Patterns of disordered eating behavior, adaptive and maladaptive perfectionism and impulsivity in early adolescents

Abstract

In this study, latent profile analysis (LPA) was used to identify naturally occurring patterns or profiles of maladaptive and adaptive perfectionism, impulsivity, and disordered eating (DE) in early adolescence (111 boys, 138 girls; M age = 13.6 years). Profile membership at age 13 was used to examine disordered eating patterns assessed at ages 11 and 12, providing insights into how symptoms had developed prior to profile formation. Using LPA, we identified five profiles: (1) high functioning (2), maladaptively impulsive (3), anxious-avoidant (4), maladaptively perfectionistic, and (5) maladaptively impulsive-perfectionistic. The maladaptively perfectionistic profile showed the highest levels of dieting, preoccupation with food, and body concerns as well as perceived sociocultural pressure to be thin, followed by the maladaptively impulsive-perfectionistic profile. Analyses of earlier DE patterns indicated that the anxious-avoidant profile consistently showed the lowest BMI and highest perceived pressure to eat, both at age 13 and at earlier assessment points, suggesting possible feeding-related difficulties. The findings confirm that maladaptive and adaptive perfectionism, impulsivity, and BMI are jointly associated with the development of DE in early adolescence, demonstrating that considering these traits together yields greater explanatory value than focusing on any single factor alone. While identifying general risk factors is essential, differentiating personality-based profiles of vulnerable adolescents may be particularly valuable for targeted early prevention.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-026-01630-w.

Plain language summary

This study examined how perfectionism and impulsivity are linked to disordered eating in early adolescence. We studied 249 boys and girls aged 13 to 14 and also looked at their eating patterns from ages 11 to 12. We identified naturally occurring groups of adolescents based on their levels of perfectionism, impulsivity, and disordered eating, and found five distinct profiles that differed in their eating-related attitudes and behaviors. Among the profiles found, adolescents with high perfectionism showed the strongest signs of restrictive eating, preoccupation with food, and concerns about their body. Another group had low body weight and reported feeling pressure from others to eat, suggesting feeding-related challenges. Some of these differences were already evident at ages 11–12, suggesting that these patterns were present across early adolescence. These findings highlight that young people can experience disordered eating in different ways. Recognising such patterns may help identify those at greater risk and improve early prevention and support.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-026-01630-w.

Mid to late adolescence is the peak onset period for disordered eating (DE) behavior [1, 2]. Accordingly, eating disorders (ED) prevalence increases from childhood to early adolescence [3, 4]. DE behavior which does not meet the full diagnostic criteria for EDs is associated with similar distress as EDs [2, 5]. Moreover, maladaptive eating attitudes and behavior predict the development of EDs [6]. The prevalence of EDs in adolescents continues to rise [7]. EDs are often chronic and marked by a fluctuating course, with only a small percentage recovering without treatment [8]. Therfore, paying attention to the earliest symptoms is vital for early treatment and, therefore, a better prognosis [9]. This highlights the urgency of early symptom recognition to mitigate the chronic and often relapsing course of EDs.

The development of EDs is associated with multiple biological, sociocultural, and psychological factors [10]. Among many other factors, personality traits, e.g., perfectionism and impulsivity, have consistently been shown to be an important link in the etiology of DE [11, 12]. Both perfectionism and impulsivity are multidimensional traits with adaptive and maladaptive sides [13–16]. Adaptive perfectionism encompasses organization and striving for high standards, while maladaptive perfectionism encompasses concern over mistakes and perceived criticism from parents [17]. Adaptive impulsivity is characterized by processing information quickly, resulting in fast thinking and responses when such a style is optimal. In contrast, maladaptive impulsivity is characterized by failure to inhibit inappropriate responses and recklessness [14]. Although impulsivity is frequently considered a cognitive function best assessed through behavioral tasks, it is essential to recognize the broader, multifaceted nature of impulsivity. Various conceptualizations emphasize emotional, cognitive, and behavioral aspects, each with its own approach to measurement [13, 18, 19]. Moreover, recent evidence highlights that impulsivity is a stable and measurable trait that predicts real-world outcomes [20].

High impulsivity is considered a risk factor for EDs in adults [21]. Strongest associations have been found with binge eating and purging, but high levels of impulsivity have also been found in EDs primarily associated with restricting type of behavior [22–25]. In children and adolescents, longitudinal studies investigating associations between impulsivity and DE have produced mixed results [26–28]. It has been suggested that the effect of impulsivity on DE depends on the interplay of other coexisting traits and factors (e.g., neuroticism, affective reactivity, and perfectionism) [26, 29]. Recent research also continues to emphasize that impulsivity plays a complex role in EDs, particularly when combined with high negative affectivity [30–32].

Perfectionism has been described as a transdiagnostic factor in adult psychopathology [33], and numerous studies have consistently associated perfectionism with DE and EDs among children, both cross-sectionally [34, 35] and longitudinally [36, 37]. A meta-analysis has also confirmed the associations between perfectionism and conscientiousness with EDs in adolescents [38]. In the case of EDs, both adaptive and maladaptive dimensions of perfectionism, as well as their combinations have been linked to DE [35, 39–41]. For example, a recent meta-analysis in children and adolescents concluded that there was a small effect between ED symptoms and perfectionistic strivings and a moderate effect between perfectionistic concerns and ED symptoms [42]. Regarding diagnostic differences, another meta-analysis found no significant differences between individuals with anorexia nervosa and bulimia nervosa in maladaptive perfectionism, suggesting that despite symptom-level differences, perfectionism is elevated across all EDs [39].

Though it is crucial to look for risk factors in general to identify adolescents at high risk for developing EDs, identifying personality-based subtypes of individuals who are vulnerable to risks could be even more important for early prevention and treatment. Personality profiles have shown clinical utility and predicted long-term functioning in adults with mental health problems [43]. Profiling is especially beneficial in EDs, as there is considerable heterogeneity in categorical ED diagnoses, complicating research and clinical work [44, 45]. Personality-based profiles can offer a way to create more homogenous subgroups, helping to reduce the difficulties mentioned above. In the ED literature, three (and in some studies up to five) personality-based subtypes have been identified across different samples and measures: high functioning (characterized by adaptive self-regulatory tendencies), overcontrolled (characterized by high inhibition, perfectionism, and rigidity), and undercontrolled (characterized by high impulsivity and disinhibited behavioral patterns) [46].

