Understanding early risk factors for eating disorder symptoms in adolescence: the role of body dissatisfaction, negative emotional reactivity and self-esteem at age 10–11 years

ABSTRACT

Objective

Risk factors associated with eating disorders (EDs) have been widely studied, although previous research has been limited to largely cross-sectional data or understanding risk factors in adult populations. Little is therefore known about the role of risk factors in early adolescence for the subsequent development of ED symptoms in adolescence.

Method

The present study examined factors identified in transdiagnostic models of EDs, in a sample of 2,372 adolescents who participated in Waves 3, 4 and 7 of the Longitudinal Study of Australian Children. The present study examined the association between body dissatisfaction, negative emotional reactivity and self-esteem in adolescents aged 10–11, and probable ED status measured via self-report at age 16–17, while adjusting for confounders measured at age 8–9 years.

Results

Negative reactivity and self-esteem at age 10–11 years were not significantly associated with ED status at age 16–17. Body dissatisfaction was associated with a decreased likelihood of meeting probable ED status at age 16–17, however this effect was small and likely not clinically significant, with only 2% of variance in probable ED status accounted for by body dissatisfaction.

Conclusion

Findings of the study suggest that transdiagnostic models of EDs may not capture the key risk factors for EDs in early adolescence.

Key points

What is already known about this topic:

1. Transdiagnostic models of EDs describe mood intolerance, interpersonal difficulties, clinical perfectionism and low self-esteem as maintenance factors for EDs

2. The validity of these contributing factors to EDs in adolescence has been supported in research using cross-sectional designs

3. Identifying ED risk factors in early adolescence can inform ED preventative programs in efforts to limit symptom progression

What this topic adds:

1. Transdiagnostic models of EDs may be limited in informing ED risk factors in early adolescence

2. ED risk factors may be sensitive to time and changes throughout adolescence

3. Alternative factors or interaction effects may be important in understanding ED risk in adolescence

KEYWORDS:

• Adolescent
• body dissatisfaction
• disordered eating
• emotion
• reactivity
• self-esteem

Introduction

Eating disorders (EDs) occur at a high rate in the population, 8.4% and 2.2% in females and males, respectively, and are also associated with a high mortality rate (Galmiche et al., 2019; Mitchison et al., 2020). The peak age for first diagnosis of an ED is 15–24 years, with evidence that problematic eating behaviour may present as early as 11–12 years of age (Micali et al., 2013; Morris et al., 2022). Once diagnosed, depending on the severity of EDs, they can be associated with poor response rates in treatment, and relapse is common (Halmi, 2013; Sala et al., 2023; Smith & Woodside, 2021). Research to date has largely focused on what maintains EDs in adulthood, with less on identifying risk factors that lead to onset in adolescence. Therefore, the aim of the present study was to examine three characteristics commonly associated with the development and maintenance of EDs in adults, body dissatisfaction, negative emotional reactivity and self-esteem (Fairburn et al., 2003), to examine if these characteristics in early adolescence (10–11 years) are associated with later experience of ED symptoms at age 16–17 years. Identifying the characteristics in early adolescence that increase risk for EDs may help reduce the prevalence of the disorder and identify key targets for prevention programs.

