Interoceptive impairments in early-stage anorexia nervosa: exploring the impact of childhood trauma and heart rate variability

Abstract

Background

Anorexia nervosa, restricting subtype (AN-R), is a severe psychiatric disorder, with abnormal interoception, autonomic nervous system disturbances, and increased exposure to childhood traumatic experiences (CTEs), frequently observed as correlates. This is the first study exploring the impact of CTEs and heart rate variability (HRV; i.e., an index reflecting parasympathetic arousal) on interoceptive accuracy (IA; i.e., the ability to track changes in bodily signals) in AN.

Methods

Twenty-five patients with AN-R within a year of onset and 25 matched healthy controls were recruited. IA was assessed through the heart beat detection task. HRV was measured before and after the task. Participants also completed the Childhood Trauma Questionnaire. We performed a to detect significant differences between groups in HRV reactivity and IA, and a linear regression to test the effect of factors of interest on IA.

Results

Patients with AN-R displayed significantly increased HRV reactivity and decreased IA compared to HCs. They also reported significantly more CTEs. Furthermore, childhood emotional neglect significantly predicts IA impairments.

Conclusions

Although the pathway linking emotional neglect to abnormal interoception in AN-R remains to be clarified, an embodiment-informed framework may show promise in the treatment of individuals with eating disorders who experienced childhood maltreatment.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-025-01255-5.

Plain language summary

Anorexia nervosa (AN) is a serious mental health disorder. It affects how people sense their body signals, including heartbeats. AN can interfere with the autonomic nervous system. It often occurs in those who experienced trauma during childhood. This study explored how childhood trauma and heart rate variability (HRV) impact interoceptive accuracy (IA), which is the ability to feel bodily signals, in people with AN. Researchers studied 25 people newly diagnosed with AN and 25 healthy individuals. They looked at IA using a specific task. They measured HRV before and after the task. They also used a questionnaire to assess childhood trauma. People with AN had lower IA. They also had more reactive HRV. Plus, they reported more childhood trauma, especially emotional neglect. Emotional neglect was also a strong predictor of IA impairments. The study shows that treatments focusing on how the body handles emotions may help people with eating disorders, especially those who faced childhood trauma.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40337-025-01255-5.

Background

Anorexia nervosa (AN) is a severe psychiatric disorder characterized by self-imposed starvation and a profound fear of weight gain. Despite an overall stable incidence rate, it represents a significant public health concern [1], resulting in the in the highest mortality rate among psychiatric disorders [2, 3], its diagnosis among children and adolescents shows an increasing trend [1].

In the restricting subtype of anorexia nervosa (AN-R), severe caloric restriction and extreme weight loss—without recurrent binge-eating or purging behaviors—are defining characteristics. These behaviors contribute to a broad spectrum of adverse health outcomes, many of which are also observed in the binge-eating/purging subtype [4]. This includes, but is not limited, to metabolic dysfunction, endocrine irregularities, and cardiovascular impairment. Such complications substantially magnify the vulnerability to various conditions and escalate the probability of early death [5]. Among cardiac complications, disturbances within the autonomic nervous system (ANS) contribute to the heightened risk of cardiovascular-related mortality and the abrupt onset of fatal events in these patients [6, 7]. The ANS links the cardiovascular and central nervous systems, flexibly regulating internal bodily processes in response to both physiological shifts and environmental cues [6]. Operating as a dynamic regulator, the ANS involves delivering sensory feedback from organs through brain areas, such as the brainstem and hypothalamus. This enables the ANS to adjust its output, and to effectively regulate the body's physiological state [6].

Heart rate variability (HRV), i.e., the variation in time intervals between consecutive heartbeats, may serve as a non-invasive measure of ANS (i.e., vagal) activity, offering insights into an individual's physiological adaptability and stress response [8]. Of note, alterations in HRV, including phasic change in cardiac vagal activity, have been consistently reported in individuals with AN-R, possibly contributing to cardiovascular dysfunction and to diverse psychopathological conditions, such as psychological inflexibility and blunted emotional awareness, which constitute core features of AN-R [8, 9].

