Temperament and Psychopathology in Early Childhood Predict Body Dissatisfaction and Eating Disorder Symptoms in Adolescence

Abstract

Eating disorders (ED) are highly impairing and dangerous conditions that typically onset in adolescence. However, very few prospective studies have examined early childhood risk factors for ED pathology. Given well-established links between temperament and psychopathology, examination of these factors could inform prevention efforts. The current multi-method, multi-informant prospective longitudinal study tested whether laboratory-observed and parent-reported temperament and psychiatric disorders at ages 3 and 6 (N = 609) predict body dissatisfaction at ages 12 and 15 and dimensional symptoms of EDs (anorexia nervosa [AN] and bulimia nervosa [BN]) at age 15 (n = 458) in a community sample. Results indicated that early childhood temperament (positive and negative emotionality, perceptual sensitivity, impulsivity, less shyness) and childhood psychopathology (anxiety, oppositional defiant, attention deficit/hyperactivity, and depressive disorders), predicted body dissatisfaction in adolescence. In addition, childhood perceptual sensitivity and oppositional defiant and depressive disorders predicted AN symptoms. Demographic characteristics (female sex, lower levels of fathers’ education, and parental marital status) in childhood predicted body dissatisfaction and AN symptoms. No temperament or psychopathology variables predicted BN symptoms. This study is an important first step toward continuing to identify areas of focus for future research on early childhood risk factors for ED symptoms and body dissatisfaction.

1. Introduction

Eating disorders (EDs) are serious mental illnesses that are associated with high mortality, morbidity, and societal and personal costs (Arcelus et al., 2011; Stice et al., 2013; Swanson et al., 2011). EDs typically onset in adolescence, between the ages of 12–15, with estimates ranging from 6.1% to 13% of female youth receiving a diagnosis of anorexia nervosa (AN), bulimia nervosa (BN), binge eating disorder (BED), or a subthreshold diagnosis (Stice et al., 2013; Stice et al., 2009). Notably, subthreshold disordered eating is associated with near similar levels of impairment as full-threshold diagnoses (Allen et al., 2013; Swanson et al., 2011). Despite high rates of EDs and associated impairment, very little is known about which early features might predict risk for later development of an ED. Early childhood temperament and psychopathology are promising factors to pursue as potential predictors of EDs given their presence prior to the average age of onset of EDs and the well-documented associations between these factors and later psychopathology in adolescence and adulthood (Clark, 2005; Klein et al., 2012). Identification of early temperamental and/or psychiatric risk factors of ED symptoms and related constructs such as body dissatisfaction could lead to the development of models of risk for EDs as well as targeted prevention efforts in childhood that minimize the likelihood of the onset of EDs and body dissatisfaction. The present study investigates children’s temperament and psychiatric disorders at ages 3 and 6 as potential predictors of body dissatisfaction at ages 12 and 15 and ED symptoms at age 15.

Most research on risk for ED symptoms has focused identifying risk factors in late childhood (ages 10–12) and early adolescence (ages 12–15) (Stice et al., 2011; Stice & Whitenton, 2002). A host of factors predict EDs and related features (Thompson & Stice, 2001), including depression, body dissatisfaction, dieting (Stice et al., 2011), anxiety (Schaumberg et al., 2019), negative emotionality (Stice et al., 2001; Stice et al., 2017), and impairment, overeating, and fasting (Stice et al., 2017). Meta-analytic reviews in late childhood and early adolescence support the role of thin-ideal internalization, negative emotionality, and lack of social support as ED risk factors (Stice, 2002). In particular, this research has identified body dissatisfaction as both a risk factor as well as an intermediate phenotype (occurring during the prodromal phase) of EDs, occurring prior to ED onset (Stice et al., 2010; Stice & Whitenton, 2002). Body dissatisfaction has also been found to magnify the strength of the associations between other risk factors and EDs (Stice & Desjardins, 2018). Research shows that risk factors for EDs also predict body dissatisfaction. For example, adiposity, perceived pressures to be thin, internalization of the thin ideal, and social support deficits predict later body dissatisfaction (Rohde et al., 2015; Stice & Whitenton, 2002). Further, body dissatisfaction can be highly impairing even in the absence of a full threshold ED illness (Mond et al., 2013). Therefore, body dissatisfaction is an essential construct to examine as both a risk factor and outcome of interest when assessing risk for EDs and related pathology and impairment. Although most research specifies late childhood/early adolescent ED risk and intermediary factors, identification of factors in early childhood, well before the average age of onset, is needed.

To date, there have been few prospective studies have examined early childhood (birth to age 8) risk factors. Female sex, body dissatisfaction, sleeping and feeding problems, familial feeding environment, parental beliefs about children’s weight, and maternal psychopathology and eating pathology predict later EDs and symptoms (Hafstad et al., 2013; Kotler et al., 2001; Marchi & Cohen, 1990; Micali, 2005; Nicholls & Viner, 2009). However, some studies have failed to find the same patterns (Hafstad et al., 2013). Although this research begins to identify some potential early risk factors, more work is needed to determine if assessments in early childhood can predict later ED symptoms and body dissatisfaction. Importantly, hypothesized risk factors have not been tested prospectively. For example, in meta-analytic work on temperament/personality and EDs, traits such as perfectionism, neuroticism, sensitivity to social rewards, and reduced extraversion are associated with EDs (Farstad et al., 2016). Yet, there is no literature testing if early temperament predicts later ED symptoms and body dissatisfaction. Given well-established links between temperament and psychopathology (Clark, 2005; Klein et al., 2012), it is important to examine the role of temperament in risk for EDs.

Furthermore, childhood psychopathology may also relate to risk for EDs. A large body of literature suggests that anxiety plays a role in the development of EDs (Levinson & Rodebaugh, 2012; Pallister & Waller, 2008; Schaumberg et al., 2019). There are high rates of co-occurrence among anxiety disorders and ED in adolescence and beyond (Bulik, 2002; Kaye et al., 2004; Pallister & Waller, 2008) and anxiety in late childhood is a risk factor (Schaumberg et al., 2019), yet empirical work delineating if early childhood anxiety relates to later EDs and body dissatisfaction is lacking. Finally, many established risk factors, such as depression, have been examined only in late childhood or early adolescence. The identification of risk factors for EDs earlier than previously identified could inform prevention efforts to reduce the likelihood of the onset of these very impairing and dangerous conditions.

