT+PA), or acceptance-based treatment with an emphasis on exercise (ABT+PA).

Results:

MVPA and percent weight loss did not significantly differ by condition at 12 or 18 months. Participants engaging in relatively higher levels of MVPA had greater long-term weight losses compared to participants engaging in lower levels of MVPA.

Conclusions:

Further clinical innovations are needed so that participants in BWL programs can more readily adopt and maintain the recommended amounts of MVPA.

TRIAL REGISTRATION:

ClinicalTrials.gov identifier: NCT02363010

Physical activity (PA) promotion is an important aspect of behavioral weight loss (BWL) programs. Independent of any weight loss it produces, PA has the potential to improve many aspects of health, including cardiorespiratory fitness, bone mineral density, blood glucose control and quality of life, and can reduce risk of cardiovascular disease, cancer, type 2 diabetes, and depression (Piercy et al., 2018). Even after a clinically significant weight loss (i.e., 5–10% of initial body weight), many individuals remain in the overweight or obese body mass index (BMI) range and/or regain weight over time, so the potential for PA to buffer against weight-related health risks is key (Wadden, Butryn, & Byrne, 2004).

A large body of non-experimental research has shown that individuals successful at weight loss maintenance typically engage in high levels of moderate-to-vigorous physical activity (MVPA). For example, in the Look AHEAD BWL trial, individuals who maintained a >10% weight loss at 48 months engaged in significantly more MVPA at both 12 and 48 months compared to those with long-term weight loss <10% (Unick et al., 2017). Experimental research examining the effects of PA on weight loss maintenance, on the other hand, is inconclusive. Two clinical trials that randomly assigned participants to PA level found support for the hypothesis that high levels of PA promote weight loss maintenance, although in both trials significant differences between conditions were not seen at long-term follow-up (Fogelholm, Kukkonen-Harjula, Nenonen, & Pasanen, 2000; Jeffery, Wing, Sherwood, & Tate, 2003; Tate, Jeffery, Sherwood, & Wing, 2007). Moreover, four other BWL trials have randomized participants to varying levels or types of PA and failed to observe expected differences between conditions in long-term weight loss (Borg, Kukkonen-Harjula, Fogelholm, & Pasanen, 2002; Jakicic et al., 2008; Leermakers, Perri, Shigaki, & Fuller, 1999; Perri et al., 1988). Notably, in all four studies, the PA prescription was lower than current recommendations for weight loss or weight loss maintenance, which may have reduced the efficacy of the exercise-focused conditions. Two of these studies also relied on self-reported PA, which is less accurate than accelerometer-measured PA, making it challenging to draw conclusions about the efficacy of these interventions for increasing PA (Perri et al., 1988; Sallis & Saelens, 2000). Additionally, adherence to PA prescriptions was low in three of the studies, with participants in PA-focused conditions typically experiencing significant increases in PA compared to non-PA-focused conditions early in the intervention (i.e., at 6-months), but experiencing declines in adherence at later follow-up (Jakicic et al., 2008; Perri et al., 1988; Borg et al., 2002). The poor adherence observed in these experimental studies is consistent with the broader literature, which indicates that a majority of participants in traditional BWL programs do not adopt or maintain prescribed amounts of PA (Avery, Flynn, Van Wersch, Sniehotta, & Trenell, 2012; Catenacci & Wyatt, 2007).

One potential reason for low adherence to PA in BWL is that the amount of treatment content and attention devoted to PA is not optimal. In most standard BWL programs, greater intervention time is focused on maintaining changes in eating behavior (Knowler et al., 2002). Some experimental studies that specifically targeted PA have allotted more intervention time and resources to PA but still may not have provided participants with necessary behavioral strategies for PA promotion and maintenance. For example, in two of the previously referenced studies, PA was promoted through supervised exercise programs, but behavioral strategies for maintaining free-living PA were not taught (Borg et al., 2002; Perri et al., 1988). Other studies have substantially reduced or eliminated treatment content related to dietary change, but with disappointing results (Borg et al., 2002; Leermakers et al., 1999), suggesting that the amount of treatment focused on eating versus PA should be carefully considered. One very simple approach to more effectively promote PA in BWL programs may be to increase the emphasis placed on building behavioral skills for long-term PA engagement, in terms of amount of intervention time and content, while still devoting sufficient attention to other necessary weight control behaviors (e.g., reducing calorie intake). The weight loss maintenance phase is a natural time to increase attention to PA because the added benefit of PA during initial weight loss is modest, and participants may need to devote the majority of their resources to learning skills for dietary change during initial weight loss.