The Transdiagnostic Cognitive Behavioural Model of Eating Disorders (Fairburn et al., 2003) describes the common core psychopathology of EDs as a dysfunctional system of evaluating self-worth based on eating habits, shape and weight and one’s ability to exert control over them, which is maintained by four main psychological processes, including mood intolerance, interpersonal difficulties, clinical perfectionism and pervasive low self-esteem. There is evidence to suggest that this model may also help to explain the onset and maintenance of ED symptoms in adolescence. For example, Curzio et al. (2018) reported that perfectionism, low self-esteem, mood intolerance and overevaluation of shape and weight (measured through levels of body dissatisfaction and shape/weight concern) were associated with ED symptoms cross-sectionally in a sample of adolescents (n = 419), mean age 15 years, seeking treatment for EDs. While this suggests that these factors are associated with disordered eating in mid-adolescence, few longitudinal studies to date have examined the role of these factors in early adolescence in the development of later ED symptoms (e.g., Espinoza et al., 2019; Kenny et al., 2022). There is evidence to suggest that body dissatisfaction, negative emotional reactivity and self-esteem may also play a role in the onset of ED symptoms in adolescence. For example, Rohde et al. (2015) studied 496 American community adolescent girls who were followed annually from age 13 to 21, and found that elevated body dissatisfaction at age 13–16 was associated with 68% increased likelihood of development of EDs in the 4-year period following each assessment. While emotion regulation has been extensively studied in relation to EDs, it has been proposed that disordered eating behaviours may represent efforts to avoid intense and aversive cognitive and emotion states as a result of individuals’ elevated emotional reactivity (García-Grau et al., 2002; Smith et al., 2017). An exploratory study found that adolescents aged 17 who had been diagnosed with an ED (n = 94) exhibited elevated emotional reactivity, hence supporting the possible influence of emotional reactivity on the development of EDs (Nock et al., 2008). Finally, core low self-esteem is consistently identified as a key factor in the development and maintenance of EDs. A large number of studies have reported that for individuals diagnosed with an ED, a pervasive negative view of the self predicts engagement in weight-controlling behaviours in an attempt to achieve a desired thin body shape. Both cross-sectional and longitudinal studies have identified an association between low self-esteem and increased risk for psychiatrist-diagnosed EDs during adolescence (Cervera et al., 2003; Gual et al., 2002). The factors of body dissatisfaction, emotional reactivity and self-esteem are therefore considered likely contributors to the development of ED symptoms, but have not yet been tested in a longitudinal study over the peak years of ED development from early to mid-adolescence.

The present study examined the association between body dissatisfaction, negative emotional reactivity and self-esteem at age 10–11 years and ED symptoms at age 16–17 years. The study examined these associations in a large community-based population-level sample using data from the Longitudinal Study of Australian Children (LSAC). Specifically, the aim of the present study was to determine whether body dissatisfaction, negative emotional reactivity and self-esteem during early adolescence (age 10–11) were associated with probable ED development in later adolescence at age 16–17. The study examined these associations while adjusting for multiple potentially confounding family factors at age 8–9, including maternal psychological distress, maternal warmth and family socioeconomic status. These environmental factors were chosen as they were supported by research to influence both risk factors and disordered eating. For example, maternal warmth has been found to be significantly and directly associated with body dissatisfaction and disordered eating (Wade et al., 2013). In addition, studies have suggested that ED risk in children may also be influenced by maternal psychological distress, as opposed to mothers suffering from EDs specifically (Micali et al., 2011; Milan & Acker, 2014). Furthermore, past research has supported that family socioeconomic status may influence the strength of ED risk factors, such as body dissatisfaction (West et al., 2019). The prospective longitudinal study design is illustrated in Figure 1.

Figure 1. Diagram illustrating study variables and the direction of associations examined.

Method

Participants and procedure

Data for this study came from the LSAC, and included children and adolescents at age 8–9, 10–11 and 16–17, and their parents from the Kinder (K) Cohort. These data were drawn from Waves 3, 4 and 7 of the LSAC, conducted in 2008, 2010 and 2016, respectively. Details of the study design and sampling procedure are described in detail elsewhere (Soloff et al., 2005). In brief, the study used a two-stage cluster sampling design in which Australian postcodes were randomly sampled and stratified by state of residence and remoteness to ensure representativeness. The study was approved by the Australian Institute of Family Studies ethics committee.

For the K cohort, 4983 children were recruited at age 4 to 5, with 4331 (87%), 4169 (84%) and 3089 (62%) retained at wave 3, 4 and 7, respectively. At wave 7, 2372 adolescents (representing 48% of the initial sample in 2004) had complete data available for all measures used in the present study across the three waves of data collection, and represent the “complete-case” sample and the focus of the present study. The sample characteristics of the adolescents who participated are shown in Table 1. “Response sample” in Table 1 represents those participants in LSAC with completed responses for each individual measure. The LSAC sample, and hence the present study’s sample, consisted of participants of male and female gender.

Table 1. Characteristics of the response and complete-case samples.