We are not aware of any conceptual and empirical models of early childhood temperament and psychopathology in predicting risk for adolescent EDs and body dissatisfaction. However, the established links between childhood temperament (Caspi et al., 1995; Caspi et al., 1996; Forbes et al., 2017; Frenkel et al., 2015) and psychopathology (Copeland, Angold, et al., 2014; Copeland, Shanahan, et al., 2014; Groenman et al., 2017; Sorcher et al., in press) and a range of symptoms and disorders (e.g., anxiety, depression, substance use disorders) in adolescence and adulthood, as well as the overlap between these symptoms/disorders and EDs (Bahji et al., 2019; Bulik, 2002; Puccio et al., 2016), suggest that childhood temperament and psychopathology are likely also related to EDs and related outcomes like body dissatisfaction as well. Empirical investigation of childhood predictors of these outcomes in early- to mid-adolescence can contribute to better understanding of risk for EDs and, therefore, more targeted prevention efforts in childhood.

Thus, the goals of the current multi-method, multi-informant prospective longitudinal study were to explore whether temperament and psychiatric disorders assessed at ages 3 and 6 predict body dissatisfaction assessed at ages 12 and 15 and symptoms of EDs (anorexia nervosa and bulimia nervosa) assessed at age 15. Sex and other demographic variables (race/ethnicity and parental marital status and education) were also examined as predictors. We hypothesized, based on prior literature, that temperamental negative emotionality (NE), anxiety and depressive disorders, and female sex would be the strongest predictors of body dissatisfaction at age 12 and age 15 and ED symptoms at age 15. However, given the lack of data, we did not have a priori hypotheses about other temperamental characteristics, such as impulsivity, anger, and discomfort, nor do we have developmental hypotheses with regard to which predictors would be associated with each outcome at each time point. Instead, we aimed to test a range of potential temperamental characteristics and psychiatric disorders with the goal of beginning to identify areas of focus for future research on early childhood risk factors for ED symptoms and body dissatisfaction.

2. Method

2.1. Participants

Parents of 3-year-old children were recruited from a suburban community as part of a longitudinal study of risk factors for psychopathology (Klein & Finsaas, 2017). An additional 50 children from underrepresented minority groups were recruited at age 6 to increase the diversity of the sample. The present study reports on data from four assessment phases of the study: early childhood (age 3: N = 559 and age 6: n = 502), early adolescence (age 12: n = 443), and adolescence (age 15: n = 458). Eligibility for the study included having a 3-year-old child, living within 20 contiguous miles of Stony Brook, NY, and at least one biological parent who could read/speak English. Children could not have any significant medical or developmental disorders. Sociodemographic characteristics for the 609 participants who entered the study are presented in Table 1.

Table 1.

Sociodemographic Characteristics of the Study Sample and Descriptives for Temperament and Psychopathology Variables in Early Childhood and Body Dissatisfaction and Eating Disorder Symptom Variables in Adolescence

Sociodemographic Characteristics (N = 609 unless otherwise noted)		n	%
Sex	Female	277	45.5
	Male	332	54.5
Race	Asian	14	2.3
	Black/African American	49	8.0
	Native American	1	0.2
	White	542	89.0
	Other	3	0.5
Ethnicity	Hispanic or Latino	60	9.9
	Not Hispanic or Latino	549	90.1
Parents’ marital status1	Married	507	89.7
	Not/never married	58	10.3
Parents’ level of education2
Mother	Four-year college degree or higher	314	53.4
	Did not have four-year college degree	274	46.6
Father	Four-year college degree or higher	266	45.2
	Did not have four-year college degree	322	54.8

Descriptives for Study Variables	M or n	SD or %	Min-Max	Reliability (α)
Laboratory-Observed Temperament
Age 3
Positive Emotionality	.00	1.80	−6.96–6.09	.87
Negative Emotionality	.00	.55	−.82–3.48	.82
Age 6
Positive Emotionality	.00	.84	−2.03–3.06	.83
Negative Emotionality	.00	.56	−.70–4.28	.71
Parent-Reported Temperament				Parent (Coparent)
Age 3
Anger/Frustration	4.50	.70	2.42–6.31	.79 (.77)
Discomfort	4.12	.72	1.83–6.00	.68 (.67)
Fear	3.91	.80	2.43–6.67	.74 (.66)
Impulsivity	4.65	.67	2.43–6.67	.75 (.65)
Perceptual Sensitivity	5.02	.64	3.02–7.00	.74 (.71)
Sadness	3.88	.60	1.83–6.13	.63 (.66)
Shyness	3.43	1.10	1.08–6.92	.92 (.91)
Age 6
Anger/Frustration	4.26	.80	1.69–6.50	.82 (.82)
Discomfort	4.02	.73	1.58–5.96	.70 (.71)
Fear	3.83	.84	1.00–5.92	.75 (.71)
Impulsivity	4.45	.66	2.42–6.50	.75 (.74)
Perceptual Sensitivity	5.05	.67	2.83–6.92	.74 (.76)
Sadness	3.85	.64	1.42–5.75	.66 (.68)
Shyness	3.22	1.15	1.00–6.69	.93 (.92)
Child Psychopathology
Age 3 3
Anxiety disorder	106	19.59	-	-
Depressive disorder	10	1.84	-	-
ADHD	11	2.03	-	-
ODD	51	9.43	-	-
Age 6 4
Anxiety disorder	81	15.70	-	-
Depressive disorder	26	5.04	-	-
ADHD	30	5.81	-	-
ODD	44	8.53	-	-
Adolescent Body Dissatisfaction and Eating Disorder Symptoms
Body dissatisfaction: Age 12	5.31	5.44	0–25	.79
Body dissatisfaction: Age 15	6.97	5.43	0–25	.79
Anorexia Nervosa symptoms	.15	.68	0–6	.69
Bulimia Nervosa symptoms	.08	.57	0–7	.79

Note.