Interoceptive accuracy (IA) generally refers to the capacity to perceive and monitor subtle change in internal bodily cues, encompassing signals from muscles, skin, joints, and visceral afferents [10]. Disordered interoception has long been recognized as a core psychopathological feature of eating disorders (EDs) [11], with individuals with anorexia nervosa (AN) often exhibiting impairments in accurately perceiving and interpreting bodily signals [12–14]. However, research on the underlying mechanisms contributing to this relationship remains limited [14]. Of note, autonomic reactivity may influence internal bodily state perception [16]. Moreover, parasympathetically driven changes in heart rate may affect interoceptive abilities, possibly via prefrontal and (para-)limbic brain region activity [17, 18]. In addition, individual differences in IA may influence emotion recognition through the ANS [16, 19, 20].

AN-R is thought to result from a complex interaction of factors [5]. Genetic predispositions may create a vulnerability that is further intensified by early life experiences and environmental risks [5, 21]. Childhood trauma (CT) has been identified as a non-specific environmental risk factor for the onset of EDs in adulthood, though the explanatory frameworks for this association are still to be clearly defined. [22]. Furthermore, individuals with EDs who have experienced childhood maltreatment tend to exhibit a more severe clinical profile, including an earlier age of onset, greater symptom severity, increased suicidal or self-harm behaviors, as well as poor long-term outcome [23, 24]. Emerging evidence suggests a significant link between early traumatic experiences and alterations in interoception, with CT hindering an adequate processing and integration of body signals [25]. This disruption may increase vulnerability to mental disorders, including AN-R, by possibly impairing the ability to accurately perceive internal states [26, 27].

Given the established influence of vagus nerve activity on interoceptive pathways [20, 28], and the potential interaction between childhood traumatic experiences and interoception [27], we aimed to explore the joint impact of phasic changes in HRV, a proxy for dynamic vagal activity, and CT on IA in a carefully selected group of individuals with AN-R within a year of onset. We hypothesized that individuals with AN-R would show reduced IA compared to age- and education-matched healthy controls (HCs), with HRV and CT serving as predictors of IA measured via the heartbeat counting task.

Methods

Participants

Twenty-five female patients, aged 14 to 30 (mean age = 21.8 ± 5.17), diagnosed with DSM-5 AN-R were consecutively recruited from outpatient clinics specialized in Feeding and Eating Disorders at the Fondazione Policlinico Universitario Agostino Gemelli IRCCS of Rome, Italy. Only patients admitted within a year of beginning weight loss or failing to achieve age-expected weight gain were considered eligible for the study to ensure a carefully selected clinical group with first-onset AN-R, thus avoiding the potentially confounding effects of prolonged malnutrition on cardiac activity [29].

We used the Structured Clinical Interview for DSM-5 Research Version (SCID-5-RV; [30]) to diagnose AN-R and identify psychiatric comorbidities. Additionally, we screened patients for Personality Disorders with the Structured Clinical Interview for DSM-5 Personality Disorders (SCID-5-PD; [31]). Diagnostic interviews were conducted at the beginning of the study by highly trained raters with a high level of inter-rater reliability (k > 0.8). The following inclusion and exclusion criteria were adopted to allow a suitable study sample: all participants needed to be in a stable medical phase of their illness, with no weight loss in the past month, and currently meet DSM-5 criteria for AN-R, without past or present comorbid psychiatric disorders or substance use. Additionally, participants had to demonstrate proficiency in written and spoken Italian and show normal cognitive function, as assessed with Raven’s progressive matrices [32]. We recruited through local online advertising 25 gender-, educational-, and age-matched healthy controls (HCs). All HCs were screened for a lifetime personal history of psychiatric disorders by means of the SCID-5-RV and the SCID-5-PD. Any other DSM-5 psychiatric disorder qualified for exclusion from the HC group. Other eligibility criteria did not differ between HCs and the AN-R group. For both AN-R and HC groups, unstable medical illnesses (such as head trauma, concussion, neurological and cardio-respiratory diseases, and diabetes), cognitive impairment, as well as the current use of medications that alter cardio-respiratory activity [33] were additional exclusion criteria. Furthermore, since regular exercise can affect HRV [33, 34], we recruited only individuals not regularly engaging in athletic or endurance sports. Similarly, to minimize the impact of malnutrition on HRV, included patients were prescribed a diet of at least 1000 cal per day for a minimum of one month prior to the start of the study. The final sample consisted of 50 individuals, 25 with AN-R within a year of onset and 25 gender-, age-, and educational level-matched HCs. This sample size was justified by reviewing previous neurophysiological studies that assessed HRV and IA in clinical groups with psychiatric disorders, where sample sizes typically ranged from 20 to 30. [8, 35–38]. In addition, we conducted sensitivity power analysis to determine the minimum detectable effect size given our fixed sample size, significance level, and desired power (shown in the Supplementary Material).