		Response sample^†		Complete-case sample^‡ (n = 2,372)
	Age (years)	n	M (SD) or %	M (SD) or %
% Girls
Age 8–9	8–9	4331	48.9%	48.7%
Age 10–11	10–11	4169	48.9%	48.7%
Age 16–17	16–17	3089	49.0%	48.7%
Indigenous status (Yes)
Age 8–9	8–9	4329	2.8%	1.8%
Age 10–11	10–11	4167	2.8%	1.8%
Age 16–17	16–17	3089	2.2%	1.7%
Body dissatisfaction (PBII) (% Want a thinner body)	10–11	4084	37.9%	35.4%
(95% CI)			(36.4–39.4)	(33.5–37.4)
Negative reactivity (SATI)	10–11	4115	2.3 (0.8)	2.3 (0.8)
Self-esteem (SDQ-II)	10–11	4089
Median (Interquartile range)			34.0 (8.0)	35.0 (7.0)
% below median			43.6%	48.8%
Probable ED criteria met (BEDT)
Anorexia Nervosa	16–17	2833	4.1%	4.3%
Bulimia Nervosa	16–17	2928	1.8%	1.6%
Binge Eating Disorder	16–17	2929	0.9%	0.8%
Any Eating Disorder	16–17	2833	6.2%	6.2%
Probable ED criteria not met (BEDT)	16–17	2833	93.8%	93.8%
Maternal psychological distress (K6)	8–9	3688
M (SD)			3.5 (3.6)	3.3 (3.3)
Range			24	23
% probable serious mental illness^§			3.0%	1.7%
Maternal warmth (CRQ)	8–9	3718
Median (Interquartile range)			4.3 (0.8)	4.3 (0.8)
% below median			44.0%	44.4%
Family socioeconomic statusv(SEIFA)	8–9	4331
1^st quartile			14.9%	12.9%
2^nd quartile			34.9%	34.1%
3^rd quartile			26.5%	27.3%
4^th quartile			23.7%	25.7%

^†Response sample is the number of participants who responded to each item for each study measure.

^‡Complete-case sample includes participants with complete data on all items across study measures.

^§K6 cut-off score for “probable serious mental illness” is 13 (Kessler et al., 2003).

Abbreviations: ED, eating disorder; M, mean; SD, standard deviation; CI, confidence interval; PBII, Pictorial Body Image Instrument; SATI, School-Age Temperament Inventory; SDQ-II, Self-Description Questionnaire II; BEDT, Branched Eating Disorder Test; K6, Kessler 6 Psychological Distress Scale; CRQ, Child Rearing Questionnaire; SEIFA, Socio-Economic Indexes for Areas.

Measures

Body dissatisfaction was measured at age 10–11 using the Pictorial Body Image Instrument (PBII; Collins, 1991). Adolescents were shown an array of seven figures depicting adolescents of the same sex, which ranged in body size from very thin to very large. The discrepancy between the participants’ indicated “actual” and “ideal” body images was used as a measure of body dissatisfaction. As literature has associated EDs with a specific desire for thinness (e.g., Fairburn et al., 2003), scores were categorised as follows: (1) Adolescents who wanted a thinner body (scores of 1 to 6), and (2) those who were satisfied with their body or wanted a larger body (scores of −6 to 0). For the present study, the group who desired a thinner body was classified as experiencing body dissatisfaction. Previous research has supported the validity and reliability of the Pictorial Body Image Instrument (Gardner & Brown, 2010).

Emotional Reactivity was measured at age 10–11 using 4 items from the School-Age Temperament Inventory which assessed negative emotional reactivity (SATI; McClowry, 1995) (e.g., “Responds intensely to disapproval (shouts, cries, etc.)”). Parents rated their child on a 5-point Likert scale from “1 = never” to “5 = always”. Reactivity scores were obtained by deriving the mean score of the four items, with a higher score representing higher negative reactivity. The original SATI has been validated in a previous study with good internal consistency (α = .80–.92; McClowry et al., 2003).

