An investigation of associations between parenting and binge eating across pubertal development in pre‐adolescent and adolescent female participants

Abstract

Objective

Puberty is a period of increased risk for the development of binge eating in female adolescents. Although developmental changes in autonomy‐seeking behaviors and body weight and shape may influence both parenting styles and binge eating during puberty, studies have yet to examine how parenting practices may be differentially associated with youth outcomes depending on developmental stage. The current study examines whether interactions between puberty and parenting are associated with higher levels of binge‐eating symptoms during/after puberty in female youth.

Methods

Analyses used cross‐sectional data from a previous study of disordered eating and puberty in 999 female youth (ages 8–16) and their parents from the Michigan State University Twin Registry. Youth self‐reported binge eating, pubertal development, and perceived parental care and overprotection. Both parents and youth reported on parent–child conflict. Mixed linear models were used to examine whether pubertal development moderates the strength of associations between parenting (parent–child conflict, parental care, and parental overprotection) and offspring binge eating.

Results

Although higher levels of parental overprotection and conflict, and lower levels of parental care were all significantly associated with binge eating, none of the associations were significantly moderated by pubertal development or age.

Discussion

The quality of the parent–child relationship is significantly associated with binge eating in female youth regardless of developmental stage, highlighting the need for targeting harmful parenting strategies during adolescent eating disorder intervention.

Public Significance

This is the first study to examine whether parenting/binge‐eating associations in female participants differ across pubertal development. In a large population‐based sample, we found lower parental care, higher parent–child conflict, and higher parental overprotection were all associated with higher levels of binge eating. Notably, associations did not differ across pubertal stage or age, suggesting that parenting is significantly associated with binge eating, regardless of developmental stage.

1. INTRODUCTION

Puberty is a critical risk period for the development of binge eating (i.e., eating a large amount of food in a short period of time accompanied by loss of control) in girls (Klump, 2013; Klump et al., 2017), as binge‐eating rates are substantially higher in post‐puberty relative to pre‐puberty (Klump, 2013). Several factors have been proposed to contribute to this risk, including increases in ovarian hormones (Forney et al., 2019; Klump et al., 2018) and increased impulsivity/negative urgency (Davis & Smith, 2018; Pearson et al., 2012). One relatively ignored set of factors has been changes in parenting and the parent–child relationship. Parenting strongly predicts academic and psychological outcomes in youth (Maccoby, 1992); controlling and harsh parenting (e.g., high in conflict and criticism) predict internalizing and externalizing disorders (Burt et al., 2005; Yap et al., 2014), while warm parenting (e.g., high in care and acceptance) is associated with improvements in these same outcomes. The vital importance of parenting is underscored by the focus on altering parenting practices as key mechanisms of effects in several interventions for youth psychopathology (e.g., Yap et al., 2016).

Importantly, developmental studies have found substantial increases in parent–child conflict, and decreases in parent–child closeness, during early adolescence and puberty in girls (Mastrotheodoros et al., 2019; McGue et al., 2005; Steinberg, 1987). Girls report more intense parent–child conflict than boys during early adolescence (i.e., ages 11–14), despite similar levels of conflict frequency (Allison & Schultz, 2004; Mastrotheodoros et al., 2020), and a recent meta‐analysis found associations between more negative aspects of parenting (e.g., low parental warmth and high parental control) and internalizing problems (i.e., anxiety and depressive symptoms) became stronger with older age in a sample of children and adolescents (M age = 11.61; Pinquart, 2017). Although changes in parenting across pubertal development have been much less studied, one study found stronger associations between parental support during a stress test and physiological stress responses in girls who were at earlier, rather than later, stages of puberty (Doom et al., 2015). Researchers have proposed explanations for age and puberty‐related changes in parenting, including youth increases in autonomy‐seeking and risk‐taking behaviors that can be associated with increased conflicts with parents and parental expectations (Laursen et al., 1998). Although fewer studies have examined changes in the effects of parental support/care, it is possible that parental support loses its salience relative to peer support as youths develop and increase their efforts to individuate and develop a sense of autonomy (Helsen et al., 2000). All of these factors may be relevant for binge eating, particularly given increased autonomy in food choices and eating (Neufeld et al., 2021) and heightened concerns and negative comments about body weight/shape (Dahill et al., 2021) during/after puberty that may further exacerbate parenting/binge‐eating associations.

Unfortunately, no studies have examined whether differences in parenting during puberty are associated with binge eating in girls. Findings suggest binge‐eating symptoms are significantly associated with low parental care (Ackard et al., 2006; Hampshire et al., 2022) and high parental control (Depestele et al., 2017; Hampshire et al., 2022; Salafia et al., 2009), although most of these studies have relied on retrospective reports of parenting and focus on older adolescents/young adults only. Two studies examined age differences in parenting/disordered eating associations across adolescence in girls. Korotana et al. (2018) found that binge eating was more strongly associated with lower parental involvement (i.e., closeness) at age 11 versus ages 14 and 17, while Spanos et al. (2010) found that binge eating was associated with higher levels of parent–child conflict at ages 14 and 17 versus age 11. Taken together, these findings suggest that parental closeness may be more important for binge eating in younger youth (and lose salience as children age), whereas parent–child conflict may exhibit stronger associations with binge eating as children age. Unfortunately, neither study examined pubertal development and thus, it remains unknown whether puberty impacts these parenting/binge‐eating associations.