The study received approval from the local Ethics Committee (Protocol ID 2014) and was conducted in agreement with the Declaration of Helsinki and its ethical principles for medical research involving human subject as developed by the World Medical Association. Participants had to provide written, informed consent after receiving full explanation of all study procedures. Participants did not receive compensation for their involvement in the study.

Procedure

Participants’ age, weight and height were noted; the latter two served for calculating body mass index (BMI). Smoking status, demographics, and clinical characteristics were recorded for each participant on admission. The study was conducted over two consecutive days. On the first day, psychometric assessments were administered, while the second day was dedicated to the heartbeat detection task, during which electrocardiogram (ECG) recordings were obtained. Additionally, to measure HRV reactivity, a 2-min resting ECG was taken before (Baseline) and after (Recovery) task completion on Day 2. Participants were instructed to abstain from caffeine, tobacco, and alcohol for 2 h before IA assessment and ECG recording [33]. Three 10 mm Ag/AgCl pre-gelled electrodes (ADInstruments, UK) were placed on participants’ wrists in an Einthoven's triangle configuration. The ECG data were converted and amplified using an eight-channel amplifier (PowerLabT26; ADInstruments UK), then displayed, stored, and analyzed with the LabChart 7.3.1 software package (ADInstruments Inc, 2011). All tasks were performed while participants were comfortably seated in a quiet, well-lit room, while relaxing and remaining as still as possible during the recordings, so to minimize motion artifacts.

Psychometric assessment

Upon arrival at the laboratory, participants completed the following psychometric scales:

-Childhood Trauma Questionnaire-Short Form (CTQ), a 28-item, self-report, retrospective, questionnaire [40] designed to investigate traumatic childhood experiences. Respondents choose from five possible answers, ranging from 'never true' to 'very often true,' based on the frequency of the events. The questionnaire evaluates five trauma types, i.e., emotional abuse or neglect, physical abuse or neglect, and sexual abuse. Scores for each trauma type range from 5 to 25, with higher scores indicating greater levels of childhood mistreatment. The CTQ has been utilized in clinical [41, 42] and non-clinical populations [43],the tool displayed high reliability [43].

-Eating Disorder Examination Questionnaire (EDE-Q). Eating disorder symptom severity was evaluated with the EDE-Q version 6.0 [44]. This is a 28-item, self-rated questionnaire assessing attitudinal features of EDs and core ED behaviors, including concerns about eating, weight, and shape, as well as dietary restraint over the past 28 days. Responses are recorded on a 7-point Likert scale,scores range from 0 to 6 for each item. The instrument generates four subscales, which are combined to create an overall severity measure. Higher scores indicate increased ED severity. The EDE-Q has shown strong reliability and validity in AN-R clinical populations [45], including those receiving nutritional rehabilitation [46].

-Depression Anxiety and Stress Scale-Short Form (DASS). We employed the DASS to assess subthreshold affective symptoms over the past week [47]. The DASS is a self-report tool consisting of 21 items, created to evaluate the severity of overall psychological distress and symptoms associated with depression, anxiety, and stress. Each item is classified into four Likert responses from 0 (“nothing”) to 3 (“Most of the time”). Scores are reported as a total ranging from 0 to 63, which is calculated by combining the scores from each of the three subscales. The DASS showed excellent psychometric properties in both non-clinical and clinical samples [48].