One important dimension for creating such profiles lies in examining how perfectionism and impulsivity, often studied separately, can co-occur within the same individuals. High levels of both traits may amplify risk for DE behaviors, underscoring the need to consider their combined influence within a broader psychological profile [30, 47, 48]. In youth, this interplay can also be reflected in the frequent co-occurrence of internalizing symptoms (e.g., anxiety, depression, somatic complaints) and externalizing behaviors (e.g., aggression, conduct problems, hyperactivity), suggesting that such trait constellations contribute to broader vulnerability patterns in psychopathology [49, 50]. Closely related to this, perfectionism also overlaps conceptually with compulsivity. Maladaptive perfectionism, with its focus on avoiding mistakes and rigid standards, can resemble compulsive attempts to manage distress through repetitive behaviors. Supporting this view, perfectionistic traits have been suggested to act as a bridge between obsessive-compulsive disorder and EDs [51]. Research showing that impulsivity and compulsivity can co-occur despite their apparent opposition [52–54] illustrates a broader principle: traits that may appear contrasting, such as perfectionism and impulsivity, can nonetheless converge in ways that shape vulnerability to DE.

The literature on personality-based subtyping and its relation to ED symptoms in children and adolescents is currently insufficient. As known to the authors, only three studies have been conducted on this topic. In a study [55] where Q-factor analysis was conducted based on clinicians’ ratings they found three ED subtypes: high- functioning/perfectionistic, emotionally dysregulated, and avoidant/depressed. In another study [36] four main cluster were found: a combined high perfectionism and impulsivity, a pure perfectionistic, a purely impulsive, and a resilient class. A more recent study on adolescents with anorexia nervosa found the best solution to describe personality heterogeneity to be the separation of two classes: a class showing higher internalizing and a class showing higher externalizing traits [56]. Although in adult samples meaningful subtypes based on perfectionism and impulsivity are well established [48, 57] much less is known whether these could be observed in adolescence. Importantly, prior adolescent studies have treated DE symptoms primarily as outcomes rather than as joint indicators within person-centered models. As adolescence is a critical developmental period [1–4] for the onset of DE, it is particularly important to investigate how these traits and symptoms co-occur during this stage.

The current study aims to profile individuals based on maladaptive and adaptive perfectionism, impulsivity, and DE behaviors/attitudes in a community sample of adolescents. The person-centered approach offers a way to describe the population’s heterogeneity, reflecting how different traits coexist in an individual. A person-centered approach describes population heterogeneity by capturing how traits co-occur within individuals. The decision to combine personality traits and DE symptoms as indicators in the LPA was driven by methodological and theoretical considerations. Theoretically, personality and psychopathology can be linked in several ways. Beyond one-way causal perspectives, two particularly relevant frameworks are the pathoplastic model, which emphasizes mutual shaping over time, and the complication model, which suggests that psychopathology can reinforce or exacerbate maladaptive traits [58, 59]. This perspective aligns with cognitive-behavioral models of DE, which emphasize the interplay of enduring traits and symptomatic behaviors [60]. Empirically, traits such as perfectionism and impulsivity prospectively predict ED risk and severity, while symptom remission has been shown to reduce perfectionism and negative urgency, supporting a bidirectional association [61, 62]. Methodologically, this approach reduces potential biases associated with the classify-analyze approach, which may underestimate the relationship between traits and outcomes due to classification uncertainty [63] and yields greater explanatory power for the profile [64]. This approach allows us to achieve a comprehensive understanding of the observed profiles, highlighting the value of integrating risk factors and outcomes in profiling.

The present study extends our previous work [48, 64] by applying an integrative person-centered approach to a younger, non-clinical population, examining adaptive and maladaptive perfectionism, functional and dysfunctional impulsivity, and DE symptoms simultaneously within the same model. By additionally exploring earlier DE patterns in relation to later profile membership, the study incorporates a developmental perspective that has been relatively underexplored in adolescent personality-based subtyping research. Building on this rationale, we formulated the following hypotheses. First, we expected at least a four-profile solution to emerge: (1) a profile characterized by low maladaptive perfectionism and impulsivity (2), a profile marked by high maladaptive perfectionism (3), a profile with high maladaptive impulsivity, and (4) a profile showing high levels of both perfectionism and impulsivity. Second, we hypothesized that these profiles would differ meaningfully in terms of anxiety, depression, and DE behaviors and attitudes in adolescents aged 11–14 years. Finally, we conducted exploratory analyses of earlier DE patterns at ages 11 and 12, based on profile membership established at age 13, as an additional way to validate the robustness and developmental relevance of the identified profiles.

Method

Participants

This study is part of a longitudinal project with three measurement waves conducted at 12-month intervals: Wave 1 in 5th grade (n = 308; 175 girls, 133 boys; M = 11.53 years, SD = 0.53), Wave 2 in 6th grade (n = 262; 153 girls, 109 boys; M = 12.56, SD = 0.52), and Wave 3 in 7th grade (n = 249; 138 girls, 111 boys; M = 13.63, SD = 0.55). Children were recruited from public schools providing secondary education in different counties in Estonia. Of the participants, 39% lived in a city, 54% in a town, and 7% in a village. Most participants (66%) reported living with both parents, 29% with one parent, and 5% in other arrangements. All participants identified as White/Caucasian. Data from Wave 3 (when participants were aged 13–14 years) were used for profiling, as this was the first wave in which personality traits were assessed. Additionally, data from all three waves were available for 212 participants (85.1% of the Wave 3 sample). Participants did not receive any compensation for taking part in this study.

Procedure

The study was approved by the Ethics Review Committee on Human Research of the University of Tartu. Written informed consent was obtained from the participants and their parents. The children had the opportunity to withdraw from the study at any moment. Questionnaires were administered in schools by research assistants during scheduled classes. Completion of the questionnaires took approximately 45 min. Before administering the questionnaires, researchers explained the content of the questionnaires and highlighted the possibility for children to ask for help if that would be needed. After filling in the questionnaires, the school nurse privately measured the children´s height and weight in the school medical room.