¹n = 565;

²n = 588;

³n = 541;

⁴n = 516;

ADHD = Attention-deficit/hyperactivity disorder; ODD = Oppositional defiant disorder

Attrition analyses examined whether participants who did and did not participate in the follow-up assessments differed on any age 3 and 6 variables. Compared to children in families who participated in the follow-up, children who did not participate had lower age 6 laboratory-observed positive emotionality (PE) and higher age 3 parent-reported anger/frustration, discomfort, fear, and sadness (see Supplementary Table 1). In addition, of the families who did not participate at follow-up, youth were more likely to be male (80/128; 62.5%) than female (48/128; 37.5%), χ² (1, N=609)=4.17, p=.04, and were more likely to have mothers who did not graduate college (65/111; 58.6%) than mothers graduated (46/111; 41.4%), χ² (1, N = 588)=7.87, p=.01.

The human subjects review committee at Stony Brook University approved the study. At each assessment, informed consent was obtained from parents, assent was obtained from children, and families were compensated for their participation.

2.2. Measures

2.2.1. Childhood Predictors

Observed Temperament at Age 3 and Age 6.

Children’s temperament was assessed using a structured laboratory observation, the Laboratory Temperament Assessment Battery (Lab-TAB)(Goldsmith et al., 1995), which consists of standardized tasks designed to elicit various emotions and behaviors related to temperament. For the present study, we coded expressions of PE and NE (including fear, sadness, and anger). Observations were videotaped through a one-way mirror and recorded for coding.

Positive and Negative Emotionality.

At age 3, 12 Lab-TAB tasks were administered; examples of tasks included “Impossibly Perfect Green Circles” in which the experimenter repeatedly asked the child to draw a circle on a large piece of paper, mildly criticizing each attempt (to elicit sadness and anger) and “Pop-up Snakes” in which the child and experimenter surprised the parent with a can of potato chips that actually contained coiled pretend snakes (to elicit PE and fear). At age 6, nine Lab-TAB tasks were administered; examples of tasks included “Mixed-Up Puzzles” in which the child was told a puzzle was “really easy,” but pieces were missing, making it an impossible puzzle to complete (to elicit sadness and anger) and “Story Time” in which the child told a story to a “story expert” (to elicit fear). See (Dyson et al., 2015) for descriptions of all tasks administered at age 3 and age 6.

Coders included undergraduate and graduate students and study staff who were assigned specific episodes and trained extensively in the coding procedures. Once coders reached at least 80% agreement with an experienced rater, they coded independently. At both time points, observations were coded for affective expressions related to temperament. Although tasks were designed to elicit specific aspects of temperament, all emotional and behavioral expressions were coded in all tasks to capture temperament across a range of situations. Each instance of facial, bodily, and vocal PE, fear, sadness, and anger was rated as low, moderate, or high in intensity. To compute scores, ratings were weighted by intensity and summed within each channel (facial, bodily, vocal) for each aspect of temperament within each episode. Ratings were averaged by channel across all episodes, standardized, and averaged across channels to yield continuous PE and NE (fear, sadness, anger) scores at ages 3 and 6. Second raters coded 35 videotapes; at age 3, the interrater intraclass correlation coefficient (ICC) was .92 for PE and .74 for NE; at age 6, the interrater ICC was .94 for PE and .81 for NE. See Table 1 for descriptives.

Parent-Reported Temperament at Age 3 and Age 6.

Temperament was also assessed using mother- and father-reports on the Child Behavior Questionnaire (CBQ) (Rothbart et al., 2001), a widely used measure to assess temperament in 3- to 7-year-old children with 194 items. The CBQ was administered at the age 3 (n=518 mothers, n=402 fathers) and age 6 (n=466 mothers, n=369 fathers) assessments. The CBQ includes narrow-band scales that can be summarized by three broad dimensions (Negative Emotionality, Extraversion/Surgency, Effortful Control). We examined a subset of the narrow-band scales to try to identify more specific associations with ED variables. The following subscales were selected given their potential relevance to risk for EDs: Anger/frustration, Discomfort, Fear, Impulsivity, Perceptual sensitivity, Sadness, and Shyness. Each subscale included 12–13 items; coefficient alphas (α) ranged from .63–92 (median=.74) at age 3 and .66–.93 (median=.75) at age 6 (see Table 1 for specific values). We averaged mother and father reports within each assessment (the median correlation between mother and father reports was .44 at age 3 and .57 at age 6). See Supplementary Table 2 for results separated by mother- and father-reports. See Table 1 for descriptives.

Psychopathology at Age 3 and Age 6.

The Preschool Age Psychiatric Assessment (PAPA) (Egger et al., 1999) was administered to parents to assess psychopathology when children were age 3 (n = 541) and age 6 (n = 516). For the present study, four broad diagnostic categories were included: 1) any anxiety disorder (specific phobia, separation anxiety, social phobia, generalized anxiety disorder, agoraphobia, selective mutism); 2) any depression disorder (major depressive disorder, dysthymia, or depression not otherwise specified [NOS]); 3) attention-deficit/hyperactivity disorder (ADHD), and 4) oppositional defiant disorder (ODD). Interviews were conducted by doctoral students in clinical psychology at the age 3 assessment and by a doctoral student and a master’s-level clinician at age 6. Interrater reliability was assessed at age 3 (n = 21) and age 6 (n = 25): kappas were excellent at age 3 (1.00 for all diagnostic categories) and acceptable to excellent at age 6 (.64–.89). See Table 1 for descriptives.

2.2.2. Adolescent Outcomes

Body Dissatisfaction: Age 12 and Age 15.

The body dissatisfaction subscale from the Minnesota Eating Behavior Survey (MEBS) (von Ranson et al., 2005) was administered to youth at age 12 and age 15 (n = 428 at both assessments). This subscale included five items and assessed youth’s self-reported displeasure with their body size and shape (e.g., “I’m always wishing I was thinner”). Each item was rated on a six-point Likert scale ranging from never to always. Higher scores reflect greater body dissatisfaction. This measure demonstrates good psychometric properties with both males and females (Klump et al., 2012; von Ranson et al., 2005). At age 12, 72.2% of youth endorsed at least one item; at age 15, 82.0% of youth endorsed at least one item. See Table 1 for descriptives.

Eating Disorder Symptoms: Age 15.

