Mindfulness-Based Group Intervention in Adolescents At-risk for Excess Weight Gain: A Randomized Controlled Pilot Study

Abstract

Objectives:

To assess feasibility/acceptability of a mindfulness-based approach to excess weight prevention in adolescents at-risk for excess weight gain. To pilot test efficacy of a mindfulness-based intervention for improving food reward sensitivity, stress-eating, executive function (EF), and BMI/adiposity.

Methods:

A pilot randomized controlled trial was conducted with 12-17y adolescents at-risk for excess weight gain based on above-average weight (body mass index [BMI]≥70%ile) or parental history of obesity (BMI≥30kg/m²). Adolescents were randomized to a mindfulness-based (n=29) or health education control group (n=25) that met for six weekly one-hour sessions. Feasibility/acceptability were determined from attendance and acceptability survey ratings. At baseline, six-week and six-month follow-up, adolescents’ perceived stress was measured with the Perceived Stress Scale, food reward sensitivity with a behavioral task, stress-eating during a laboratory test meal, and EF with the parent-reported Behavior Rating Inventory of Executive Function and NIH Toolbox. At the same intervals, BMI indices and body fat by air displacement plethysmography were assessed in a fasted state.

Results:

Median session attendance was 6:6 sessions in both conditions; program acceptability ratings were above-average. Compared to health education, adolescents in mindfulness had lower food reward sensitivity at six-months (Cohen’s d=.64, p=.01). There were no between-condition differences in BMI (mindfulness vs. health education_Δsix-months 95%CI 0.20, 1.52 kg/m² vs. 0.21, 1.62 kg/m²) or adiposity (−3.64, −0.61% vs. −4.31, −1.04%) changes.

Conclusions:

A mindfulness-based group intervention is feasible/acceptable among adolescents at-risk for excess weight. In this pilot sample, mindfulness and health education were equivocal for BMI/adiposity outcomes. Future trials with a larger, adequately-powered sample and longer-term follow-up are necessary to test efficacy of a mindfulness-based intervention for food reward sensitivity, stress-eating, EF, and stabilizing growth trajectories in youth at-risk for adult obesity.

Health-related behavioral patterns established during adolescence influence body mass index (BMI) and metabolic trajectories into adulthood (Alberga, Sigal, Goldfield, Prud'homme, & Kenny, 2012; Bjerregaard, et al., 2018; Simmonds, Llewellyn, Owen, & Woolacott, 2016). Standard lifestyle-based programs including a combination of diet and exercise to prevent excess weight gain have demonstrated insufficient effectiveness in adolescents (Kobes, Kretschmer, Timmerman, & Schreuder, 2018). Adolescence is a critical period for excess weight gain, referring to gaining too much weight for one’s health over time (Todd, Street, Ziviani, Byrne, & Hills, 2015). Alternative approaches targeting malleable developmental characteristics of adolescence, particularly reward sensitivity and executive function (EF), may more effectively mitigate excess weight gain (Alberga, et al., 2012).

Neurobiological models of obesity propose that heightened sensitivity to palatable food as a reward, in combination with an insufficiency in the top-down control necessary for regulating reward-driven eating, predispose individuals to excess weight gain (e.g., Volkow, Wang, Tomasi, & Baler, 2013). Adolescence is a hallmark developmental period for these dynamic, interacting processes (Chambers, Taylor, & Potenza, 2003). One key feature of adolescence is the maturation of EF (Diamond, 2013). EF refers to neurocognitive processes that govern goal-directed control over cognition, behavior, and emotion (Diamond, 2013). Key indices of EF include attention, inhibitory control, and working memory (Diamond, 2013). EF shows age-related improvements during the adolescent period and also can be enhanced with training (Diamond, 2013). Moreover, evidence from cross-sectional and longitudinal studies suggests that executive dysfunction is associated with excess weight gain, likely by increasing behaviors that negatively affect energy balance, such as overeating (Goldschmidt, Hipwell, Stepp, McTigue, & Keenan, 2015; Reinert, Po'e, & Barkin, 2013). The relative immaturity of EF in adolescence is coupled with heightened sensitivity to reward (Van Leijenhorst, et al., 2010). Simultaneously, adolescents are gaining autonomy from parents, including with respect to food choices (Alberga, et al., 2012; Birch & Fisher, 1998), during a developmental period that is also associated with heightened perceived stress (Felton, et al., 2017). Increased reliance on food as a reward and the manifestation, or worsening, of eating in response to stress (“stress-eating”) promote excess weight gain in adolescents (Ibrahim, Thearle, Krakoff, & Gluck, 2016; Stice & Yokum, 2016)—highlighting these factors as important potential mechanisms linking food reward sensitivity, stress-eating, and excess weight gain. Consequently, intervening to lessen food reward sensitivity and improve EF in adolescents at-risk for excess weight gain may supply them with more effective methods of self-regulation and in turn, provide a prevention strategy for excess weight gain with developmentally-sensitive timing.

Mindfulness-based interventions may be poised to alter reward sensitivity and improve EF, and as a result, prevent excess weight gain through better self-regulation, particularly over eating (Garland, 2016; Rogers, Ferrari, Mosely, Lang, & Brennan, 2017). Mindfulness is a multifaceted construct that centers on cultivating attentiveness to present-moment experience, on purpose, without judgment (Kabat-Zinn, 1991). Mindfulness-based interventions include a variety of activities (e.g., psychoeducation, guided meditation, interactive/experiential activities, and mindful movement/yoga) geared toward increasing present-moment attention to thoughts, emotions, and bodily sensations (Broderick, 2013; Kabat-Zinn, 1991). Reviews of preliminary data from randomized controlled trials in adults with obesity suggest that mindfulness-based interventions reduce emotional distress, decrease stress-eating, prevent continued excess weight gain, and/or result in weight loss (Dunn, et al., 2018; Rogers, et al., 2017). Although the explanatory mechanisms have not been elucidated, mindfulness-based training may positively alter stress-eating and weight outcomes through altered reward sensitivity (Garland, 2016) and/or improved EF (Allen, et al., 2012; Hendrickson & Rasmussen, 2013, 2017; Leyland, Rowse, & Emerson, 2018).