Measures

Demographic information

Participants provided demographic information at each wave of data collection. This included age, gender, current school grade (5th-7th grade across waves), and place of residence (city, town, or village). Participants also reported their family living arrangement (living with both parents, one parent, or other arrangements) and self-identified race.

Body mass index

BMI (kg/height m²) was calculated based on participants’ actual weight and height, measured by standardized procedures described above.

Attempts to change one’s body

A single question was used to assess attempts to change one’s body weight: “Have you tried to change your body weight?“ (options: “Yes, I have tried to gain weight“, “No, I have not done anything to change my weight“, “Yes, I have tried to lose weight”).

Self-report questionnaires

Children’s Eating Attitude Test (ChEAT) [65] is a 26-item self-report questionnaire measuring children’s eating attitudes and behaviors. The Estonian version has 18 items and consists of four subscales: Food preoccupation (how much children think they can control thinking about food, e.g., “I think about food a lot of the time”), Dieting (restricting food intake and foods high in calories, e.g., “I stay away from foods with sugar in them”), Body concerns (worrying about body weight and shape, e.g., “I think a lot about wanting to be thinner”) and Pressure to eat (social pressure to eat, e.g., “I feel that others would like me to eat more”). Items were answered on a 6-point Likert scale (from “always “to “never”). The Cronbach a for total score ranged from 0.82 to 0.87 in three data waves and from 0.71 to 0.91 for subscales. Correlations between the same subscales within in three assesment waves stayed in the range of 0.41–0.57 (lowest for Dieting and highest for Pressure to eat subscales). The lowest correlation between waves was observed for the Dieting subscale, while the highest was found for the Pressure to Eat subscale.

Scale for maladaptive and adaptive impulsivity was constructed based on the constructs of functional (quick and appropriate thinking and response style when such a style is optimal) and dysfunctional (excessive haste and restlessness) impulsivity [14]. Items from the short version of the International Personality Item Pool (IPIP) [66, 67] that would reflect these constructs were used, and the wording was simplified for children. The items were assessed on a five-point Likert scale (from “I completely agree “to “I don’t agree at all”). Two subscales were derived: Maladaptive impulsivity with three items (“I give promises I cannot keep”, “I make decisions without thinking them through“, and “I make my plans at the last minute”) (inter-item correlation from 0.25 to 0.45 across waves, Cronbach a = 0.60)), and Adaptive impulsivity with four items (“Nothing usually scares me”; “I know how to get things done”, “I don’t have to regret my actions “, “I stay calm in difficult situations” (inter-item correlation from 0.22 to 0.34, Cronbach a = 0.61)).

Frost Multidimensional Perfectionism Scale (FMPS) [17]. The Estonian adaptation [68] is based on the original multidimensional model proposed by Frost and colleagues. The full Estonian version includes additional subscale (Personal standards); however, for the present adolescent sample, we used a 21-item subset comprising the Organization, Concern over mistakes, and Parental criticism subscales. Statements were answered on a 5-point Likert scale (from “strongly disagree” to “strongly agree”). Internal consistency in the present sample was acceptable: Organization (Cronbach a = 0.73), Concern over mistakes (Cronbach a = 0.84), and Parental criticism (Cronbach a = 0.72).

Perceived Sociocultural Pressure Scale (PSPS) [69] assesses perceived pressure to be thin or muscular from friends (e.g., “I have felt pressure from my friends to lose weight”), family (e.g., “Family members tease me about my weight or body shape”), and media (e.g., “I have felt pressure from media (e.g., TV, magazines) to lose weight”). Item wording differed by gender, such that pressure to be thin was assessed for girls, whereas pressure to be muscular was assessed for boys. Internal consistency was acceptable (Cronbach’s α = 0.73 for girls; α = 0.71 for boys). The scale consisted of nine items for both boys and girls, assessed on a 3-point scale (“never”, “rarely”, “often”).

Children’s Depression Inventory (CDI) [70] (Estonian version [71]) is a 27-item self-report questionnaire assessing depression in children and adolescents. The scale consists of groups of statements, in which children choose the statement that most accurately reflects them. The total score of the scale was used for the current study (Cronbach a = 0.84).

State and Trait Anxiety Inventory for Children (STAI-C) [72] was used for measuring trait anxiety only. The scale consists of 21 items, assessed on a 3-point scale (from “almost never “to “often”) (Cronbach a = 0.90).

Data analysis

Latent profile analysis (LPA) with robust likelihood maximum method was performed in program Mplus version 6.12 [73]. Subscales measuring perfectionism and functional and dysfunctional impulsivity were treated as indicator variables in the model. Considering the development of personality in adolescence and the pathoplastic relationship between personality and ED [58], we also included DE behaviors and attitudes in the LPA as indicators (see Introduction for rationale). 1000 random sets of starting values were used in the initial stage, and 250 optimizations in the final stage to avoid converging on a local solution. 100 bootstrap draws were performed, each with two sets of random starting values and one final stage optimization for the model with one less class. For the alternative model, 50 sets of random starting values and 20 final stage optimizations were used (for guidelines see: Muthén & Muthén, 1998–2017). Model selection was based on (1) Bayesian Information Criterion (BIC) [74], (2) Akaike Information Criterion (AIC) [75], (3) Sample-Size Adjusted BIC (SSABIC) [76]. Lower values of the before-mentioned indicators indicate better model fit. Entropy was used to estimate the accuracy with which models classify individuals (values ranging from 0 to 1, higher values indicating greater accuracy). Lastly, the Bootstrap Likelihood Ratio test (BLRT) [77] and Lo-Mendell-Rubin test (LMR) [78] were used to compare if the improvement in the model was statistically significant when one more class was included in the model. Item-level missing data were addressed using the expectation-maximization (EM) method for up to 10% of missing data from a scale. When too many items were missing in a scale to apply EM without significantly increasing the risk of bias, full information maximum likelihood (FIML) estimation was used in LPA to handle scale-level missing data. Missingness of one indicator out of the nine included in the LPA was allowed.

