Brief Report: Shared weight and dietary changes in parent-child dyads following family-based obesity treatment

Abstract

Objectives

The primary objective was to determine whether children and their participating parents undergoing family-based behavioral treatment (FBT) for obesity show similar dietary changes following treatment, and if so, whether these shared dietary changes explain the similarity in weight change within the parent-child dyad.

Methods

Data come from a randomized controlled trial of 148 parent-child dyads who completed FBT and were followed over a two-year maintenance phase. Energy-dense, nutrient-poor foods (‘RED’ foods) and fruit and vegetable intake were assessed across time.

Results

Maintenance of lower RED food intake following FBT predicted weight maintenance in children and in parents (ps<.01), and dietary and weight changes were correlated within parent-child dyads (ps<.01). Most interesting, the similarity in long-term weight maintenance between children and their parents was predicted by the similarity in long-term changes in RED food intake between children and their parents (p<.001).

Conclusions

These findings point to the important role of maintaining low energy-dense, nutrient-poor food intake for long-term weight maintenance in children and parents. Furthermore, these results suggest that the correlation between parent and child weight maintenance can be explained in part by similar long-term changes in energy-dense, nutrient-poor food intake.

Family-based behavioral treatment (FBT) is an effective childhood obesity treatment that targets the child and a participating parent (Epstein, Valoski, Wing, & McCurley, 1994). Several studies have shown that a child’s long-term weight loss in response to FBT is predicted by the participating parent’s long-term weight loss (Boutelle, Cafri, & Crow, 2012; Epstein, Paluch, Beecher, & Roemmich, 2008; Wrotniak, Epstein, Paluch, & Roemmich, 2004). This correlated weight response might reflect shared genetics underlying appetitive phenotypes (Epstein, Dearing, & Erbe, 2010). It could also reflect changes to the shared environment that influence both parent and child. For example, modeling positive health behavior and altering the home environment could encourage sustained changes in energy-balance behaviors in both parent and child (Wrotniak, Epstein, Paluch, & Roemmich, 2005).

There also might be shared behavioral changes within the parent-child dyad that serve as a proximate cause for the concordance in FBT response; however, no previous research has explored this possibility. If such shared behavioral changes exist, treatment might help parents and their children accomplish common goals related to behavior change rather than work on separate, distinct goals. Shared dietary changes (e.g., decreased energy-dense, nutrient-poor foods intake, and increased fruit and vegetable intake) are a likely candidate (Epstein et al., 2008). Parent and child dietary intake is correlated cross-sectionally (Wang, Beydoun, Li, Liu, & Moreno, 2011), but it is unknown whether long-term dietary changes in response to FBT are also correlated within the parent-child dyad. The current study measured children’s and participating parents’ relative weight and self-reported dietary intake at several time points over a two-year maintenance phase following FBT. It was predicted that long-term weight maintenance would be correlated between the child and participating parent, and that this correlation would be explained in part by similar self-reported dietary changes over the maintenance phase.

Methods

Participants

Participants were 148 parent-child dyads who completed FBT and were randomized to maintenance treatment (MT) conditions. Children were overweight or obese (body mass index [BMI]≥85^th percentile for age and sex (Kuczmarski et al., 2000); BMI z-score=2.20±0.30), aged 7–12 years (9.9±1.3), mostly female (n=102; 68.9%) and mostly White/non-Hispanic (n=102; 68.9%). Participating parents were primarily overweight or obese (BMI=34.8±6.2 kg/m²), aged 27–66 years (42.7±6.3), and mostly female (n=121; 81.8%). The average family fell into the mid- to-upper socioeconomic status (SES) range, based on parental occupation and level of education (Hollingshead, 1975). Children and parents provided written informed assent and consent, respectively. The study received Institutional Review Board approval.

Procedures and Measurement

Study details can be found elsewhere (reference removed for blind review), and additional information on the measures can be found in the Online Supplement. Briefly, after completing a 5-month course of FBT (20 weekly sessions), families were randomized to one of three 4-month maintenance conditions: two active MTs (both consisting of 16 weekly sessions, one focused on behavioral maintenance skills [n=50] and the other on social/self-perceptual factors [n=50]), or no further treatment (n=48). All participants were followed between months 9 (post-MT) and 29 (2 years post-FBT), during which time no further treatment was delivered.