Few studies have evaluated mindfulness-based interventions for eating or excess weight gain in adolescents, and available data have illustrated mixed results for feasibility and acceptability (Barnes & Orme-Johnson, 2012; Daly, Pace, Berg, Menon, & Szalacha, 2016; Shomaker, et al., 2017). In a randomized pilot study of adolescents with overweight/obesity, program retention during a mindfulness-based intervention focused upon mindful eating was relatively weak (57%) (Daly, et al., 2016); among program completers, those who received mindfulness-training had lower BMI at post-intervention and one-month follow-up than adolescents who received nutrition/exercise knowledge (Daly, et al., 2016). Other studies have shown better retention and acceptability, as well as decreased insulin resistance or blood pressure compared to a control intervention, with mindfulness-based interventions delivered to adolescents at-risk for cardiometabolic disease (Barnes & Orme-Johnson, 2012; Shomaker, et al., 2017). In adolescent samples not selected specifically for overeating or excess weight gain problems, mindfulness-based interventions have demonstrated feasibility, acceptability, and positive preliminary effects on EF (Tang, Yang, Leve, & Harold, 2012).

The first objective of the current study was to assess the feasibility and acceptability of Learning to BREATHE—a six-week, six-session adolescent mindfulness-based group (Broderick, 2013)—among adolescents at-risk for excess weight gain. The hypothesis was that there would be evidence of feasibility and acceptability based upon ≥80% attendance (≥5:6 sessions) and above-average program ratings. The second objective was to explore the effects of this mindfulness intervention, compared to a time-matched health education control group providing knowledge about healthy living, on changes in adolescents’ perceived stress, food reward sensitivity, stress-eating, EF, BMI, and adiposity at six-week (immediate post-treatment) and six-month follow-up. Due to the pilot nature of the study, the hypothesis was that there would be patterns of decreased perceived stress, less food reward sensitivity, less stress-eating, improved EF, less excess weight gain, and less increases in body fat in adolescents randomized to mindfulness versus health education.

Methods

Participants

Volunteers were recruited through letters to Northern Colorado-area families, flyers in schools and physician offices, informational community sessions, newspaper/radio advertisements, and e-mails to community list serves. Materials invited adolescents who may be at-risk for gaining too much weight to participate in a group designed to decrease stress and promote healthy growth. Similar to previous behavioral clinical trials for the prevention of excess weight gain (e.g., Tanofsky-Kraff, et al., 2017), girls and boys were eligible to participate if they were: (i) 12-17y, (ii) BMI ≥70^th percentile for age/sex or had two biological parents with reported obesity (BMI ≥30 kg/m²), (iii) in good general health, based upon a health history conducted with a parent and including no major medical problems (e.g., type 1 or 2 diabetes, musculoskeletal problems), (iv) free of psychiatric symptoms that would impede compliance and necessitate treatment (e.g., suicidal behavior), (v) not taking medication affecting weight, mood, and/or eating (e.g., antidepressants, insulin sensitizers, stimulants), (vi) not involved in current weight loss treatment or psychotherapy, and (vii) not pregnant. Females taking oral contraceptives were included. Selection criteria for this pilot study were intended a priori to recruit generally healthy adolescents at-risk for excess weight gain. Children and adolescents whose BMI exceeds the 50^th percentile have been found to be 5 times more likely than youth with a BMI<50^th percentile to develop overweight in young adulthood, with more increased risk among youth whose BMI exceeds the 75^th percentile (Field, Cook, & Gillman, 2005). Likewise, parental obesity confers greater risk of obesity in adolescent offspring (Whitaker, Wright, Pepe, Seidel, & Dietz, 1997).

Procedures

The current study was a randomized controlled pilot and feasibility study with interventions conducted in parallel ( NCT03085160). Procedures took place at an outpatient, pediatric research laboratory at Colorado State University. Recruitment started in October 2014 and ended in May 2015. The Institutional Review Board of Colorado State University approved all procedures. Following a phone screen to estimate eligibility, participants attended a screening appointment to determine eligibility and collect baseline assessments. Parents/guardians and adolescents provided written consent and assent, respectively, after having the study described to them in detail. Eligible participants were randomized to a mindfulness-based group or health education control group, which both met for six one-hour sessions, once per week (Table 1). Randomization was generated by an electronic program (randomization.com) with permuted blocks and stratified by sex, race/ethnicity, age, and weight. After determining eligibility, the study coordinator assigned participants to interventions. From May 2015 to March 2017, five cohorts were run in parallel on separate days during non-school hours. Following the intervention phase, adolescents returned for two follow-up visits, one six-weeks (immediate post-treatment) and the other six-months after the intervention. Follow-ups took place between July 2015 and November 2017. Evaluators of follow-up assessments were not consistently blinded to intervention allocation. Adolescents were compensated for their time.

Table 1.

Key content of the mindfulness-based group and health education control group sessions

Week	Mindfulness Group	Health Education Group
1	Body: Psychoeducation on mindfulness; mindful listening and mindful eating activities; breath awareness practice; body scan	Didactic information on domestic violence
2	Reflections: Psychoeducation and activities on thoughts and their connection to feelings; mindfulness of thoughts practice	Didactic information on alcohol, drugs, and tobacco
3	Emotions: Psychoeducation and activities on emotion and their connection to the body; mindfulness of emotions practice	Didactic information on nutrition and body image
4	Attention: Psychoeducation and activities on stress; mindful movement practice	Didactic information on depression and suicide
5	Tenderness: Psychoeducation and activities on self-compassion and selfcare; loving-kindness mindfulness practice	Didactic information on gang violence and non-violent conflict resolution
6	Habits: Activities on continuation and maintenance of mindfulness practices in daily living; mindful writing, listening, and speaking activities Empowerment: Gain an inner edge (overall program goal)	Didactic information on sun safety

Note: Each week for six weeks, the groups met for one 1-hour session. The mindfulness-based curriculum was derived from Learning to BREATHE (Broderick, 2013).