Other statistical analyses were performed in SPSS Statistics version 23. Classes were compared on the measures included in LPA by Welch’s ANOVA using Games-Howell’s post hoc for assessing pairwise differences. For the Pressure to be thin scale, one-way ANOVA and Hochberg’s post hoc were used. Chi-square test of independence was used to compare gender distribution and attempts to change one’s body weight. To assess changes in BMI and ChEAT scores, two-way repeated-measures ANOVA was conducted, with time as a within-subject variable and profile membership as a between-subject variable. The Greenhouse-Geisser correction was used when the assumption of sphericity was violated. A series of one-way repeated measure ANOVAs were also conducted to assess the main effect of time in the emerged profiles separately.

Results

Latent profiles based on disordered eating behaviors, maladaptive and adaptive impulsivity, and perfectionism

The LPA included boys and girls from Wave 3 (ages 13–14 years; n = 249). A series of one- to seven-profile models were estimated based on the indicator variables: ChEAT, FMPS, and maladaptive and adaptive impulsivity.

Statistical fit indices of each solution are presented in Table 1. The five-profile model was selected based on fit indices and considering theoretical meaningfulness. Considering that BIC started to increase in the six-class model and although some fit indices also decreased in six and seven-class models, the change was relatively small. In addition, Entropy can be considered good enough for accurate distinction while not deducting any small but theoretically meaningful classes.

Table 1.

Fit indices for 1–7 class solution

Classes	Free parameters	LL	AIC	BIC	Adjusted BIC	Entropy	BLRT	LMR
1	18	−5584.20	11204.40	11267.71	11210.65	-	-	-
2	28	−5471.17	10998.35	10096.83	11008.07	0.868	0.00001	0.191
3	38	−5409.65	10895.30	11028.96	10908.50	0.937	0.00001	0.052
4	48	−5376.87	10848.84	11017.68	10865.52	0.920	0.00001	0.701
5	58	−5342.04	10800.09	11004.10	10820.24	0.837	0.00001	0.100
6	68	−5315.21	10766.41	11005.60	10790.03	0.858	0.00001	0.453
7	78	−5294.46	10744.92	11019.28	10772.01	0.878	0.00001	0.662

Best fitting model is in bold

LL log likelihood, AIC Aikaike information criterion, BIC Bayesan information criterion, BLRT Bootstrap Likelihood Ratio test, LMR Lo-Mendell Rubin test

Statistical differences on indicator variables, BMI, depression, and anxiety

The final five-profile solution is depicted in Fig. 1. Descriptive statistics and ANOVA statistics can be found in Table 2. BMI, depression, and anxiety scores were included in ANOVA for external validation of the emerged classes.

Fig. 1.

z-scores for dimensions of perfectionism, impulsivity and disordered eating attitudes and behaviors in the five profile solution

Table 2.

Descriptive statistics and differences between profiles in indicator variables (ChEAT subscales, perfectionism, impulsivity) and BMI, depression and anxiety scores

Variable	High functioning (n = 129); 61 boys, 68 girls	Maladaptively impulsive (n = 54); 33 boys, 21 girls	Anxious-avoidant (n = 25); 5 boys, 20 girls	Maladaptively perfectionistic (n = 25); 6 boys, 19 girls	Maladaptively impulsive and perfectionistic (n = 16) 6 boys, 10 girls	Welch´s ANOVA
	M (SD)	M (SD)	M (SD)	M (SD)	M(SD)	Welch´s F (df1, df2)	p	est. ω2
ChEAT total	10.14 (7.16)(e, d)**e*	12.79 (7.87)(c, d)**	23.01 (10.80)(a, b,d)**	37.71 (7.93)(a, b,c, e)**	19.56 (11.64)a*d**	66.13 (4, 53.63)	< 0.001	0.51
Dieting	2.57 (2.50)d**	2.28 (2.47)d**	2.04 (2.51)d**	10.12 (4.06)(a, b,e, e)**	3.63 (2.92)d**	21.36 (4, 55.57)	< 0.001	0.25
Food preoccupation	1.38 (1.57)(b, c)*d**	2.49 (2.35)(a, d)**	4.27 (3.62)a*	4.96 (3.08)a**b*	3.07 (3.30)	13.00 (4, 51,29)	< 0.001	0.16
Body concerns	4.61 (5.09)d**	5.85 (5.57)d**	6.74 (6.45)d**	19.33 (6.48)(a, b,c, e)**	8.88 (7.14)d**	28.37 (4, 54.70)	< 0.001	0.31
Pressure to eat	1.57 (1.89)c**	2.17 (2.03)c**	9.96 (2.81)(a, b,d, e)**	2.75 (3.01)c**	4.00 (3.43)c**	50.58 (4, 52.48)	< 0.001	0.44
Perfectionism
Organization	10.86 (1.93)b**	7.17 (2.20)(a, c,d)**e*	9.76 (1.98)b**	10.16 (2.34)b**	9.96 (2.37)b*	27.87 (4, 55.90)	< 0.001	0.30
Concern over mistakes	0.85 (1.36)c*(d, e)**	1.46 (1.56)d*e**	2.16 (2.08)a*e**	3.60 (2.72)a**(b, e)*	7.44 (2.58)(a, b,c)**d*	30.62 (4, 52.55)	< 0.001	0.32
Parental criticism	0.77 (0.94)(b, e)** (c, e)*	3.13 (0.15)(a, e)**c*	1.72 (1.49)e**(a, b)*	2.36 (1.82)a*e**	5.75 (1.48)(a, b,c, d)**	65.94 (4, 52.47)	< 0.001	0.51
Impulsivity
Maladaptive	3.44 (2.06)(b, e)**d*	5.72 (1.96)a*	4.71 (2.07)e*	5.08 (2.06)(a, e)*	7.50 (2.53)a**(c, d)*	19.08 (4, 56.04)	< 0.001	0.23
Adaptive	10.75 (2.54)b*	9.21 (2.59)a*	9.72 (2.73)	9.32 (2.69)	11.32 (3.74)	4.51 (4, 55.67)	0.003	0.05
Other variables
BMI	20.50 (3.39)c**	20.95 (4.18)c**	17.98 (1.78)(a, b,d)**e*	22.64 (4.83)c**	20.15 (1.91)c*	10.14 (4, 62.29)	< 0.001	0.13
Depression	7.20 (4.95)(b, c,e, d)**	12.91 (5.12)a**	13.96 (4.97)a**	13.76 (6.66)a**	18.94 (7.90)a**	23.36 (4, 53.18)	< 0.001	0.26
Anxiety	10.66 (6.08)(b, c)**(a, e)*	15.06 (7.26)a*	19.60 (6.91)a**	19.28 (6.24)a**	19.38 (11.19)a*	18.02 (4, 55.15)	< 0.001	0.22