Given the above, the purpose of the current study was to examine whether associations between parenting and binge eating differ across puberty in a large, population‐based sample of females in pre‐adolescence and adolescence. Given previous findings suggesting stronger associations between parental involvement/closeness and binge eating early in development (Korotana et al., 2018), and stronger conflict/binge‐eating associations later in development (e.g., Spanos et al., 2010), we hypothesized that associations between parental care and binge eating would be stronger in pre/early puberty, whereas associations between binge eating and conflict would be stronger in mid/late puberty,

2. METHODS

2.1. Participants

The current study used an archival, cross‐sectional sample of 999 female twins (N = 500 twin pairs, ages 8–16 years, M = 11.75, SD = 2.03) and their parents from the Twin Study of Mood, Behavior, and Hormones during Puberty (TSMBH; Klump et al., 2018). The current research was reviewed and approved by the insitutional review board (IRB) at Michigan State University. Data collection occurred between 2008 and 2014. The TSMBH sample was recruited from the larger, population‐based Michigan State University Twin Registry (MSUTR; Burt & Klump, 2019), which recruits twins using birth records (see Burt and Klump (2019) for method description). Response rates of the MSUTR (56%–85%) and the TSMBH (65%) are on par or better than other twin studies using similar recruitment methods (Burt & Klump, 2019; Klump et al., 2018). MSUTR twins are demographically representative of the Michigan population with respect to race/ethnicity and socioeconomic status (Burt & Klump, 2019).

Because the TSMBH's primary aim involved examining phenotypic and genetic effects of ovarian hormones on disordered eating during puberty, the following inclusion criteria were required for participation: (1) no hormonal contraceptive use within the past 3 months; (2) no psychotropic or steroid medications within the past 4 weeks; (3) no pregnancy or lactation within the past 6 months; and 4) no history of genetic/medical conditions known to influence hormones or appetite/weight. The TSMBH sample was representative of Michigan in regard to race/ethnicity (80.6% identified as White, 7.8% identified as African American/Black, 3.8% identified as Hispanic/Latinx, 0.6% identified as Asian, 0.2% identified as Native American/Alaskan Native, and 7% identified as Multiracial) and annual household income (5.6% reported an annual household income of under $20,000, 12.8% $20,000–$40,000, 15.2% $40,000–$60,000, 26.2% $60,000–$100,000, 34.7% over $100,000, and 5.4% missing data).

Because the parent study was a genetic study, data collection focused on biological parents, regardless of relationship status or whether they lived with twins. Current relationship status of the biological parents is reported in Table 1. Notably, when a biological parent was not available, or when the twins had limited contact with the biological parent, step‐parents or other caregivers (e.g., partners) were invited to participate in the study. Although the TSMBH only required that one parent participate, a number of families (38%) had more than one parent/caregiver in the study. Most commonly, the biological mother participated (96% of families), but in many cases, the biological father or step‐father participated as well (see Table 1). However, even if a parent did not participate in the study, the twins still reported on the parenting styles of that biological parent or a step‐parent (if there was limited contact with the biological parent; see Table 1). Importantly, additional analyses were conducted to ensure that family composition (e.g., married vs. not married) or the participation/nonparticipation of some parents did not unduly influence results. Primary study findings regarding the moderation of parenting/binge‐eating associations by puberty were identical to those reported below in the full sample (see Tables S1–S3). Thus, only findings for the full sample are presented below.

TABLE 1.

Information about family composition, the number of parents participating in data collection, and the caregivers the twins reported on for the parenting assessments

Variable	N (% of twins)
Parent participation in data collection (N = 498 families)
Both biological mother and father participated	376 (37.6%)
Biological mother only	590 (59.1%)
Biological father only	24 (2.4%)
Biological mother and step‐father	4 (.4%)
Neither parent participated	2 (.2%)
Twin report on caregivers (N = 975 twins)
Twin reported on biological mother	956 (95.7%)
Twin reported on biological father	877 (97.8%)
Twin reported on step‐mother	2 (.2%)
Twin reported on step‐father	26 (2.6%)
Relationship status of participating parents (N = 495 biological mothers, 222 biological fathers, 1 step‐father)
Married	564 (80.8%)
Divorced	70 (10%)
Widowed	7 (1%)
Living with significant other	18 (2.5%)
Never married and not living with partner	31 (4.3%)
Missing information	18 (3.6%)

Note: Relationship status was reported separately by each parent, and information regarding whether the relationship status was in reference to the other biological parent (e.g., married = biological parents married to each other) or to another person (e.g., married to a step‐parent) was not assessed—only the overall relationship status, without reference to who the partners were, was collected by the parent study.

2.2. Measures

2.2.1. Demographic information

Parents reported on twin race/ethnicity, household income, parent occupation, and parent level of education. Because twins were recruited using birth records, participant sex reflects the sex assigned at birth provided on their birth certificates. Gender identity was not assessed. Ethnicity was measured with the question, “Would you describe your twins' ethnicity as Hispanic or Latino (no/yes)?”. Options for race included “White,” “Black or African American,” “Asian,” “American Indian or Alaska Native,” “Native Hawaiian or Other Pacific Islander,” “More than One Race,” and “Unknown or Not Reported.” Participants were able to select all that applied for race.