Interventions

Mindfulness-based group.

Learning to BREATHE is a curriculum derived from mindfulness-based stress reduction (Kabat-Zinn, 1991) and adapted for adolescents with experiential activities and guided discussions to teach standard mindfulness skills (Broderick, 2013). Examples include breath awareness, body scanning, mindful eating, sitting meditation, loving-kindness practice, and gentle yoga. The original curriculum was designed to offer flexibility to facilitators in delivery timing and selection of exercises. A manualized version of BREATHE was used for consistency in timing and content, but the content was minimally modified from its original format. For instance, in session 1, a brief justification (~1 minute) of how program participation may help adolescents to maintain a healthy weight over time was added. The amount of intervention time spent on eating was not increased from the standard program. Brief (~10 minutes/day) homework was assigned for practicing skills in daily life. Adolescents were given meditation audio-recordings, a yoga mat, a meditation cushion, homework log, and worksheets. They reported homework completion at sessions 2-6 to facilitators. The intervention was co-facilitated by Master’s graduate students in Marriage and Family Therapy who attended a workshop with the developer and reviewed/practiced material with the lead investigator, a licensed clinical psychologist. To ensure intervention fidelity and leader competence, facilitators received weekly clinical supervision on audio-recorded sessions from a licensed clinical psychologist with expertise in mindfulness-based interventions with adolescents.

Health education.

Health education was drawn from a didactic program, “Hey Durham” (Bravender, 2005), as a control condition matched for instruction time and designed to parallel health knowledge presented in a middle/high school health class. Group sessions incorporated handouts, videos, and presentations. As in our previous trials (Shomaker, et al., 2017; Shomaker, et al., 2016), sessions covered six topics: alcohol/drug use, nutrition/body image, domestic violence, gang violence/non-violent conflict resolution, sun safety, and major depression/signs of suicide. All topics were didactic, and no direct counseling was provided. The section on nutrition/body image provided basic information on healthy eating and unhealthy eating (e.g., extreme dieting). This segment did not overlap with the content on mindful eating in the Learning to BREATHE intervention. Health education was co-facilitated by Master’s graduate students in Food Science and Human Nutrition, Public Health, and Prevention Science.

Program acceptability.

For the primary outcome of feasibility and acceptability, total session attendance was described. Also, at six-week follow-up, participants completed an adapted program acceptability questionnaire (Hunsley, 1992). This 8-item survey assessed adolescents’ perceptions of overall program participation, facilitators, and intervention sessions.

Assessments

Assessments of secondary outcomes were collected at three visits: baseline, six-week follow-up, and six-month follow-up. Measurements were typically collected in the order of fasting measures of BMI/body composition, followed by a standardized breakfast shake at approximately 9:00 am, behavioral tasks including food reward sensitivity, surveys/interviews, and the stress-eating laboratory test meal procedure, with the meal served at approximately 12:00 pm. At the outset of each visit, a research staff member confirmed that the participant had nothing to eat or drink other than water after 10:00 pm the night prior. The following key constructs were assessed:

Perceived stress.

The widely-used and psychometrically-sound 10-item Perceived Stress Scale was administered to measure adolescents’ perceptions of stress (Cohen & Williamson, 1988; Kupst, et al., 2015). The total score is calculated as the sum of all items, rated on a 5-point Likert scale from 0=Never to 4=Very Often, with higher values reflecting greater stress (Range: 0-40). Consistent with previous reports of this measure’s use in adolescent samples (Carlozzi, et al., 2010; Quach, Jastrowski Mano, & Alexander, 2016), internal reliability in the current sample was good (α=.85).

Food reward sensitivity.

A behavioral task was administered to determine the relative reinforcing value or reward sensitivity of a preferred palatable snack food as compared to an alternate reward (Epstein, Leddy, Temple, & Faith, 2007; Goldfield, Epstein, Davidson, & Saad, 2005). The task was administered following a standardized breakfast shake comprised of 33% of each adolescent’s estimated daily energy needs and administered to reduce hunger. In this task, the adolescent is asked a series of questions about whether they would prefer to work for a preferred, highly-palatable snack food (e.g., chips, cookies, or candy of their choice) or money (a quarter). The “work” required to obtain the reward is clicking a handheld tally counter a prescribed number of times. The first question begins with an equal amount of work to obtain either reinforcer (20 presses). With each subsequent item, the work to obtain the food reward increases on a fixed ratio progressive reinforcement schedule, while the work to obtain the monetary reward is held constant. The work and the rewards are hypothetical. Adolescents are shown the food, money, and counter and instructed to use the counter to get a feel for the work, but when the assessment begins, they are asked to imagine the work and no rewards are administered to the participant. The outcome is the switch point—when the participant switches from choosing the snack to the money on two consecutive turns. Higher values represent greater food reward sensitivity. Youth’s performance on this task has been related to weight gain (Hill, Saxton, Webber, Blundell, & Wardle, 2009) and poorer response to behavioral weight loss treatment (Best, et al., 2012).

Stress-eating.

Total energy intake was measured from a multi-item lunch food array served directly after the Trier Social Stress Test (TSST), a laboratory stressor that has been validated to evoke a physiological and subjective stress response (Gunnar, Talge, & Herrera, 2009; Yim, Quas, Cahill, & Hayakawa, 2010). A TSST-version modified for adolescents was used (Yim, et al., 2010). Following rest (~20-minutes), adolescents were asked to prepare and deliver a speech in front of an evaluator and video-camera, and subsequently to complete a challenging mental arithmetic task. Evaluators were trained to remain neutral rather than supportive. Directly following the TSST, adolescents were invited to consume ad libitum from a ~4,500-kcal array of chicken nuggets, white bread, barbeque sauce, peanut butter, grape jelly, chocolate candies, cookies, milk, apple juice, and water. On average, adolescents reported moderate hunger prior to eating (M=4.7, SD=3.0) on a visual analog scale ranging from 0=Not at all hungry to 10=Extremely hungry). Participants were instructed to eat as they would at a normal meal and to open the door when finished. Foods were weighed on an electronic scale before and after eating, and weight (g) differences before and after eating were calculated. Energy intake was determined from data available from food manufacturers. Similar paradigms to study children’s and adolescents’ stress-eating in the laboratory have demonstrated associations of energy intake with psychological characteristics and weight (Tanofsky-Kraff, Haynos, Kotler, Yanovski, & Yanovski, 2007).

