The Healthy Homes/Healthy Kids 5‐10 Obesity Prevention Trial: 12 and 24‐month outcomes

Abstract

Background:

Pediatric primary care is an important setting for addressing obesity prevention.

Objective:

The Healthy Homes/Healthy Kids 5‐10 randomized controlled trial evaluated the efficacy of an obesity prevention intervention integrating pediatric primary care provider counseling and parent‐targeted phone coaching.

Methods:

Children aged 5 to 10 years with a BMI between the 70th and 95th percentile and their parents were recruited from pediatric primary care clinics. Participants received well‐child visit provider counseling about obesity and safety/injury prevention and were then randomized to a 14‐session phone‐based obesity prevention (OP; n = 212) or safety and injury prevention contact control (CC; n = 209) intervention. The primary outcome was 12 and 24‐month child BMI percentile.

Results:

There was no overall significant treatment effect on child BMI percentile. Caloric intake was significantly lower among OP compared with CC participants at 12 months (P < .005). In planned subgroup analyses, OP condition girls had significantly lower BMI percentile (P < .05) and BMI z‐score (P < .02) at 12 and 24 months relative to CC girls and were less likely to be overweight (38.0% vs 53.0%, P < .01) or (obese 3.4% vs 8.8%, P < .10) at follow‐up.

Conclusions and Relevance:

An obesity prevention intervention integrating brief provider counseling and parent‐targeted phone counseling did not impact 12 and 24‐month BMI status overall but did have a significant impact on BMI in girls.

1 |. INTRODUCTION

Childhood obesity prevention (OP) is a public health priority.¹ Pediatric primary care is an important setting for OP because a large majority of families have contact with care providers who are important, trusted sources of information about health risks and prevention behaviors.² Despite potential for high impact, implementation of recommendations for routine primary care obesity screening and counseling has been suboptimal.^3–7 Barriers include concerns about effectiveness and perceived reluctance of parents to discuss weight‐related issues.^4,5 Conversations about obesity and its associated risks can be difficult to engage in with parents, and providers have often had limited training and experience addressing these issues. Moreover, time constraints in busy pediatric practice settings are a significant barrier.

A growing body of literature describes primary care‐based obesity interventions that aim to leverage the influence of pediatric care providers and overcome previously identified barriers. These studies support the feasibility of primary care‐based interventions but provide inconsistent evidence for intervention efficacy, with conclusions limited by heterogeneity with respect to intervention modality and study methodology.^6,7 Moreover, a limited number have focused on prevention,^8–21 few have targeted elementary school-aged children,^10,12,13 and the majority have been pilot studies or nonrandomized trials.^11–14,20

Emerging evidence suggests that motivational interviewing (MI) is a promising approach for addressing childhood obesity in primary care settings.^22–25 The optimal strategies for augmenting pediatric care provider counseling during well‐child visits, however, have yet to be identified. The largest pediatric primary care‐based obesity intervention trial to date, the Brief Motivational Interviewing to Reduce Body Mass Index (BMI²) study, was a cluster‐randomized trial evaluating the efficacy of two MI interventions varying in intensity relative to usual care (Group 1).²⁵ One of the MI interventions included up to four MI sessions with the pediatric primary care provider (Group 2) and the other included the Group 2 provider intervention and up to six MI sessions with a registered dietician (RD) with only one of these sessions required to be in‐person (Group 3). Results showed a significant reduction in BMI percentile among Group 3 relative to Group 1 index children. The authors note, however, that a major limitation was the low rate of session completion by the RDs and recommend further exploration of utilizing telephone‐based counseling as an adjunct to pediatric primary care‐based obesity interventions.