Another approach to increasing PA in the context of BWL is to teach skills that address PA barriers in novel ways. Acceptance-based treatment (ABT) aims to foster a mindful awareness of unpleasant internal experiences (e.g., thoughts, feelings, urges, physical sensations), and a willingness to accept this discomfort in the service of engaging in behaviors that are consistent with long-term goals and values, even when doing so is difficult. Research on the efficacy of ABT for PA promotion is limited, but extant studies show promising results. In the context of brief interventions (e.g., 4 sessions), ABT has produced significant increases in PA (Goodwin, Forman, Herbert, Butryn, & Ledley, 2012; Remmert et al., 2018), and appears to outperform control conditions (Butryn, Forman, Hoffman, Shaw, & Juarascio, 2011; Tapper et al., 2009). Studies are needed that specifically examine the effects of ABT for PA during weight loss maintenance, with larger sample sizes and longer follow-up.

Consistent with efforts towards precision medicine in obesity (Frühbeck, Kiortis, & Catalan, 2018), research also is needed to understand whether particular individuals might benefit from one of these novel approaches to treatment. For example, individuals with the highest BMIs or those who have a sub-optimal response to initial treatment in terms of change in weight or PA might be especially good candidates for an experimental approach. Given that little is understood about sex differences in PA research, sex also should be examined as a moderator.

To address the aforementioned gaps in the literature, the current study provided all participants with 6 months of standard BWL for induction of weight loss (Phase I), and compared participants randomized to one of three interventions during the weight loss maintenance phase that followed (Phase II): continued standard behavioral treatment (BT), behavioral treatment with an emphasis on physical activity (BT+PA), and acceptance based treatment with an emphasis on physical activity (ABT+PA). This study was designed to test the following hypotheses: 1) interventions that had an intensive focus on PA (i.e., BT+PA and ABT+PA) would result in higher amounts of MVPA and greater long-term weight loss (primary outcomes), and greater improvements in cardiorespiratory fitness and waist circumference (secondary outcomes), compared to standard behavioral treatment (BT), and 2) among the interventions that had an intensive focus on PA (BT+PA and ABT+PA), outcomes would be superior when skills for PA promotion were taught from an acceptance-based theoretical framework (i.e., ABT+PA), versus a traditional behavioral approach (i.e., BT+PA). Several potential moderators of the effect of condition on MVPA or long-term weight loss were tested as a secondary aim. Finally, to further understand MVPA dose-response, exploratory analyses determined if weight loss was concurrently or prospectively predicted by 6-, 12-, and 18-month MVPA level, specified as <150 min/week (the minimum amount recommended by national guidelines), 150–249 min/week, or ≥250 mins/week (the amount prescribed in the program).

Method

Design and Randomization

This study was a three-arm, randomized controlled trial. The study was approved and monitored by the Drexel University institutional review board and preregistered at clinicaltrials.gov (Clinical Trials Identifier: NCT02363010). Participants provided informed consent during enrollment, which was conducted by the research coordinator and research assistants. The statistician conducted randomization using a random numbers table. Participants, counselors, and other study staff were blind to allocation, with no access to allocation data during Phase I. During Phase II, three different forms of behavioral treatment were tested and only outcomes assessors remained blinded to condition. An independent officer completed data and safety monitoring.

Participants

Participants were recruited from the community for this study from 2014 to 2016. Inclusion criteria were: BMI 27 – 45 kg/m² (measured in clinic), age 18–70 years, and completion of all steps in the enrollment process. Exclusion criteria were: medical or psychiatric conditions that could pose a risk during lifestyle modification or significantly limit the ability to begin a program of PA; history of bariatric surgery; current use of weight-affecting medication; weight loss of 5% or more in the past 6 months; current pregnancy, lactation, or plans to become pregnant during the study period; participation in or plan to participate in another weight loss program during the study period; having an immediate family member or household member participating in the study. Participants received compensation for assessment visits as follows: $25 for baseline, $25 for month 6, $25 for month 12, and $50 for month 18.

During project planning, estimates of the necessary sample size varied from 300 to 350, depending on attrition. At project launch, a final sample of 320 participants was selected, with the assumption of up to 30% attrition, to provide 81% power to detect a medium effect size of weight loss or MVPA. Power analyses accounted for a three-level hierarchical structure, with repeated measures for each participant in a group within one of the three intervention arms, assuming a within-subject correlation of 0.5 and a within-group correlation of 0.05 (estimated from pilot data). A medium effect size was chosen because 1) an effect of that size may be needed to justify a departure from standard behavioral treatment (McGough & Faraone, 2009), 2) previous research comparing ABT to BT has observed a medium effect size for a primary outcome (Forman et al., 2016), and 3) medium or larger effect sizes have been observed for the comparison of weight loss in those engaging in high versus low MVPA (Tate, Jeffery, Sherwood, & Wing, 2007).