BMI body mass index, ChEAT Children´s Eating Attitude Test, a statistically significant differences from high functioning group, b statistically significant differences from the maladaptively impulsive group, c statistically significant differences from the anxious-avoidant group, d statistically significant differences from the maladaptively perfectionistic group, e statistically significant differences from the maladaptively impulsive and perfecionistic group; ^* - p < 0.05; ^**- p < 0.001

The five classes were labeled as follows: (1) high functioning, (2) maladaptively impulsive, (3) anxious-avoidant, (4) maladaptively perfectionistic, and (5) maladaptively impulsive and perfectionistic.

Participants within the high-functioning class demonstrated low levels of maladaptive impulsivity and perfectionism, as well as the lowest scores on ChEAT subscales, while reporting high levels of organization and adaptive impulsivity.

Participants in the maladaptively impulsive class showed the lowest level of organization, moderately high levels of maladaptive impulsivity, and high scores of perceived parental criticism. Their DE behavior and attitudes were comparable to participants in the high-functioning class. However, they reported significantly higher levels of depression and trait anxiety.

Participants in the third class, named anxious-avoidant, were similar to participants in the high-functioning class regarding perfectionism and impulsivity levels, but reported the highest levels of perceived pressure to eat, high scores in the Food preoccupation subscale and also high trait anxiety. We also found that participants within the anxious-avoidant class had the lowest BMI compared to all the other classes.

The fourth class was named maladaptively perfectionistic. Participants in that class had moderate levels of maladaptive perfectionism, the highest levels of body concerns, dieting, preoccupation with food, and high trait anxiety.

Lastly, participants the maladaptively impulsive and perfectionistic class (hereinafter: maladaptively impulsive-perfectionistic) had the highest maladaptive impulsivity and perfectionism levels. Although DE behaviors in this class were mild to moderate, these participants reported the highest levels of depression and anxiety. This difference was primarily driven by girls within this class, who reported higher levels of both depression and anxiety than boys (see Supplementary material, Table 1).

A chi-square test of independence was conducted between gender and profile membership. Significant differences emerged (χ2_{(4, 249)} = 17.08, p = 0.002), the association between gender and profile membership was moderate (Cohen, 1988), Cramer’s V = 0.262. Post hoc analyses using standardized residuals with +/−1.96 indicated that the differences were not significant in pairwise comparisons. However, these were close to significant (standardized residual +/−1.8) in the anxious-avoidant class (more girls than boys in this class) and the moderately impulsive class (more boys than girls in this class). Mean values in the five profiles for both genders can be found in Supplementary material Table 1. Due to the very small gender-differentiated sample size in some groups (n=minimally 5), further statistical comparisons between genders were not conducted.

Perceived pressure to be thin/muscular and attempts to change one’s body weight

There were some significant differences in perceived sociocultural pressure to be thin/muscular in emerged profiles for the whole sample F_(4,234)=13.37, p < 0.001, for the girls F_(4,126) = 13.43, p < 0.0001, and for the boys F_(4,103) = 2.89, p = 0.026; partial η² = respectively 0.19, 0.29, 0.11. Post hoc analyses indicated that participants in the maladaptively perfectionistic class reported significantly higher perceived pressure than participants in all other classes. This was particularly pronounced among girls, for whom profile differences were stronger, whereas among boys overall differences were smaller and post hoc comparisons did not reach statistical significance. Mean scores by profile and gender, as well as post hoc analysis are presented in Table 3.

Table 3.

Percentage of individuals based on attempts to regulate body weight among profiles

	High functioning (n = 128) 61 boys, 67 girls	Maladaptively impulsive (n = 54) 33 boys, 21 girls	Anxious -avoidant (n = 25) 5 boys, 20 girls	Maladaptively perfectionistic (n = 24) 6 boys, 18 girls	Maladaptively impulsive and perfectionistic (n = 15) 6 boys, 9 girls
Attempts to change one´s body weight
Have tried to gain body weight	3.9% (n = 5)	7.4% (n = 4)	32.0% (n = 8)	4.2% (n = 1)	6.7% (n = 1)
Have not done anything to change body weight	64.8% (n = 83)	53.7% (n = 29)	28.0% (n = 7)	4.2% (n = 1)	40.0% (n = 6)
Have tried to lose weight	31.3% (n = 40)	38.9% (n = 21)	40.0% (n = 10)	91.7% (n = 22)	53.3% (n = 8)

Perceived sociocultural pressure to be thin or muscular	M(SD)	M(SD)	M(SD)	M(SD)	M(SD)
Whole sample	3.32 (2.06)d*	4.17 (3.07)d*	4.09 (1.98)d*	7.35(2.90)(a, b,c, e)*	4.56 (2.85)d*
Girls	3.40 (1.98)d*	4.24 (2.93)d*	3.67 (1.72)d*	7.76 (2.73)(a, b,c)*	5.60 (2.76)
Boys	3.24 (2.16)	4.13 (3.20)	5.60 (2.30)	6.17 (3.31)	2.83 (2.23)

a statistically significant differences from high functioning group, b statistically significant differences from the maladaptively impulsive group, c statistically significant differences from the anxious-avoidant group, d statistically significant differences from the maladaptively perfectionistic group, e statistically significant differences from the maladaptively impulsive and perfecionistic group; ^* - p < 0.01

A chi-square test of independence was conducted to examine the relationship between profile membership and attempts to change one´s body weight. Significant differences emerged (χ2_{(4, 246)} = 58.58, p < 0.0001), with moderate association, Cramer´s V = 0.345. Post hoc analyses using standardized residuals with +/−1.96 indicated that more individuals in the maladaptively perfectionistic class had tried to lose weight compared to the other classes, and more individuals in the anxious-avoidant class had tried to gain weight. Frequencies for all classes are presented in Table 3.