Executive function (EF).

A parent/guardian completed the reliable and validated Behavior Rating Inventory of Executive Function (BRIEF), measuring everyday behavior problems associated with key domains of EF (Gioia, Isquith, Guy, Kenworthy, & Baron, 2000). Eight scales are computed as the sum of their respective items, with responses on a 3-point Likert scale from 0=Never to 2=Often: (i) inhibit (10-items)—inhibitory control and impulsivity (Range: 0-20; α=.87 in the current sample), (ii) shift (8-items)—ability to make transitions and problem-solve flexibly (Range: 0-16; α=.82), (iii) emotional control (10-items)—self-regulation of emotions (Range: 0-20; α=.93), (iv) initiate (8-items)—ability to begin a task and problem-solve independently (Range: 0-16; α=.80), (v) working memory (10-items)—ability to hold information online to complete a task/accomplish a goal (Range: 0-20; α=.91), (vi) plan/organize (12-items)—ability to manage current/future task demands (Range: 0-24; α=.91), (vii) organization of materials (6-items)—orderliness of materials (Range: 0-12; α=.87), and (viii) monitor (8-items)—interpersonal awareness and work-checking habits (Range: 0-16; α=.80). Higher scores reflect more problematic executive dysfunction (Gioia, et al., 2000). The BRIEF has demonstrated high internal reliability, test-retest reliability, and convergent/discriminant, concurrent, and predictive validity (Gioia, Kenworthy, & Isquith, 2010).

In addition, the National Institutes of Health (NIH) Toolbox Flanker Inhibitory Control and Attention Test and the NIH Toolbox List Sorting Working Memory Test were administered to adolescents as behavioral measures of EF (Weintraub, Dikmen, et al., 2013). In the Flanker task, individuals press a button to indicate the direction of a middle stimulus (arrow) flanked by either congruent or incongruent stimuli. In the List Sorting task, participants must immediately recall and sequence visual and oral stimuli. For both NIH Toolbox tasks, raw scores were converted to age-adjusted scores, which have a mean of 100 and a standard deviation of 15. Higher scores on the Flanker test reflect better (more positive) inhibitory control and attention. Higher scores on the List Sorting test reflect better (more positive) working memory. These tasks have demonstrated excellent psychometric properties in children and adolescents (Weintraub, Bauer, et al., 2013).

BMI/adiposity.

Weight (kg) was measured on a digital scale in a fasted state with shoes removed. Height (cm) was measured in triplicate by stadiometer and averaged. Raw BMI (kg/m²) was computed and BMI-z and percentile were derived from CDC 2000 growth charts (Kuczmarski, et al., 2002). Degree of excess gain was determined as ΔBMI_actual less ΔBMI_expected as derived from growth charts. Percent body fat was estimated using air displacement plethysmography (BodPod, Life Measurement Inc., Concord, CA), which shows good test-retest reliability and convergence with dual-energy x-ray absorptiometry (Vicente-Rodriguez, et al., 2012).

Analytic plan

A pilot sample size of N=30-50 is recommended for informing a two-arm main trial in which small-to-moderate standardized effects are anticipated (Whitehead, Julious, Cooper, & Campbell, 2016). After winsorizing outliers on all variables to 1.5 times the interquartile range <25^th or >75^th percentile (3.6% of all data points), variables approximated a normal distribution. Baseline information was compared between mindfulness versus health education using independent samples t-tests and chi-square. Study flow was described following CONSORT guidelines. Baseline characteristics were examined as predictors of attrition. To address the primary aim of feasibility/acceptability, median session attendance was compared between conditions with a non-parametric test; average homework completion was described in the mindfulness condition. Acceptability ratings were compared between conditions with t-tests and chi-square. For the secondary aim, ANCOVAs were tested with the dependent variable of baseline to six-week or baseline to six-month changes in perceived stress, food reward sensitivity, stress-eating, EF, BMI, and adiposity, with multiple imputed data with the intent-to-treat sample (i.e., all randomized participants). As a sensitivity analysis, analyses with completers were highly similar (Supplemental Tables 1-3). Condition was the independent variable, and baseline level of the respective outcome, baseline age, and sex were covaried. For the secondary outcome of stress-eating, state hunger prior to eating also was considered as a covariate, but it was not included in the models presented because it was neither related to stresseating (p>.16) nor significantly altered the results. Effect sizes were estimated with Cohen’s d (~.20: small, ~.50: medium, ≥.80: large). Statistical significance was considered p≤.01 to correct for multiple tests and lessen the likelihood of type I error. Due to the pilot nature of the study, trend-level effects (p≤.05) also are described (Leon, Davis, & Kraemer, 2011).

Results

Figure 1 displays study flow. Fifty-four adolescents were randomized to the mindfulness-based intervention (n=29) or health education control (n=25). Table 2 presents baseline characteristics by condition. There were no significant differences in demographic characteristics or key variables between conditions (p-values >.23).

Figure 1.

Study flow

Table 2.

Baseline descriptive characteristics of adolescents who participated in the mindfulness-based group intervention and health education control group program

Characteristic	Mindfulness Group	Health Education Group	p
Female, n (%)	16 (55%)	14 (56%)	.95
Race/ethnicity, n (%)			.61
Non-Hispanic White	19 (66%)	18 (72%)
Hispanic	8 (28%)	7 (28%)
American Indian	1 (3%)	0 (0%)
Asian	1 (3%)	0 (0%)
Weight status, n (%)
Lean	6 (21%)	5 (20%)	.43
Overweight	5 (17%)	8 (32%)
Obesity	18 (62%)	12 (48%)
Age, years	13.97 (1.42)	14.49 (1.72)	.23

Note: Mean (SD) unless otherwise noted as percentage.