The use of brief in‐office provider counseling followed by phone sessions with a health coach is a promising approach for facilitating high levels of participation in the context of busy family and work schedules. The Healthy Homes/Healthy Kids (HHHK 5‐10) study²⁶ was a randomized clinical trial evaluating the efficacy of a primary care‐based OP intervention aimed at 5 to 10‐year old children who were at risk for obesity defined as between the 70^th and the 95^th percentile for gender and age on the CDC Growth Charts. The intervention combined pediatric primary care provider counseling during a well‐child visit and phone coaching sessions with parents to support them in making home environmental changes to promote healthy eating, activity patterns, and child weight status. The theoretical and clinical underpinnings of the intervention used a combination of social cognitive theory (SCT), which posits that behavior is determined jointly by knowledge, attitudes, behavioral skills, and environmental factors that facilitate implementing those skills,^27,28 MI,^29–33 which emphasizes the importance of participant self‐determination and direction setting in the change process, and goal setting adherence enhancement strategies.^34–36 Furthermore, our approach incorporated the need for flexibility and adaptation of intervention strategies given anticipated changes in children’s development and context over the intervention period and recognition of the complexity and multiple environmental contexts (eg, community‐level, built environment) that influence the development and maintenance of obesity.^37,38 This paper describes the primary and secondary 12 and 24 month outcomes of the HHHK 5‐10 trial.

2 |. METHODS

2.1 |. Study design and participants

The HHHK 5‐10 study was a two‐arm individually randomized clinical trial. The study protocol was reported previously.²⁶ The study was approved by the HealthPartners Institutional Review Board. The intervention period was 12 months. Outcomes were assessed during home visits at baseline, 12 months, and 24 months. Participants were enrolled in the study between May 2010 and September 2012 from 20 pediatric primary care clinics in the greater Minneapolis‐St. Paul area. Follow‐up assessments were completed in November 2014.

Eligibility criteria were (1) 5 to 10‐year‐old child attending a well‐child visit conducted by a pediatric or family practice care provider; (2) child at risk for obesity defined as BMI between the 70th and 95th percentile for age and gender on the CDC growth charts; (3) no medical problems that would preclude study participation (eg, a chromosomal abnormality, kidney disease, Type I diabetes, lupus, or cancer); (4) no steroid medication use more than 1 month; and (5) child not participating in another health‐related research study.

For recruitment, the electronic medical record was queried to identify age and BMI‐eligible children with upcoming well‐child visits. After review by study staff and the primary care provider, an invitation letter was sent to the parents of the child with the scheduled well‐child visit. Study staff conducted follow‐up phone calls to assess interest and conduct a brief screening with parents/primary caregivers who were interested in participating. A home visit was scheduled, and child height and weight were measured to confirm eligibility. Written parental/primary caregiver informed consent and child assent were then obtained and baseline measures administered by study staff. The parent/primary caregiver who signed the informed consent and completed the baseline measures then participated in the phone coaching calls and subsequent measurement visits.

2.2 |. Randomization

After baseline measures and well‐child visit completion, participants were randomized in to treatment group using a 1:1 randomization schedule in blocks or sets of 10 to ensure research staff could not influence randomization by adjusting enrollment order.

2.3 |. Intervention

The HHHK 5‐10 study included an OP intervention and a contact control (CC) intervention focused on general health, safety, and injury prevention. Social cognitive theory (SCT)^27,28 provided the theoretical basis for both interventions, and the interventions were also informed by the literature on the participant‐centered MI counseling style.^29,31–33

All families received tailored guidance from their primary care provider regarding the child’s BMI percentile and obesity and prevention topics. Study staff provided an HHHK flipchart to facilitate intervention delivery and HHHK pamphlets which included relevant anticipatory guidance. Following the well‐child visit, families were randomized to the OP or CC condition. Families in both conditions received a treatment group‐specific workbook, six biweekly phone coaching calls from a trained phone coach over the first 3 months, and eight monthly phone coaching calls during the remainder of their first year of the study. OP arm behavioral target areas based on pediatric obesity guidelines³⁹ included limiting sugar‐sweetened beverage consumption, encouraging fruit and vegetable consumption, limiting television and other screen time, eating breakfast daily, limiting restaurant eating, encouraging family meals, and limiting portion size. The CC intervention focused on home safety and injury prevention, fire safety, bicycle safety, and sun protection. Phone session content for each condition included parent self‐assessment and behavior change prioritization, goal setting, and adherence strategies. During the first call, the phone coach focused on establishing rapport with the parent/caregiver, briefly reviewed each behavioral target, and began the process of goal setting by eliciting the parent/caregiver’s assessment of their family’s current status and interest in making changes in each of the behavioral target areas. During the second and subsequent calls, the phone coach worked with the parent to discuss progress on goals set, problem solve with the parent to overcome barriers to goal adherence, and identify new goals as appropriate. This approach allowed us to tailor the content of the phone calls to accommodate family differences with respect to the number and type of relevant problem areas and parent willingness to work on a goal in a particular area. To accommodate these variations and enhance parent motivation to make changes, we worked with parents to identify the problem areas and goals most relevant to them, while also ensuring that we address the core household recommendations over the course of the phone coaching calls.