Shared Intervention Components

Phase I (Months 1 to 6) consisted of 16 closed-group sessions, held on a weekly (8 sessions) and then bi-weekly (8 sessions) basis, with approximately 12 participants in each group. During Phase I, all participants received BWL treatment designed to induce 10% weight loss, with materials adapted from Look AHEAD (2006) and the Diabetes Prevention Program (2002). Participants were instructed to keep daily records of dietary intake. Calorie intake was emphasized as the key determinant of weight loss. Stimulus control, problem solving, goal setting, and social support skills were taught. Participants were instructed to gradually self-monitor and increase free-living MVPA, with a goal of maintaining 250 minutes per week of MVPA by 6 months and beyond. Participants were instructed to conduct MVPA in bouts of 10 minutes or more. Participants self-reported weekly MVPA minutes and average calorie intake during each session’s group check-in, and counselors also provided brief written feedback on self-monitoring records on which exercise and dietary intake were reported. Counselors had doctoral-level psychology training. In Phase I and II, all group sessions began with private measurement of weight.

Phase II Interventions (Months 7–18)

Group sessions (14 total) continued in Phase II, beginning with 7 weekly sessions followed by 4 bi-weekly sessions. The final three sessions were held in Months 12, 15, and 18. Each participant also had a 15-minute phone call with a counselor between the quarterly sessions (three calls total) to promote continued engagement. Supplementary Table A details the intervention content in each arm.

Behavioral Therapy (BT)

The BT condition in Phase II continued to be based on Look AHEAD (2006) and Diabetes Prevention Program (2002) materials. Sessions were designed to apply traditional behavioral skills such as problem solving and goal setting to the challenges of long-term lifestyle modification. Approximately two-thirds of intervention content and session time was designed to be applied to eating behavior, with a secondary emphasis on PA.

Behavioral Therapy with Physical Activity Emphasis (BT+PA)

In the BT+PA condition, approximately two-thirds of session time and content was designed to focus on PA, with one-third of session time focused on eating behavior. The intervention was created by adapting material from the Look AHEAD (2006) and Diabetes Prevention Program (2002) protocols to be PA-focused, and by incorporating techniques from Michie’s behavior change taxonomy (Michie et al., 2011). For example, when a session on “maintaining motivation” was conducted, exercises and discussion in session were focused primarily on enhancing motivation for PA. As another example, the application of goal setting skills was primarily focused on PA goals. Progress towards PA goals, particularly the barriers or facilitators of such progress, was reviewed in each session in greater detail than occurred in the BT condition. PA, rather than eating behavior, was the primary target for problem solving skills. Group leaders frequently encouraged development of implementation intentions for PA.

Acceptance-Based Behavioral Therapy with Physical Activity Emphasis (ABT+PA)

The amount of emphasis on promoting PA was designed to be similar in BT+PA and ABT+PA. However, in ABT+PA, acceptance-based behavioral skills were taught, rather than traditional behavioral skills. This approach was adapted primarily from an acceptance-based weight loss protocol (Forman & Butryn, 2016). A key goal was to increase awareness of internal experiences that shape PA behaviors. The approach validated the sense that many aspects of PA are “uncomfortable,” meaning that it can be difficult to tolerate the thoughts (e.g., “I would rather be doing something else”), emotions (e.g., boredom), urges (e.g., to avoid or end exercise), or physical sensations (e.g., sweating) that occur while one attempts to engage in PA or while making PA-related decisions. Participants learned how to respond to internal experiences with a stance of non-judgmental acceptance, which enables flexibility (i.e., the ability to engage in a wide range of behaviors, regardless of the accompanying internal experiences). Ultimately, acceptance was intended to promote long-term persistence in PA. Values clarity, which is integral to the use of acceptance skills, included the ability to consider the ways in which being physically active enables pursuit of what is most important in one’s life (e.g., being physical fit can make travel or community service more feasible). Participants were encouraged to use their “long-term mind” to have a heightened awareness of their values at moments of PA-related decision making, rather than being driven by transient internal experiences.

Measures

Assessments were conducted at Months 0, 6, 12, and 18.

Demographics.

Race, ethnicity, sex, age, and education were self-reported at baseline.

Moderate-to-Vigorous Physical Activity.