The Kiddie Schedule for Affective Disorders and Schizophrenia Present and Lifetime Version (K-SADS-PL) (Kaufman et al., 1997) was administered when youth were age 15 to n = 458 children and one of their parents. Parents (first) and children (second) were interviewed separately by doctoral students in clinical psychology and a master’s-level clinician supervised by a clinical psychologist and child psychiatrist. Interviewers were unaware of other data. DSM-IV anorexia nervosa (AN) and bulimia nervosa (BN) symptoms were assessed in the prior three years. Combined parent and child reports were used to derive summary ratings for each symptom. Symptoms were rated on a 3-point scale (0=not present; 1=subthreshold; 2=threshold) and summed to create an AN scale and a BN scale. The present study includes dimensional symptom scale scores as only 7 (1.5%) adolescents met criteria for an ED diagnosis, and because eating disorders are likely best conceptualized dimensionally (Wildes & Marcus, 2013). Interrater reliability was not computed because there was insufficient variability in ED symptoms in the reliability sample of 25 cases. The range of AN symptoms in the sample was 0–6 symptoms; 6.6% of the sample reported one or more AN symptoms. The range of BN symptoms was 0–7 symptoms; 2.9% of the sample reported one or more BN symptoms. See Table 1 for additional descriptives.

2.3. Data Analytic Plan

Zero-order correlations were conducted among all continuous predictors (laboratory-observed and parent-reported early childhood temperament assessed at age 3 and age 6) and outcomes (body dissatisfaction assessed at age 12 and age 15; AN and BN symptoms assessed at age 15); point-biserial correlations were conducted for dichotomous predictors (demographic variables; age 3 and age 6 diagnoses). To examine unique associations between demographic, temperament, and psychopathology in early childhood and body dissatisfaction and ED symptoms in adolescence, significant bivariate predictors were entered into the same multivariate linear regression model with each relevant outcome. Finally, to examine relations between all significant predictors (i.e., all predictors of each DV with significant associations) and each outcome, three sets of multivariate linear regressions were run. Across both sets of regressions, Full Information Maximum Likelihood (FIML) was used to account for attrition and missing data and we report explanatory power (R²) for each multivariate model. In addition, we calculated effect size r [sqrt((t²)/((t²) + (n + number of predictors − 1))] for each predictor in the multivariate analyses.

3. Results

3.1. Demographic predictors of body dissatisfaction and ED symptoms.

Several demographic variables were significantly associated with adolescent body dissatisfaction and ED symptoms (see Table 2). Specifically, females reported more body dissatisfaction at age 15 and more symptoms of AN (but not BN) than males. Children of parents who were not married in early childhood reported greater body dissatisfaction at age 15 than children of parents who were married. Finally, youth with fathers who did not graduate college reported greater body dissatisfaction at age 12 and age 15 and more BN symptoms at age 15 compared to youth with fathers who did graduate college. There were no significant associations with child race/ethnicity and mothers’ level of education.

Table 2.

Correlations Between Demographic Variables, Temperament, and Psychopathology in Early Childhood and Body Dissatisfaction and Eating Disorder Symptoms in Adolescence

	Body Dissatisfaction and Eating Disorder Symptoms
	Age 12	Age 15
Early Childhood Predictors	Body Dissat.	Body Dissat.	AN Sx	BN Sx
Demographic variables assessed at age 3 1
Sex2	.09	.26 ***	.15 **	.09
Child Race/Ethnicity3	.02	.01	−.05	−.04
Parental Marital Status4	−.08	−.16**	−.02	.02
Mothers’ Level of Education5	−.06	−.05	−.05	−.02
Fathers’ Level of Education5	−.22***	−.25***	−.09	−.11*
Laboratory-Observed Temperament
Age 3
Positive Emotionality	−.01	−.08	−.07	.02
Negative Emotionality	.08	.01	−.02	−.02
Age 6
Positive Emotionality	.16 **	.01	.02	−.02
Negative Emotionality	.12 *	.03	.02	.01
Parent-Reported Temperament
Age 3
Anger/Frustration	.01	−.03	.03	−.03
Discomfort	−.03	.02	−.01	.02
Fear	−.02	.00	−.01	.01
Impulsivity	.05	.06	.00	−.01
Perceptual Sensitivity	.01	.11 *	.13 *	.03
Sadness	−.02	−.08	−.02	.01
Shyness	−.08	−.12*	−.05	−.07
Age 6
Anger/Frustration	.08	.04	.06	.01
Discomfort	.08	.08	.01	.06
Fear	.07	.05	.01	−.01
Impulsivity	.11 *	.05	.05	.08
Perceptual Sensitivity	−.04	.01	.15 **	.03
Sadness	.07	.05	.04	−.02
Shyness	−.07	−.06	−.04	−.08
Child Psychopathology
Age 3
Anxiety disorder	.12 *	.06	.03	−.03
Depressive disorder	.02	−.03	−.03	−.02
ADHD	−.02	−.06	−.01	−.01
ODD	14 **	.02	.14 **	.06
Age 6
Anxiety disorder	.04	−.05	−.04	−.06
Depressive disorder	.04	.08	.11 *	.02
ADHD	.12 *	−.03	.00	.06
ODD	.13 **	.12 *	.07	.01

Note. Ns ranged from 393–458.

¹Variables were reported at the age 6 assessment for the 50 families that joined the study at that time;

²Male = 1; Female = 2;

³White/non-Hispanic = 1; non-white and/or Hispanic = 2;

⁴Parents not married = 1; Parents married = 2;

⁵Mother/Father did not graduate four-year college = 1; Mother/Father graduated four-year college = 2;

Body Dissat. = body dissatisfaction; AN = anorexia nervosa; BN = bulimia nervosa; Sx = symptoms; ADHD = Attention-deficit/hyperactivity disorder; ODD = Oppositional defiant disorder;

^*p < .05;

^**p < .01;

^***p < .001

3.2. Temperament predictors of body dissatisfaction and ED symptoms.

Early childhood laboratory-observed and parent-reported temperament assessed at ages 3 and 6 were significantly associated with adolescent body dissatisfaction and ED symptoms (Table 2). Specifically, greater laboratory-observed PE and NE at age 6 predicted body dissatisfaction at age 12. Greater parent-reported impulsivity at age 6 was associated with body dissatisfaction at age 12, and lower levels of parent-reported shyness at age 3 were associated with body dissatisfaction at age 15. Finally, greater parent-reported perceptual sensitivity at age 3 predicted body dissatisfaction at age 15. In addition, greater parent-reported perceptual sensitivity at age 3 and age 6 predicted AN symptoms at age 15. A similar pattern was identified with separate mother- and father-reported temperament variables (Supplementary Table 2).