2.4 |. Measures

2.4.1 |. Child body mass index (BMI) percentile and Z‐score

Child weight and height were measured by trained and certified data collectors using a Seca 876 flat scale and Seca 217 stadiometer (Seca Corp., Hanover, MD). Weight and height were measured twice with a third measure taken if the first two measurements differed by more than 0.2 kg for weight or more than 1.0 cm for height. Measurements were averaged, and BMI percentile and z‐score were computed using the 2000 CDC growth chart data.^40,41 A second data collector performed height and weight measurements for 8% of the participant visits across the three time points (baseline, 12 months, and 24 months). The interrater reliability is excellent, ICC > 0.99, among all participants who were selected to be measured twice, as well as within each of the three time points.

2.4.2 |. Dietary intake, physical activity, and sedentary behavior

Diet Recall

A single multi‐pass 24‐hour recall was administered with parent/child dyads by staff certified to use the Nutrition Data System for Research software versions 2009 to 2011 (NDSR, Nutrition Coordinating Center, University of Minnesota).⁴² Recalls were analyzed using NDSR version 2013 software to estimate total energy intake, percent calories from fat, and fruits and vegetables and sugar‐sweetened beverage servings.

2.4.3 |. Accelerometry

Physical activity was measured using ActiGraph GT1M accelerometers (ActiGraph LLC, Pensacola, FL). Children were asked to wear the accelerometers for 7 days during waking hours, except during water activities. At least four valid monitoring days, defined as eight or more hours of wear time, were required for inclusion in analyses. To estimate minutes spent in moderate‐to‐vigorous physical activity (MVPA), data were aggregated into 1‐minute epochs; cut points were defined using the Evenson et al (2008) equations.⁴³

2.4.4 |. Screen time

Time spent viewing television and media was assessed with four items.⁴⁴ Parents reported the amount of time their child watched TV and used other media (video games, computer games, or a computer for something other than school work) on an average weekday and weekend day. Item response options were 0 hours (coded as 0), less than 1 hour (0.5), 1 hour (1), 2 hours (2), 3 hours (3), 4 hours (4), and 5 or more hours (5) per day. Separate estimates of the time per day spent watching TV and playing video games/using a computer were calculated as the sum of the weekday estimate multiplied by 5 and the weekend day estimate multiplied by 2, divided by 7. Total screen time per day was estimated as the sum of average daily TV time and video game/computer time.

2.4.5 |. Demographic characteristics

Parent‐reported demographic characteristics included child sex, ethnicity, and race and their own age, sex, ethnicity, race, marital status, employment status, and educational attainment.

Treatment fidelity and satisfaction

Well‐child visit content received was assessed via phone survey with parents 1 to 2 weeks post well‐child visit. Parents reported whether their provider talked about BMI percentile, whether they received the HHHK pamphlet, and whether the provider addressed physical activity, sedentary behavior, healthy eating, and safety/injury prevention issues.