Participants were instructed to wear ActiGraph (Pensacola, FL) GT3X tri-axial, solid state accelerometers for all waking hours for 7 consecutive days at each assessment point. ActiLife software was used with cutpoints defined by Troiano et al. (2008) to identify bouts of 10 minutes or more in which PA of moderate-to-vigorous intensity was detected. Data were considered valid and included in data analyses if the participant wore the accelerometer for at least 10 hours each day for 4 or more days.

Weight and Height.

Participants’ weights were measured by research staff with a Tanita® model WB-3000 digital scale, with participants wearing light street clothing. Height was measured with a stadiometer. Measurements were taken two times and averaged.

Half-Mile Walk Time.

After a 30 second warm up period, participants were asked to walk 0.5 miles on a treadmill in the lab. Participants were shown how to adjust the speed of the treadmill and were instructed to walk as briskly as they could, adjusting the speed as they wished throughout the task. Participants were told that the task was a measure of physical fitness. Shorter times indicated a greater level of cardiorespiratory fitness (Mayorga-Vega, Bocanegra-Parrilla, Ornelas, & Viciana, 2016).

Waist Circumference.

Waist circumference was measured horizontally at the umbilicus. Measurements were taken two times and averaged if the measurements were discrepant by less than 2.0 cm (measurements were re-taken if the discrepancy was greater).

Treatment Fidelity.

All treatment sessions were audio-recorded, and 10% of recordings were rated for counselor fidelity to the treatment manual. The treatment manual was organized so that each session included a check-in, two or three topics of discussion or activities designed to build a particular behavior skill, and assignment of skill builders. Each of these sections of the session was rated on a 1–10 scale for adherence, with adherence defined as the extent to which the material was delivered as specified in the manual, including pacing of each topic. BT sessions also were given a rating for the extent to which the counselor avoiding PA contamination (i.e., limited the focus on PA to that specified in the manual) and BT and BT+PA conditions were rated for avoiding ABT contamination.

Data Analysis Plan

Data were analyzed in R (Team, 2013) and SPSS Version 25 (IBM Corp., 2017). Data distributions and assumptions were examined before conducting any formal statistical test. Analyses were conducted on an intent-to-treat basis with multiple imputation via the mice package in R (van Buuren, 2011), such that all participants enrolled in the trial were included in analyses (BT: n = 110, BT-PA: n = 105, ABT-PA: n = 105), with two exceptions noted for exploratory analyses. We performed the analyses for each of the imputed datasets separately, and combined the test statistics across datasets based on Rubin’s rule as implemented in the miceadds package in R (Robitzsch, 2017). To account for the zero-inflated distribution of MVPA, compound Poisson linear regression models (cplm; Zhang, 2013) were used to test the treatment effect on MVPA at 12 and 18 months separately, controlling for MVPA measured at earlier time points. The effect of treatment on percent weight loss at 12 and 18 months were examined separately using general linear models controlling for weight measured at earlier time points. General linear models also were used to examine the treatment effect on half-mile walk time and waist circumference. Moderation analyses were used to examine whether the effect of treatment condition on MVPA or percent weight loss depended on sex, baseline BMI, baseline MVPA, weight loss in Phase I or change in MVPA during Phase I. Additional exploratory analyses using general linear models were performed to further understand the relationship between specific levels of MVPA (regardless of condition) and weight loss outcomes among participants (n = 271) who provided data at 6 months; this subsample was chosen to reduce missing data to allow for greater precision in estimating dose-response for this analysis.

Results

Sample (N = 320) characteristics were as follows: 21.9% of participants identified as male; 3.8% of participants identified as Hispanic or Latino, 70.0% White or Caucasian, 25.0% Black or African American, 2.8% more than one race, 1.6% Asian, and 0.6% American Indian or Alaska Native. Education level was as follows: 5.0% high school graduate or less, 17.8% associate’s or technical degree or partial college, 33.1% bachelor’s degree, 44.1% graduate degree. As shown in Table 1, conditions did not differ by demographic characteristics, treatment dose, counselor fidelity ratings, or retention. Baseline characteristics did not significantly predict 18-month retention status, with the exception of age (B = 0.04, z = 2.19, p = 0.03) and gender (B = −1.07, z = −2.21, p = 0.03), such that younger participants and women were more likely to have missing data. The prediction of retention status by age and gender did not differ by condition. No serious adverse events occurred. See Figure 1 for the CONSORT diagram.

Table 1.

Comparison of Conditions on Baseline Demographic Characteristics and Treatment Factors.

