Systematic review and meta-analysis of combined cognitive-behavioral therapy and physical activity and exercise interventions for pediatric chronic disease

Abstract

Objective

Cognitive-behavioral (CBT) interventions combined with either a physical activity (CBT+PA) or exercise intervention (CBT+Ex) are becoming more common in pediatric populations. Considering the independent effects of PA and exercise on health and psychological outcomes, it is unclear whether CBT alone differs from CBT+PA or CBT+Ex in efficacy. The main objective of this systematic review and meta-analysis of randomized clinical trials (RCTs) was to assess the efficacy of CBT+PA and CBT+Ex interventions in pediatric chronic disease.

Method

This review included RCTs in children (≤18 years) with a chronic condition, a CBT+Ex or CBT+PA intervention, and an objective measure of PA&Ex. Seven databases were searched using MeSH terms and key terms and included studies published before July 1, 2023. Abstracts were reviewed for inclusion by two independent reviewers, data was extracted by three independent reviewers. Risk of bias (RoB 2) and study quality were coded. Random effect meta-analyses of differences in between-group change in PA&Ex were conducted.

Results

Eligible studies (k = 5) reported outcomes for a combined 446 children. A small, nonsignificant overall effect was found (d = 0.10, 95% CI −0.16, 0.35) indicating intervention groups (CBT+PA or CBT+Ex) increased engagement in PA&Ex more than comparator groups (CBT). Additional analyses were inconclusive due to the small number of eligible studies.

Discussion

Additional RCTs are needed with integrated PA&Ex interventions targeting pediatric chronic disease. Future trials should report more detailed PA&Ex data. The full protocol for this analysis was prospectively registered in Open Science Framework (project ID: osf.io/m4wtc).

The U.S. Centers for Disease Control and Prevention (CDC) recommends that children and adolescents (age 6–17 years old) participate in ≥60 min of moderate-to-vigorous intensity physical activity (MVPA) every day (Chaput et al., 2020; Piercy & Troiano, 2018). These guidelines focus on Physical Activity, which is defined as “any bodily movement produced by skeletal muscles that results in energy expenditures” (Caspersen et al., 1985, 126). A subcategory of physical activity (PA), Exercise is a more narrowly defined construct and includes movement that is “planned, structured, repetitive, and purposive in the sense that improvement or maintenance of one or more components of physical fitness is an objective” (Caspersen et al., 1985, 128).

Regular PA and exercise (PA&Ex) is associated with numerous short- and mid-term health benefits in children and adolescents that persist into adulthood (Hosker et al., 2019), including improved health-related quality of life (HRQOL) (Marker et al., 2018), decreased risk of obesity, diabetes, cardiorespiratory health concerns, and cardiometabolic disorders (Baranowski et al., 1997; Jáuregui et al., 2020; Malik et al., 2021; Rowland, 2001; Saunders et al., 2016). Despite the vast evidence supporting PA&Ex, children and adolescents globally do not meet PA&Ex guidelines (Kann et al., 2018) and this problem may be amplified in youth with a chronic disease or health condition (Saunders et al., 2016; West et al., 2019). Current PA&Ex guidelines focus on youth without chronic disease. However, efforts have increased to introduce PA&Ex as an intervention target for some common chronic pediatric diseases (West et al., 2019). The integration of PA&Ex into health interventions aligns with contemporary intervention models for pediatric chronic disease, which often combine multiple evidence-based treatments that directly target the child or adolescent, including psychological, physical, and pharmacological interventions (Bronfenbrenner, 1986; Cohen et al., 2017; Wilson & Lawman, 2009).

Combined cognitive-behavioral therapy (CBT) and PA&Ex interventions (i.e., CBT+PA&Ex) have been implemented or are being evaluated in many pediatric health conditions, including obesity (Vlaev et al., 2021), diabetes (Gutierrez-Colina et al., 2023), cardiovascular disease (Annesi, 2022), chronic pain (Kashikar-Zuck et al., 2018), juvenile idiopathic arthritis (Armbrust et al., 2015), cystic fibrosis (Havermans & Willem, 2019), and cancer (Brennan et al., 2013). CBT interventions for youth with chronic disease include skills-focused approaches to target maladaptive thoughts and behaviors, coping strategies, disease self-management, and problem solving (Moore et al., 2019; Morey & Loades, 2021; Thompson et al., 2011). Concurrently, an exercise as medicine approach may be used to supplement and reinforce CBT, prevent disease development, affect disease pathways and mechanisms, and be used to manage behaviors and symptoms associated with disease (Lee et al., 2012; Thornton et al., 2016; West et al., 2019).