2.2.2. Binge eating

The seven‐item binge‐eating scale from the Minnesota Eating Behaviors Survey (MEBS; von Ranson et al., 2005) assessed current levels of thoughts of engaging in or engaging in binge eating and secretive eating. Although twin self‐reports and parent reports of the child's binge eating were available for analyses, only the twin reports were used in analyses due to poor convergence between parent and youth reports on binge eating in this sample (Mean ICC = .15; see Vo et al., 2019). The MEBS was developed for use in youth ages 9 and above, and studies show it is appropriate for use in pre‐pubertal children (Luo et al., 2016). Internal consistencies have ranged from .65 to .75 in past work (von Ranson et al., 2005) and were .68 in the current sample (Vo et al., 2021). The MEBS also shows good criterion‐related validity, where girls with bulimia nervosa score significantly higher than controls on the binge‐eating subscale (von Ranson et al., 2005).

The Minnesota Eating Behavior Survey (MEBS; previously known as the Minnesota Eating Disorder Inventory [M‐EDI]) was adapted and reproduced by special permission of Psychological Assessment Resources, Inc., from the Eating Disorder Inventory (collectively, EDI and EDI‐2) by Garner et al. (1983) by the Psychological Assessment Resources, Inc. Further reproduction of the MEBS is prohibited without prior permission from Psychological Assessment Resources, Inc.

2.2.3. Parenting

The 25‐item Parental Bonding Instrument (PBI; Parker et al., 1979) assessed twins' perceived parenting received by mothers and fathers. The PBI was only completed by the twins and thus, no parental reports were available. The PBI generates care (i.e., parental warmth) and overprotection subscales (i.e., parental overcontrol). The PBI was originally developed for use in adults who rated how well each item described their parents, as remembered from childhood up to age 16. The TSMBH asked youth to rate parental behaviors as remembered during their entire life up to the point of completing the survey. Prior studies in adolescent samples had participants rate items similarly and have demonstrated good internal consistency for both scales (α's = .84–.87; Rigby et al., 2007). Internal consistency in the current sample ranged from .80 to .84 for parental care and .68 to .70 for parental overprotection. Test–retest reliability of the PBI in the original validation sample was .76 and .63 for the care and overprotection subscales, respectively (Parker et al., 1979).

In addition to individual care and overprotection scales, the PBI generates quadrants of parenting styles including: (1) affectionate constraint (high care and high overprotection), (2) optimal parenting (high care and low overprotection), (3) affectionless control (low care and high overprotection), and (4) neglectful parenting (low care and low overprotection). Quadrants were examined to better understand whether overall profiles of parenting predict binge‐eating symptoms differently than individual parenting dimensions. PBI quadrants were developed by calculating a mean split on care and overprotection subscales to generate “low” and “high” classifications of each, which were then paired together in each quadrant (e.g., low care and high overprotection as “affectionless control”). The original validation study established cutoffs with adults reporting retrospectively (Parker et al., 1979). Since the current study included pre‐adolescents/adolescents, we established cutoffs for care and overprotection using sample‐specific means on averaged twin reports of maternal and paternal behaviors (M care = 28.80; M overprotection = 14.04).

The 12‐item conflict subscale from the Parental Environment Questionnaire (PEQ; Elkins et al., 1997) assessed discontent or hostility within the parent–child relationship. Twins and parents provided reports on parent–child conflict, and reports described current levels of conflict in the individual parent–child relationship. The PEQ was originally developed for use in adolescents ages 11–17 but has been used in children as young as age 6 (Waller et al., 2018). All subscale items were reverse‐scored and summed to ensure that higher scores indicated greater conflict. The conflict subscale has shown good internal consistency in child samples ages 6–10 (α = .85–.87; Waller et al., 2018), adolescent samples ages 10–18 (α = .90; Klahr et al., 2011), and in the current sample (α = .89–.90). The conflict subscale has shown moderate‐to‐high correlations with self‐reported parent–child conflict measured with the Family Environment Scale (r = .55; Elkins et al., 1997).

2.2.4. Pubertal development

The Pubertal Development Scale (PDS; Petersen et al., 1988) asked twins to rate current changes in body hair, skin, growth spurts, and breast development on a 4‐point scale: (1) “Not yet started showing changes”; (2) “Have barely started showing changes”; (3) “Changes are definitely underway”; and (4) “Changes seem completed”. Onset of menses was coded as absent (1) or present (4). PDS items were summed and averaged to develop pubertal development scores. Similar to past studies, we minimized missing data on pubertal status by using maternal report on the PDS when youth report was missing (Klump et al., 2017, 2018). Past studies have shown acceptable internal consistency in adolescent girls (α = .76–.83; Petersen et al., 1988) and high correlations between PDS scores and clinician ratings of pubertal development (r's = .61–.67, Petersen et al., 1988). Good internal consistency was found in this sample (α = .84; Vo et al., 2021).

2.2.5. Body mass index

Body mass index (BMI) was calculated (kg/m²) using twin height and weight measured with a wall‐mounted ruler and digital scale, respectively.