Note: Mindfulness-based group: n=29; Health education group: n=25.

Six-week retention was 93.1% in mindfulness and 76.0% in health education (p=.08). Six-month retention was 82.8% in mindfulness and 72.0% in health education (p=.34). Comparing adolescents lost to attrition with those retained, there were no differences in sex, race/ethnicity, baseline age, BMI/adiposity, food reward sensitivity, or stress-eating. Those lost to six-week follow-up had less EF-emotional control problems than those retained (p=.005). There were no other significant baseline differences for six-week attrition nor any differences for six-month attrition.

Feasibility/acceptability

One adolescent in the mindfulness-based intervention attended zero sessions; five adolescents in health education attended no sessions. These families reported scheduling conflicts prior to the intervention. In both conditions, median attendance was 6:6 possible sessions (p=.29). Homework completion among adolescents randomized to the mindfulness-based intervention averaged 63% adherence to assigned activities (SD=23%; Range: 0-100%).

Table 3 displays acceptability ratings. All adolescents (100%) who participated in the mindfulness-based intervention perceived the program would help others like them, compared to 89% in health education who perceived the program would help others (p=.09). The majority reported the programs addressed their general concerns (mindfulness: 92% versus health education: 89%, p=.70). Average acceptability ratings in both interventions exceeded a score of 3=Average. There were two between-condition differences. Adolescents in health education perceived helpfulness of facilitators as 4.53 (SD=.51) on a scale from 1=Not at all helpful to 5=Extremely helpful, whereas adolescents in mindfulness perceived helpfulness as 3.88 (SD=.99; p=.01). In the mindfulness condition, adolescents rated how supported they felt by facilitators as 3.96 (SD=.60), on a scale from 1=Definitely not to 5=Definitely yes, which trended toward being lower as compared to 4.32 (SD=.48) in health education (p=.04).

Table 3.

Program acceptability ratings of adolescents who participated in the mindfulness-based group and health education group

Item	Mindfulness Group	Health Education Group	p
	n (%)
Addressed my concerns (yes)	24 (92%)	16 (89%)	.70
Group would help others like me (yes)	25 (100%)	16 (89%)	.09
	M (SD)
Felt comfortable to open up	3.11 (1.11)	3.68 (1.00)	.08
Group leaders were helpful	3.88 (.99)	4.53 (.51)	.01
Enjoyed coming to sessions	3.65 (.89)	3.68 (.75)	.91
Mood improved versus before group	3.65 (.56)	3.53 (.51)	.44
Supported by group leaders	3.96 (.60)	4.32 (.48)	.04
Related to other group members	3.73 (.83)	3.84 (.76)	.65

Note: Continuous items were rated on a Likert scale from 1 to 5, with 1 representing the poorest response and 5 representing the most positive response.

Note: n=26 in the mindfulness-based group and n=19 in the health education group completed program acceptability ratings.

Intervention differences in secondary outcomes

Table 4 summarizes changes in perceived stress, food reward sensitivity, and stress-eating. Accounting for initial perceived stress, baseline age, and sex, changes in perceived stress did not significantly differ by condition at either follow-up (p-values ≥.06). Within condition, adolescents in the mindfulness-based group reported decreased perceived stress at post-treatment [95%CI_{Δperceived stress} −3.90, −.53], but not six-months [−3.17, .67], whereas those in health education reported decreased perceived stress at six-months [−6.06, −1.92].

Table 4.

Changes in perceived stress, food reward sensitivity, and eating in response to stress over the study period in the intent-to-treat sample

	Mindfulness Group	Health Education Group
Outcome	Mean (SE)	Mean (SE)	Between-Condition Δ Effect‡	Cohen’s d	p
Perceived stress
Baseline	15.38 (1.24)	15.06 (1.33)	--		--
Six-weeks	13.08 (1.41)*	15.39 (1.52)	−.36, 4.63	.45	.09
Six-months	14.71 (1.46)	11.55 (1.57)*	−5.56,.09	−.47	.06
Food reward sensitivity
Baseline	50.35 (11.67)	53.36 (12.57)	--		--
Six-weeks	59.82 (12.52)*	82.65 (13.48)*	−12.13, 22.81	.16	.55
Six-months	37.99 (9.72)	61.67 (10.47)*	2.63, 22.53	.64	.01
Stress-eating, kcal
Baseline	745.51 (70.59)	779.25 (76.03)	--		--
Six-weeks	846.14 (72.92)	910.98 (78.53)	−151.32, 82.87	−.14	.57
Six-months	786.00 (66.92)	989.09 (72.08)*	−2.33, 276.30	.49	.05

Note: Mean (SE) are unadjusted values.

Note:

^‡95% CI for health education group (n=25) compared to mindfulness-based group (n=29) between-condition difference on change in outcome at six-weeks and six-months. Cohen’s d refers to the standardized effect size of the between-condition difference, and the p-value refers to the significance of the between-group condition effect. All between-condition effects are adjusted for the respective baseline level of each outcome, sex, and baseline age.

Note:

^*Refers to within-condition change from baseline of p<.05.

Change in food reward sensitivity at six-months differed between mindfulness and health education, with a moderate effect size (Cohen’s d=.64, p=.01). Within condition, adolescents in the mindfulness-based intervention showed no change from baseline [95%CI_{Δfood reward} −4.12, 9.36], whereas those in health education showed significant increases in food reward sensitivity at six-months [7.93, 22.47].

Accounting for baseline stress-eating, baseline age, and sex, there was a trend-level difference between conditions in change in stress-eating from baseline to six-months (d=.49, p=.05). Adolescents in health education increased stress-eating [95%CI_{Δstress-eating} 38.58, 241.82 kcal], whereas adolescents in the mindfulness intervention showed no change [−91.11, 97.54 kcal].