Self-esteem was measured at age 10–11 using 8 items from the Marsh Self-Description Questionnaire II (SDQ-II; Marsh et al., 1991) (e.g., “In general, I like the way I am”). Adolescents responded to the items on a 5-point Likert scale from “1 = false” to “5 = true”. Self-esteem scores were obtained by summing responses, with a higher score indicating a higher self-esteem. The SDQ-II scales have demonstrated good internal consistency (α = .83–.91; Marsh et al., 1991). Due to the negatively skewed distribution of adolescents’ summed scores, scores were categorised into two groups through performing a median split: below average self-esteem (scores ranging 8–34) and average or above self-esteem (scores ranging 35–40).

Maternal psychological distress was measured at age 8–9 using the Kessler 6 Psychological Distress Scale (K6; Kessler et al., 2003). The K6 consists of six items that ask about a participant’s feelings over the past 4-week period. Items were answered on a 5-point Likert scale from “1 = all of the time” to “5 = none of the time”. Mothers’ responses were scored following the scoring protocol and a threshold value of 13 was used to identify mothers with “No probable serious mental illness” (scores <13) and those with “Probable serious mental illness” (Kessler et al., 2003).

Maternal warmth was measured at age 8–9 using the parental warmth scale that was derived from the Child Rearing Questionnaire (CRQ; Paterson & Sanson, 1999). The parental warmth scale included 6 of the original 10 items, and measures how often parents express affection, have warm and intimate times with the child and feel close to the child. Items were answered on a 5-point Likert scale from “1 = never/almost never” to “5 = always/almost always”. Mothers’ responses were used to derive mean scores of the six items, with a higher score representing more maternal warmth. This scale has demonstrated a good internal consistency (α = 0.84; Kemmis-Riggs et al., 2020). Due to the negatively skewed distribution of mean scores, scores were categorised into two groups through performing a median split to maintain interpretability. Groups represented below average warmth (scores ranging 1.0–4.2) and average or above warmth (scores ranging 4.3–5.0).

Family socioeconomic status was measured at age 8–9 using the Socio-Economic Indexes for Areas (SEIFA) Index of Advantage/Disadvantage (Trewin, 2001). This measure was developed by the Australian Bureau of Statistics (ABS) and is a census-based measure of socio-economic advantage and disadvantage in an area based on variables including income, occupation and education. Lower scores on this measure indicated more disadvantage and higher scores indicated more advantage. For ease of interpretation, ABS recommends examining SEIFA scores in quantiles (e.g., quartiles). The SEIFA index has demonstrated good validity based on analyses using Australian census (Trewin, 2001).

Probable ED diagnoses were measured at age 16–17 using the Branched Eating Disorder Test (BEDT; Selzer et al., 1996). The BEDT is suitable for use in community samples and is designed to allow for the identification of partial-syndrome EDs, defined as the fulfilment of two out of three DSM symptoms for AN and BN, and one single symptom for BED. The instrument has shown high agreement with the Eating Disorder Examination (EDE; sensitivity 100%, specificity 99% and positive predictive value 70%) using an Australian community sample of adolescent girls (Selzer et al., 1996). The BEDT has been used to assess probable (or partial-syndrome) ED diagnoses according to DSM-IV (American Psychiatric Association, 1994) criteria (Patton et al., 2008). However, minor revisions were made by LSAC to the original BEDT to ensure consistency with DSM-5, and to clarify wording (Hughes et al., 2019). The BEDT consists of nine stem items branching out to a maximum of 31 items. Items ask participants about their engagement in particular behaviours and attitudes over the past 4 weeks (e.g., “In the last four weeks have you felt at any time that you have lost control of your eating or felt you ate too much?”), and the frequency and duration of engagement in the endorsed behaviours or attitudes. The identification criteria for probable AN, BN and BED diagnoses were adopted from Bisset et al. (2019) (See Appendix, Table A1). Probable ED was identified if adolescents met any criteria for AN, BN or BED.

Statistical analysis

Data were analysed using the Statistical Package for the Social Sciences (SPSS, Version 27).

Separate univariate logistic regression analyses were conducted to examine the association of body dissatisfaction, negative emotional reactivity and self-esteem with probable ED status. Multivariate logistic regression analyses were conducted to examine any changes in the reported associations after adjusting for potential reciprocal effects among the risk factors and the effects of potential confounders.