2.3. Statistical analyses

2.3.1. Data preparation

Following study conventions for the TSMBH, scores on all measures were prorated if ≤10% of items were missing; if over 10% of items were missing, the final scores were coded as missing. MEBS scores were log‐transformed to account for positive skew. All variables were standardized before analyses. Multicollinearity statistics were examined to ensure findings were not unduly influenced by correlations between predictors in the model. Tolerance scores were all above .10 and VIF scores were below 2, indicating no multicollinearity concerns.

As noted above, twin reports were available for all parenting variables, but parental report was only available for the PEQ conflict scale. Studies in other areas of psychopathology frequently combine twin and parent reports into single parenting measures (see Burt et al., 2005, 2007), as correlations in perceived parenting tend to be in the medium‐to‐large effect size range (r's > .30). To examine whether this was the case in the current sample, Pearson correlations were calculated to examine similarity in perceived parenting across the variables and different reporters in this study (see Table S4). All correlations were in the same range as those reported in previous work (Burt et al., 2005, 2007) and indicated that family members viewed the parenting quite similarly (r's = .66–.68 for twin reports of parenting received by mothers versus fathers; r's = .76–.78 for maternal and paternal reports of conflict on the PEQ; r's = .36–.49 for twin and parent reports of conflict on the PEQ).

Consequently, twin and parent reports of parenting were combined into a single index, following methods used in previous work (Burt et al., 2005, 2007). We first averaged twin reports of conflict with the mother and father to form a twin report on both parents. We then averaged mother and father reports of conflict together to form a combined parent report. Finally, twin‐reported and parent‐reported conflict scores were averaged together, producing one overall, parent–child conflict score to use in analyses. When combining these reports, up to two missing reports were allowed to maximize the number of participants with conflict data. Nonetheless, to ensure that our use of average scores did not unduly influence results, we conducted parent–child conflict models separately using only twin report averages or parent report averages, and models that included the individual twin, mother, and father reports (not averaged scores) all in the same model. Results were identical to those reported herein for our primary analyses of the moderation of parenting/binge‐eating associations by pubertal status (see Tables S5 and S6).

2.3.2. Statistical models

Mixed linear models (MLMs) were used for analyses given the nonindependence of the family data, where data from individual respondents (level 1) were nested within family (level 2). We did not include twin zygosity (i.e., identical versus fraternal) as an additional family factor in the model, as zygosity was not a significant moderator of puberty/parenting associations with binge eating (see Table S7). To account for the number of models examined, we used a Bonferroni‐adjusted p value (p < .0008).

Models were conducted separately for each parent–child parenting variable. We conducted all models with and without BMI as a covariate, given associations between higher BMI and both higher levels of binge eating (West et al., 2019) and more advanced pubertal stage (Bini et al., 2000). In addition to these primary models, we conducted exploratory analyses examining whether associations between puberty, parenting, and binge eating differ across race/ethnicity, given lack of data on racially and ethnically diverse populations in the eating disorders field and calls to routinely consider these important contextual variables in the study of risk factors (Mikhail & Klump, 2021). No studies have examined differences in parenting/disordered eating associations, despite evidence that parenting style effects may differ across race/ethnicity (e.g., Finkelstein et al., 2001; Pinderhughes et al., 2001). Only individuals reporting “White/Caucasian” (N = 805, 80.6%), “Black/African American” (N = 78, 7.8%), “Hispanic/Latinx” (N = 38, 3.8%), or “More than one race” (N = 70, 7%) were included in these analyses, given the small number of respondents for other race/ethnicity categories (N = 8 total, 0.8%). Race/ethnicity was coded using effect codes which allowed us to examine the difference between the group coded as “1” and the unweighted grand mean of the total sample.

3. RESULTS

3.1. Descriptive statistics

Descriptive statistics for all variables are presented in Table 2. The full range of scores was present for binge eating and pubertal development measures, and there was adequate variability in parenting and parenting styles (38.5% optimal, 18.7% affectionate constraint, 28.6% affectionless control, 11.2% neglectful, and 2.9% missing). The distributions of binge eating and parenting scores were consistent with other population‐based (Ong et al., 2018; Spanos et al., 2010) and community samples (Tanofsky‐Kraff et al., 2007).

TABLE 2.

Descriptive statistics

Variable	Mean	SD	Observed range	Possible range
MEBS binge eating	0.88	1.32	0–7	0–7
PBI care composite	28.80	5.57	10–36	0–36
Twin report on mother	29.45	5.61	7–36	0–36
Twin report on father	28.14	6.42	3–36	0–36
PBI overprotection composite	14.04	5.39	0–32.88	0–39
Twin report on mother	14.38	5.66	0–32	0–39
Twin report on father	13.65	5.99	0–35.75	0–39
PEQ conflict composite	20.60	5.45	12–45.09	12–48
Twin report on mother	20.04	7.72	12–48	12–48
Twin report on father	19.48	7.96	12–48	12–48
Mom report on twin	21.33	6.74	12–45.82	12–48
Dad report on twin	21.63	6.59	12–44.73	12–48
PDS final average scores	2.23	0.90	1–4	1–4
BMI	19.49	4.47	10.70–46.55	–

Note: Although log‐transformed values were used for MEBS Binge Eating and Eating in the Absence of Hunger Total Score, raw means, standard deviations, and ranges are listed here for descriptive purposes.