Limitations of extant literature and rationale for this review

The extent of the literature describing the independent effects of CBT and PA&Ex is unclear and limited in several ways. First, studies describing the combined effects of CBT and PA&Ex rarely delineate the differences between PA and exercise; and researchers and clinicians should be cognizant of the semantics of these terms and measurement considerations when evaluating and describing published interventions. Second, little is known about the differences in the efficacy or effectiveness of CBT interventions when combined with PA vs. exercise. Third, existing reviews and meta-analyses have been disease-specific and have not attempted to evaluate the efficacy of combined effects of CBT and PA&Ex interventions on behavioral and physical outcomes that crosscut chronic disease states in pediatric populations (Bernard et al., 2018). While CBT+PA&Ex interventions may be advantageous for some disease groups, we are unaware of previous reviews that examine the treatment effects of these combined interventions across different chronic disease populations. There is a need to better understand for whom these interventions work best, especially considering the increased prevalence of combined CBT and PA&Ex interventions. Lastly, it is unclear whether CBT+PA&Ex intervention efficacy may be moderated by other variables such as treatment delivery modality or contraindications associated with a specific chronic disease condition.

Objectives

The goal of this systematic review and meta-analysis was to evaluate and delineate PA&Ex intensities and levels to achieve the following aims: (1) assess the efficacy of combined CBT and PA&Ex interventions for improving PA&Ex levels, (1a) across pediatric chronic diseases among children and adolescents, and (1b) in comparison to a control or comparison group; (2) evaluate if interventions that utilize CBT+PA differ from CBT+exercise (CBT+Ex); and (3) explore whether intervention effects differ by the type of chronic disease and modality of administration (e.g., unsupervised vs. supervised, in-person vs. hybrid vs. telehealth). It is noted that while other outcomes (i.e., long-term follow-up assessment, discrete PA intensity levels, METs, and HRQOL) and variables of interest (i.e., PA&Ex device used, algorithms used, and mode of intervention delivery) were proposed in our protocol, they were not included in the final meta-analysis due to insufficient data and sample sizes.

Method

Literature search strategy

The full protocol for this analysis was prospectively registered in Open Science Framework on October 16, 2023 (Black et al., 2023), and is available publicly. A systematic search of PubMed, EmBase, CINAHL, APA PyscINFO, Proquest Dissertations, Scopus, and Web of Science was conducted on November 3, 2023 by a medical librarian (A.G.). Databases were searched using MeSH terms, key words, and Boolean operators. MeSH terms included: “psychosocial interventions,” “cognitive restructuring,” “acceptance and commitment therapy,” “physical activities,” “physical fitness,” and natural language (i.e., key terms) including “childhood cancer,” “transcendental meditation,” “chronic illness,” and “kinesiotherapeutic intervention.” The full search strategy is available in online supplementary material 1.

Study eligibility criteria and selection

To be included in this review, studies had to be randomized clinical trials (RCTs) which tested an integrated CBT intervention combined with a PA&Ex intervention in children and adolescents (0–18 years of age) diagnosed with a chronic disease, which is defined as a chronic condition with a duration of at least 3 months and requiring ongoing management (Moore et al., 2019). Studies must have been published before July 1, 2023. CBT interventions were defined as psychological interventions designed to change thoughts and behaviors with the goal of improving patient functioning and treatment outcomes (Cuijpers et al., 2013). CBT interventions evaluate, challenge, and modify dysfunctional and unhelpful thoughts and beliefs to promote behavior change. The goals of CBT are achieved via a combination psychoeducation and skills-based coping, management, and thought and behavior-change strategies (Evershed, 2011). PA and/or exercise interventions were defined as any intervention or strategy that encouraged, recommended, or prescribed physically exertional movement above and beyond a patient’s current levels. Chronic illnesses included but were not limited to: “wolfram syndrome,” “diabetes mellitus” types 1 and 2, “obesity,” “epilepsy,” “anemia, sickle cell,” and “lupus erythematosus, systemic.”

Both active comparator (e.g., alternative active treatment, equivalence groups, usual care, and standard of care or optimized standard of care) and control conditions (e.g., nontreatment, wait-list, placebo, or attention controlled) were included, including those whose primary goals were to evaluate the effect of their intervention compared to non- or minimal treatment, as well as those that chose to evaluate the additional treatment effects provided by the intervention (Freedland et al., 2019). A liberal requirement of at least 10 participants in each trial arm at postintervention was required to maximize the generalizability of included studies.

To reduce the potential impact of publication bias, studies from both peer-reviewed journals and grey literature (e.g., information outside of traditional publishing, including dissertations and conference abstracts) were included in the search strategy and were considered for inclusion in the analyses. Only studies available in English, the language spoken by the current research team, were included.

Selection of studies

Each abstract was independently reviewed in duplicate. Our process differed slightly from our protocol, as multiple authors (W.R.B., D.A.W., B.F., C.R.B., C.D.P. & L.V.K.) and research coordinators (Noor Qureshi, S.D.) reviewed abstracts simultaneously after random assignment in Covidence (Covidence, 2023), a software for data screening and extraction. The original protocol proposed that each step be completed in phases, where the studies would be reviewed by members of the team (i.e., W.R.B., S.D., and Noor Qureshi), and then reviewed by the remainder of the study team. However, this strategy was impractical when using the random assignment framework in Covidence. The random assignment method resulted in a stronger methodology as it is intended to help reduce bias of study ratings and reviews.