As part of post-hoc analyses, analyses compared the level of body dissatisfaction, reactivity and self-esteem for adolescents at age 16–17 who did and did not meet probable ED criteria for each ED type. This was followed by separate multivariate regression analyses for each ED type, to examine whether the three factors were associated with each ED type when adjusting for confounding variables. Conducting these post-hoc analyses allowed the examination of whether effects from the main analyses were replicated for each ED type, and whether any effects were specific to certain ED types.

Results

Sample characteristics and descriptive statistics

Table 1 shows that the scores from the response and complete-case samples were generally consistent. The complete-case sample consisted of slightly fewer mothers who experienced psychological distress (by 1.3%) and slightly fewer adolescents who experienced body dissatisfaction (by 2.5%), though the 95% confidence intervals for these differences overlapped. Despite slight differences in sample characteristics, the response and complete-case samples had consistent proportions of adolescents who met probable status in any EDs and in the specific diagnoses (i.e., AN, BN and BED).

All subsequent analyses were conducted with the complete-case sample. At age 10–11, 35.4% of the adolescents wanted a thinner body, 7.9% wanted a larger body and 56.7% indicated no dissatisfaction at all. Analysis showed that there was no significant difference in the proportion of boys and girls wanting a thinner body at age 10–11 (girls = 34.4%, 95% CI: 31.6–37.2%; boys = 36.4%, 95%CI: 33.7–39.2%; see Table 1).

At age 16–17, 6.2% of the adolescents (n = 148) met probable criteria for any of the three EDs and 4.3% (n = 101), 1.6% (n = 38) and 0.8% (n = 20) met probable AN, BN and BED, respectively. Table 2 shows that the proportion of underweight adolescents at age 16–17, identified based on adolescents’ BMI scores (BMI < 18.5), was higher in the probable met group (68.9%, 95% CI: 60.8–76.3%) than the probable not met group (5.2%, 95% CI: 4.3–6.2%).

Table 2. Distribution of characteristics comparing individuals meeting probable ED criteria to those who did not meet the criteria (n = 2372).

	Probable ED group (n = 148)	No ED group (n = 2,224)
	M (SD), 95%CI or % (95% CI)	M (SD), 95%CI, or % (95% CI)
Body dissatisfaction (% Want a thinner body)	20.3% (14.1–27.7%)	36.4% (34.4–38.5%)
Negative emotional reactivity	2.3 (0.8), (2.2–2.4)	2.3 (0.8), (2.2–2.3)
Self-esteem (% below median cut-off)	45.9% (37.7–54.3%)	49.0% (46.9–51.1%)
Maternal psychological distress (% probable serious mental illness)	2.7% (0.7–6.8%)	1.7% (1.2–2.3%)
Maternal warmth (% below median cut-off)	49.3% (41.0–57.7%)	44.1% (42.0–46.2%)
Family socioeconomic status
1^st quartile	16.9% (11.2–23.9%)	12.6% (11.2–14.0%)
2^nd quartile	32.4% (25.0–40.6%)	34.2% (32.2–36.2%)
3^rd quartile	23.6% (17.1–31.3%)	27.6% (25.7–29.5%)
4^th quartile	27.0% (20.1–34.9%)	25.6% (23.8–27.5%)
BMI (% underweight)	68.9% (60.8–76.3%)	5.2% (4.3–6.2%)

Abbreviations: ED, eating disorder; M, mean; SD, standard deviation; CI, confidence interval.

Table 2 shows the distribution of risk factors and potential confounders for adolescents who were above the probable ED threshold and those who were not at age 16–17. The proportion of adolescents who experienced body dissatisfaction, as defined by wanting a thinner body, was higher in the no ED group, than the probable ED group. This pattern of effect was not due to the categorisation of the body dissatisfaction scores, with continuous scores showing a similar pattern of effect (see Appendix, Table A2). However, the size of this difference was small and unlikely to be clinically meaningful. There were no differences in the mean reactivity scores at age 10–11 between the probable ED and no ED groups. The proportion of adolescents who had a self-esteem score lower than the median cut-off point was slightly higher, by 3.1%, in the no ED group than the probable ED group, though the 95% confidence intervals for this difference were overlapping.