Heartbeat detection task

IA was evaluated during the heartbeat detection task [49]. The Schandry task is the most frequently adopted method for evaluating IA [50]. In this task, participants were instructed to silently count their own heartbeats across four randomly presented time intervals each consisting of 25 s, 35 s, 45 s, and 100 s. The start-point of each interval was heralded by an initial audio-visual cue, which was followed by a stop cue indicating its end. Subsequently, participants reported the number of heartbeats they perceived. No feedback regarding the upcoming duration of trials or performance quality was provided. Participants were not allowed to utilize any aids or strategies, such as feeling the pulse on their wrist. To reduce the use of estimation-based strategies, participants were instructed to count only the heartbeats they actually felt, without counting or guessing [50, 51]. IA was computed based on the absolute difference between the estimated and actual recorded number of heartbeats using the formula: $1/4\sum (1-(|\text{recorded beats}-\text{counted beats}|)\text{recorded beats})$. This formula yields an IA score ranging from 0 to 1, where a score of 1 indicates perfect heartbeat tracking and scores closer to zero suggest poorer interoception. Findings support the discriminant validity of IA scores derived from the heartbeat detection task [52].

ECG analysis

The ECG was sampled at 1 kHz. The analysis of the ECG R-R time series before and after the heartbeat detection task first involved the elimination of the first 10 acquired data-points due to their non-stationary conditions generating their signals. Subsequently, for each curve, an outlier and ectopic beat removal analysis was performed, followed by linear interpolation and Malik filter interpolation, respectively. If the impact of outlier and ectopic beat removal is found to exceed a threshold (e.g., 10–20% deviation from the expected HRV range), we discarded the affected segments to avoid erroneous conclusions.

Using HRV analysis package on Python, we removed outliers from R-R time series and we calculated RMSSD, defined as: $RMSSD=\sqrt{\frac{1}{N-1}{\sum }_{i=1}^N{({\mathit{RR}}_{i+1}-{\mathit{RR}}_i)}^2}$. RMSSD is a time-domain metric reflecting vagal influences on cardiac chronotropy and, unlike other HRV parameters, is relatively unaffected by respiration [33].

HRV reactivity was defined as the change in absolute HRV values from baseline to recovery, allowing us to quantify the variation in cardiac vagal activity between a resting state and a post-event condition.

Statistical analysis

Demographic and clinical (scores on the CTQ, DASS, and EDE-Q) characteristics of patients with AN-R and HCs were compared using the chi-squared (χ²) test for categorical measures and Student’s T-test for continuous variables. To identify significant differences in phasic changes in HRV and interoception between individuals with AN-R and HCs, we conducted a multivariate analysis of covariance (MANCOVA). HRV reactivity and IA were the dependent variables, with group (AN-R vs. HCs) as the independent factor, and DASS total score, age, smoking status, and BMI as covariates. This analysis was essential due to evidence suggesting that affective symptoms, age, smoking, and BMI influence the autonomic activity of the heart [33] and interoceptive parameters [14]. f²(V) was provided for effect size [small effect: f²(V≈0.01; medium effect: f²(V)≈0.06; large effect: f²(V)≈0.16; [53]]. In case of significance of the initial model, we performed a series of one-way analyses of covariance (ANCOVA) to examine group differences on the dependent variables. To further confirm significant differences between individuals with AN-R and HCs in resting-state HRV, we conducted supplemental analyses that considered repeated measures of the autonomic activity of the heart from baseline to recovery (shown in the Supplementary Material). In the second part of our analysis, we conducted a multiple linear regression on the overall sample to predict the severity of IA impairments. This model included factors that demonstrated significant between-group differences in the univariate and bivariate analyses.