Disordered eating behaviors and BMI across earlier assessment points (ages 11–14)

DE and BMI data from the earlier measurement waves - Wave 1 (ages 11–12) and Wave 2 (ages 12–13) - were examined to assess patterns preceding the Wave 3 (ages 13–14) profile classification. Repeated-measures ANOVA was conducted with Wave 3 profile membership as the between-subjects factor and time as the within-subjects factor.

The mean scores for BMI over three waves in the profiles are depicted in Fig. 2. For BMI, there was a main effect for time F_{(1.85, 389,85)=}103.34, p < 0.001, partial η² = 0.329), meaning that BMI increased in all classes over time, but there was also a time and class interaction effect F_{(7.39, 389,85)=}2.29, p = 0.024, partial η² = 0.042). Differences in BMI between classes can already be seen at age 11. While the maladaptively perfectionistic class had the highest BMI, the anxious-avoidant class constantly had the lowest BMI.

Fig. 2.

BMI in five profiles over three waves

The mean scores of the ChEAT subscales are depicted in Fig. 3 and mean scores of ChEAT and BMI for three waves in Supplementary material Table 2.

Fig. 3.

Mean scores of Children´s Eating Attitude Test in five profiles over three waves. Minimum and maximum scores for every subscale: Body concerns 0–30, Restricting 0–15, Food preoccupation 0–20, Pressure to eat 0–15

For the Dieting subscale, there was a statistically significant time and class interaction effect F_{(7.38, 361.71)} = 10.09, p < 0.001, partial η² = 0.171. Dieting decreased over time in the maladaptively impulsive-perfectionistic class F_{(2, 28)} = 4.55, p = 0.019, partial η2 = 0.245, the high functioning class F_{(1.77, 183.74)} = 10.70, p < 0.001, partial η² = 0.093, the maladaptively impulsive F_{(2, 80)} = 6.73, p = 0.002, partial η² = 0.144, and the anxious-avoidant class F_{(2, 40)} = 7.47, p = 0.002, partial η² = 0.272, while increased in the maladaptively perfectionistic class F_{(2, 36)=}13.32, p p < 0.001, partial η² = 0.425. Time and class interaction remained significant (p < 0.001) after controlling for BMI.

There was also a significant time and class interaction effect F(7.44, 344.06) = 2.85, p = 0.006, partial η² = 0.058) for Body concerns subscales. There was a significant decrease in Body concerns in the high-functioning class F_{(2, 194)} = 3.48, p = 0.033, partial η² = 0.035, but no change in the other classes, respectively maladaptively impulsive-perfectionistic class (p = 0.521, partial η² = 0.043), the maladaptively impulsive (p = 0.178, partial η² = 0.040), the anxious-avoidant (p = 0.982, partial η² = 0.001) and the maladaptively perfectionistic class (p = 0.076, partial η² = 0.180). Time and class interaction effect remained significant (p = 0.003) after controlling for BMI, but there was also a significant interaction effect between BMI and time F_{(1.87, 343.77)} = 4.15, p = 0.019, partial η² = 0.022) indicating that changes in body concerns across early adolescence also varied as a function of BMI. Given the small effect size and the exploratory nature of these repeated-measures analyses, this interaction should be interpreted cautiously.

For Food preoccupation, there was also a time and profile membership interaction effect F_{(7.88, 341.12)} = 2.07, p = 0.038, partial η² = 0.046), respectively an increase in the maladaptively perfectionistic class F_{(2, 28)} = 3.95 p = 0.031, partial η² = 0.220, and no change in the maladaptively impulsive-perfectionistic (p = 0.725, partial η² = 0.069), the high-functioning (p = 0.789, partial η² = 003), anxious-avoidant (p = 0.061, partial η² = 0.144), and the maladaptively impulsive class (p = 0.497, partial η² = 0.017). Time and class interaction effect remained significant (p = 0.034) after controlling for BMI.

For Pressure to eat, time and profile membership interaction effect was also significant F_{(7.73, 400.23)} = 6.66, p < 0.001, partial η² = 0.114). There was an increase in Pressure to eat scores in the anxious-avoidant class F_{(2, 44)} = 8.35, p = 0.001, partial η² = 0.275, but no change in any other classes, respectively the maladaptively impulsive-perfectionistic (p = 0.054, partial η² = 0.176), the high functioning (p = 0.075, partial η² = 0.024), the maladaptively impulsive (p = 0.587, partial η² = 0.012) and the maladaptively perfectionistic class (p = 0.909, partial η² = 0.005). Again, time and class interaction effect remained significant (p < 0.001) after controlling for BMI.

Discussion

The current study aimed to classify boys and girls in their early adolescence based on dimensions of perfectionism, impulsivity, and DE behaviors and attitudes.

The emerged five-profile model and statistical approaches for subtyping

The five profiles found had significant similarities with the previously reported undercontrolled (the maladaptively impulsive and maladaptively impulsive-perfectionistic), overcontrolled (the maladaptively perfectionistic), and high functioning (the high functioning) classes found in adults; see review by [43, 46]. With the exception of the anxious-avoidant group, the profiles also overlap with those reported in Boone et al. [47] community-based adolescent sample. The anxious-avoidant profile did not fully map onto previously established classes but shared features with the anxious-depressed group described in Thompson-Brenner et al. [55] study, particularly in its association with internalizing problems. To our knowledge, this is the first study to identify five distinct profiles specifically within an adolescent population.

The chosen analytical strategy may have influenced the identification of classes, particularly the differentiation of the anxious-avoidant profile. One factor is our decision to incorporate ED symptoms alongside personality traits in the LPA. This approach enriches profile identification by capturing the interplay between enduring traits and symptomatic behaviors, rather than treating them as separate domains. It also acknowledges that some individuals may show elevated trait levels without concurrent DE symptoms (or vice versa), with difficulties manifesting in other areas. As a result, this method yields greater explanatory power for conceptualizing the overall phenotype compared to analyses that do not include symptoms as indicators, which is also relevant in light of our earlier work showing that profiles derived from personality traits alone may have limited explanatory power for DE [64]. This is further illustrated by Christian et al. [79] who conducted an LPA on an undergraduate sample using perfectionism and impulsivity but excluding ED symptoms. Their four-class model explained only 8% of the variance in binge eating and 5% in restricting, while showing stronger associations with other conditions such as social anxiety and ADHD.