Univariate and multivariate analyses

Table 3 shows the regression analyses for the associations between risk factors at age 10–11 and probable ED at age 16–17, in unadjusted and adjusted models. In Model 1, body dissatisfaction showed a significant association with probable ED at age 16–17 (p < .001), and the associated odds ratio indicated that body dissatisfaction was associated with 56% decrease in odds of meeting probable ED at age 16–17. This was, however, a small effect with only 2% variance in ED status explained by body dissatisfaction. Negative emotional reactivity and low self-esteem were not significantly associated with increased odds of meeting probable ED criteria at age 16–17. Adjusting for the influence of the other risk factors (Model 2) did not change the pattern of effects. In the final model (Model 3), which adjusted for all variables including confounding variables, the pattern of effect was consistent.

Table 3. Unadjusted and adjusted multiple regression analyses for the associations between body dissatisfaction, negative reactivity and self-esteem at age 10–11 and probable ED status at age 16–17 (n = 2372).

	Model 1 – unadjusted			Model 2 – adjusted			Model 3 – adjusted
	OR	95% CI	p	OR	95% CI	p	OR	95% CI	p
Body dissatisfaction	0.44	0.30–0.67	<.001***	0.45	0.30–0.67	<.001***	0.44	0.29–0.67	<.001***
Negative emotional reactivity	1.02	0.82–1.26	0.89	1.03	0.83–1.27	0.82	1.00	0.81–1.24	1.00
Self-esteem	0.89	0.63–1.24	0.48	0.97	0.69–1.36	0.84	0.95	0.68–1.34	0.77

Abbreviations: OR, odds ratio; CI, confidence interval; R², Nagelkerke R Square; Bold and ***, p < .001.

Model 1 presents the association between the risk factors and outcome, without adjustment for other risk factors.

Model 2 adjusts for all risk factors in the same model.

Model 3 additionally adjusts for maternal psychological distress, maternal warmth and family socio-economic status.

Post-hoc analyses

In order to examine the observed pattern of effect between body dissatisfaction and probable ED further, we conducted analyses separately by disorder type to examine whether this pattern was consistent across the disorder types. The results of disorder-specific associations are presented in Appendix Table A3. Adjusted regression models indicated that body dissatisfaction was associated with a decrease in odds of meeting probable ED at age 16–17 for AN (only – not significant for BN or BED), albeit of small effect and unlikely to be clinically meaningful. Similar to the main analyses, mean negative emotional reactivity scores were highly consistent between probable ED and no ED groups across AN, BN and BED. The effect of self-esteem suggested that there was 2.51 times the likelihood of meeting probable BN for adolescents with low self-esteem than those without low self-esteem.

Given the unexpected finding that those with a desire for a thinner body showed decreased likelihood of meeting probable ED, further analyses were conducted (see Appendix, Table A2). Results suggested that the present effect was not due to the method of categorisation of body dissatisfaction scores.

Discussion

The aim of the present study was to investigate whether body dissatisfaction, negative reactivity and self-esteem in adolescents aged 10–11 years were associated with probable ED status when adolescents reached age 16–17, while controlling for maternal psychological distress and warmth, and family socioeconomic status. Body dissatisfaction at age 10–11, defined as wanting a thinner body, was found to be associated with less likelihood of meeting probable ED at age 16–17. While this effect was weak, it was also contrary to expectations and was driven by a higher number of adolescents meeting the criteria for probable AN at age 16–17 reporting lower levels of body dissatisfaction than those without an ED. The pattern of effect for those who met probable BN and BED, while not statistically significant, was consistent with the hypothesis that higher levels of body dissatisfaction at age 10–11 years would be associated with a higher risk of ED at age 16–17 years.

Negative reactivity and self-esteem at age 10–11 were not significantly associated with probable ED at age 16–17. However, post-hoc analyses found that self-esteem had a specific significant effect for probable BN during late adolescence. Overall, body dissatisfaction, negative reactivity and self-esteem collectively only accounted for 2% of variance in adolescents’ probable ED status at age 16–17 when adjusted for confounding factors. This suggests that these factors may not be important risk factors for the development of EDs over the adolescent years, and that there may be differing risk factors in early adolescence across different types of EDs.