Cohen’s f² was provided for local effect sizes [f² ≥ 0.02, f² ≥ 0.15, and f² ≥ 0.35 represent small, medium, and large effect sizes, respectively [53]]. The significance level was set at 5%. Possible multicollinearity among variables of interest was assessed using variance inflation factor (VIF) indicators. All statistical analyses were conducted using JASP.

Results

As an effect of matching, participants with AN-R and HCs did not differ significantly for age, educational level or living conditions, occupation and smoking status. Patients presented with significantly lower BMI with respect to HCs. Individuals with AN-R and HCs significantly differed for family history of psychiatric disorders. Moreover, patients reported significantly more childhood traumatic experience compared to HCs, including emotional abuse and neglect, physical abuse and neglect, and sexual abuse, as well as significantly higher DASS and EDE-Q total scores (Table 1).

Table 1.

Clinical, demographic, and psychometric data of the sample.

	AN-R (N = 25)	HC (N = 25)	Overall	df	χ2 or t	p
Demographic and clinical characteristics
Age (M ± SD)	21.8 ± 5.1	24.2 ± 3.5	23.0 ± 4.5	48	-1.9	0.058
Body Mass Index (M ± SD)	15.2 ± 1.5	21.1 ± 2.0	18.2 ± 3.4	48	-11.5	< .001
Years of education (M ± SD)	15.5 ± 2.5	16.6 ± 2.0	16.0 ± 2.3	48	-1.7	.091
Overall weight loss (M ± SD)	8.9 ± 3.1
Duration of illness in months (M ± SD)	4.5 ± 3.0
Living alone (n%)	2(8.0)	2(8.0)	4(8.0)	1	0.0	1.0
Single child (n%)	2(8.0)	4(16.0)	6(12.0)	1	0.7	0.380
Early weaning from breastfeeding (n%)	2(8.0)	0(0.0)	2(4.0)	1	2.0	0.149
Occupation (n%)	20(80)	19(76)	39(78.0)	1	0.1	0.773
Smoking (n%)	9(36.0)	12 (48.0)	21(42.0)	1	0.7	0.390
Family history of psychiatric disorders (n%)	6(24.0)	0(0.0)	6(12.0)	1	6.8	0.009
Psychometric assessment
CTQ-Emotional Neglect (M ± SD)	11.4 ± 5.9	6.8 ± 3.2	9.1 ± 4.8	48	3.8	< .001
CTQ-Emotional Abuse (M ± SD)	9.3 ± 4.3	6.3 ± 1.4	7.8 ± 3.5	48	3.3	0.002
CTQ-Physical Neglect (M ± SD)	6.9 ± 2.9	5.5 ± 1.0	6.2 ± 2.3	48	2.2	0.027
CTQ-Physical Abuse (M ± SD)	6.2 ± 2.7	5.0 ± 0.2	5.6 ± 2.0	48	2.2	0.026
CTQ-Sexual Abuse (M ± SD)	7.6 ± 4.9	4.9 ± 1.1	6.3 ± 3.8	48	2.7	0.010
DASS-21 (M ± SD)	57.3 ± 28.3	22.1 ± 16.6	39.7 ± 29.1	48	5.3	< .001
EDE-Q (M ± SD)	3.2 ± 1.5	1.0 ± 1.0	2.1 ± 1.6	48	5.0	< .001

Significant results in bold characters. Abbreviations: AN-R = Anorexia Nervosa-restrictive subtype; CTQ = Childhood Trauma Questionnaire; DASS-21 = Depression, Anxiety and Stress Scale; EDE-Q = Eating Disorder Examination Questionnaire; df = degrees of freedom; HC = healthy controls M = mean; p = statistical significance; SD = standard deviation; t, Student’s t; χ² = chi-squared test.