Abbreviations: BMI, body mass index; MEBS, Minnesota Eating Behaviors Survey; PBI, Parental Bonding Instrument; PDS, Pubertal Development Scale; PEQ, Parental Environment Questionnaire.

3.2. Pearson correlations

As expected, there were significant correlations between parenting measures and binge eating (p's < .0008; see Table 3), although effect sizes tended to be in the small‐to‐moderate range (e.g., r = .18–.21). More advanced pubertal development was also associated with higher binge‐eating scores (r = .12; p < .0008) although it was only inconsistently associated with parenting (r's = .06–.24, p's = .09 to <.0008).

TABLE 3.

Pearson correlations between binge eating, parenting variables, pubertal development, age, and BMI

Variable	1	2	3	4	5	6
1. MEBS binge eating	–
2. PBI care composite	−.19*	–
3. PBI overprotection composite	.18*	−.52*	–
4. PEQ conflict composite	.21*	−.47*	.33*	–
5. Pubertal status	.12*	−.11*	.06	.24*	–
6. Age	.05	−.07	−.01	.19*	.80*	–
7. BMI	.14*	−.13*	.03	.12*	.49*	.41*

Abbreviations: BMI, body mass index; MEBS, Minnesota Eating Behaviors Survey; PBI, Parental Bonding Instrument; PEQ, Parental Environment Questionnaire.

^* p < .0008.

3.3. Multilevel models

As shown in Tables 4 and 5, there were several significant main effects of parenting on binge‐eating scores in the MLMs. Lower levels of parental care and higher levels of both parental overprotection and parent–child conflict were significantly associated with higher levels of binge eating (p's < .0008). Likewise, in terms of the PBI parenting styles, groups characterized by nonoptimal styles of parenting (i.e., the affectionate constraint, affectionless control, and neglectful parenting groups) generally reported significantly higher binge‐eating scores than the optimal parenting group (see Table S8 showing follow‐up group comparisons).

TABLE 4.

Multilevel models examining associations between parenting, puberty, and binge eating, with and without BMI

Variable	Binge eating
PBI parental care (with BMI)	β	SE	Df	t	p
Intercept	.002	.04	489.20	0.04	.965
Care	−.16	.03	943.55	−4.86	<.0008
Puberty	.10	.04	695.23	2.80	.005
Care × Puberty	.03	.03	952.42	1.09	.277
PBI parental care (without BMI)	β	SE	df	t	p
Intercept	.02	.04	546.85	0.46	.644
Care	−.15	.03	938.22	−4.50	<.0008
Puberty	.06	.04	760.45	1.64	.101
BMI	.08	.04	822.08	2.05	.040
Care × BMI	−.06	.03	942.71	−1.85	.065
Puberty × BMI	−.04	.04	867.53	−1.14	.254
Care × Puberty	.06	.04	948.95	1.84	.066
PBI parental overprotection (without BMI)	β	SE	df	t	p
Intercept	−.01	.04	488.77	−0.14	.885
Overprotection	.17	.03	947.88	5.37	<.0008
Puberty	.11	.04	695.07	3.04	.002
Overprotection × Puberty	−.01	.03	936.24	−0.46	.649
PBI parental overprotection (with BMI)	β	SE	df	t	p
Intercept	.01	.04	549.92	0.25	.806
Overprotection	.17	.03	944.99	5.40	<.0008
Puberty	.07	.04	758.06	1.67	.095
BMI	.10	.04	809.21	2.52	.012
Overprotection × BMI	.07	.04	940.68	1.86	.064
Puberty × BMI	−.03	.04	852.38	−0.88	.378
Overprotection × Puberty	−.04	.03	935.11	−1.14	.254
PEQ parent–child conflict (without BMI)	β	SE	df	t	p
Intercept	−.003	.04	499.63	−0.07	.945
Conflict	.19	.04	841.44	5.49	<.0008
Puberty	.07	.04	701.44	1.98	.048
Conflict × Puberty	−.02	.03	884.41	−0.65	.519
PEQ parent–child conflict (with BMI)	β	SE	df	t	p
Intercept	.02	.04	549.50	0.40	.690
Conflict	.18	.04	827.56	5.18	<.0008
Puberty	.05	.04	759.68	1.20	.229
BMI	.07	.04	796.91	1.60	.111
Conflict × BMI	.05	.03	931.57	1.53	.126
Puberty × BMI	−.05	.04	875.61	−1.27	.204
Conflict × Puberty	−.04	.04	902.27	−1.12	.265

Note: Scores were standardized before analyses.

Abbreviations: BMI, body mass index; PBI, Parental Bonding Instrument; PEQ, Parental Environment Questionnaire.

TABLE 5.