A second factor is the use of LPA itself, as compared to alternative strategies such as cluster analysis. LPA capitalizes on the dimensional structure of the data and reduces researcher bias in profile selection, thereby making it possible to detect groups not anticipated a priori. Together, these choices may help to explain why the anxious-avoidant profile emerged in our study, and they highlight how integrating traits and symptoms can advance both research and intervention by better reflecting the complexity of adolescent DE. Finally, it is important not to dismiss smaller or less common subgroups. Bohane et al. (2017) emphasize that they may still capture clinically meaningful differences or predict future outcomes. Similarly, it is also argued that while the precise number of personality trait-based profiles remains controversial, incorporating less well-documented classes can yield novel insights [81]. Together, these perspectives highlight the value of integrating traits and symptoms when profiling, and they help explain why the anxious-avoidant profile, though less typical, emerged as a distinct group in our study.

Disordered eating behavior and attitudes, perfectionism, and impulsivity in the five profiles

The above-discussed differences in statistical approaches may explain discrepancies in findings across studies regarding DE levels within identified profiles. In Boone et al.’s [47] study, individuals in the combined maladaptive perfectionism and impulsivity class reported the highest level of DE, while no differences between participants in the purely overcontrolled and undercontrolled classes were found. In our study, the highest level of DE behavior was present in the maladaptively perfectionistic profile, followed by the maladaptively impulsive-perfectionistic profile. This finding aligns with a study conducted in an adult sample of individuals with EDs, where the highest level of DE was found in the combined “healthy” and “unhealthy” perfectionism class [57]. In our previous study, which we conducted on patients with EDs and healthy controls, the highest level of DE was present in the combined dysfunctional perfectionism and impulsivity class. Still, high levels were also reported in the purely perfectionistic class [48]. This divergence between adult and adolescent findings may reflect developmental stage and sample composition differences. During early adolescence, impulsivity may operate as a broader transdiagnostic vulnerability linked to emotional and behavioral dysregulation, rather than being specifically tied to eating-related behaviors (further discussed below).

These differences emphasize the important role of impulsivity and perfectionism while studying DE behavior. In the maladaptively impulsive class, the level of DE was low and did not differ from the high-functioning class. That could explain the mixed results from previous research [26–28] regarding the relationship between impulsivity and DE and indicate that impulsivity alone is not a risk factor for the emergence of EDs in adolescence. On the other hand, high or even moderate perfectionism can be strongly associated with DE behavior. However, the maladaptively impulsive class may have problem behaviors in different domains - highly expected would be externalizing disorders, e.g., ADHD, behavior problems, etc.

Somewhat surprisingly, the maladaptively impulsive-perfectionistic profile did not have the highest level of DE. Instead, this profile was mainly characterized by emotional problems (high levels of depression and proneness to high anxiety). Several explanations might exist for that. Firstly, it may be that this class has individuals who express DE behavior, but their main difficulty lies in emotion regulation. Secondly, it may be that eating pathology is still developing in individuals belonging to this profile, and more severe difficulties could be identified at a later age. Thirdly, there may be other factors influencing DE. For example, in our sample, BMI and perceived pressure to be thin/muscular were highest in the maladaptively perfectionistic class. This suggests that perfectionism and impulsivity alone may not fully account for early DE risk, and that their interaction with factors such as BMI and sociocultural pressures may be particularly important. Supporting this, it has indeed been found that highly anxious preadolescents, whose BMI fell within the overweight or obesity range, are at an increased risk to develop DE in adolescence and young adulthood [82].

An interesting finding was the emergence of the anxious-avoidant class. These individuals resembled the high-functioning class in perfectionism and impulsivity but differed in reporting higher perceived pressure to eat, elevated trait anxiety, and consistently lower BMI from ages 11 to 14, based on earlier data examined through profile membership at age 13. This pattern suggests that restrictive eating in this group may be less driven by weight or shape concerns and more by anxiety or discomfort associated with eating itself. Such features show parallels with avoidant-restrictive food intake disorder (ARFID), which is recognized in DSM-5 and ICD-11 as involving restrictive eating due to anxiety, sensory sensitivity, or low interest in food rather than body image concerns [83–85]. Consistent with our findings, children with ARFID often report greater parental pressure to eat [86] and such pressure has been linked to picky eating and later DE behaviors [87–89]. Unfortunately, no specific measures to identify ARFID symptoms or picky eating were used in our study. Future research should incorporate comprehensive assessments, including ARFID-specific measures.

Disordered eating differences across profiles from ages 11 to 14

The differences between classes in body concerns and pressure to eat were already observable at ages 11–12, whereas differences in dieting and preoccupation with food became more pronounced at ages 12–13. Thus, differences in body-related concerns were observable earlier in development, whereas distinctions in eating-related behaviors and food-related preoccupations became more pronounced somewhat later. Importantly, since profiles were identified based on age-13 data, these earlier findings should be understood not as developmental trajectories but as a way of validating the profiles by examining whether distinctions were also evident at previous time points. For example, individuals in the maladaptively perfectionistic class were already more concerned about their bodies at the age of 11–12 years than individuals in other classes. Similarly, individuals in the anxious-avoidant class reported higher pressure to eat at the same age, as compared to other classes. This aligns with findings from a study, which examined developmental trajectories of eating pathology and found that girls with more pathological trajectories (defined as subgroups showing persistently higher levels of eating pathology over time) showed elevated DE behaviors as early as age 11 [90]. It has been reported that early onset of DE behaviors is likely to predict ongoing behavior [91], making it crucial to study which factors are associated with transition from subthreshold to full-threshold EDs. Miskovic-Wheatley et al. (2023) also conclusively underscored that factors such as a younger age of presentation and a shorter duration of illness at the first presentation are associated with more positive long-term outcomes, supporting the need for early intervention.