The MANCOVA revealed a significant global effect [f²(V) = 0.235, Pillai’s Trace V = 0.19, F = 4.30, df = 2, p = 0.021)] of the variables of interest on the two groups. Multivariate normality was confirmed by the Shapiro–Wilk Multivariate Normality Test (W = 0.98; p = 0.78). Subsequent univariate ANCOVAs showed that patients with AN-R had a significant increase in HRV reactivity [F = 4.77, p = 0.035; AN-R: 3.15 ± 1.46; HCs: 2.31 ± 1.15], as well as significantly decreased IA (F = 5.0, p = 0.030; AN-R: 0.42 ± 0.24; HCs: 0.58 ± 0.11) compared to HCs, even after adjusting for DASS total score, age, smoking status, and BMI. Notably, none of these covariates were significant (Table 2). According to the multiple linear regression model, childhood emotional neglect significantly predicts interoceptive impairments on the heart beat detection task, with CTQ emotional neglect subscale score negatively associated with overall IA values (t = -4.17; p < 0.001; Table 3). Fit measures (R² = 0.47, R² adjusted = 0.30, F = 2.88), suggest that the model accounts for approximately 47% of the variance in IA, indicating a moderate-to-strong fit between the observed and predicted values. The model showed no significant multicollinearity, as indicated by lower than 4 VIF values of all variables of interest [54].

Table 2.

Analysis of covariance for HRV reactivity and IA by DASS-21, BMI, age, and smoking status as covariates

		Mean square	df1	df2	F	p
HRV reactivity	Group	7.68	1	36	4.7	0.035
	DASS-21	0.62	1	36	0.3	0.537
	BMI	0.10	1	36	0.0	0.800
	Age	3.96	1	36	2.4	0.125
	Smoking	5.73	1	36	3.5	0.067
IA	Group	0.19	1	36	5.0	0.030
	DASS-21	0.00	1	36	0.1	0.788
	BMI	0.00	1	36	0.0	0.834
	Age	0.01	1	36	0.3	0.541
	Smoking	2.69	1	36	0.0	0.933

Significant results in bold characters. Abbreviations: BMI = body mass index; DASS-21 = Depression Anxiety and Stress Scale; df1 = Degrees of freedom between groups; df2 = Degrees of freedom within groups; F = value of variance of the group means; HRV = Heart Rate Variability; IA = Interoceptive Accuracy

Table 3.

Multiple linear regression: effect of predictors on IA total score

			95% Confidence Interval
Predictor	Estimate	SE	Lower	Upper	t	p	f2
BMI	0.51	0.01	− 0.04	0.00	−1.32	0.195	0.030
Family history of mental disorders	0.23	0.14	− 0.05	0.52	1.63	0.112	0.046
EDE-Q	0.00	0.02	− 0.03	0.04	0.02	0.840	0.001
DASS-21	−0.00	0.00	−0.00	0.00	−1.09	0.283	0.037
HRV reactivity	−0.02	0.02	− 0.07	0.02	− 1.12	0.270	0.021
CTQ-Emotional Neglect	− 0.02	0.01	− 0.04	− 0.01	− 4.17	<0.001	0.401
CTQ-Emotional Abuse	0.00	0.01	− 0.01	0.02	0.56	0.579	0.005
CTQ-Physical Neglect	− 0.01	0.01	− 0.04	0.01	− 1.09	0.282	0.020
CTQ-Physical Abuse	− 0.05	0.03	− 0.11	0.01	− 1.70	0.098	0.050
CTQ-Sexual Abuse	− 0.00	0.01	− 0.03	0.02	− 0.42	0.672	0.003

Significant results in bold characters. Abbreviations: BMI = Body Mass Index; CTQ = Childhood Trauma Questionnaire; DASS-21 = Depression, Anxiety and Stress Scale; EDE-Q = Eating Disorder Examination Questionnaire; HRV = Heart Rate Variability; p = statistical significance; SE = standard error; t = Student’s t statistic; f² = Cohen’s f