3.3. Psychopathology predictors of body dissatisfaction and ED symptoms.

Early childhood psychopathology assessed at ages 3 and 6 was significantly associated with adolescent body dissatisfaction and ED symptoms (Table 2). Children with diagnoses of anxiety disorders and ODD at age 3 and ADHD and ODD at age 6 reported greater body dissatisfaction at age 12. In addition, ODD at age 6 predicted body dissatisfaction at age 15. Finally, age 3 ODD and age 6 depression diagnoses predicted AN symptoms at age 15.

3.4. Multivariate models with demographic, temperament, and psychopathology predictors of body dissatisfaction and ED symptoms.

Significant demographic, laboratory-observed and parent-reported temperament, and psychopathology variables were entered into the same multivariate model for each relevant outcome using FIML. There were no bivariate temperament or psychopathology predictors of BN symptoms in the primary analyses; therefore no models were run with this DV.

In a model with significant predictors of age 12 body dissatisfaction (fathers’ education, laboratory-observed age 6 PE and NE, age 6 parent-reported impulsivity, age 3 anxiety disorder and ODD, age 6 ADHD and ODD), lower levels of fathers’ education, age 6 laboratory-observed NE, and age 3 anxiety disorder remained significant predictors; the model explained 9.6% of variance in age 12 body dissatisfaction. In a model with significant predictors of age 15 body dissatisfaction (sex, parental marital status, fathers’ education, age 3 parent-reported perceptual sensitivity and shyness, age 6 ODD), female sex, unmarried parents, lower levels of fathers’ education, lower levels of age 3 parent-reported shyness, and age 6 ODD remained significant predictors of body dissatisfaction at age 15 (model R² = 16.8%).

In a model with significant predictors of AN symptoms (sex, age 3 and age 6 parent-reported perceptual sensitivity, age 3 ODD, age 6 depressive disorders), female sex, age 3 ODD, and age 6 depression remained significant predictors of AN symptoms (model R² = 6.5%).

Finally, three additional models were run to examine associations among all significant predictors in early childhood and each adolescent outcome (Table 4). In the model with age 12 body dissatisfaction, lower levels of fathers’ education and age 3 anxiety disorder remained as significant unique predictors (model R² = 10.9%). In the model with age 15 body dissatisfaction, female sex, unmarried parents, lower levels of fathers’ education, greater levels of perceptual sensitivity at age 3, lower levels of perceptual sensitivity at age 6, and lower levels of shyness at age 3 were significant (model R² = 19.6%). In the model with AN symptoms, female sex, age 3 ODD, and age 6 depression each continued to uniquely predicted AN symptoms at age 15 (model R² = 7.4%).

Table 4.

Multivariate models with all significant demographic, temperament, and psychopathology predictors of each body dissatisfaction and ED symptoms outcome

Age 12 Body Dissatisfaction
Demographic & early childhood predictors	B	SE	z	p	r
Sex	.91	.52	1.77	.077	.07
Parental marital status	−.74	.91	−.82	.415	.03
Fathers’ education	−1.91	.53	−3.58	<.001	.15
Age 6 laboratory-observed PE	.29	.34	.83	.405	.03
Age 6 laboratory-observed NE	.82	.45	1.84	.066	.08
Age 3 parent-reported perceptual sensitivity	.33	.54	.61	.541	.03
Age 3 parent-reported shyness	−.19	.31	−.61	.545	.03
Age 6 parent-reported impulsivity	.18	.54	.33	.742	.01
Age 6 parent-reported perceptual sensitivity	−.79	.52	−1.52	.128	.06
Age 3 anxiety disorder	1.68	.71	2.38	.017	.10
Age 3 ODD	1.58	1.00	1.59	.113	.06
Age 6 depressive disorder	−1.27	1.36	−.94	.348	.04
Age 6 ADHD	1.34	1.23	1.09	.277	.04
Age 6 ODD	1.33	1.08	1.23	.217	.05
Age 15 Body Dissatisfaction
Sex	2.85	.50	5.71	<.001	.23
Parental marital status	-2.54	.89	-2.86	.004	.12
Fathers’ education	−2.19	.51	−4.31	<.001	.17
Age 6 laboratory-observed PE	−.55	.33	−1.67	.095	.07
Age 6 laboratory-observed NE	.09	.44	.21	.831	.01
Age 3 parent-reported perceptual sensitivity	1.25	.50	2.51	.012	.10
Age 3 parent-reported shyness	−.83	.29	−2.90	.004	.12
Age 6 parent-reported impulsivity	.14	.50	.29	.773	.01
Age 6 parent-reported perceptual sensitivity	−1.21	.49	−2.45	.014	.10
Age 3 anxiety disorder	1.20	.70	1.73	.084	.07
Age 3 ODD	.53	.91	.59	.558	.02
Age 6 depressive disorder	1.00	1.24	.81	.420	.03
Age 6 ADHD	−2.01	1.14	−1.77	.077	.07
Age 6 ODD	1.34	.99	1.36	.174	.06
Age 15 Anorexia Nervosa Symptoms
Sex	.17	0.06	2.91	.004	.12
Parental marital status	−.03	.11	−.22	.823	.01
Fathers’ education	−.07	.07	−.94	.348	.03
Age 6 laboratory-observed PE	−.04	.04	−.94	.348	.04
Age 6 laboratory-observed NE	.01	.06	.26	.793	.01
Age 3 parent-reported perceptual sensitivity	.06	.06	1.04	.298	.04
Age 3 parent-reported shyness	−.04	.04	−.97	.330	.04
Age 6 parent-reported impulsivity	.02	.06	.25	.799	.01
Age 6 parent-reported perceptual sensitivity	.07	.06	1.11	.269	.05
Age 3 anxiety disorder	.00	.08	.01	.989	.00
Age 3 ODD	.33	.11	2.90	.004	.12
Age 6 depressive disorder	.29	.15	2.00	.045	.08
Age 6 ADHD	−.17	.14	−1.23	.218	.05
Age 6 ODD	.03	.12	.21	.830	.01

Note. N = 609; PE = Positive Emotionality; NE = Negative Emotionality; ADHD = Attention-deficit/hyperactivity disorder; ODD = Oppositional defiant disorder. Significant associations are in bold font.