The finding that body dissatisfaction did not positively predict adolescents meeting probable ED was not consistent with previous research (e.g., Prnjak et al., 2021; Rohde et al., 2015). However, research to date has typically followed adolescents up over 3–4 years, as opposed to the 7-year follow-up in the present study (Prnjak et al., 2021). It is possible that body dissatisfaction may be a robust predictor of EDs in the shorter term but is not a reliable predictor when testing the association over a longer period of time across adolescence. Indeed, studies have supported the fluctuating nature of body dissatisfaction throughout adolescence (Bucchianeri et al., 2013), which may also be influenced by factors such as changes in friendships (Webb & Zimmer-Gembeck, 2014). Therefore, it is possible that adolescents who were initially satisfied with their bodies at the start of early adolescence may have experienced subsequent changes in body size and had less capacity to manage subsequent dissatisfaction throughout adolescence, thus subsequently meeting probable ED at age 16–17. Alternatively, it is also possible that adolescents who indicated body dissatisfaction at the start of early adolescence were able to subsequently engage in activities that helped them reach a satisfactory body size, which in turn would protect them from meeting probable ED during late adolescence.

Inconsistent with the present study, previous research has supported that negative emotional reactivity increases the risk of ED in adolescents (Hochgraf et al., 2017). The only study, to the authors’ knowledge, that investigated this association longitudinally in adolescents (Juarascio et al., 2016) reported that negative emotional reactivity in adolescence significantly predicted an increase in ED attitudes concerning weight and eating behaviours over the subsequent 6 years. A key difference between this study and the present study is that the present study measured subsequent probable ED status, rather than ED attitudes. Juarascio et al. (2016) study used a seven-item measure, The College Eating Disorders Screen (Nowak et al., 2003), to measure ED attitudes (e.g., “I believe I am fatter than most people say I am”), which only showed low to moderate correlations with disordered eating behaviours, as opposed to the BEDT used in the current study which is a more comprehensive (40-item) measure of both ED attitudes and disordered eating behaviours. Therefore, discrepancy between results suggests that negative emotional reactivity may not predict disordered eating in late adolescence. Moreover, Juarascio et al. (2016) study did not adjust for confounders, therefore discrepancy between results may also suggest that when accounting for relevant family factors, negative emotional reactivity does not accurately predict ED status later in adolescence.

The current study also found no association between self-reported self-esteem at age 10–11 and later ED symptomology. In contrast, a recent meta-analysis of 13 studies (n = 8893) focussed on adolescents (Mage = 14.2 years) found that self-esteem was associated with later eating disorder risk, although the overall effect was small (Colmsee et al., 2021). Furthermore, the time period of follow-up was a significant moderator, with shorter follow-up time periods (e.g., less than 2 years) associated with greater effect. It is possible, therefore, that low self-esteem measured later in adolescence may be able to predict ED development across diagnoses within the subsequent approximate 2 years, but low self-esteem earlier in adolescence may not predict transdiagnostic ED development over a longer period into late adolescence. However, as the current study did find that low self-esteem during early adolescence predicted probable BN specifically during late adolescence, this suggests that BN development may be more sensitive to global self-esteem at an early age compared to AN and BED.