	BT	BT+PA	ABT+PA	p
Female, n (%)	90 (81.8)	78 (74.3)	82 (78.1)	0.41
Age, M (SD)	51.3 (11.4)	54.3 (9.01)	52.7 (10.4)	0.10
African American, n (%)	28 (25.5)	26 (24.8)	26 (24.8)	0.69
Sessions attended (including makeups), M (SD)	24.16 (7.94)	26.33 (6.02)	25.38 (7.28)	0.08
Counselor fidelity, on 1–10 scale, M (SD)	9.48 (0.27)	9.44 (0.67)	9.62 (0.29)	0.72
Retention
6 Months, n (%)	99 (90.0)	96 (91.4)	91 (89.4)	0.54
12 Months, n (%)	74 (67.3)	83 (79.0)	75 (71.4)	0.15
18 Months, n (%)	70 (63.6)	79 (75.2)	75 (71.4)	0.17

Note. p-values are from Chi-Square analyses for categorical variables or ANOVA analyses for continuous variables.

Figure 1.

CONSORT Flow Chart of Participants

Table 2 and Supplementary Figures A and B show change over time in outcomes by condition. As expected, conditions did not differ in MVPA at baseline or 6 months or Phase I weight loss (all p-values > .05). Compound Poisson general linear models found that MVPA was not significantly different by condition at 12 months (BT+PA versus BT: B = 0.12, p = 0.41, d = 0.13; ABT+PA versus BT: B = 0.04, p = 0.46, d = −0.01; ABT+PA versus BT+PA: B = −0.07, p = 0.82, d = −0.14) or 18 months (BT+PA versus BT: B = −0.10, p = 0.42, d = −0.06; ABT+PA versus BT: B = −0.22, p = 0.28, d = −0.20; ABT+PA: versus BT+PA B = −0.07, p = 0.82, d = −0.15), controlling for baseline MVPA. Descriptive analyses were run to characterize adherence. At 12 months, 21.8% of BT participants, 20.0% of BT+PA participants, and 16.0% of ABT+PA participants were meeting the MVPA prescription (≥250 min/week of MVPA). At 18 months, ≥250 min/week of MVPA was observed in 13.2% of BT participants, 13.0% of BT+PA participants, and 7.6% of ABT+PA participants. Results of general linear models showed no significant group differences in weight loss at 12 months (F(2, 390) = 0.39, p = 0.68, η_p² = 0.004) or 18 months (F(2, 318) = 2.20, p = 0.11, η_p²= 0.019), controlling for baseline weight. There were no significant group differences in the time to walk a half-mile at 12 months (F(2, 60) = 2.94, p = 0.06, η_p² = 0.03) or 18 months (F(2, 35) = 1.23, p = 0.30, η_p² = 0.02), controlling for baseline half-mile walk time. Finally, waist circumference did not significantly differ by condition at 12 months (F(2, 402) = 0.53, p = 0.59, η_p² = 0.005) or 18 months (F(2, 264) = 0.56, p = 0.57, η_p² = 0.006), controlling for baseline waist circumference.

Table 2.

Comparison of Conditions over 18-month Intervention.

	BT M(SD)	BT+PA M(SD)	ABT+PA M(SD)
BMI (kg/m2)
Baseline	35.1 (4.3)	34.5 (4.6)	35.8 (5.3)
% Weight loss
Baseline	N/A	N/A	N/A
6 Months	10.5 (5.6)	10.2 (5.0)	9.9 (4.7)
12 Months	12.3 (8.9)	12.8 (8.2)	11.6 (7.8)
18 Months	10.2 (11.9)	12.4 (10.1)	9.1 (10.3)
MVPA (min/week)
Baseline	64.2 (91.9)	64.2 (78.6)	48.6 (73.6)
6 Months	133.9 (124.5)	143.0 (121.0)	128.8 (115.7)
12 Months	129.1 (117.5)	144.6 (124.9)	127.3 (126.9)
18 Months	110.0 (134.6)	102.5 (124.0)	85.9 (101.7)
Half-mile walk time (sec)
Baseline	537.1 (98.2)	576.6 (157.3)	562.2 (101.8)
6 Months	503.9 (80.8)	528.2 (103.7)	514.1 (73.1)
12 Months	487.7 (74.7)	512.2 (96.5)	500.5 (77.5)
18 Months	498.0 (80.9)	518.4 (107.6)	517.0 (103.7)
Waist circumference (cm)
Baseline	42.6 (5.5)	42.6 (4.7)	41.8 (4.5)
6 Months	39.0 (5.1)	39.7 (4.8)	38.9 (4.3)
12 Months	38.6 (5.2)	38.8 (4.9)	38.2 (4.6)
18 Months	39.2 (5.8)	38.9 (5.3)	38.5 (4.7)

Moderation analyses were conducted to determine if the effect of condition on the two primary outcomes (MVPA or percent weight loss) depended on other variables. As shown in Table 3, the effect of condition on 18-month MVPA did not depend on sex, baseline BMI, baseline MVPA, weight loss during Phase I, or change in MVPA during Phase I. The effect of condition on 18-month percent weight loss did not depend on sex, weight loss during Phase I, or change in MVPA during Phase I. Baseline BMI was a significant moderator of 18-month percent weight loss. To understand this pattern of results, conditions were compared within each tertile of baseline BMI (Supplementary Figure C). Among participants in the upper tertile of baseline BMI (> 37 kg/m²), those randomized to BT+PA had significantly greater 18-month weight losses than those randomized to BT (p = .03).