4. Discussion

The present study is, to our knowledge, the first prospective longitudinal investigation of temperament and psychopathology in early childhood (ages 3 and 6) and demographic factors predicting eating disorder (ED) symptoms (age 15) and body dissatisfaction (ages 12 and 15) in adolescence. This study included multiple informants (mother, father, adolescent) and multiple methods (laboratory observations, diagnostic interviews, questionnaires) to assess risk for EDs in a large community sample. Given most ED symptoms develop in adolescence, it is crucial to identify early predictors prior to ED symptom onset to clarify the role of predictors versus correlates of EDs.

We found associations between demographic variables (sex, parental marital status, fathers’ lower level of education), laboratory observations (NE, PE) and parent reports of temperament (perceptual sensitivity, less shyness, impulsivity) and diagnostic interviews assessing diagnoses (anxiety, depression, ADHD, ODD) with ED symptoms and body dissatisfaction. No temperament or psychopathology variables predicted BN symptoms. Some variables predicted body dissatisfaction at age 12 and not age 15 (and vice versa); for example, PE, NE, impulsivity, and ADHD predicted body dissatisfaction at age 12 but not age 15, whereas female sex, perceptual sensitivity, and shyness predicted body dissatisfaction at age 15 but not at age 12. These discrepancies may be due to developmental differences between early and mid-adolescence (e.g., biological/pubertal differences); additional research is needed to examine outcomes at different points in development.

4.1. Associations between demographic variables, body dissatisfaction, and ED symptoms.

Female sex, parental marital status (being unmarried), and lower levels of fathers’ education were associated with body dissatisfaction and ED symptoms, and remained significant predictors in multivariate models that included temperament and psychopathology variables. As hypothesized, females were more likely than males to report body dissatisfaction and AN symptoms, consistent with a multitude of research showing EDs are more common in females (Mitchison & Hay, 2014). In the present study, females reported, on average, five times as many AN symptoms as males.

In addition, children of parents who were not married in early childhood reported greater body dissatisfaction at age 15 compared to children of parents who were married. Children of fathers who did not graduate college reported greater body dissatisfaction at age 12 and age 15, and more BN symptoms compard to children of fathers who did graduate college; there were no associations between mothers’ education and any outcomes. It is possible that if fathers are often primary wage-earners, their education may be more directly tied to the family’s level of resources than mothers’ education. Similarly, children in families with unmarried parents (e.g., in single-parent households) may have fewer resources. Findings have been mixed with regard to sociodemographic characteristics and EDs. For example, in the National Comorbidity Survey-Replication Adolescent Supplement study, parental marital status and education were not associated with EDs in adolescents (Swanson et al., 2011); the authors did not appear to examine mothers’ and fathers’ level of education separately, or measure body dissatisfaction, which may account for the discrepancy with our results in which lower levels of fathers’ (but not mothers’) educational attainment was a robust predictor of body dissatisfaction. Also in contrast to our findings, several large population-based studies in Sweden have found greater parental educational attainment predicted EDs (Ahrén et al., 2013; Goodman et al., 2014). Finally, consistent with our findings, lower family socioeconomic status (a composite index of mothers’ and fathers’ education levels and income) was associated with more parent-reported disordered eating symptoms (even when accounting for neighborhood disadvantage) in a sample of nearly 3,000 female twins ages 8–17 (Mikhail et al., 2021). Although further clarification about the role of sociodemographic factors in the development of EDs is needed, the present study identified female sex and lower levels of fathers’ education and, to a lesser extent, parental marital status (the latter two factors possibly indicative of family socioeconomic status/level of resources) as unique predictors of body dissatisfaction and ED symptoms.

4.2. Associations between early childhood temperament and body dissatisfaction and ED symptoms in adolescence.

Greater levels of laboratory-observed higher PE and NE at age 6 predicted body dissatisfaction at age 12, though only NE (not PE) was associated with body dissatisfaction in one multivariate model. The positive bivariate association between PE and body dissatisfaction is inconsistent with the study hypothesis and research implicating low PE/extraversion as related to risk for eating disorders (Farstad et al., 2016; Glashouwer et al., 2014; Harrison et al., 2014; Haynos et al., 2020) and risk for depression and anxiety (Kotov et al., 2010; Naragon-Gainey et al., 2009; Olino et al., 2010), conditions commonly comorbid with EDs (Bulik, 2002; Kaye et al., 2004; Pallister & Waller, 2008). As EDs can continue to emerge over adolescence into young adulthood (Stice, 2002), the unexpected positive relation with PE may change with further development. Conversely, the association between NE and age 12 body dissatisfaction is consistent with the study hypothesis and the large literature on NE and psychopathology (Kotov et al., 2010; Tackett et al., 2013) and risk for EDs (Stice, 2001; Stice et al., 2017).

Temperament was also assessed when children were age 3 and age 6 using parent reports. Specifically, lower parent-reported shyness at age 3 predicted age 15 body dissatisfaction and parent-reported impulsivity at age 6 predicted age 12 body dissatisfaction. The finding that lower shyness predicted greater body dissatisfaction is inconsistent with literature that links social anxiety and EDs (Levinson & Rodebaugh, 2012; Pallister & Waller, 2008; Schaumberg et al., 2019). However, shyness and related constructs (e.g., behavioral inhibition) are likely heterogenous and not always predictive of later anxiety (Dyson et al., 2011; Klein & Mumper, 2018). In addition, this inverse association between early shyness and later body dissatisfaction, is consistent with our finding that more laboratory-observed PE predicted greater body dissatisfaction. It may be that youth who are more extraverted and less shy are more socially engaged with their peers, which could lead to more social comparison and self-consciousness about their bodies. Impulsivity could also be related to risk for the development of positive urgency (proneness to behave impulsively while experiencing positive emotions), which may be related to eating disorder risk (Claes et al., 2015; Farstad et al., 2016).