Each author independently determined eligibility by screening the title and abstracts of studies identified by the search. Research coordinators (S.D., Noor Qureshi) co-coded 10 abstracts with W.R.B. Additionally, abstracts in question were reviewed and discussed in biweekly research meetings, and on an ad hoc basis. Full-text articles were obtained for the remaining abstracts and potential review articles. Forty-eight review articles were evaluated for studies which could be included, resulting in 11 potential additional studies for consideration for inclusion in this meta-analysis; however, none were included for full-text review. Full-text review was conducted by W.R.B., S.D., and Noor Qureshi, utilizing the same strategy as the abstracts. Though a protocol was in place to resolve questions or discrepancies, this was not needed to determine which proceeded to data extraction.

Selected outcomes

This review evaluated PA&Ex outcomes across treatment as continuous data; as such, only studies with at least baseline and posttreatment data were included. Posttreatment was defined either as the time-point specified by the research team of the study being reviewed, or as the most immediate time-point after the conclusion of treatment. Only one study included follow-up data for PA&Ex. Studies that objectively measured PA via accelerometry were included in this review. Other objective measures of PA were also included if deemed to be reliable and relevant to the field and if discrete intensity-level PA data was provided. Reviewed publications were searched for device-based measures of PA. Numerous PA intensity levels were included in the search, including sedentary behavior, light-intensity, moderate-intensity, and vigorous and high intensity (Norton et al., 2010) (see online supplementary material 2). METs and MET algorithms were considered a priori but were not included in the studies in this analysis. HRQOL was also included in the protocol and extracted if available, but not included in the meta-analysis due to only one study including an HRQOL measure (Braam et al., 2018). Subsequently, this study focuses on PA&Ex as the outcome. Other exploratory variables included activity type (CBT+PA vs. CBT+Ex) and modality of delivery. The full data-extraction form is provided in online supplementary material 3.

Data extraction, risk of bias, and study quality

After identifying studies for inclusion, independent data extraction was conducted (see online supplementary material 3) and risk of bias (RoB) assessment was completed by five study authors (W.R.B., D.A.W., B.F., J.L.J., and S.D.) using the Cochrane RoB version 2 tool (RoB 2) (Sterne et al., 2019). RoB was judged as (a) low RoB, (b) some concerns, and (c) high RoB. Overall study quality was also assessed and recorded per the Cochrane GRADE system (i.e., Very low, Low, Moderate, and High) (W.R.B.). Data extraction was conducted in triplicate, first by W.R.B. and then secondary and tertiary extractions were conducted by the remaining team members. This approach varied from the proposed protocol to expedite the study progress and to provide greater confidence in the accuracy of extracted data. As a final quality control measure, all data extracted from articles were linked with team members’ initials to ensure questions about individual data points were directed toward the correct team member. As an additional step not specified a priori, after data extraction, studies were reviewed by two study authors with expertise in meta-analyses and PA interventions (L.V.K., C.D.P.). This process was used to resolve questions around study inclusion/exclusion based on provided data. See the PRISMA flow chart (Figure 1) for the status of identified studies (Page et al., 2021), as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., 2020). Of note, team review did result in a change to how one included study was coded. Jackson et al. (2022) described both the use of PA-focused session content, combined with exercise prescription that “focused on increasing the frequency, intensity, and duration of PA engagement” (p. 4). While this was initially coded as an exercise adjunct, follow-up discussions with the Jackson research group noted that their exercise prescription was generally nonspecific and tended to focus on PA behaviors. Thus, the present research team re-coded this as a PA intervention.

Figure 1.

PRISMA flow diagram.

Analyses

Effect size calculation

We calculated standardized mean differences (SMDs; Cohen’s d) to estimate differences in between-group changes (i.e., difference in change scores). We used SMDs, rather than absolute mean differences, because of the anticipated differences in the types of assessment metrics employed across studies. While Hedges g was initially proposed in the protocol to calculate the SMD, as it is more robust to small sample sizes which are common in studies of clinical populations (Hedges, 1981), all samples were greater than 20, making it appropriate for our analysis to utilize Cohen’s d rather than Hedges g. Where studies did not report within-group standard deviations necessary to calculate between-group changes, a correlation of 0.5 was used (Lipsey & Wilson, 2001). Where pre- and postsample sizes differed, postsample sizes were used to calculate SMD.