Table 3.

Moderation of the Effect of Condition on Outcomes at 18 Months

Outcome: MVPA at 18 months	ABT+PA versus BT			BT+PA versus BT			ABT+PA versus BT+PA
Moderator	B	95% CI	p	B	95% CI	p	B	95% CI	p
Sex	−0.24	−1.09, 0.61	0.66	−0.37	−1.17, 0.43	0.46	0.17	−0.63, 0.98	0.84
Baseline BMI	0.04	−0.04, 0.11	0.46	0.06	−0.02, 0.13	0.16	−0.03	−0.10, 0.04	0.54
Baseline MVPA	0.002	−0.002, 0.006	0.42	0.003	−0.001, 0.006	0.24	−0.001	−0.005, 0.003	0.69
Weight loss during Phase I	0.02	−0.01, 0.04	0.35	0.01	−0.20, 0.03	0.62	0.01	−0.02, 0.04	0.56
Change in MVPA during Phase I	0.05	−0.24, 0.34	0.87	−0.08	−0.40, 0.23	0.89	0.11	−0.24, 0.46	0.84
Outcome: percent weight loss at 18 months	ABT+PA versus BT			BT+PA versus BT			ABT+PA versus BT+PA
Moderator	B	95% CI	p	B	95% CI	p	B	95% CI	p
Sex	−0.54	−7.94, 6.85	0.98	−1.60	−8.74, 5.54	0.78	1.34	−5.52, 8.19	0.78
Baseline BMI	0.69	−0.09, 1.46	0.08	0.88	0.09, 1.67	0.02	−0.22	−0.88, 0.44	0.51
Weight loss during Phase I	−0.06	−0.03, 0.19	0.70	−0.05	−0.28, 0.17	0.80	<−0.01	−0.27, 0.26	0.98
Baseline MVPA	−0.01	−0.04, 0.03	0.61	0.01	−0.03, 0.04	0.67	−0.02	−0.05, 0.01	0.30
Change in MVPA during Phase I	−0.02	−0.05, 0.02	0.29	0.01	−0.02, 0.03	0.60	−0.03	−0.06, 0.01	0.15

Exploratory analyses also were conducted to examine 18-month outcomes only among the participants who achieved a clinically significant weight loss at 6 months (i.e., ≥5% weight loss, which was achieved at 6 months by 84.1% of BT, 84% of BT+PA, and 84.3% of ABT+PA). No differences were found in 18 month MVPA (BT+PA versus BT: B = −0.20, p = 0.32, d = −0.13; ABT+PA versus BT: B = −0.16, p = 0.45, d = −0.16; ABT+PA versus BT+PA: B = 0.07, p = 0.67, d = −0.02) or weight (p = 0.33, η_p² = 0.015).

Exploratory analyses found concurrent and predictive relationships between MVPA and weight loss, as shown in Table 4. For example, participants engaging in >250 min/week of MVPA at 6 months had significantly greater weight loss at 12 months, compared to those engaging in <150 min/week of MVPA at 6 months (14.5% vs. 10.9%). Participants engaging in ≥250 min/week of MVPA at 18 months had significantly greater weight loss at 18 months, compared to those engaging in <150 min/week of MVPA at 18 months (16.0% vs. 9.3%).

Table 4.

Percent Weight Loss at 6, 12 and 18 Months Predicted by Category of MVPA at 6, 12, and 18 Months