Phone coaches completed post‐session questionnaires assessing session length and percent of session time spent on behavioral target areas (0%, 1%‐10%, 10%‐25%, 25%‐50%, and 50%‐100%). The percent of session time spent on each target area was rated for all sessions except sessions 1 and 6, during which the phone coach and parent completed an assessment covering all target areas. To ascertain the number and percent of parents who spent time on each behavioral target area at least once during the course of their phone coaching sessions, we categorized the percent of session time spent on each behavioral target area (0% versus 1%‐10%, 10%‐25%, 25%‐50%, and 50%‐100%). To estimate the minutes spent discussing behavioral targets during each session, the midpoint of each response choice was multiplied by the session length. These estimates were summed across sessions to calculate the total time spent on each target area. The percent of intervention time spent discussing each behavioral target was calculated as the sum of the minutes spent on each target area across sessions divided by the sum of the length of each session. Parent satisfaction and perceived helpfulness of the intervention were assessed via 12‐month survey items.

2.4.6 |. Analysis plan and sample size justification

The primary analysis evaluated the efficacy of the OP intervention relative to the CC intervention in preventing unhealthy child weight gain. A general linear mixed model predicted BMI percentile (BMI%) from randomized treatment group, the time at which BMI% was observed, and the treatment by time interaction using all available BMI% values from all randomized children (intent‐to‐treat approach). Treatment group (CC, OP) and time (baseline, 12 months, 24 months) were treated as fixed between and within subjects effects, respectively. A random intercept was estimated for each child to account for the statistical dependency among repeated BMI% measures within children. The same analytic approach was used for the secondary outcome measures. Generalized linear mixed models were used to predict daily fruit and vegetable servings (log link, Poisson distribution) and the consumption of any sugar‐sweetened beverages (logit link, binomial distribution). The analytic approach was modified for two outcomes that denoted whether children were overweight (BMI% ≥ 85) or obese (BMI% ≥ 95) at the 12 and 24‐month follow‐ups. In these models, 12 and 24‐month overweight or obesity status (logit link, binomial distribution) was predicted from fixed between subjects effects for treatment group and mean‐centered baseline BMI% in addition to the estimated random intercept. A planned contrast tested whether 12 and 24‐month observations were significantly different among OP relative to CC children, accounting for their baseline difference.

Secondary planned analyses assessed whether gender or age modified the treatment by time effect. Parameters quantifying the main and interactive effects of one modifier were added to the primary analytic model. The planned contrast was estimated within time and age or gender and used to describe patterns of treatment heterogeneity for three‐way interactions that approached statistical significance (P < 0.10).

An a priori power analysis evaluated the likelihood that randomizing N = 400 children equally to OP and CC would be sufficient to detect a clinically meaningful 3% to 5% between groups difference in 12 and 24‐month BMI% under a range of assumptions about the likely design effect magnitude. The intraclass correlation of BMI% values within children was assumed to be large (ICC = .60‐.70). Anticipated retention was 80% and based on pilot data we estimated SD_BMI% ≈ 6‐10. The range in the minimum detectable standardized effect for the planned contrast was Cohen’s d ≈ 0.56‐0.64 when SD = 6‐8, which corresponds to a difference of BMI% = 3.3‐5.1. Baseline BMI% descriptive statistics were consistent with our assumptions (ICC = 0.62 and SD = 6.9). Thus, the primary analyses were sufficiently powered to detect a clinically meaningful BMI % difference.

3 |. RESULTS

3.1 |. Baseline characteristics and retention

Baseline characteristics are presented in Table 1. The mean age of children was 6.6 years (SD = 1.7), and 208 (49.4%) were girls. Most children were non‐Hispanic white (69.1%), and the average BMI percentile was 84.9 (SD = 6.9). The majority of participating parents were females, married or living as married, and employed full or part‐time. The average parent BMI was 28.5 (SD = 6.2). The children were similar across treatment groups on most measured characteristics although children of Hispanic ethnicity were over‐represented in the CC group.

TABLE 1.