A summary of changes in EF is presented in Table 5. There were no significant differences between conditions, but a number of trend-level patterns in parent-reported EF. Accounting for baseline emotional control, baseline age, and sex, adolescents in the mindfulness-based group decreased emotional control problems at six-weeks [95%CI_{Δemotion control} −2.94, −.76] compared to adolescents in health education [−1.40, 1.04] (Cohen’s d=.57, p=.05). The between-condition difference in emotional control was attenuated at six-months (d=.55, p=.08). Additionally, adolescents in the mindfulness-based intervention showed a trend toward greater decreases in planning/organization problems at six-weeks [95%CI_{Δplan/organize} −1.96, .19] compared to health education [−.43, 2.04] (d=.56, p=.04), a trend-level effect that was further attenuated at six-months (d=.45, p=.13). Adolescents in the mindfulness-based intervention also showed a trend toward greater decreases in EF-self-monitoring problems than those in health education at six-weeks [95%CI_Δmonitor −1.68, .04 vs. −.49, 1.40] (d=.56, p=.05) and six-months [−1.92, .10 vs. −.42, 1.65] (d=.62, p=.05). There were no other significant or trend-level patterns between conditions in EF. Within condition, adolescents in the mindfulness group had decreased EF-inhibition problems at six-months, whereas those in health education showed no change [95%CI_Δinhibition −2.30, −.48 vs. −1.58, .41], with no between-condition difference (d=.35, p=.24).

Table 5.

Changes in executive function (EF) over the study interval in the intent-to-treat sample

	Mindfulness Group	Health Education Group
Outcome	Mean (SE)	Mean (SE)	Between-Condition Δ Effect‡	Cohen’s d	p
EF: Inhibit
Baseline	4.74 (.71)	5.36 (.76)	--		--
Six-weeks	4.38 (.88)	6.19 (.94)	−.73, 2.54	.30	.28
Six-months	.84 (1.94)*	3.03 (2.09)	−.53, 2.15	.35	.24
EF: Shift
Baseline	6.12 (.57)	5.63 (.61)	--		--
Six-weeks	5.52 (.60)	6.90 (.65)	−.47, 2.24	.36	.20
Six-months	7.14 (1.45)	4.49 (1.56)	−.96, 1.62	.15	.62
EF: Emotion control
Baseline	8.04 (1.06)	7.89 (1.14)	--		--
Six-weeks	5.97 (.95)*	9.34 (1.02)	−.02, 3.37	.57	.05
Six-months	3.97 (4.90)	11.07 (5.27)	−.19, 3.24	.55	.08
EF: Initiate
Baseline	7.00 (.59)	7.26 (.64)
Six-weeks	6.95 (.80)	8.71 (.86)	−.44, 1.41	.32	.30
Six-months	5.22 (.96)	7.71 (1.04)	−.18, 1.92	.46	.10
EF: Working memory
Baseline	6.99 (.80)	8.43 (.86)
Six-weeks	6.32 (1.07)	9.91 (1.15)	−1.16, 1.88	.14	.64
Six-months	9.63 (4.07)	3.58 (4.39)	−1.23, 1.76	.10	.73
EF: Plan/organize
Baseline	9.77 (.96)	10.33 (1.04)
Six-weeks	8.91 (1.15)	13.58 (1.24)	.07, 3.31	.56	.04
Six-months	14.57 (7.31)	5.05 (7.87)	−.41, 3.14	.45	.13
EF: Materials
Baseline	7.76 (.58)	7.36 (.54)
Six-weeks	6.92 (1.00)	6.26 (.93)	−.47, 2.04	.35	.22
Six-months	7.17 (1.79)	3.81 (1.66)	−1.13, 1.65	.11	.71
EF: Monitor
Baseline	6.86 (.60)	7.38 (.64)
Six-weeks	6.39 (.70)	8.42 (.75)	.00, 2.56	.56	.05
Six-months	3.64 (1.26)	7.84 (1.36)	.01, 3.05	.62	.05
Flanker
Baseline	105.94 (1.65)	104.53 (1.78)
Six-weeks	107.36 (1.93)*	109.88 (2.08)*	−3.65, 4.16	.03	.90
Six-months	104.88 (3.92)	105.82 (4.22)	−2.98, 5.65	.16	.54
List Sort
Baseline	104.96 (2.28)	104.13 (2.46)
Six-weeks	109.31 (2.90)*	103.10 (3.12)*	−5.52, 2.73	−.17	.51
Six-months	113.57 (3.60)*	111.77 (3.87)*	−5.92, 2.88	−.19	.50

Note: Mean (SE) are unadjusted values.

Note:

^‡95% CI for health education group (n=25) compared to mindfulness-based group (n=29) between-condition difference on change in outcome at six-weeks and six-months. Cohen’s d refers to the standardized effect size of the between-condition difference, and the p-value refers to the significance of the between-group condition effect. All between-condition effects are adjusted for the respective baseline level of each outcome, sex, and baseline age.

Note:

^*Refers to within-condition change from baseline of p<.05.

There were no between-condition differences in either of the NIH Toolbox behavioral measures of EF. Accounting for covariates, adolescents in the mindfulness and health education conditions both improved Flanker-inhibitory control/attention at six-weeks [95%CI_ΔFlanker .49, 5.73 vs. .49, 6.25] but not at six-months [−2.35, 3.41 vs. −1.34, 5.07]. Similarly, adolescents in mindfulness and health education both improved List Sort-working memory at six-weeks [95%CI_{ΔList Sort} 3.18, 8.74 vs. 1.54, 7.60] and six-months [4.68, 10.58 vs. 2.86, 9.37].

There were no between-condition differences in any BMI or adiposity measure (Table 6). Adolescents in both mindfulness and health education conditions increased raw BMI (kg/m²) at six-months [95%CI_ΔBMI .20, 1.52 kg/m² vs. .21, 1.62 kg/m²], with no difference between-conditions (p=.91). BMI-z and percentile were stable over six-month follow-up. Within both mindfulness and health education conditions, adolescents significantly lowered percent body fat at six-week follow-up [95%CI_{Δbody fat%} −2.55, −.07% vs. −3.91, −1.25%] and six-month follow-up [−3.64, −.61% vs. −4.31, −1.04%]; those in health education also had lower fat mass from baseline to six-weeks [−2.59, −.18 kg], with no between-condition difference in adiposity at either interval (p-values ≥.17).