Meta-analysis

Restricted maximum likelihood random-effects (RML) models were employed to estimate pooled effects of interventions (Higgins et al., 2023). The use of RML models differed from the proposed protocol’s use of DerSimonian–Laird models, given that studies had higher sample sizes than initially anticipated. In line with common convention, the direction of effects was corrected such that effects favoring the intervention were positive (e.g., the intervention group experienced a larger desirable change pre-to-post relative to the comparator) (Borenstein et al., 2009; Lipsey & Wilson, 2001). In line with convention, Cohen’s d scores are considered small if d < 0.20, medium if d = 0.50, and large if d ≥ 0.80 (Fritz et al., 2012). Forest plots of effect sizes are displayed in online supplementary material 4. In addition to the overall effect of combined CBT and PA&Ex, exploratory analyses are presented for areas of heterogeneity, including comparisons of intervention to either treatment as usual, or active control, activity type (i.e., MVPA, sedentary time), intervention type (i.e., CBT+PA, CBT+Ex), disease group (i.e., overweight vs. other), and modality (i.e., in-person delivery vs. virtual vs. hybrid).

Risk of publication bias

A funnel plot and Eggar’s regression were proposed in the protocol to investigate publication bias. However, convention recommends analysis to detect publication bias be performed only when there are 10 or more studies, and so tests for publication bias were not conducted (Page et al., 2023). Similarly, tests for heterogeneity were proposed in the protocol, but given the low number of eligible studies, these tests were likely to produce unstable estimates and were not conducted (Page et al., 2023).

Results

Study selection

In total, 1,101 unique articles were identified, and 66 studies were included in full-text screening. From these studies, eight were included for data extraction (Braam et al., 2018; Brennan et al., 2013; Jackson et al., 2022; Jelalian et al., 2010, 2019; Pbert et al., 2016; Sepúlveda et al., 2020; Skjåkødegård et al., 2022), and five in the final meta-analysis (Braam et al., 2018; Jackson et al., 2022; Pbert et al., 2016; Sepúlveda et al., 2020; Skjåkødegård et al., 2022). Per Figure 1 (i.e., PRISMA flow chart), the most common reason for exclusion from data extraction was that the article did not include PA measurement that met the protocol’s definition. Of the eight manuscript study teams, seven were emailed for additional data or clarification on reporting—one study author responded to provide additional data for inclusion in this analysis. Other studies were excluded because they did not include posttreatment data (Brennan et al., 2013), had low sample size (Jelalian et al., 2010), or did not report descriptive data for PA (Jelalian et al., 2019).

Study characteristics

For the five studies included in the analysis, a demographic summary is provided in Table 1. The chronic disease conditions represented in these five studies include: cancer (n = 1), congenital heart disease (n = 1), overweight/obesity (n = 2), and severe obesity (n = 1). The mean age of participants ranged from 9.9 to 16.4 years, and overall there were 423 participants included across the studies; sample sizes ranged from 60 individuals (Jackson et al., 2022) to 114 (Skjåkødegård et al., 2022). The combined sample was 53% female; individual study sex ratios ranged from 45% (Sepúlveda et al., 2020) to 62% (Pbert et al., 2016). There was notable variability in the reporting of study demographics; they are summarized in Tables 1 and 2 for each study. Two studies did not report any race or ethnicity data (Braam et al., 2018; Skjåkødegård et al., 2022) and a third was conducted internationally, and measured ethnicity according to whether individuals identified as “Latin American,” “Spanish,” or “Other” (Sepúlveda et al., 2020). Of those who do report race and ethnicity, 63% (Pbert et al., 2016) and 87% (Jackson et al., 2022) were White.

Table 1.

Overview of studies.

First author, year	Design	Age (mean, SD)	Total sample (N)	Percent female (total, intervention, comparator)	Disease condition	Quality*	PA outcome
Braam et al., 2018	Individually randomized parallel-group trial	13.2, 3.1	68	46.0, 47.0, 45.0	Cancer—solid tumor and hematological	High†	Mean counts per minute**
Jackson et al., 2022	Individually randomized parallel-group trial	16.3, 1.0	60	46.7, 41.9, 51.7	Congenital Heart Disease	Moderate	Change in MVPA and sedentary behavior
Pbert et al., 2016	Cluster-randomized parallel-group trial	16.39, 1.20	111	62.1, 63.0, 61.4	Overweight/Obese	Moderate	Percent of time spent in MVPA
Sepúlveda et al., 2020	Individually randomized parallel-group trial	9.87, 1.29	70	45.0, 40.7, 48.3	Overweight/Obese	Low	Minutes per day of sedentary, light, moderate, and vigorous activities
Skjåkødegård et al., 2022	Individually randomized parallel-group trial	12.6, 3.04	114	57.6, 61.0, 56.4	Severe obesity	Moderate	Percent of time spent in MVPA

^*All studies started quality ratings at “High,” as all included studies are RCTs. Studies were downgraded if found to be at a “high” risk of bias via ROB2. Sepúlveda et al., (2020) was downgraded to “low” for the purpose of this study, as their study was designed to evaluate effects of a different intervention component; thus, both groups included the same amount of PA intervention.

^**Mean counts per minute is defined as a score encompassing horizontal, vertical, and depth of motion, with higher scores indicating higher levels of activity.

^†It is noted that while overall study methodology was rated “high,” reporting of their primary outcome is atypical when compared to other studies in this analysis.

Note. PA=physical activity; MVPA=moderate-to-vigorous intensity physical activity.