Baseline characteristics, overall, and by treatment arm

	Total	Obesity Prevention	Safety/Injury Prevention
N randomized	421	212	209
Child participants
Female, n(%)	208 (49.4)	101 (47.6)	107 (51.2)
Age in years, M (SD)	6.6 (1.7)	6.6 (1.6)	6.6 (1.7)
Non-Hispanic White, n(%)	289 (69.1)	154 (73.0)	135 (65.2)
Hispanic, n(%)	29 (6.9)	8 (3.8)	21 (10.1)
BMI a percentile, M (SD)	84.9 (6.9)	84.7 (6.9)	85.0 (7.0)
BMI a 70–84th percentile, n(%)	206 (48.9)	105 (49.5)	101 (48.3)
BMI a 85–95th percentile, n(%)	215 (51.1)	107 (50.5)	108 (51.7)
BMI a in kg/m2, M (SD)	17.8 (1.3)	17.8 (1.3)	17.9 (1.4)
BMI a z-score, M (SD)	1.08 (0.31)	1.07 (0.31)	1.09 (0.32)
Parent/primary caregiver participants
Female, n(%)	391 (92.9)	198 (93.4)	193 (92.3)
Age in yrs, M (SD)	37.5 (6.5)	37.7 (6.3)	37.3 (6.8)
Non-Hispanic White, n(%)	330 (79.0)	173 (82.0)	157 (75.9)
Hispanic, n(%)	15 (3.6)	3 (1.4)	12 (5.8)
BMI a in kg/m2, M (SD)	28.6 (6.3)	28.8 (6.4)	28.4 (6.1)
College or graduate degree, n(%)	299 (71.5)	155 (73.5)	144 (70.0)
Married/living as married, n(%)	338 (80.9)	179 (84.8)	159 (76.8)
Full or part-time employment, n(%)	350 (83.7)	172 (81.5)	178 (86.0)

Of the 421 children randomized, 363 (86.2%) were retained at 12 months and 367 (87.2%) at 24 months (Table 1). Retention rates did not differ by treatment arm.

3.2 |. Treatment fidelity

The majority of parents reported that their child’s provider discussed BMI% (85%) and healthy eating and physical activity topics (82%) at the well‐child visit. While 79% of parents reported receiving the HHHK pamphlet, only 42.0% of parents reported that their provider used the HHHK flipchart. More than half (56%) of parents reported that the well‐child visit conversation helped them think about changes to make at home.

As shown in Figure 1, nearly two‐thirds of parents in each group completed all 14 sessions. The average session length was 24.3 (SD 9.3) and 20.2 (SD = 8.7) minutes in the OP and CC groups, respectively. Eighty‐five percent of OP group parents reported that their coach helped them to make changes at home and 80% found the support from their coach was helpful.

FIGURE 1.

Healthy Homes/Healthy Kids Modified CONSORT Diagram

Table 2 presents information regarding the behavioral target areas discussed during OP sessions. The majority of parents discussed each behavioral target area at least one time. Examination of proportion of time spent discussing behavioral target areas across phone sessions showed that OP group parents spent an average of 48.7% (SD = 20.4) of their total time discussing diet‐related target areas, 26.9% (SD = 16.2) discussing physical activity‐related target areas, and 7.7% (SD = 11.0) of their time across sessions discussing limiting media use. The average number of minutes spent discussing diet, physical activity, and media use target areas was 111.2 (60.7), 60.4 (42.8), and 15.8 (20.4) minutes, respectively.

TABLE 2.

Descriptive information regarding behavioral target areas discussed during obesity prevention intervention phone sessions