Discussion

In line with our hypothesis, our findings suggest that patients with AN-R showed reduced IA compared to HCs. Accordingly, individuals with AN-R may struggle not only with recognizing specific visceral cues related to hunger and satiety, but also with a general reduction in their ability to accurately perceive bodily signals [55]. Interoception in AN-R has been traditionally assessed through self-report questionnaires such as the Eating Disorder Inventory (EDI; [56]). Specifically, individuals with AN-R generally score higher on the “interoceptive deficits” subscale of the EDI compared to HCs, indicating an overall diminished interoceptive sensitivity. This was found to predict ED symptom onset in multiple studies [57, 58]. Conversely, impaired IA in AN has been shown in few studies [15, 59, 60] but not in others [35, 61–63]. However, considerable variability exists across studies in aspects such as including cases with the AN binge-purge subtype along restrictive AN, different ranges of illness duration, psychiatric comorbidities, and the specific measures of IA used, thus complicating direct comparisons [14]. While EDI interoceptive deficit was also found in individuals with AN, there was no correlation between IA performance and EDI interoceptive deficit scores, thus suggesting that the subjective evaluative interoceptive experience and objective IA do not overlap in AN. While these results, derived from measures of IA at rest, may indicate a trait-like condition of reduced interoceptive perception in AN, there are also intriguing findings suggesting that patients with AN may perceive interoceptive cues in critical arousing/aversive contexts, such as meal anticipation, more intensely, rather than in a more dampened way [65]. Similarly, there is evidence for increased neural activation of insula and anterior cingulate cortex while looking at food pictures in individuals with acute AN, suggesting that interoceptive signal processing may be intensified in critical situations that impact relevant symptom areas [66]. Consistently with this conceptual framework, impaired perception of bodily cues in AN may be interpreted as a possible coping strategy of a phasic aversive interoceptive signaling [58].

Most neurophysiological studies indicate that patients with AN typically show significantly higher tonic HRV than control groups, reflecting a parasympathetic/sympathetic imbalance with parasympathetic dominance and reduced cardiac sympathetic modulation, especially at the onset of the illness [7, 9]. Notably, individuals with AN showed the highest RMSSD compared to other mental illnesses [8]. This pattern may represent a physiological adaptation to prolonged starvation [67], and may ultimately lead to marked resting bradycardia and increased HRV. Over time, and with increasing severity of the disorder, the pattern of cardiac activity shifts from being primarily controlled by the vagus nerve to being dominated by the sympathetic nervous system [29]. However, in our supplementary analysis, we did not observe in our sample such difference in baseline (i.e., resting) HRV. Notably, existing evidence suggests that refeeding restores HRV levels to those observed in healthy individuals [68, 69]. This finding aligns with animal studies on caloric restriction and unrestricted feeding, suggesting that HRV may serve as a non-invasive physiological marker to complement existing assessment tools in identifying AN-related cardiac dysfunction and monitoring treatment progress [9]. For instance, a study involving individuals with AN currently undergoing nutrition rehabilitation reported no differences in resting HRV [69]. Accordingly, the fact that patients with AN in our study received specific nutritional interventions may have limited our ability to detect significant differences in baseline HRV.

Compared to tonic HRV, relatively little is known about phasic HRV in AN. According to various lines of evidence, phasic cardiac vagal activity is thought to reflect self-regulatory efforts [70]. In line with this hypothesis, tasks that require suppression or reappraisal of negative emotions have been associated with increased phasic HRV [71, 72]. Similarly, phasic cardiac vagal control has been consistently shown to increase during recovery following various forms of stressful emotional induction [70]. Of note, heightened emotional reactions to body-related stimuli were observed in individuals with AN, both at subjective and neurophysiological levels [73, 74]. For instance, meta-analytic evidence indicates that emotion-related neural networks are engaged in processing bodily stimuli in individuals with AN, suggesting that negative emotional arousal is linked to a cognitive processing bias toward body-related stimuli [76]. Thus, in line with this conceptual framework, the increased phasic HRV observed among patients with AN may result from active efforts to self-regulate distressing emotional arousal triggered by body-related cues. Alternatively, the elevated HRV reactivity observed in the AN group during the heartbeat detection task could be due to difficulties in focusing on body-related cues, as attention-demanding or challenging stimuli are known to suppress HRV [77].