Our study suggests that prevention efforts should be implemented by ages 11–12 years or earlier, focusing on preoccupation with body appearance and weight as the core issues of DE. Furthermore, raising awareness of early eating-related attitudes and rigid weight-related beliefs within educational settings may facilitate earlier recognition of emerging concerns and foster supportive environments [93, 94]. Across all profiles, psychoeducation on healthy coping strategies and the promotion of a balanced relationship with food and body image may reinforce adaptive behaviors. For example, nearly all adolescents in the maladaptively perfectionistic profile reported efforts to lose weight by age 13 and experienced significant sociocultural pressure to attain thin or muscular ideals. Targeted interventions to reduce body concerns and foster body neutrality could thus be particularly beneficial.

In the earlier assessment waves, the maladaptively impulsive-perfectionistic profile resembled the maladaptively perfectionistic profile in dieting behavior. Over time, however, dieting appeared to increase in the perfectionistic group but decrease in the impulsive-perfectionistic group. One possible interpretation is that adolescents high in both impulsivity and perfectionism may initially attempt restrictive eating but struggle to sustain it due to elevated impulsivity, as higher impulsivity has been associated with binge–purge forms of eating pathology (e.g., BN, BED, and the binge–purge subtype of AN) [95–97]. While speculative, this interpretation aligns with both longitudinal [98, 99] and meta-analytic findings [100] indicating that restrictive ED presentations or AN can transition into binge-purge forms or behaviors. Early intervention targeting perfectionism and restrictive eating behaviors could be crucial in altering the trajectory of DE, potentially preventing the shift to more severe, impulsive behaviors. These observations underscore the potential value of early interventions that target perfectionism and restrictive eating. Supporting this, interventions aimed at reducing perfectionism have been shown to decrease both perfectionistic traits and DE [101].

Personality is still developing and changing during early adolescence. Therefore, deviations in levels of reported personality traits may reflect incompletely developed regulation strategies, as discussed by Boisseau et al. [102], which may stabilize and change with age [103, 104]. For example, while generally extraversion, agreeableness, openness, and conscientiousness increase during adolescence, there can be short term declines in adolescents’ openness and increase in neuroticism between between ages 11 to 13 years [105]. In addition, DE affects personality development in early adolescence. For example, prolonged (semi)starvation and dieting have been shown to increase rigidity and emotion dysregulation [106]. This reflects the pathoplastic relationship between personality and DE, where traits like perfectionism and impulsivity not only influence but are influenced by DE behaviors. Longitudinal studies are needed to assess the stability of personality-based profiles over time and the stability of individual membership in profiles (e.g., using latent transition analysis).

Practical implications toward earlier support opportunities

In addition to the general ideas previously presented about the timing and focus of prevention efforts, the identification of distinct profiles in adolescence provides valuable information for tailoring intervention strategies for DE. For example, recognizing adolescents with high maladaptive perfectionism may highlight the need for early strategies targeting rigid thinking patterns, while those with anxious-avoidant tendencies might benefit from approaches that reduce anxiety around food and foster positive eating experiences, including exposure-based techniques. Adolescents with maladaptively impulsive-perfectionistic tendencies, who show elevated emotional problems alongside DE, may benefit from interventions aimed at enhancing emotion regulation, such as skills drawn from dialectical behavior therapy [107, 108]. These examples remain preliminary but illustrate how personality-informed profiling could complement existing screening and prevention tools by flagging different constellations of vulnerability. At the same time, while our findings point toward directions for refining early identification and intervention strategies, distinct profiles do not preclude shared underlying processes. Future research should examine whether these profiles converge on common mechanisms (such as emotion regulation difficulties) which could provide transdiagnostic targets for more efficient and broadly applicable interventions. In this way, personality-informed profiling and a focus on shared mechanisms can serve as complementary rather than conflicting approaches.

Limitations and future directions

This study has some important limitations to consider. Firstly, replication studies for the identified profiles are needed, especially given the small but distinctive classes that emerged. Additionally, assessing other psychopathological issues beyond those currently included (such as ADHD and conduct problems) could enrich the findings. Future studies could incorporate symptoms of other disorders, particularly ARFID or selective eating, in LPA to support a more transdiagnostic approach. Third, while we examined earlier symptom patterns based on the emerged profiles, future studies assessing intraindividual stability by evaluating profiles across multiple time points are needed. Future follow-up assessments could help clarify whether individuals in the sample later develop diagnosable EDs, given that diagnostic measures were not incorporated in the present study. Fourth, an additional limitation concerns the use of raw BMI rather than BMI-for-age z-scores. Because exact age in months was not available in the dataset, precise z-score conversion was not feasible and findings involving BMI should be interpreted with caution. Future research should incorporate standardized BMI-for-age metrics when available. Lastly, the reliance on self-report questionnaires to assess constructs presents another limitation. Self-report data may introduce biases, such as social desirability or inaccurate self-assessment, potentially impacting the validity of the results. Future studies could benefit from a multi-method and multi-informant approach, including clinician-administered or parental assessments, along with behavioral measures, to enhance the robustness of the findings.

Conclusions

Our study demonstrates that meaningful profiles of maladaptive and adaptive perfectionism, impulsivity, and disordered eating can already be identified in early adolescence. Through latent profile analysis, we identified a five-profile solution: high-functioning, maladaptively impulsive, anxious-avoidant, maladaptively perfectionistic, and maladaptively impulsive-perfectionistic. These profiles differed not only in concurrent anxiety, depression, and BMI, but also showed distinct patterns of disordered eating at earlier assessment points (ages 11 and 12), providing validation for their distinctiveness. Rather than any single factor in isolation, the combined influence of personality traits, BMI, and sociocultural pressures appears to be especially relevant in shaping early disordered eating. This knowledge can inform the design of prevention and intervention programs that are better attuned to the vulnerabilities and needs of adolescents during this critical developmental period.

Author contributions

Both authors contributed to this paper significantly and have approved the final article. KS was involved in conceptualisation, data collection and curation, formal analysis, investigation, methodology, visualisation and writing – original draft. KA was involved in conceptualisation, data curation, investigation, methodology, project administration, resources, supervision and writing – review and editing.

Funding

The authors did not receive support from any organisation for the submitted work.

Data availability

The data that support the findings of this study are available from the authors upon reasonable request.

Declarations

Ethics approval and consent to participate

Approval was obtained from the Research Ethics Committee of the University of Tartu. The procedures used in this study adhere to the tenets of the Declaration of Helsinki. Informed consent was obtained from all individual participants included in the study as well as from their parents.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.