Multilevel models examining associations between binge eating and parenting style, controlling for BMI

Variable	Binge eating
PBI affectionate constraint	β	SE	Df	t	p
Intercept	.04	.04	575.16	0.85	.396
Affectionate constraint	.13	.04	919.65	3.16	.002
Puberty	.06	.04	789.39	1.59	.113
BMI	.12	.04	848.46	2.70	.007
Affectionate constraint × BMI	.02	.05	910.60	0.41	.686
Puberty × BMI	−.03	.04	854.42	−0.76	.450
Affectionate constraint × Puberty	.003	.05	899.00	0.06	.955
PBI affectionless control	β	SE	df	t	p
Intercept	.03	.04	554.75	0.77	.441
Affectionless control	.16	.04	942.86	4.02	<.0008
Puberty	.06	.04	761.30	1.58	.115
BMI	.11	.04	804.51	2.59	.010
Affectionate control × BMI	.06	.04	944.00	1.30	.193
Puberty × BMI	−.04	.04	854.58	−1.07	.284
Affectionless control × Puberty	−.05	.04	941.96	−1.22	.224
PBI neglectful	β	SE	df	t	p
Intercept	.06	.04	622.92	1.48	.140
Neglectful	.18	.05	939.97	3.67	<.0008
Puberty	.05	.04	820.62	1.07	.286
BMI	.12	.04	866.11	2.64	.009
Neglectful × BMI	.01	.06	924.11	0.18	.859
Puberty × BMI	−.03	.04	874.28	−0.90	.367
Neglectful × Puberty	−.04	.06	918.43	−0.76	.446

Note: Scores were standardized before analyses. Parenting style was effect‐coded for affectionate constraint, affectionless control, and neglectful parenting, with optimal parenting serving as the reference group in each group. For example, the effect code for affectionate constraint coded participants reporting affectionate constraint parenting as “1,” affectionless control and neglectful parenting as “0,” and optimal parenting as “−1.”. Effect codes compared effects of the parenting style being examined with the effects across all the parenting styles.

Abbreviation: BMI, body mass index.

Nonetheless, counter to study hypotheses, there were no significant puberty x parenting interactions for any of the individual parenting variables or the PBI parenting styles (all p's = .066–.955; see Tables 4 and 5). In addition, in our exploratory models, all two‐way (e.g., race/ethnicity × parenting) and three‐way interactions (race/ethnicity × pubertal development × parenting) were nonsignificant (all p's = .093–.974; see Table 6). Taken together, these data suggest that although parenting is significantly associated with binge eating scores, associations are similar across pubertal development and across the race/ethnicities examined herein.

TABLE 6.

Multilevel models examining associations between race/ethnicity, puberty, and parenting

Predictor	Binge eating
PBI parental care
Black/African American versus Overall Sample	β	SE	df	t	p
Intercept	.01	.06	555.52	0.09	.928
Care	−.20	.05	918.02	−4.34	<.0008
Puberty (PDS)	.10	.06	837.76	1.63	.104
Black versus overall	.02	.07	553.21	0.24	.811
Care × PDS	.03	.05	914.08	0.66	.507
Care × Black versus Overall	−.07	.05	918.02	−1.48	.138
PDS × Black versus Overall	.002	.07	838.94	0.03	.974
Care × PDS × Black versus Overall	−.01	.05	916.27	−0.21	.835
Hispanic/Latinx versus Overall Sample	β	SE	df	t	p
Intercept	.005	.07	515.08	0.07	.948
Care	−.23	.06	947.84	−4.16	<.0008
Puberty (PDS)	.11	.07	715.08	1.64	.102
Hispanic versus Overall	.01	.08	509.74	0.19	.853
Care × PDS	.004	.06	939.83	0.07	.948
Care × Hispanic versus Overall	−.10	.06	948.85	−1.68	.093
PDS × Hispanic versus Overall	.02	.07	696.70	0.23	.820
Care × PDS × Hispanic versus Overall	−.05	.07	944.47	−0.67	.505
Multiracial versus Overall Sample	β	SE	df	t	p
Intercept	.04	.06	510.35	0.62	.538
Care	−.21	.05	948.26	−4.13	<.0008
Puberty (PDS)	.11	.06	738.47	1.96	.050
Multiracial versus Overall	.06	.07	505.87	0.88	.380
Care × PDS	.02	.05	946.92	0.46	.647
Care × Multiracial versus Overall	−.08	.06	943.37	−1.37	.172
PDS × Multiracial versus Overall	.01	.06	722.61	0.23	.815
Care × PDS × Multiracial versus Overall	−.02	.05	948.84	−0.37	.713
PBI parental overprotection	Binge eating
Black/African American versus Overall Sample	β	SE	df	t	p
Intercept	−.003	.07	566.19	−0.06	.954
Overprotection	.21	.06	940.78	3.76	<.0008
Puberty (PDS)	.09	.07	820.41	1.41	.160
Black versus Overall	.01	.07	564.20	0.13	.900
Overprotection × PDS	−.01	.05	928.63	−0.19	.848
Overprotection × Black versus Overall	.05	.05	941.55	0.90	.369
PDS × BvO	−.02	.07	818.62	−0.24	.807
Overprotection × PDS × Black versus Overall	.01	.06	925.69	0.11	.914
Hispanic/Latinx versus Overall Sample	β	SE	df	t	p
Intercept	.02	.07	490.93	0.32	.746
Overprotection	.20	.06	906.78	3.29	.001
Puberty (PDS)	.10	.07	697.64	1.44	.149
Hispanic versus Overall	.04	.07	489.54	0.52	.602
Overprotection × PDS	.01	.06	940.81	0.12	.901
Overprotection × Hispanic versus Overall	.04	.07	914.68	0.57	.569
PDS × Hispanic versus Overall	−.01	.07	686.26	−0.08	.933
Overprotection × PDS × Hispanic versus Overall	.03	.06	942.33	0.45	.654
Multiracial versus Overall Sample	β	SE	df	t	p
Intercept	.03	.06	515.13	0.51	.609
Overprotection	.20	.05	943.96	3.93	<.0008
(Puberty) PDS	.11	.05	731.75	1.96	.050
Multiracial versus Overall	.06	.07	511.04	0.85	.398
Overprotection × PDS	−.003	.04	938.28	−0.07	.944
Overprotection × Multiracial versus Overall	.05	.05	942.84	0.88	.379
PDS × Multiracial versus Overall	.005	.06	719.83	.08	.935
Overprotection × PDS × Multiracial versus Overall	.02	.05	935.84	.35	.725
PEQ parent–child conflict	Binge eating
Black/African American versus Overall Sample	β	SE	df	t	p
Intercept	.08	.06	542.02	1.27	.207
Conflict	.22	.06	877.54	3.86	<.0008
Puberty (PDS)	.04	.06	796.47	0.57	.571
Black versus Overall	.11	.07	541.70	1.57	.117
Conflict × PDS	−.07	.05	908.73	−1.31	.191
Conflict × Black versus Overall	.04	.06	883.90	0.57	.571
PDS × Black versus Overall	−.05	.07	798.68	−0.68	.498
Conflict × PDS × Black versus Overall	−.07	.06	914.14	−1.18	.238
Hispanic/Latinx versus Overall Sample	β	SE	df	t	p
Intercept	.07	.07	511.39	1.04	.299
Conflict	.21	.07	842.05	3.09	.002
Puberty (PDS)	.04	.07	723.46	0.62	.537
Hispanic versus Overall	.09	.08	507.32	1.24	.214
Conflict × PDS	−.07	.06	889.97	−1.11	.268
Conflict × Hispanic versus Overall	.02	.08	848.84	0.26	.798
PDS × Hispanic versus Overall	−.03	.08	707.95	−0.43	.665
Conflict × PDS × Hispanic versus Overall	−.06	.07	893.18	−0.92	.360
Multiracial versus Overall Sample	β	SE	df	t	p
Intercept	.08	.06	505.68	1.32	.189
Conflict	.16	.05	814.93	3.12	.002
Puberty (PDS)	.06	.06	729.04	1.13	.259
Multiracial versus Overall	.11	.07	501.91	1.68	.093
Conflict × PDS	−.04	.05	868.11	−0.83	.405
Conflict × Multiracial versus Overall	−.04	.06	816.42	−0.72	.472
PDS × Multiracial versus Overall	−.01	.06	717.17	−0.14	.887
Conflict × PDS × Multiracial versus Overall	−.03	.05	868.61	−0.53	.596