Child and parent weight and height were measured by trained research assistants using a calibrated balance beam scale and stadiometer. Child BMI z-score was calculated using age- and sex-specific CDC normative data (Kuczmarski et al., 2000). For parents, BMI (kg/m²) was calculated. Children and parents’ dietary intake was measured via self-report in the form of four-day paper-based food diaries (including at least one weekend day) prior to each assessment. Food diaries have been shown to accurately capture children’s energy intake (Bandini et al., 2003). For an individual’s data to be included, diaries for all four days were required. Across time points, 20% of participants with weight data did not provide complete food diaries. Two dietary variables were derived from participants’ food diaries. To assess nutrient-poor, energy-dense foods, the number of RED food servings per day were quantified (based on the Traffic Light Diet (Epstein, 2003)), which include (A) having >5g of fat per serving, (B) cereal with ≥25% of calories from fat, or ≥30% of calories from sugar, and (C) non-nutritive foods of any kind, regardless of fat, sugar, or calories (e.g., sugar-free Jell-O, sugar-free gum, diet soda, oil-based salad dressing). Fruits could be dried, fresh, or frozen, but did not include fruit juices. Vegetables included all forms except vegetable chips or vegetable juices. Fruits and vegetables were combined into a single variable due to the low frequency of either separately. The numbers of RED foods and of fruits and vegetables were divided by the total foods and beverages consumed per day to express each as a proportion of total food/beverage intake. Research assistants, who were not interventionists in the study, were trained together by one of the coauthors (DSMV) to code the food diaries using concrete criteria from a food reference guide (also provided to study participants). Coders practiced on a set of diaries and to ensure inter-rater reliability, DSMV provided feedback and resolved discrepancies.

Statistical Analyses

Six latent growth curve models (LGCMs) were constructed to estimate change in relative weight, fruit and vegetable intake, and RED food intake during the maintenance phase, separately for parents and children. Additional information on the construction of the LGCMs and on statistical power for the main analyses can be found in the Online Supplement. Linear regression analyses used the slopes derived from the LGCMs for the following analyses. First, it was determined whether weight maintenance was predicted by changes in fruits and vegetables and in RED food, separately for parents and children. Second, it was determined whether change in a child’s BMI z-score was predicted by the participating parent’s change in BMI. Third, it was determined whether changes in a child’s dietary intake were predicted by the participating parent’s change in dietary intake. Finally, similarity scores were computed from the slope scores, which quantified how similar the weight and dietary changes were within a parent-child dyad by subtracting the parent-standardized slope from the child-standardized slope. Similarity scores close to zero indicate high similarity of change. These similarity scores were used to determine whether the similarity in dietary changes between children and parents predicted the similarity in weight change between children and parents. All regression models included MT condition, age, sex and minority status, and SES. Standardized estimates (β) are presented.

Results

Complete descriptive statistics for the primary study variables can be found in the Online Supplement. Briefly, at the beginning of the maintenance phase (month 5), mean±SD child BMI z-score was 1.99±0.38, and parent BMI was 32.8±5.9. The proportion of RED foods in the diet for children and parents was 0.30±0.15 and 0.23±0.13, respectively. The proportion of fruits and vegetables was 0.18±0.09 and 0.23±0.09, respectively. At the end of the maintenance phase (month 29), child BMI z-score was 2.03±0.46, and parent BMI was 33.8±6.1. The proportion of RED foods in the diet for children and parents was 0.43±0.17 and 0.35±0.15, respectively. The proportion of fruits and vegetables was 0.17±0.10 and 0.22±0.10, respectively.

The six regression models are summarized in Tables 1 and 2. Model 1 showed that long-term change in children’s self-reported RED food intake independently predicted children’s weight maintenance (p=.007), whereas long-term change in fruits and vegetables did not (p=.98). Similarly, model 2 showed that long-term change in parents’ self-reported RED food intake independently predicted parents’ long-term weight maintenance (p<.001), whereas long-term change in fruits and vegetables did not (p=.14). Models 3 to 5 revealed that weight and dietary changes during the maintenance phase were significantly correlated within the parent-child dyad. First, long-term maintenance of parent weight predicted long-term maintenance of child weight (p=.003, see Model 3). Second, change in parent RED food intake predicted change in child RED food intake (p<.001, see Model 4). Third, change in parent fruit and vegetable intake predicted change in child fruit and vegetable intake (p=.002, see Model 5). Finally, model 6 shows that the correlation in weight change between the child and participating parent could be explained in part by the correlation in RED food changes within the dyad (p<.001) but not by the correlation in fruit and vegetable changes within the dyad (p=.21). Sensitivity analyses suggested that shared method variance (i.e., the fact that parents completed a subset of children’s food diaries) was not the cause for the primary results (see Online Supplement for details).

Table 1.