Table 2.

Race and ethnicity reporting.

First author, year	Black (N)	White (N)	Mixed race/multiracial (N)	Other race (N)	Hispanic/Latino ethnicity (N)	Other ethnicity (N)
Braam et al., 2018	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
Jackson et al., 2022	5	52	1	1	1	N/A
Pbert et al., 2016	22	70	15	4	35	N/A
Sepúlveda et al., 2020	Not reported	Not reported	Not reported	Not reported	Not reported	Latin American: 38 Spanish: 28 Other (i.e., Moroccan): 4
Skjåkødegård et al., 2022	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported

Study details on the trial and treatment components and study design are included in Table 3 and PA&Ex methods and measurement are listed in Table 4. The included studies were heterogeneous in multiple design components. Comparator groups varied and included treatment as usual (Braam et al., 2018; Skjåkødegård et al., 2022), exercise prescription and tracking (Jackson et al., 2022), and psychoeducation (Pbert et al., 2016). One study included an active intervention condition excluding the treatment component of interest (Sepúlveda et al., 2020), which may have suppressed the treatment effects of PA&Ex, as both treatment arms consisted of the same core set of CBT and PA strategies. Interventions frequently included CBT components such as structured psychoeducation, problem solving, coping skills, self-talk and thought challenging, and consideration of treatment barriers. PA and exercise programming included PA education, and specific and individualized exercise prescription. Only two studies (Braam et al., 2018; Pbert et al., 2016) consisted of active exercise adjuncts with separate times or sessions with planned exercises and activities; all other studies included passive recommendations and guidance. Only two studies (Jackson et al., 2022; Skjåkødegård, 2022) detailed the criteria used to preprocess the accelerometry data (e.g., valid wear time, intensity cut-points, data extraction algorithms) which are important for understanding wear-time and categorization of activity intensities.

Table 3.

Description of interventions.

First author, year	Intervention/comparator	Treatment length	Modality	Randomization procedure	PAE components	CBT components
Braam et al., 2018	Quality of Life in Motion (QLIM)/Usual care	12 weeks	In person—group	Block randomization occurred after completion of baseline measures. Strata included age, gender, cancer type, and treatment phase.	Two 45-min sessions per week Individualized aerobic exercise circuit Gradual increases in physical activity intensity and peak heart rate targets each month Home-based high-intensity exercise program Booster session performed 3- to 6-months postintervention	One 60-min session every 2 weeks Psychoeducation pertaining to: Expression of feelings Self-perception Coping skills Aimed to improve psychosocial function by targeting: HrQoL Self-perception Behavior problems Depressive symptoms Psychoeducation aimed to increase ability to cope with effects related to disease Participants given home exercises on session topics if deemed necessary Two psychosocial education sessions for parents
Jackson et al., 2022	Congenital Heart Disease Physical Activity Lifestyle (CHD-PALS)/Exercise prescription and Fitbit tracker	20 weeks	Telehealth—individual	Block randomization occurred using a 1:1 allocation stratified by number of minutes spent in MVPA during baseline assessment (<21 min/day or ≥ min/day). Blocks varied randomly by sizes of four and six.	Individualized exercise prescription Increases in physical activity duration and intensity at least twice Physical activity monitoring with Fitbit	Eight 30-min online sessions with health coach Education pertaining to positive impact of physical activity Discussing physical activity’s pros and cons Discussion on subjective physical activity norms Assessment of peer and family attitudes and perceptions of PA engagement Strategies to overcome perceived PA barriers Goal setting
Pbert et al., 2016	Lookin’ Good Feelin’ Good: school nurse intervention and after-school exercise program/12 informational visits with school nurse	8 months	In person—individual	Pair matched cluster randomization based on school enrollment numbers. Schools were given a randomly generated number with the school receiving a lower number assigned to the treatment condition.	Encouraged at minimum 1 hr of exercise most days Participants kept diet and PA log that was reviewed at each session Pedometer provided to support increasing physical activity Information provided on applying F.I.T.T (frequency, intensity, time, and type of activity) principles Encouraged reduction of sedentary behavior	Six 30-min weekly sessions Structured goal setting Assessment of progress toward goals Education on strategies for motivation Problem solving education based on Social Cognitive Theory Stopping negative thoughts Changing participants’ problem cues Strategies to handle stress and social events Sleep hygiene Support for healthy behavior change
Sepúlveda et al., 2020	ENTREN-F programme with family intervention/ENTREN programme without family intervention	6 months	In person—individual	Random assignment to either psychosocial family workshop programming or individual psychological workshop.	Physical activity and nutrition knowledge sessions Goal setting	Six 2-hr parent sessions Nine 2-hr child sessions Three 2-hr parent and child combined sessions Two-hour follow-up session at 6 months Psychoeducation Stages of behavior change education Barriers assessment Goal setting Motivational interviewing Nutrition knowledge Emotion and cognitions Behaviors and physical functioning Relaxation skills Cognitive styles and flexibility Identifying self-talk
Skjåkødegård et al., 2022	Family-Based Social Facilitation Treatment/Treatment as usual in outpatient clinic with nurse	17 weeks	In person—mixed individual and group	Randomization occurred in 1:1 ratio, with sequentially numbered nontransparent envelopes used to hide randomization scheme from study staff.	Physical activity guidance based on the Traffic Light Diet	Seventeen weekly sessions of unspecified length Utilized family-based behavioral treatment for obesity Engaging socioecological support systems Education on resolving external factors that impact eating habits Self-monitoring Goal setting Planning and reward systems Stimulus control Emotion regulation strategies Reframing negative automatic thoughts Communication and interpersonal skills Parenting strategies Healthy modeling from parents Healthy methods of self-evaluation and self-assertion