	6-month MVPA				12-month MVPA				18-month MVPA
	Low (n = 164)	Medium (n = 58)	High (n = 49)	Comparisons	Low (n = 162)	Medium (n = 55)	High (n = 54)	Comparisons	Low (n = 203)	Medium (n = 38)	High (n = 30)	Comparisons
Percent weight loss Month 6	9.26%	11.76%	11.73%	H versus L B = 2.47, p = .002; M versus L B = 2.50, p = .001; H versus M B = −0.03, p = 0.97
Percent weight loss Month 12	10.92%	14.14%	14.49%	H versus L B = 3.57, p = .009; M versus L B = 3.23, p = .016; H versus M B = 0.34, p = 0.83	10.61%	14.27%	15.21%	H versus L B = 4.61, p = .002; M versus L B = 3.66, p = 01; H versus M B = 0.95, p = 0.63
Percent weight loss Month 18	9.71%	12.57%	11.38%	H versus L B = 1.67, p = 0.39; M versus L B = 2.85, p = 0.11; H versus M B = −1.19, p = 0.60	9.44%	11.95%	12.91%	H versus L B = 3.47, p = .09; M versus L B = 2.51, p = 0.22; H versus M B = 0.96, p = 0.72	9.34%	12.62%	16.04%	H versus L B = 6.70, p = .03; M versus L B = 3.28, p = 0.23 H versus M B = 3.42, p = 0.42

Note. These exploratory analyses were conducted among participants (n = 271) providing data at 6 months, to reduce amount of missing data and increase precision of dose-response estimates. Low (L) MVPA is <150 mins/week. Medium (M) MVPA is 150–249 mins/week. High (H) MVPA is ≥250 mins/week.

Discussion

Most BWL programs are not sufficiently effective at promoting high levels of PA, despite the clear health benefits of exercise and the possibility that high levels of PA may improve long-term weight loss. This study tested new ways of promoting PA in BWL. Results indicated that weight loss maintenance interventions that had an intensive focus on PA (i.e., the BT+PA and ABT+PA conditions) did not result in higher amounts of MVPA or greater percent weight loss at 12 or 18 months, compared to traditional intervention (i.e., BT). Among the interventions that had an intensive focus on exercise, MVPA and percent weight loss at 12 and 18 months were similar whether skills for PA promotion were taught from an acceptance-based theoretical framework (i.e., ABT+PA) or traditional behavioral orientation (i.e., BT+PA). Secondary analyses found that the effect of condition on 18-month weight loss depended on baseline BMI, with better weight loss in BT+PA versus BT among patients with higher baseline BMIs. Across conditions, participants engaging in higher levels of MVPA had greater weight losses compared to participants engaging in lower levels of MVPA.

The average increases from baseline that were observed in MVPA were large, but still many participants were not reaching the level of MVPA typically recommended for health benefits (150 min/week), and even fewer participants engaged in the prescribed level of MVPA that is thought to be necessary for long-term prevention of weight regain (250 min/week). The findings from this study are consistent with four others in which experimental interventions designed to increase PA and thus improve weight loss maintenance failed to do so (Borg et al., 2002; Jakicic et al., 2008; Leermakers et al., 1999; Perri et al., 1988) and further illustrate the challenge of adoption and maintenance of PA among individuals receiving BWL intervention. Our findings of poor adherence to prescribed PA are consistent with the BWL literature. For example, Tate et al. (2007) found that only 11.9% of participants assigned to complete a high level of PA (2500 kcal/wk energy expenditure) were consistently doing so at 12, 18, and 30-month follow-up assessment.

In the present study, several explanations for the similarity of outcomes observed in the comparison and experimental conditions can be considered. Because the experimental interventions in this study apparently were not successful at improving MVPA outcomes, it is not surprising that long-term weight losses were similar across conditions. However, it also is possible that differences in MVPA occurred but were not detected, as accelerometers sampled only a brief period of behavior (i.e., 1 week), and noise could be introduced by aberrations in factors such as weather, illness, or stress. Similarly, the treadmill-based half-mile walk test was chosen as a measure of cardiorespiratory fitness because it has higher feasibility than alternative assessment methods, but this method is inferior to direct measurement of maximum oxygen uptake, and thus it is possible that differences in fitness occurred but were not detected.

The theoretical approach in either experimental condition may not have been a sufficiently strong fit for PA promotion (i.e., perhaps the factors that must be targeted to increase PA are not fully understood). Alternatively, it is possible that that the approach used in either experimental condition has promise, but that intensity, dose, or timing of the BT+PA or ABT+PA treatments was not optimal. Perhaps the treatments did not differ from BT substantially enough. Undetected problems with treatment delivery also may have occurred, as fidelity was assessed in only 10% of sessions.

Forman et al. (2016) and Lillis et al. (2016) previously have shown that acceptance-based treatment outperformed behavioral treatment in weight loss outcomes when delivered with the traditional, primary focus on dietary change (Forman et al., 2016). Focusing acceptance-based treatment on PA appears to have eliminated that weight loss advantage, perhaps because the dose of ABT in the current study was inadequate or the application of acceptance-based skills to eating behavior (versus PA) was insufficient. This study was designed such that the three arms differed in the amount of intervention time addressing eating behavior. This may have contributed to the null results; perhaps one or both experimental arms did increase PA, but the benefits of this were offset by poorer eating behavior. This study did not measure dietary intake, and thus it is unknown to what extent the reductions in eating-focused intervention time may have been problematic.