Summary of regression analyses predicting children’s BMI z-score maintenance by their dietary changes (Model 1), parents’ BMI maintenance by their dietary changes (Model 2), and children’s BMI z-score maintenance by parents’ BMI maintenance (Model 3)

	Model 1: Predicting Children’s BMI z-score Maintenance		Model 2: Predicting Parent’s BMI Maintenance		Model 3: Predicting Children’s BMI z-score Maintenance
	β	P value	β	P value	β	P value
Step 1 (Covariates only)
Participant sex (1 = Female)	.03	.727	−.06	.530	.03	.727
Participant race (1 = non-white)	−.04	.656	−.08	.373	−.04	.656
Participant Age	−.04	.649	−.03	.747	−.04	.649
Family SES	−.08	.321	.03	.705	−.08	.321
BSM vs. Control	−.17	.075	.03	.799	−.17	.075
SFM vs. Control	−.22	.023	.01	.945	−.22	.023
Step 2 (Predictors of interest added)
Participant RED food slope	.25	.007	.32	<.001	--	--
Participant FV slope	.003	.978	−.14	.142	--	--
Parent BMI maintenance slope	--	--	--	--	.24	.003
Change in explained variance for step 2	ΔR2 = .05		ΔR2 = .13		ΔR2 = .06
Total explained variance for model	R2 = .10		R2 = .15		R2 = .11

Note: BSM = behavioral skills maintenance treatment. FBT = family-based treatment. FV = fruits and vegetables. SES = socio-economic status. SFM = social facilitation skills maintenance treatment.

Table 2.

Summary of regression analyses predicting children’s RED food maintenance (Model 4), children’s fruit and vegetable maintenance (Model 5), and the similarity in parent and child weight maintenance (Model 6)

	Model 4: Predicting Children’s RED Food Maintenance		Model 5: Predicting Children’s FV Maintenance		Model 6: Predicting Similarity in Weight Maintenance Within Parent-Child Dyad
	β	P value	β	P value	β	P value
Step 1 (Covariates only)
Child sex (1 = Female)	−.13	.105	−.15	.059	.12	.140
Child race (1 = non-white)	.24	.003	−.001	.993	−.05	.581
Child Age	.05	.578	−.18	.025	−.001	.992
Family SES	−.13	.120	−.03	.683	−.10	.232
BSM vs. Control	−.03	.712	.39	<.001	−.17	.076
SFM vs. Control	.02	.862	.11	.211	−.18	.059
Step 2 (Predictors of interest added)
Parent RED food slope	.47	<.001	--	--	--	--
Parent FV slope	--	--	.26	.002	--	--
Similarity in RED food slopes	--	--	--	--	.40	<.001
Similarity in FV slopes	--	--	--	--	−.11	.214
Change in explained variance for Step 2	ΔR2=.21		ΔR2=.05		ΔR2=.16
Total explained variance for model	R2 = .31		R2 = .21		R2 = .21

Note: BSM = behavioral skills maintenance treatment. FBT = family-based treatment. FV = fruits and vegetables. SES = socio-economic status. SFM = social facilitation skills maintenance treatment.

Discussion

The current study identified behavioral mechanisms underlying the correlation in weight change within child-parent dyads involved in FBT. Dietary changes were also correlated within dyads, and most interesting, the similarity of changes in RED food intake explained why children and their parents tended to show similar weight maintenance patterns following treatment completion. Thus, dyads with highly similar patterns of change in RED food intake were likely to have highly similar patterns of weight maintenance. These findings have important implications for achieving long-term weight maintenance following FBT. First, by showing that maintenance of lower RED food intake uniquely predicted weight maintenance in parents and children, the findings emphasize that the behavioral changes important to weight loss are also important to long-term weight maintenance (Casazza et al., 2013). Second, the findings suggest that clinicians should help parents and their children to identify and implement common strategies related to minimizing RED food intake. For example, RED foods might need to be removed from all areas of the home that the both parent and child inhabit. Third, these findings suggest that obesity treatments that target the parent only (e.g., Golan & Crow, 2004) might be effective at producing weight loss in the untreated child if the program targets the reduction of RED food intake by both parent and child.

There are limitations to the current work. Dietary intake was reported rather than directly observed and precluded examining other important dietary variables (e.g., caloric intake, macronutrients). Some parents completed a proportion of their children’s food diaries, leading to non-independence of data in these instances. Physical activity was not examined, which may also be important to weight maintenance. Several dyads did not provide complete data across all time points. Finally, the relative homogeneity of the sample (mostly female, Caucasian, and mid-upper SES) limits the generalizability of the results. The major strength of this study is the tracking of self-reported dietary changes in children and parents across a two-year maintenance period. These results suggest long-term maintenance can be achieved in children and parents if they maintain a low proportion of RED foods in the diet, which should be a focus of treatment. Moreover, the results provide novel evidence that the correlated weight change within the parent-child dyad may be attributed to similar dietary changes following treatment completion. Future studies should examine additional behavioral changes (e.g., physical activity) that might be correlated within parent-child dyads and explore the basic genetic and environmental factors that might underlie these shared changes relevant to long-term weight maintenance.