Note. CBT = cognitive-behavioral; HrQoL = health-related quality of life; MVPA = moderate-to-vigorous intensity physical activity; PA = physical activity.

Table 4.

Physical activity and exercise assessment and measurement.

First author, year	Posttreatment	Follow-up	Activity (physical activity or exercise)	PA measurement device	PAE algorithm	PA outcome	Days included	Minimum wear time
Braam et al., 2018	4 months	12 months	Exercise	Actical Accelerometer, B Series	Not specified	Mean counts per minute	4 consecutive days (3 weekday and 1 weekend)	Not specified
Jackson et al., 2022	∼22 weeks	N/A	Physical activity	Actigraph Model wGT3X-BT	Choi and Evenson	Change in MVPA and sedentary behavior	4 days including at least one weekend day	8 hr/day
Pbert et al., 2016	8 months	N/A	Exercise	Actigraph Model GT1M	Trost; Not specified	Percent of time spent in MVPA	7 days	Not specified
Sepúlveda et al., 2020	6 months	12 months	Physical activity	Actigraph model GT3X	Not specified	Minutes per day of sedentary, light, moderate, and vigorous activities	7 consecutive days	Not specified
Skjåkødegård et al., 2022	6 months	12 months	Physical activity	Actiwatch 2	Ekblom, Nyberg, Ekelund, and Marcus	Percent of time spend in MVPA	4 days	10 hr/day

Note. PA = physical activity; MVPA = moderate-to-vigorous intensity physical activity.

RoB and study quality

On the RoB, four studies were judged to be at high risk and one was judged to be of some concern (see Figure 2). When considering subcategories of bias, the most common areas of concern pertained to domain 3 (missing outcome data). Specifically, in these studies, missing data were not evaluated for bias via sensitivity analysis to evaluate potential baseline differences between those who completed posttreatment assessments vs. those who did not. Additional areas of concern were in domain 5 (selection of the reported outcome) and usually occurred in studies with multiple time-point measurements. Per RoB 2 guidelines, studies were automatically considered high risk if they included follow-up data (as multiple measurement points) as part of their primary trial reports, or included multiple subdomains of PA (e.g., different intensity levels) without specifying a primary outcome. A rating of high risk in any category automatically classifies a study as high risk for overall bias; given that PA and exercise studies should include multiple levels of PA&Ex as part of standard reporting, they may be particularly prone to high bias ratings. For this study, if authors clearly stated the type of PA used for their primary outcome (e.g., MVPA, counts), they were not considered to have multiple assessments of this variable. No adverse events were reported among the studies included in this analysis; however, only two studies (Braam et al., 2018; Jackson et al., 2022) specifically noted a lack of adverse events. It is unclear whether the lack of adverse events reported in the other studies was due to the absence of events or a lack of monitoring for adverse events. Overall, studies in this analysis were of moderate quality; however, this analysis was unable to assess publication bias, which could affect ratings in future meta-analyses.

Figure 2.

Risk of bias.

Analysis of pooled effect sizes

Our protocol initially proposed the evaluation of PA&Ex intensities (e.g., light, moderate, vigorous) across studies; however, we were not able to evaluate PA&Ex intensities due to sparce data reporting. Only one study provided sufficient data to calculate the effect size of each intensity (Sepúlveda et al., 2020). Summative measures of PA&Ex (e.g., minutes of MVPA, counts, percent MVPA) were reported in eight eligible studies, with five of these studies reporting sufficient, interpretable data to be included in the meta-analysis. Subsequently, MVPA was the primary outcome for this meta-analysis. Using random effect meta-analysis, the combined estimated effect was 0.10 (95% CI −0.16, 0.35) (online supplementary material 4, Figure S1). This is a small, statistically nonsignificant effect favoring the intervention groups (i.e., participants in the intervention groups increased their engagement in PA&Ex more than comparator groups); however, given the nonsignificance of the effect, the conclusion of any group difference is questionable. Given the small number of studies included in the analyses, this body of evidence should be considered with caution. Effect sizes by study are shown via Forest plots in online supplementary material 4.