Perceptual sensitivity, which refers to the extent to which the child has a tendency to detect negligible or low intensity stimuli in the environment, was also a significant predictor. Specifically, parent-reported perceptual sensitivity at age 3 was associated with body dissatisfaction at age 15 and AN symptoms, and perceptual sensitivity at age 6 was associated with AN symptoms. Associations between early childhood perceptual sensitivity and AN symptoms have not, to our knowledge, been demonstrated longitudinally, but they are consistent with growing research that has identified interoceptive sensitivity, as well as interoceptive deficits, as a maintenance and risk factor for EDs (Brown et al., 2020; Eshkevari et al., 2014; Jenkinson et al., 2018; Merwin et al., 2010). In the multivariate model with all early childhood predictors, higher levels of perceputal sensitivity at age 3 continued to predict body dissatisfaction at age 15; conversely, lower levels of perceptual sensitivity at age 6 predicted body dissatisfaction at age 15. Given that this model included all significant predictors of any adolescent outcome (Table 4) and that there was no bivariate association between age 6 perceptual sensitivity and age 15 body dissatisfaction (Table 2), this association may reflect a supressor effect. Together, the seemingly contradictory findings may underscore that interoceptive deficits can include both heightened and/or blunted sensitivity. Our findings suggest that such perceptual sensitivity, previously documented as a predictor of risk for, presecence of, and recovery from EDs (Jenkinson et al., 2018; Leon et al., 1995) in older youth and adults, can be identified at a very young age and predict ED-related outcomes. Researchers should assess perceptual awareness in young children to identify whether these findings can be replicated. If so, a next step would be to develop and evaluate preventative interventions focused on normalizing bodily perceptions.

4.3. Associations between early childhood psychopathology and body dissatisfaction and ED symptoms in adolescence.

Age 3 anxiety disorder and ODD and age 6 ADHD and ODD predicted body dissatisfaction at age 12. In addition, age 6 ODD predicted body dissatisfaction at age 15, and age 3 ODD and age 6 depression predicted AN symptoms. In the multivariate models, early childhood anxiety and ODD uniquely predicted body dissatisfaction, and ODD and depression uniquely predicted AN symptoms even after accounting for other early childhood predictors.

The findings that ODD (e.g., defiance, arguing with others) and depression (e.g., irritability, sadness), along with laboratory-observed NE, predicted ED outcomes are consistent with the growing literature on the role of early childhood mood dysregulation/irritability as an important transdiagnostic risk factor for psychopathology (Brotman et al., 2017; Dougherty et al., 2015; Evans et al., 2017; Hawes et al., 2020; Humphreys et al., 2019; Stringaris et al., 2009; Vidal-Ribas et al., 2016). These results are also in line with cross-sectional and prospective evidence of associations between depression and EDs (Bulik, 2002; Puccio et al., 2016; Stice et al., 2011). These findings support the notion that early mood and behavioral difficulties may, in part, predispose some individuals to EDs; more research is needed to develop and test whether preventative interventions focused on mood and behavior dysregulation in young children limit the development of EDs in at-risk individuals. The lack of associations between parent-reported temperamental anger/frustration suggest that more severe expressions of mood difficulties identified via diagnostic interview may be more predictive of ED pathology than temperament. It is also possible that the diagnostic interview assessed these behaviors in more depth compared to the parent-reported questionnaires, or that the behavioral rather than the mood component of ODD was more predictive.

Anxiety disorders at age 3 predicted body dissatisfaction at age 12 in the bivariate and multivariate models. Anxiety is likely to be an important predictor given cross-sectional and prospective evidence of associations between anxiety and EDs in other studies (Levinson et al., 2018; Levinson & Rodebaugh, 2012, 2016; Levinson et al., 2017; Pallister & Waller, 2008; Schaumberg et al., 2019). Further, specific aspects of anxiety and related constructs that we did not measure (e.g., interoceptive fears; food-related fears; intolerance of uncertainty) may also be likely to increase risk for EDs (Schaumberg et al., 2021).

Age 6 ADHD predicted body dissatisfaction at age 12. This association was not significant in the multivariate model, perhaps due to overlap with ODD and parent-reported impulsivity in the model. There is prospective and cross-sectional evidence of links between ADHD and eating pathology/EDs (Biederman et al., 2007; Bleck et al., 2015; Levin & Rawana, 2016). In addition, body dissatisfaction has been identified as a moderator of the association between symptoms of ADHD and disordered eating (Curtin et al., 2013). Therefore, ADHD may reflect potential risk for EDs in combination with other risk factors, but more data are needed to identify the role of early childhood ADHD in risk for EDs.

4.4. Multivariate Models with All Significant Predictors

We examined unique associations between all significant early childhood predictors and ED symptoms and body dissatisfaction as well as the explanatory power of the models in multivariate analyses. Lower levels of fathers’ education and age 3 anxiety disorder continued to predict age 12 body dissatisfaction when accounting for all other predictors; this model explained 10.9% of the variance in age 12 body dissatisfaction. Female sex, unmarried parents, lower levels of fathers’ education, greater levels of perceptual sensitivity at age 3, lower levels of perceptual sensitivity at age 6, and lower levels of shyness at age 3 continued to predict 15 body dissatisfaction, explaining more variance (19.6%) compared to the other models. Finally, female sex, age 3 ODD, and age 6 depression each continued to uniquely predicted AN symptoms at age 15 (7.4% of the variance explained).

Across these models, and consistent with most research as mentioned above, lower levels of fathers’ education and female sex emerged as robust unique predictors. Further, early childhood temperamental characteristics (perceptual sensitivity, shyness) were prominent predictors of age 15 body dissatisfaction, whereas early childhood psychiatric disorders were more relevant predictors for age 12 body dissatisfaction (childhood anxiety) and AN symptoms (childhood depression, ODD). These differences could possibly reflect that links between psychiatric disorders in early childhood, body dissatisfaction in early adolescence, and AN symptoms are indicative of a more severe profile or pathway compared to links between temperamental characteristics in early childhood and body dissatisfaction in mid-adolescence. Body dissatisfaction can generally increase over adolescence, which was observed in the present study, though individual-level trajectories show varying patterns (Rodgers et al., 2016; Wang et al., 2019). In addition, the spectrum model of temperament/personality and psychopathology in childhood suggests that these constructs lie on a similar dimension with temperament reflecting a less severe end of the continuum and psychopathology reflecting a more severe end of the continuum (Tackett, 2006). Together, evidence of normative increases in body dissatisfaction and theories of the temperament-psychopathology spectrum support the speculation that the associations between a more severe predictor, psychiatric disorders, and earlier body dissatisfaction (and AN symptoms) may reflect a more concerning pattern versus associations between a less severe predictor, temperament, and later, more common body dissatisfaction identified in this sample.