	Percent of Parents Who Discussed Target Area At Least One Time % (N)	Minutes Spent Discussing Each Target Area Mean (SD)	Percent of Total Time Spent Discussing Each Target Area Mean (SD)
Keep healthy fruit and vegetable options around the house	97.4 (n = 187)	39.8 (34.4)	17.7% (13.4)
Limit salty/high fat snacks, sweets, and sugared drinks	88.5 (n = 170)	17.2 (16.7)	7.3% (6.9)
Eat family meals together	84.9 (n = 163)	20.5 (24.2)	9.4% (11.2)
Portion control, food intake awareness	70.8 (n = 136)	19.7 (25.2)	7.6% (9.0)
Eat a healthy breakfast	65.1 (n = 125)	7.1 (n = 11.0)	3.3% (5.2)
Limit eating out at restaurants	51.6 (n = 99)	7.0 (n = 13.6)	3.5% (7.0)
Be physically active together as a family and support your child’s activity level	92.7 (n = 178)	30.5 (n = 26.3)	13.9% (11.7)
Keep active play equipment around the house	76.6 (n = 147)	7.6 (n = 9.2)	3.6% (5.0)
Help your child make every day an active day	67.7 (n = 130)	22.3 (n = 33.4)	9.4% (12.7)
Limit media use	82.8 (n = 159)	15.8 (n = 20.4)	7.7% (11.0)
Be balanced in your parenting	50.5 (n = 97)	8.4 (13.0)	3.7% (6.4)
Maintaining healthy changes	74.5 (n = 143)	18.9 (n = 17.7)	7.5% (7.1)

3.3 |. Primary outcome

Table 3 displays descriptive information for primary and secondary outcomes at baseline, 12, and 24 months by treatment arm as well as the treatment by time P‐value and the treatment group effect at 12 and 24 months controlling for baseline. Mean BMI percentile (BMI%) values decreased significantly from baseline (M = 84.9, SD = 6.9) through12 (M = 81.9, SD = 10.6) and 24 months (M = 80.9, SD = 12.3), P < .001. Counter to expectations that BMI% would be 3% to 5% lower among OP relative to CC children at 12 and 24 months, the difference was only −0.48% (95% CI: −1.94%, 0.98%), and the treatment group by time effect was not statistically significant (P < 0.72). A similar pattern was observed for BMI-z score (P < 0.71).

TABLE 3.

Descriptive statistics, M (SD) or %, for study outcomes by treatment arm and time, type 3 treatment by time interaction P‐values, and predicted difference (95% CI) in average 12 and 24‐month outcomes among obesity prevention relative to contact control participants

	Baseline	12 Months	24 Months	Interaction P‐Valuea	12, 24 Difference (95% CI)b
N
Obesity prevention	212	181	180
Contact control	209	183	187
BMI percentile
Obesity prevention	84.7 (6.9)	81.4 (10.5)	80.6 (11.9)	0.72	−0.48 (−1.94, 0.98)
Contact control	85.0 (7.0)	82.3 (10.7)	81.3 (12.7)
BMI z‐score
Obesity prevention	1.07 (0.31)	0.97 (0.42)	0.96 (0.49)	0.71	−0.02 (−0.07, 0.04)
Contact control	1.09 (0.32)	1.01 (0.42)	0.98 (0.48)
Moderate to vigorous PA (minutes per day)
Obesity prevention	51.6 (30.2)	38.4 (19.7)	39.1 (21.4)	0.13	0.37 (−4.31, 5.04)
Contact control	52.4 (29.8)	41.5 (21.9)	36.0 (21.3)
Moderate to vigorous PA (percent of wear time)
Obesity prevention	6.8 (4.0)	5.0 (2.6)	5.0 (2.7)	0.13	0.08 (−0.53, 0.69)
Contact control	6.8 (4.0)	5.4 (2.8)	4.6 (2.7)
Total energy intake (kcals)
Obesity prevention	1787 (527)	1677 (423)	1890 (558)	0.003	−83 (−198, 32)
Contact control	1753 (595)	1836 (512)	1829 (539)
Calories from fat (percent)
Obesity prevention	30.5 (6.7)	30.4 (6.9)	30.7 (6.8)	0.31	−1.19 (−2.77, 0.39)
Contact control	29.3 (7.0)	30.5 (7.1)	30.5 (7.4)
Fruit and vegetable intake (servings per day)
Obesity prevention	2.8 (2.4)	3.0 (2.2)	2.6 (2.1)	0.32	0.02 (−0.42, 0.54)
Contact control	2.8 (2.5)	2.7 (2.0)	2.7 (2.4)
Sugar‐sweetened beverage intake (1+ serving per day)
Obesity prevention	56.0%	54.9%	61.9%	0.41	0.4 (−13.0, 11.1)
Contact control	60.0%	64.2%	61.9%
Screen time (hours per day)
Obesity prevention	2.3 (1.2)	2.1 (1.2)	2.3 (1.1)	0.61	0.07 (−0.10, 0.24)
Contact control	2.4 (1.3)	2.1 (1.2)	2.3 (1.2)

^aTime by Tx P‐value Type 3.