Exploratory analyses were conducted to evaluate whether the intensity of the comparator group may have suppressed the overall combined estimated effect. Two studies compared a treatment to a treatment-as-usual control (Braam et al., 2018; Skjåkødegård et al., 2022), the effect of treatment was 0.23 (95% CI −0.27, 0.74) (online supplementary material 4, Figure S2), favoring active treatment but not significantly so. Three studies (Jackson et al., 2022; Pbert et al., 2016; Sepúlveda et al., 2020) included active treatment comparators, either as psychoeducation based (Pbert et al., 2016), enhanced usual care (Jackson et al., 2022), or another active treatment (Sepúlveda et al., 2020). The effect of treatment was 0.05 (95% CI −0.24, 0.35) (online supplementary material 4, Figure S3), nearly a null effect.

Moderator effects of PA or exercise type

Four studies conducted PA-focused interventions (Braam et al., 2018; Jackson et al., 2022; Sepúlveda et al., 2020; Skjåkødegård et al., 2022), with three reporting sufficient, interpretable data to be included in a meta-analysis (see online supplementary material 4, Figure S5). Models indicate a small, null effect favoring the intervention (0.15, 95% CI −0.23, 0.54); this should be considered inconclusive and interpreted with caution given the small number of studies included (Pigott, 2012).

Two studies conducted exercise interventions (Braam et al., 2018; Pbert et al., 2016) with both providing sufficient information to be included in the meta-analysis (online supplementary material 4, Figure S6). The summary effect was very small and not statistically significant (0.06, 95% CI −0.33, 0.45) and should be considered inconclusive given the low number of studies.

Exploratory analyses

In accordance with our protocol, additional analyses on the potential role of chronic disease (overweight and obesity vs. other conditions) and modality of administration (one-on-one) are presented in online supplementary material 4 (Figures S7–S9). In short, while limited by previously noted small sample sizes, the efficacy of these combined interventions did not differ by chronic disease or modality of administration. In summary, estimates of treatment efficacy should be considered null and inconclusive due to the low number of studies.

Discussion

The current meta-analysis found that CBT+PA&Ex interventions differed a small, nonsignificant amount from their comparator groups on summative measures of PA&Ex. Clinically, in context of the broader literature, this would suggest that while both CBT (Moore et al., 2019; Morey & Loades, 2021) and PA&Ex (Anderson & Durstine, 2019; Riner & Sellhorst, 2013) are individually effective intervention components for improving PA&Ex as part of treatment programs for chronic disease management, there was not sufficient evidence to support an additive effect of PA&Ex components to CBT interventions in children and adolescents with chronic disease when examining PA&Ex as an outcome. This would suggest that CBT interventions that include PA&Ex components do not yet demonstrate a clear benefit for improving PA&Ex; however, we draw this conclusion with extreme caution for multiple reasons.

Considerations for methods and data reporting

This systematic review and meta-analysis highlights shortcomings in how PA interventions and outcomes are described and reported. For studies without active PA&Ex, it is unknown whether PA&Ex strategies were implemented (Sepúlveda et al., 2020; Skjåkødegård et al., 2022). Also, given that just two studies utilized active exercise (Braam et al., 2018; Pbert et al., 2016), our results may be confounded by whether the PA&Ex intervention was passive or active. However, this would be consistent with the definitions of PA and exercise, in which exercise interventions are defined as being more planned, purposeful, and structured (Caspersen et al., 1985). Further, data preprocessing (i.e., the methods and formulas used to prepare raw data for analysis) and PA analysis methods were generally under reported. Without the inclusion of preprocessing decisions and algorithms, the ability to judge the rigor of the science and generalize findings may be limited (Herrmann et al., 2014; Troiano et al., 2014). Specifically, algorithms used for determining activity intensity levels should also be included (Migueles et al., 2019; Troiano et al., 2014; Trost et al., 2011). This information allows other researchers to evaluate not only the appropriateness of the methods for the study sample, but also can inform how the data are pooled and used in future meta-analyses and reviews. Further, team science should include PA experts who are integrated into the early development of the studies to ensure that PA constructs are measured appropriately, and that data are being processed, analyzed, and reported correctly (Bennett & Gadlin, 2012). The lack of specificity in some of these studies may be due to the use of PA&Ex as an intervention component, without the intention of studying it as a health mechanism. This is problematic for behavioral clinical trials in that it limits the ability of these studies to evaluate whether the intervention produces a clinically significant benefit on the behavioral risk factor, a key component of contemporary behavioral clinical trials (Powell et al., 2020). Suggestions for improved reporting practices of PA&Ex methods and outcomes are summarized in Table 5.

Table 5.

Summary of physical activity and exercise reporting suggestions.