Table 6.

Changes in body mass index (BMI) indices over the study period in the intent-to-treat sample

	Mindfulness Group	Health Education Group
Outcome	Mean (SE)	Mean (SE)	Between-Condition Δ Effect‡	Cohen’s d	p
BMI, kg/m2
Baseline	27.24 (.89)	27.44 (.96)	--		--
Six-weeks	27.35 (.99)	27.65 (1.06)	−.61, .68	.03	.91
Six-months	27.95 (1.09)*	28.05 (1.18)*	−.92, 1.04	.04	.91
BMI, z-score
Baseline	1.62 (.11)	1.56 (.11)	--		--
Six-weeks	1.58 (.11)	1.57 (.12)	−.10, .14	.10	.77
Six-months	1.58 (.13)	1.50 (.14)	−.15, .15	−.01	.97
BMI, percentile
Baseline	92.99 (1.37)	91.52 (1.47)	--		--
Six-weeks	92.70 (1.32)	92.15 (1.42)	−.75, .95	.07	.82
Six-months	92.60 (1.43)	91.06 (1.54)	−1.20, 1.01	−.05	.87
BMI gain, actual v. expected
Six-weeks	−.24 (.22)	.06 (.23)	−.67, .70	.01	.97
Six-months	−.94 (.39)*	−1.21 (.42)	−.88, 1.02	.04	.88
Body fat, kg
Baseline	25.68 (1.71)	25.65 (1.84)
Six-weeks	25.98 (1.94)	26.72 (2.09)*	−2.83,.47	−.40	.26
Six-months	25.42 (2.75)	26.49 (2.96)	−2.69, 1.78	−.11	.69
Body fat, %
Baseline	35.66 (1.41)	34.38 (1.52)
Six-weeks	34.63 (1.38)*	32.71 (1.48)*	−3.09, .55	−.38	.17
Six-months	32.90 (1.92)*	32.29 (2.06)*	−2.84, 1.74	−.13	.64

Note: Mean (SE) are unadjusted values.

Note:

^‡95% CI for health education group (n=25) compared to mindfulness-based group (n=29) between-condition difference on change in outcome at six-weeks and six-months. Cohen’s d refers to the standardized effect size of the between-condition difference, and the p-value refers to the significance of the between-group condition effect. All between-condition effects are adjusted for the respective baseline level of each outcome, sex, and baseline age.

Note:

^*Refers to within-condition change from baseline of p<.05.

Discussion

The primary objective of this randomized controlled pilot study was to assess the feasibility and acceptability of a six-session mindfulness-based group intervention delivered to adolescents at-risk for excess weight gain. The secondary objective was to explore if this mindfulness-based intervention showed patterns toward altering adolescents’ perceived stress, EF, food reward sensitivity, stress-eating, BMI, and adiposity, relative to a time-matched health education control group.

The majority of adolescents came to all sessions. One-hundred percent of participants perceived the mindfulness-based program would help others like them. Acceptability ratings were above-average, providing support for the notion that adolescents generally liked the program. Adolescents who participated in health education felt more supported by facilitators than adolescents in the mindfulness intervention. Facilitators did not alternate between mindfulness and health education; thus, it is not possible to ascertain whether ratings differed because of facilitator characteristics or program characteristics. Behavioral health/psychological interventions are typically expected to evoke some discomfort, which may be a necessary part of making change, yet maintaining rapport is also important (Eubanks-Carter, Muran, & Safran, 2015). In the setting of a mixed-sex group format, adolescents—who primarily exhibited overweight/obesity in this sample—may have felt more comfortable with facilitators presenting didactic material on health knowledge, mirroring more familiar content of a classroom setting, than with practicing mindfulness-based exercises that promote attention to mind-body in front of same- and other-sex peers. To further enhance acceptability of mindfulness-based interventions for adolescents at-risk for excess weight gain, it may be valuable to increase activities to encourage rapport-building among group members and facilitators (Eubanks-Carter et al. 2015). In future research, process evaluation methods may offer a helpful means of pinpointing the most acceptable facilitator skills in teaching mindfulness to adolescents at-risk for excess weight gain.

Reported homework completion averaged 63%, comparing favorably to past studies (Quach, Gibler, & Jastrowski Mano, 2017) and suggesting that adolescents in this study were practicing mindfulness skills in daily life. Although it is often assumed clinically that home practice is beneficial, at-home practice effects on mindfulness-based intervention outcomes have shown mixed results in adults (Carmody, Baer, E, & Olendzki, 2009; Vettese, Toneatto, Stea, Nguyen, & Wang, 2009). A larger investigation is required to test to what extent homework completion relates to psychological, behavioral, and/or weight-related outcomes in adolescents at-risk for excess weight gain. If at-home practice is important for outcomes, then developing strategies to make practicing more accessible—such as, by integrating phone and internet technologies—may be beneficial. Using new technologies also could permit more objective assessment of homework completion rather than reliance on self-report. Similarly, determination of the types of take-home practices that are most acceptable to adolescents would be valuable; our experience is that brief, informal tasks (e.g., take a mindful breath, practice a daily activity with mindful attention, practice a brief body scan when stressed) are more often completed than formal meditation time (e.g., listening to an audio-recording) (Dalager, et al., 2018).