^bPlanned contrast value, CI.

As shown in Figure 2, planned analyses indicated that gender moderated the time by treatment effect on BMI% (p_{tx*time*gender} = 0.02) and BMI z‐score (p_{tx*time*gender} = 0.002). BMI% was lower at 12 and 24 months among girls (P < 0.05) but not boys (P = 0.28) in the OP compared with the CC group. BMI z‐score was significantly lower at 12 and 24 months among girls (P < 0.02) but not boys (P = 0.11) in the OP compared with the CC group. Similarly, gender moderated the likelihood that children in the OP compared with the CC would be overweight (P < 0.02) or obese (P < 0.03). Girls in the OP compared with the CC group were less likely to be overweight (P < 0.01) and marginally less likely to be obese (P < 0.10) at 12 and 24 months. In contrast, boys were similarly likely across treatment groups to be overweight (P = 0.52) or obese (P = 0.17).

FIGURE 2.

BMI percentile patterns by gender

3.4 |. Secondary outcomes

Minutes of MVPA declined from baseline (M = 52.0, SD = 30.6) to 12 months (M = 40.1, SD = 21.2) and was maintained at 24 months (M = 38.2, SD = 22.5). The change over time was statistically significant (P < 0.001), but the predicted treatment by time interaction was not (P < 0.13).

The treatment by time interaction for total energy intake was significant (P < .003). Total energy intake decreased from baseline to 12 months and then increased from 12 to 24 months among OP group participants. Among CC group participants, total energy intake increased from baseline to 12 months and was maintained at 24 months. The net effect of this pattern was that total energy intake was lower among OP than CC participants at 12 and 24 months although not statistically significant (−83; 95% CI: −198, 32). There were no time or treatment by time effects for child fruit and vegetable intake or sugar‐sweetened beverage intake.

Screen time decreased from baseline (M = 2.4, SD = 1.3) to 12 months (M = 2.1, SD = 1.2, P = 0.001) and returned to baseline values at 24 months (M = 2.3, SD = 1.2, P = 0.18 relative to baseline); these changes were not differential across treatment groups (interaction P = 0.61).

The moderator analyses for secondary outcomes showed that the time by treatment effect on total energy intake was marginally different for boys and girls (P < 0.10). Total energy intake was lower at 12 and 24 months among girls in the OP group than in the CC group (P < 0.02) but was not different by the treatment group among boys at 12 and 24 months (P = 0.63). Gender did not moderate the time by treatment effect on other secondary outcomes.

4 |. DISCUSSION

The HHHK 5‐10 study evaluated the efficacy of a pediatric primary care‐based OP intervention integrating brief pediatric primary care provider counseling and phone coaching to support parents in making home environment changes. Despite high treatment fidelity, the intervention did not significantly impact 12 or 24‐month BMI percentile, BMI z‐score, physical activity, or screen time. There was some evidence, however, that the intervention led to reductions in total energy intake at 12 months. Additionally, a planned subgroup analysis showed that the intervention was moderately effective among girls but not boys. Girls in the OP group had a lower BMI percentile, lower BMI z‐score, and were less likely to be in the overweight category relative to girls in the CC group at 12 and 24 months.