Note: Scores were standardized before analyses. Race/ethnicity was effect‐coded in analyses.

Abbreviations: PBI, Parental Bonding Instrument; PDS, Pubertal Development Scale; PEQ, Parental Environment Questionnaire.

Because results for interactions with puberty were so uniformly nonsignificant, we conducted several post hoc analyses to ensure findings were not unduly influenced by other factors. First, we explored whether associations differed across age rather than pubertal development. We conducted MLMs with age as the moderator, and then we also examined the effects of pubertal development with age regressed out of the puberty score. Results from both analyses were identical to our primary models, that is, there were no significant age × parenting interactions or puberty (with age regressed out) × parenting interactions (see Tables S9 and S10).

Second, we were interested in whether there were differences in findings depending on whether the twins and their parents viewed their relationship similar or different. We examined this question by first calculating difference scores between parent and twin reports on conflict for mothers and fathers separately. We then calculated the percentiles of the difference scores, with low percentiles indicating more similar scores. The 50th percentile was used as a cutoff to create a dichotomous dummy‐coded variable indicating either high levels of agreement (coded 1) or low levels of agreement (coded 0). We then conducted three‐way interaction models (i.e., puberty × parent–child conflict × high level of agreement) to examine whether higher levels of agreement between parent and twin reports on parent–child conflict changed findings. The three‐way interaction between conflict, puberty, and parent–child agreement on level of conflict was nonsignificant in all cases (p's > .144; see Table S11), suggesting no differences in findings by parent–child agreement on parent–child conflict ratings.

Finally, because full‐threshold binge eating can be less common in younger youth, we also examined our study hypotheses using dysregulated eating symptoms that are more common in younger populations and predict the later development of binge eating (i.e., eating the absence of hunger and emotional eating; Balantekin et al., 2017; Tanofsky‐Kraff et al., 2007). Results did not differ from primary models, as the parenting x puberty interactions remained nonsignificant (see Tables S13 and S14).

4. DISCUSSION

This was the first study to examine whether pubertal development moderates associations between binge eating and parenting. Findings suggested that although lower parental care, higher parental overprotection, and higher parent–child conflict were each associated with increased binge eating, pubertal development did not significantly impact these associations. Importantly, findings did not differ when examining age, family composition, twin/parent concordance in perceptions of parenting, other types of dysregulated eating, or race/ethnicity. Overall, these findings suggested that parenting/binge‐eating associations are present across development and across key demographic characteristics in pre‐adolescent and adolescent girls.