In terms of secondary outcomes, there were no significant differences in perceived stress. Instead, adolescents randomized to the mindfulness-based intervention had significantly less six-month food reward sensitivity—operationalized as relative reinforcing value of a preferred snack food—compared to health education. At six-months, there also was a trend-level pattern for adolescents in mindfulness to have less stress-eating—measured objectively as energy intake from a laboratory test meal after a stress induction—compared to adolescents in health education. The effects on food reward sensitivity and stress-eating were moderate in size. Whereas adolescents within the health education condition increased food reward sensitivity and stress-eating over the six-month interval, adolescents within the mindfulness-based intervention showed no change in either characteristic. The observed increases within health education are consistent with observational studies indicating increases in eating in response to cues other than hunger, including stress, as children and adolescents develop (Francis, Ventura, Marini, & Birch, 2007). The current results also fit with the notion that restructuring of reward processing may be one neurobiological mechanism that underlies mindfulness-based interventions, such that reward sensitivity to palatable junk food becomes relatively less salient than alternative, healthier rewards (Garland, 2016; Volkow, et al., 2013). These preliminary findings support the possibility that mindfulness-based training tailored for adolescents to teach general mindfulness skills may offer benefits for food reward sensitivity and eating in response to stress or negative emotions, which have been predictive of excess weight and/or adiposity gain (Stice & Yokum, 2016; Stojek, et al., 2017).

There were no significant between-condition effects on parent-reported or behavioral measures of EF, but several patterns were observed. Compared to adolescents in health education, adolescents randomized to the mindfulness-based intervention showed trend-level improvements in several domains of parent-reported EF, including fewer emotional control and planning/organizational problems at six-weeks and fewer self-monitoring problems at six-weeks and six-months. The effect sizes for these differences were moderate, and there was some attenuation in the strength of effects at the six-month follow-up as opposed to at the immediate post-treatment follow-up. The degree to which adolescents continued mindfulness practices during the follow-up period was not assessed, but could be helpful in understanding the duration of EF effects in future research. The current patterns are consistent with the existing literature suggesting that experimental mindfulness inductions may offer salutary, acute effects on some domains of EF, such as attention (Leyland, et al., 2018). Because this pilot study was not powered to detect statistically significant differences between conditions, the benefits of mindfulness-based training on EF in adolescents at-risk for excess weight gain remain inconclusive. Indeed, there were no between-condition differences on behavioral measures of EF, and adolescents in both conditions showed some improvements in inhibitory control/attention and working memory. Future research with a larger sample using multi-modal assessments of EF, including neuroimaging techniques, would allow for a more systematic understanding of specific EF domains potentially altered via mindfulness training. In theory, increasing cognitive control and attention would be anticipated to alter circuitry to more effectively regulate food reward sensitivity in adolescents at-risk for excess weight gain (Garland, 2016; Volkow, et al., 2013).

There were no significant or trend-level condition effects on BMI or adiposity at six-weeks or six-months. Raw BMI increased at six-months, as would be expected in developing adolescents (Kuczmarski, et al., 2002); BMI-z and percentile were stable. In both conditions, percent body fat decreased by 2-3% from baseline to six-months. Of note, the mindfulness-based intervention was not intended for weight loss. The goal was selective prevention of excess weight gain by providing mindfulness training to reduce food reward sensitivity and improve EF. This focus appealed to adolescents and their families and may be more acceptable to adolescents than a mindful eating/weight loss focus (Daly, et al., 2016). Longer-term follow-up is required to determine if the observed, acute changes in food reward sensitivity and patterns of moderate effects for stress-eating and EF subsequently alter BMI trajectories over a longer interval. In prevention-seeking samples, effects of psychosocial interventions on growth may become more apparent over a longer time (Tanofsky-Kraff, et al., 2017).

Study strengths include the developmentally-sensitive and targeted prevention approach to excess weight gain in at-risk adolescents. However, effect sizes should be considered cautiously in small pilot studies because of problems with reproducibility (Kraemer, Mintz, Noda, Tinklenberg, & Yesavage, 2006). Yet, using a randomized, controlled design with a health education control means that the observed, moderate effects for mindfulness-training are present after accounting for the instruction time and non-specific support provided by a group. Additional strengths include objective assessment of BMI from fasting weight, inclusion of body composition, administration of a behavioral task to assess food reinforcement, a laboratory test meal to objectively assess stress-eating, and parent-reported and behavioral assessments of EF.

Generalizability is limited to a convenience sample of adolescents at-risk for excess weight gain, which was determined for this study as a BMI at or exceeding the 70^th percentile or having two parents with obesity. Although the majority of youth (~80%) had overweight or obesity at baseline, it is possible that using more restrictive selection criteria (e.g., BMI at or exceeding the 75^th or 85^th percentile) may have yielded different results. Additionally, there was limited representation of all racial/ethnic minorities disproportionately affected by obesity. Although perceptions of general acceptability of the interventions were assessed, presenting concerns were not quantitatively measured nor whether program participation addressed concerns specific to stress or weight. This information may be valuable to gather in the future, particularly with respect to the possibility of eventual implementation or dissemination in healthcare or community settings. Although retention for follow-up assessments did not significantly differ between conditions, six-month drop-out was 17% in mindfulness and 28% in health education. Further, despite no other differences in missing data patterns, baseline EF-emotional control problems were lower among those lost to six-week follow-up. Fidelity and competence in administration of the mindfulness-based intervention were closely monitored by a clinician with relevant expertise; yet, in future larger trials, collection of quantitative ratings of adherence would be important to strengthen scientific rigor. Likewise, assessors of outcomes were not consistently blinded for this pilot study, and this would be an important modification for a subsequent trial. Adolescents were followed through six-months after the interventions; however, longer follow-up would be ideal for assessing impact on growth. Further, future studies may wish to examine metabolic outcomes. Mindfulness-based approaches have shown potential for reducing metabolic risk factors in at-risk adults (Raja-Khan, et al., 2017) and adolescents (Shomaker, et al., 2017), potentially through pathways such as altering cortisol, inflammation, macronutrient intakes, sleep, and/or physical activity/sedentary behavior.

Findings from the current randomized controlled pilot study contribute to the development of alternative methods for addressing adolescent excess weight gain. A mindfulness-based program tailored to adolescents may positively affect food reward sensitivity and potentially offer some benefits for stress-eating and EF. To what extent these factors lead to sustained prevention of excess weight gain in at-risk adolescents—alone or in combination with other support systems—requires determination in a larger trial.