Strengths and limitations

The strengths of the present study include the use of a longitudinal prospective design, and the use of data from a large representative sample of Australian adolescents from the community, while accounting for confounding family factors. However, results from the study should be interpreted in the light of several limitations. For example, the study did not test for other factors consistently identified to contribute to EDs, including interpersonal difficulties and perfectionism. Therefore, it is not clear whether these factors can predict development across ED types in late adolescents, and hence provide partial support for the predictive utility of models incorporating these factors transdiagnostically across EDs. The study also did not examine the effect of testing confounding family factors as moderators on the results or examine whether a change in risk factors from early to late adolescence could predict meeting probable EDs. These analyses represent opportunities for future research. Moreover, the study was limited by the use of a brief self-reported measure of eating disorder symptoms, which was only able to identify probable EDs. However, as adolescents with subclinical EDs often progress into threshold disorders (Stice et al., 2009), this measure may accurately identify adolescents likely to meet full diagnostic criteria. It should be noted, however, that the proportion of participants meeting probable BED was significantly lower than typically identified in prevalence studies (Kjeldbjerg & Clausen, 2023), thus potentially highlighting limitations of the BEDT as an indicator of BED risk. A further limitation is that some measures used in the present study were shorter subscales extracted from larger measures, thus it was not possible to have participants complete comprehensive evaluations of the constructs of interest (i.e., negative reactivity and self-esteem). Furthermore, some studies suggest that figure rating scales are not sufficient for measuring body dissatisfaction, as the scales may assess overall weight and shape dissatisfaction, whereas body dissatisfaction likely encompasses dissatisfaction in additional aspects of appearance (e.g., skin condition, particular parts of the body; Vander Wal & Thelen, 2000). Nevertheless, the measures chosen were valid and reliable instruments, as such results do provide information on the associations between body dissatisfaction, negative emotional reactivity and self-esteem and subsequent probable ED development. Furthermore, the examination of gender differences was outside the scope of the present study. Given that there may be slightly varied mechanisms of ED risk factors between the genders (e.g., Dakanalis et al., 2015), future studies may benefit from studying the genders separately when examining ED risk in adolescence.

Implications

This study is one of few studies to examine the risk factors for EDs longitudinally in early adolescence. The findings suggest that further research is needed to understand how best to identify in early adolescence those at increased risk for the development of EDs in late adolescence. This may involve testing risk factors, especially body dissatisfaction and self-esteem, at multiple time points throughout adolescence to understand when these factors may predict ED development during the peak age of onset in late adolescence. It may also be important to examine other factors, including those from the transdiagnostic model and other models for disordered eating, in predicting ED development. In the context of the present findings, it is likely that early intervention for EDs should focus on the provision of universal preventative interventions, as it is not yet clear which factors identify those at increased risk for EDs in later adolescence. While the present study suggested that transdiagnostic models may be less appropriate for identifying early adolescents at risk of EDs, they remain as important frameworks for psychological interventions, including enhanced cognitive behavioural therapy (CBT-E) for EDs (Atwood & Friedman, 2020), to limit the maintenance of symptoms for those diagnosed with EDs.

Conclusion

In conclusion, the current study showed that when examining eating disorder risk using the transdiagnostic model, negative reactivity and self-esteem in early adolescence did not predict risk of developing probable ED in late adolescence, although there was evidence of self-esteem being a specific risk factor for probable bulimia nervosa development. The study suggests that these elements of the transdiagnostic models of eating disorders may be limited in their ability to explain risk factors in early adolescence for the development of eating disorders in late adolescence. Overall, as one of few longitudinal studies to examine eating disorder risk factors in early adolescence, the study provides useful information which suggests that identifying risk for eating disorder in early adolescence is a complex picture. Possible risk factors may be sensitive to time and changes throughout adolescence, and may also interact with other factors to predict future symptomology, thus highlighting the need for research that explores the trajectory of risk factors and ED symptom development across adolescence.

Author contributions

Kiu Lam Chan: Conceptualisation, Methodology, Formal analysis, Writing – original draft, Visualisation; Alyssa Sawyer: Conceptualisation, Methodology, Supervision, Writing – reviewing and editing; Amanda Taylor: Conceptualisation, Methodology, Supervision, Writing – reviewing and editing. All authors have approved the final article.

Acknowledgements

This paper uses unit record data from Growing Up in Australia, the Longitudinal Study of Australian Children (LSAC) conducted by the Australian Government Department of Social Services (DSS). The findings and views reported in this paper, however, are those of the authors and should not be attributed to the Australian Government, DSS, or any of DSS’ contractors or partners. doi:10.26193/F2YRL5.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/00049530.2023.2260488.

Data availability statement

The data that support the findings of this study are available from the Australian Data Archive (ADA) and the National Centre for Longitudinal Data (NCLD). Restrictions apply to the availability of these data, which were used under licence for this study. Data are available through Dataverse, an online platform which facilitates access to its longitudinal datasets, with the permission of ADA and NCLD.

Additional information

Funding

The author(s) reported that there is no funding associated with the work featured in this article.