The lack of significant differences between parenting/binge‐eating associations across puberty and age was surprising given shifts in parenting behavior across these variables in past work (Mastrotheodoros et al., 2019; Steinberg, 1987), and differences in parenting associations with mental health more generally across age (e.g., anxiety and depressive symptoms; Pinquart, 2017). Reasons for differing findings are unclear, although results from a meta‐analysis for internalizing disorders suggested that age differences in parenting/mental health outcomes may be small in magnitude (Pinquart, 2017). Although our overall sample size was large (N = 999), sample sizes in the more extreme groups (e.g., pre‐puberty N = 232, post‐puberty N = 27) were more modest. Previous studies also often examined larger samples within discrete age periods (e.g., ages 11, 14, and 17; N's = 446–468 twins examined over time at each age; Korotana et al., 2018; Spanos et al., 2010). If age and pubertal differences are small in magnitude, the lack of significant findings in the current study could be explained by the need for larger sample sizes at each age or pubertal stage. However, given how robust our findings were to our many post hoc analyses, it seems likely that we would not have observed significant differences in parenting/binge‐eating associations in our sample, even with a larger sample size.

Nonetheless, we did observe significant main effects for parenting behaviors/styles that were consistent with past studies in older adolescents and adults (Hampshire et al., 2022). Specifically, we found lower care, higher overprotection, and higher parent–child conflict, in addition to “nonoptimal” parenting styles (i.e., affectionless control, affectionate constraint, and neglectful parenting) were all associated with higher levels of binge eating. The association between affectionless control (similar to authoritarian parenting, i.e., low responsiveness, high demandingness; Baumrind, 1991) and binge eating was particularly strong (see Table S9). Past studies have shown that this parenting style is most strongly associated with eating disorders (Monteleone et al., 2020) and other negative outcomes (e.g., depression; Fox et al., 2021), potentially through transmission of poor emotion regulation skills (Shaw & Starr, 2019) and increased feelings of loneliness resulting from having a negative parent–child relationship (Abebe et al., 2012; Southward et al., 2014; Yan et al., 2018). Interestingly, a longitudinal study by Spanos et al. (2010) found baseline levels of binge eating predicted later increases in parent–child conflict (rather than the reverse), suggesting that parenting behaviors are partially influenced by the child's behavior. Examination of these type of child effects on parenting and etiologic shifts was beyond the scope of this study, but future studies should examine these processes and other mechanisms that may underlie parenting/binge‐eating associations.

Although this study had several strengths (e.g., large population‐based sample, multiple measures of binge eating and parenting), there were also some limitations. Our data were cross‐sectional, and parenting/binge‐eating associations may change across pubertal maturation within an individual rather than across individuals at different pubertal stages. It will be important for future longitudinal research to examine directionality of associations given that previous studies have shown that engaging in binge eating can lead to later increases in “negative” parenting (e.g., high parent–child conflict; low parent involvement) rather than negative parenting leading to increased binge eating (Korotana et al., 2018; Spanos et al., 2010).

We examined binge‐eating symptoms with a self‐report questionnaire in a community‐based sample. Self‐report questionnaires can result in an overreporting of binge‐eating behaviors in nonclinical samples (Berg et al., 2011), making the generalizability of our findings to clinical populations unclear. Although many eating disorder symptoms exist on a continuum (Luo et al., 2016), and a wide range of binge‐eating behaviors were reported in our sample (see Table 2), additional studies using interview‐based measures and clinical samples are needed.

Although we collected twin and parent reports on the PEQ conflict scale, we only had twin self‐reports on the PBI. Self‐reports of parenting can be influenced by twins' characteristics and emotional states (Mastrotheodoros et al., 2020; Parent et al., 2014). Moreover, the PBI care and overprotection items asked twins to describe parenting received over their lifetime, which may have made it more difficult to detect pubertal differences in parenting/binge‐eating associations. However, given that results were the same for analyses of the PEQ conflict scale (that measured only current levels of conflict), it seems unlikely that our findings were unduly influenced by the PBI lifetime timeframe. Finally, our study focused on general parenting rather than parental feeding practices (e.g., using food as reward) that have been linked to binge eating in past work (Braden et al., 2014; Matheson et al., 2015). Future research should examine whether feeding practices, in addition to general parenting behaviors, show differential associations with binge eating across development using multiple subjective (e.g., perceived parenting) and objective (e.g., live observations) measures.

AUTHOR CONTRIBUTIONS

Carolina Anaya: Conceptualization; formal analysis; visualization; writing – original draft; writing – review and editing. Kelly L. Klump: Conceptualization; data curation; funding acquisition; investigation; methodology; project administration; resources; supervision; writing – review and editing. S. Alexandra Burt: Funding acquisition; supervision; writing – review and editing.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

Supporting information

APPENDIX S1 Supplemental Material

Anaya, C. , Burt, S. A. , & Klump, K. L. (2022). An investigation of associations between parenting and binge eating across pubertal development in pre‐adolescent and adolescent female participants. International Journal of Eating Disorders, 55(12), 1811–1823. 10.1002/eat.23818

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon request.