It is notable that across conditions, percent weight loss averaged 10.6% at 18 months. Participants were, on average, successful at maintaining a clinically significant weight loss, regardless of treatment condition, and the opportunity for an experimental intervention to prevent substantial regain was limited. Participants may have been highly motivated to maintain their initial weight loss or learned sufficiently effective weight management skills in all conditions.

Power to detect small effects was limited, and it is possible that with a larger sample some observed differences would have reached statistical significance. Attrition was high as a result of participants being lost to follow-up (perhaps in part due to the burden of the assessment, which included wearing an accelerometer and completing the treadmill task), and missing data limit confidence in statistical tests. Outcomes may have differed among participants who did versus did not provide assessment data. The study was designed to allow all participants to remain in the study in Phase II (with the rationale that even those with poor initial weight loss might benefit from the Phase II experimental treatments), and randomization occurred at baseline with Phase I blinding so that participants would not need to be assigned to new groups in Phase II, which was thought to be a risk for attrition and to potentially undermine the power of the therapeutic and peer relationships formed in Phase I. Retention may have been better and variability in outcomes reduced if randomization occurred at the end of Phase I, and only for participants with clinically significant Phase I weight loss.

Analyses that collapsed across conditions found higher PA levels were correlated with and predictive of better weight loss maintenance, which is consistent with other research. For example, Jakicic, Marcus, Lang and Janney (2008) found that individuals who successfully maintained a 10% weight loss at 24 months reported higher levels of PA than those with smaller weight losses, despite no observed differences by assigned PA condition. It remains unclear to what extent MVPA has a direct, causal effect on long-term weight outcomes, as participants who are most compliant with dietary prescriptions also may be those who are most compliant with PA. Additional experimental research using supervised PA may be needed to provide greater clarity about the causal effects of PA on long-term weight loss.

Exploratory analyses found that participants with a higher BMI achieved greater long-term weight loss in BT+PA versus in BT. Participants with the highest BMIs may experience particular challenges in PA engagement related to their weight, such as physical discomfort or shame while exercising, or more generally might have a lower level of PA enjoyment (which may have contributed to their higher level of weight gain over time) (Ball, Crawford, & Owen, 2000; Mensinger & Meadows, 2017). BT+PA may have been more effective than standard intervention at helping participants learn new ways of responding to these challenges. However, BMI did not significantly moderate the effect of condition on MVPA. It is possible that measurement error in MVPA may have occurred; little rigorous research has compared the validity of hip-worn tri-axial accelerometry among the various classes of obesity, and in theory greater adipose tissue in the hip area could reduce measurement validity. It also is possible that participants with the highest BMIs applied the skills taught in the BT+PA to changes in their eating behavior in a way that promoted greater long-term weight loss. Finally, the number of moderator analyses conducted was high, raising the risk of Type I error.

This study had important limitations. As noted, attrition was high, power to detect small effects was inadequate, dietary intake was not measured, measurements of physical activity outcomes had weaknesses, and conditions differed in the extent to which eating behavior was addressed. In addition, representation of men in the sample was poor, leaving open the possibility that different results may have been observed (including for testing of sex as a moderator) if men and women had been equally represented in the sample. The majority of the sample identified as either non-Hispanic White or Black; Hispanic individuals and those of other racial groups were underrepresented, which limits generalizability of results. Representation of young adults, older adults, and those with low levels of education also was limited. The hypotheses in this study were tested in the context of a high-intensity BWL program in an academic research clinic. Interventions tested in lower-intensity, remote, or community-based formats may yield different effects.

In conclusion, this study, consistent with other research, showed that individuals who were able to adopt and maintain high levels of MVPA achieved the greatest long-term weight loss. However, the interventions that had a primary focus on PA promotion after initial weight loss were no more effective at promoting PA than standard behavioral treatment, and weight loss outcomes at end of treatment did not differ by treatment arm. These findings underscore the need for theoretical and clinical innovations in the strategies used for PA promotion in BWL programs. Such programs may need to identify more creative ways to enhance motivation for PA, target PA-related decision-making, or address barriers to PA; determine how to best use technology (including just-in-time prompts or activity sensors); or build community or workplace partnerships that can support the integration of PA into daily routines and environments. Continued research also may be needed to refine the optimal prescription for type, duration, frequency, and intensity of PA. Until BWL programs can more effectively facilitate adoption and maintenance of high levels of MVPA, many participants will fail to fully experience the benefits of exercise.