These results are consistent with previous literature indicating that OP trials, including those conducted in the context of pediatric primary care,⁷ have had a modest impact on child BMI and associated health behaviors.^45,46 Moreover, considerable response heterogeneity has been observed, with many studies finding significant results in some subgroups in the absence of overall treatment effects. For example, a number of obesity intervention trials have observed significant treatment effects among girls, but not boys^8,47; some have observed significant effects among children with a higher BMI,^48,49 and others have observed differential effects in children from different racial/ethnic backgrounds.^50–53

Although subgroup analysis results should be interpreted with caution,^54,55 the frequency with which subgroup differences in the childhood obesity intervention literature have been observed warrants further investigation. Obtaining a better understanding of response heterogeneity and increasing knowledge of what works and for whom could improve the efficacy of childhood obesity interventions. For example, several trials, including HHHK, have found treatment group effects in girls only, whereas no study that we are aware of has published results suggesting that only boys have benefited from intervention. Exploring potential reasons for these differential results could generate hypotheses regarding strategies to better tailor intervention strategies for different subgroups. The HHHK behavioral targets may have been more salient for parents of girls given that higher levels of weight concern are reported by parents of girls compared with boys.⁵⁶ Given that rates of obesity among boys are similar to those observed among girls, obtaining a better understanding of parent perceptions of obesity‐related behaviors among boys and motivation for making positive changes in dietary intake and activity patterns could inform intervention development. Most parent‐targeted obesity interventions enroll mothers as the primary parent with limited engagement of fathers or a second parent or primary caregiver.⁵⁷ Despite considerable research documenting the influence of father involvement on child development, surprisingly little is known about the role of fathers in the development of child obesity‐related behaviors and how the influence of fathers may vary according to the sex of the child. There is some evidence, for example, that maternal factors differentially impact boys and girls dieting and weight concern. Learning more about the perspective of a second caregiver, including fathers, could be informative and investigating strategies to engage a second primary caregiver could strengthen intervention effects overall, including for boys specifically.⁵⁷

Study results also raise questions about optimal intervention dose, content, and therapeutic approach. Perhaps, allowing parents the choice of goals for their children was not optimal given that parents may not have chosen behavioral targets that would most strongly impact their child’s growth. Parents frequently focused on fruit and vegetable intake, which only impacts BMI if more energy dense options are displaced.⁵⁸ Relatedly, parents spent minimal time discussing screen time, a strong obesity correlate,⁵⁹ and half as much time discussing physical activity compared with dietary intake topics, another area for improvement given the decline in MVPA observed in this and other studies.⁶⁰ Utilizing a more directive counseling approach could have led parents to work on behavioral targets most likely to influence child BMI. However, it is also possible that such an approach, in contrast to a participant‐centered MI approach, would have undermined parent engagement. Although there is some evidence that MI childhood obesity interventions are effective,^25,61 not all studies utilizing this approach have yielded significant results.^8,12,62 Variations in the quality of how MI is delivered⁶³ and the intervention strategies that accompany it may be contributing to this result heterogeneity. This warrants further study given that most published childhood obesity intervention trial results papers provided limited detail about intervention fidelity and process evaluation data.⁶⁴ Future research should explore optimizing behavioral targets and strategies to guide parents toward the most impactful areas of change.

Finally, study results raise questions about intervention delivery modality. Although phone delivery facilitated participation, engagement level could have been greater with an in‐person or videoconference intervention. Evidence suggests that telemedicine approaches facilitate attendance, but it is not yet clear that they can outperform in‐person approaches.⁶⁵ Mobile phone‐health interventions are also emerging as a promising strategy in the childhood obesity intervention literature.⁶⁶ Identifying strategies to enhance parent engagement across treatment modalities is a necessary, although not sufficient, step toward increasing the efficacy of childhood obesity interventions.

Study limitations include the relatively homogenous sample with respect to race/ethnicity and parent sex and the use of a single dietary recall. Despite these limitations, study implementation demonstrated the feasibility of conducting a large‐scale trial in a pediatric primary care setting which included obtaining buy‐in from care providers and clinic leadership. The study leveraged primary care setting strengths and provided follow‐up support without increasing burden on the primary care clinic.

Study results suggest that an intervention combining pediatric primary care provider counseling and parent‐targeted phone coaching is feasible and has some promise for preventing obesity among girls. Importantly, results point toward research directions to strengthen intervention efficacy, including optimizing intervention dose, guiding parents to address behaviors that will have the most impact, and exploring strategies to enhance effects among important subgroups, including boys.