4.5. Study Limitations

The findings should be interpreted with consideration of several limitations. First, given very few cases of ED diagnoses, diagnoses could not be included. In addition, very few associations with BN may reflect a later age of onset for BN versus AN. Future research should examine early childhood predictors of both symptoms and diagnoses in later adolescence and early adulthood when more cases of ED are observed (Stice, 2002). Given our ultimate goal to improve early detection of ED cases and related pathology, it is necessary to examine predictors in a non-clinical sample where base rates tend to be low. Second, only 20% of youth in the sample were from ethnic minority backgrounds; additional work is needed to investigate whether the associations identified in the present study would be found in more diverse samples. Third, attrition analyses revealed potential for bias in that youth who were male and had mothers who did not graduate college as well as those with lower laboratory PE scores and higher parent-reported anger/frustration, discomfort, fear, and sadness were less likely to participate in the follow-up assessments; these differences suggest that the remaining sample of participants may reflect a group at relatively lower risk. Our regression analyses included estimation of missing data to address this issue. Fourth, consistent with the concept of multifinality and the prevalence of comorbid mental health conditions, the temperament and psychopathology predictors identified in this study may not be specific to ED outcomes and would also predict other mental health difficulties. Fifth, the magnitudes of many associations were small. Given so little prospective research in this area, we sought to identify as many potential associations as possible. In addition, given the lengthy time span between assessments (the longest being from age 3 to 15), multiple informants (mothers, fathers, adolescents, and laboratory-observed behavior), and that various factors other than temperament and psychopathology (e.g., physiology; parenting behavior) contribute to ED risk, small effects are not surprising; restricting the p value to account for multiple tests would limit our ability to detect small but potentially important associations. Future work can identify whether particular predictors/combinations of predictors confer risk specific to EDs and account for more variance in risk. Future work can also examine whether the stability of constructs as well as mediating effects over time influence the outcomes of interest. Sixth, the median inter-rater agreement between mother and father ratings of children’s temperament was in the moderate range (.44 at age 3 and .57 at age 6). Although these levels are consistent with the average level of agreement identified in research using the same measure (Rothbart et al., 2001) and with most parent-report measures (Achenbach et al., 1987; De Los Reyes & Kazdin, 2005), stronger agreement could further enhance the validity of the findings. Seventh, some measures of youth temperament had modest internal consistency estimates that may have reduced power to identify significant associations with later ED dimensions. Finally, the few studies that investigated childhood predictors of EDs included some constructs that were not assessed in this study (e.g., childhood feeding problems, familial feeding environment).

4.5. Conclusions and Future Directions

Few prospective studies have examined whether risk factors in early childhood predict ED symptoms and body dissatisfaction; no studies, to our knowledge, have utilized a multi-method, multi-informant prospective approach in a large community sample starting in the preschool period to investiage risk for both ED symptoms and body dissatisfaction. In this study, we found associations between demographic variables (sex, parental marital status, fathers’ level of education), laboratory observations (NE, PE) and parent reports of temperament (e.g., perceptual sensitivity) and diagnostic interviews assessing diagnoses (e.g., ODD, depression) with ED symptoms and body dissatisfaction. This study is an important first step toward continuing to identify areas of focus for future research on early childhood risk factors for ED symptoms and body dissatisfaction. More research is needed on the contribution of demographic factors, perceptual sensitivity, PE, NE, impulsivity, anxiety, oppositionality, and depression in early childhood toward the development of EDs, as well as the role of intermediary effects. We demonstrated that factors related to risk for EDs can be identified prospectively in early childhood; it is possible that other behaviors related to EDs (e.g., perfectionism) and contexts (e.g., parent-child interactions) could be assessed in early childhood to detect additional vulnerabilities relevant to the recognition and prevention of EDs. The current study provides initial data on risk factors and provides a strong basis for additional assessment. Future research should consider incorporating additional assessment of ED-related behavior in childhood, such as food fear, fear of weight gain, deficits in interoceptive processing, as well as specific ED symptoms, such as binge eating, loss of control eating, and restriction. A more fine-grained assessment approach starting at an early age can identify specific risk factors worthy of intervention as well as chart more complex biopsychosocial developmental pathways that predict the onset of eating disorders. This work is essential to ultimately minimize the prevalence and the individual, familial, and societal burden of such serious and impairing disorders.

Table 3.

Multivariate models with significant bivariate demographic, temperament, and psychopathology predictors of each body dissatisfaction and ED symptoms outcome

	Age 12 Body Dissatisfaction
Demographic & early childhood predictors	B	SE	z	p	r
Fathers’ education	−1.97	.53	−3.73	<.001	.150
Age 6 laboratory-observed PE	.46	.33	1.41	.157	.06
Age 6 laboratory-observed NE	.89	.44	2.01	.045	.08
Age 6 parent-reported impulsivity	.28	.48	.59	.553	.02
Age 3 anxiety disorder	1.54	.70	2.21	.027	.09
Age 3 ODD	1.33	.99	1.34	.294	.06
Age 6 ADHD	1.28	1.21	1.06	.290	.04
Age 6 ODD	1.16	1.03	1.13	.260	.05
	Age 15 Body Dissatisfaction
Sex	2.53	.49	5.15	<.001	.21
Parental marital status	−2.38	.89	−2.69	.007	.11
Fathers’ education	−2.14	.51	−4.20	<.001	.17
Age 3 parent-reported perceptual sensitivity	.57	.40	1.42	.156	.06
Age 3 parent-reported shyness	−.55	.24	−2.33	.020	.10
Age 6 ODD	2.07	.89	2.33	.020	.10
	Age 15 Anorexia Nervosa Symptoms
Sex	.18	.06	2.80	.005	.11
Age 3 parent-reported perceptual sensitivity	.07	.06	1.11	.267	.05
Age 6 parent-reported perceptual sensitivity	.08	.06	1.28	.200	.05
Age 3 ODD	.32	.11	3.01	.003	.12
Age 6 depressive disorder	.29	.14	2.07	.038	.08

Note. N = 609; PE = Positive Emotionality; NE = Negative Emotionality; ADHD = Attention-deficit/hyperactivity disorder; ODD = Oppositional defiant disorder. Significant associations are in bold font.

Highlights.

• We examined early childhood predictors of body dissatisfaction and ED symptoms.

• Child temperament and psychopathology predict body dissatisfaction and AN symptoms.

• Child NE, shyness, anxiety, and ODD predict body dissatisfaction.

• Child perceptual sensitivity, depression and ODD predict AN symptoms.

• Female sex and less paternal education predict body dissatisfaction and AN symptoms.