Recommendation	Rationale
Intervention-related recommendations
Clear description of activity components and type (PA vs. Ex)	Will allow for clearer understanding of research team intention with the PA&E activities and classification of intervention type.
Reporting recommendations
Primary PA&E level or intensity (e.g., sedentary, MVPA)	Specification of the primary variable of focus, especially if done in an a priori registered protocol, may alleviate potential multiple measurement concerns for risk of bias.
Specify units of measure (e.g., average minutes per day, % MVPA)	Improved data aggregation; Aiding in interpretation of results and clinical meaningfulness.
Report Mean, SD, and sample size for PA&E outcomes for all available time points	Provide a more complete picture of PA&E changes over time; aid in future meta-analyses.
Device-based procedure reporting recommendations
Measurement Device(s) and wear-time inclusion requirements (e.g., number of days, hours per day, inclusion of weekend days)	Improved data aggregation; Potential indicator of data quality and completeness, and subsequent generalizability.
Algorithms for cut-point estimates and epoch selection	Evaluation of appropriateness of device-based methods and subsequent quality of the results
Data cleaning and processing procedures	Aid in re-producibility of methods in future studies and replication of results;
Report additional PA&E outcomes, outside of primary aim (e.g., other intensity levels: sedentary, light, moderate, vigorous, very vigorous)	Aid in future data abstraction for meta-analyses and provide a more complete picture of the physical activity profile of participants in the study.

Note. PA = physical activity; Ex = exercise; PA&E = physical activity and/or exercise; MVPA = moderate-to-vigorous intensity physical activity.

Limitations of the evidence and review process used

While the included studies were of moderate quality, they are generally of high RoB for different reasons. This raises questions about the validity of the current study’s results. Several factors related to the review process and extracted data may have contributed to the small number of included studies and heterogenous effects, limiting power for a priori analyses (Valentine et al., 2010) such as: (a) inclusion of RCTs published as of July 1, 2023, where more recent publications were not included; (b) prioritizing device-based measures of PA&Ex; (c) inclusion of such varied comparator groups, which likely affect treatment effect sizes (Freedland et al., 2019); (d) unintended age range restrictions via the focus on CBT studies; and (e) an unequal representation of obesity/overweight studies. Other pediatric chronic disease populations may respond differently to CBT+PA&Ex interventions. Also, given the focus on combined CBT interventions, this study may have missed interventions directed at younger children that did not include cognitive components, which may not be appropriate for young children (Grave & Blissett, 2004).

Conclusions and future considerations

Few studies met inclusion criteria for full-text analysis, suggesting that rigorous, randomized study designs utilizing objective PA&Ex and CBT are lacking. It is therefore difficult to isolate the independent effects of PA&Ex to CBT from these studies, and few definitive conclusions can be stated, limiting comparisons that can be made to the broader literature. Under these considerations of caution, there did not appear to be a clear effect for CBT+PAE interventions. Future reviews may consider broadening the range of interventions, including non-RCT studies, considering additional types of psychosocial variables (e.g., depressive symptoms, anxiety), or narrowing the comparator search to active treatments that include CBT components. Such a review would provide a more complete picture of both the overall effect of both CBT+PA and CBT+Ex compared to CBT interventions alone. It would also be helpful to see how findings from pediatric interventions compared to adult interventions; an adult-focused review in this area may be warranted. Future intervention studies should also consider including more complete reporting of PA and exercise outcomes via supplementary tables or appendices. This would allow authors to focus on aspects of their PA data that are most relevant to the clinical population and intervention, while enabling data to be included in future meta-analyses.

Supplementary material

Supplementary material is available online at Journal of Pediatric Psychology (https://academic.oup.com/jpepsy/).

Data availability

There are no original data to share as part of this study.

Author contributions

William R. Black (Conceptualization [lead], Data curation [equal], Investigation [lead], Methodology [supporting], Project administration [lead], Supervision [lead], Writing—original draft [lead], Writing—review & editing [equal]), Lauren von Klinggraeff (Conceptualization [equal], Formal analysis [lead], Methodology [lead], Writing—original draft [supporting], Writing—review & editing [supporting]), Dave A. White (Conceptualization [supporting], Methodology [equal], Writing—original draft [equal], Writing—review & editing [equal]), Bethany Forseth (Data curation [equal], Investigation [supporting], Writing—original draft [supporting], Writing—review & editing [equal]), Jamie L. Jackson (Conceptualization [supporting], Investigation [supporting], Project administration [supporting], Supervision [supporting], Writing—review & editing [equal]), Carolyn R. Bates (Conceptualization [supporting], Data curation [supporting], Investigation [supporting], Writing—review & editing [supporting]), Christopher D. Pfledderer (Conceptualization [supporting], Data curation [supporting], Formal analysis [supporting], Methodology [equal], Supervision [equal], Writing—review & editing [supporting]), Sidney Dobbins (Data curation [equal], Project administration [equal], Writing—review & editing [equal]), Kristen R. Hoskinson (Formal analysis [supporting], Investigation [supporting], Methodology [supporting], Writing—review & editing [equal]), Alison Gehred (Data curation [lead], Methodology [supporting], Supervision [supporting]), and Ann McGrath Davis (Conceptualization [supporting], Project administration [supporting], Supervision [supporting], Writing—review & editing [supporting])

Funding

None declared.

Conflicts of interest: None declared.