Consistently with previous findings [78, 79], patients with AN-R reported significantly more childhood traumatic experiences, compared to HCs. In addition, increased emotional neglect, but not HRV reactivity, was found to significantly predict impaired IA, as measured by the heartbeat detection task, in our sample. To our knowledge, no other study has reported this relation in EDs. Childhood maltreatment and neglect are recognized a non-specific risk factor for ED, including AN [78, 80]. The potential connection between a history of childhood neglect and the development of ED symptomatology later in life appears to be rooted in impaired emotion regulation mechanisms, including emotional awareness [58, 79, 81]). Individuals who suffer from childhood emotional neglect may have inadequate emotional reference patterns, including early experiences of being comforted and soothed by their caregivers, and thus greater difficulty in interpreting, understanding, and modulating their own emotional experiences [78]. As a consequence, they are compelled to develop external mechanisms, including eating symptoms, such as dietary restriction, to create a sense of control and to provide modulation over overwhelming affective states [82, 83].

The significance of interoception in emotion perception and regulation has been recognized since the earliest theories of emotion, which postulated a close relationship between the magnitude of an individual’s sensitivity to bodily signals and the experience of emotions [16, 85, 86]. In addition, emerging evidence suggests that increased emotional awareness is associated with higher IA, as measured by the heartbeat detection task, in healthy individuals [19, 88]. Similarly, empirical evidence originating from infant research suggest that both the recognition of one's own emotions and the awareness of one's bodily sensations are mediated by the caregiver's ability to identify and reflect these back to the child (Tronick, 2007; [89]). This occurs within an embodied process of mutual influence, which ultimately leads to the development of mechanisms for self-regulation of both emotional and somatic experiences over time [90]. Accordingly, individuals who experienced neglectful parenting during childhood may disregard internal bodily experiences by focusing, instead, on external stimuli because of a history of CT. In line with this conceptual framework, research indicates that interoception is influenced by social interactions, including parenting styles [91], with increased parental rejection of negative emotions linked to impaired interoceptive processes in youth [92].

Contrary to our hypothesis, HRV values did not predict IA performance in our sample. To date, only one study [35] has examined the relationship between HRV (at rest) and IA in AN. This study found that higher HRV was linked to poorer performance on the heartbeat counting task in AN, but the opposite trend was observed in HCs. Although we identified increased HRV reactivity during the heartbeat counting task alongside reduced IA in patients with AN-R, we could not replicate this association. To deepen the understanding of the relationship between HRV and IA in EDs, larger studies are required. Additionally, future research should explore how these variables interact across various emotion induction paradigms.

Collectively, our findings support an embodied conceptualization of AN, wherein CT is conceptualized as an event, or a series of interpersonal events, that disrupt the coherence of the self, resulting in a sense of alienation from one’s own body and emotions. Aligned with this perspective, our findings underscore the importance of psychotherapeutic interventions that address the personal significance ascribed to ED symptoms. This therapeutic approach necessitates a comprehensive exploration of the interpersonal dynamics and contexts that precede disordered eating behaviors, with the aim of identifying underlying emotional disturbances and disrupted bodily awareness. By fostering insight into these processes, such interventions can support the development of more adaptive coping mechanisms. Although standardized treatment protocols remain underdeveloped, approaches that promote the recognition, differentiation, and integration of internal states hold promise for improving therapeutic outcomes.

Conclusions

Before providing the summary of our findings, we duly acknowledge certain potential limitations that could have affected the generalizability of our results. First, the relatively small sample size limits the generalizability of our findings to the broader population of individuals with AN. Second, to ensure conceptual and methodological validity, we enrolled a sample of AN-R patients in a stable phase of illness without psychiatric comorbidities. While this approach, which excluded individuals with severe psychiatric comorbidities, may have limited generalizability by introducing selection bias, it can also be considered a strength of the study. In addition, the severity of eating symptom distress reported by patients in our sample is comparable to that observed in studies involving individuals with AN and comorbid anxiety, mood, and personality disorders [93, 94]. Third, the Schandry task has been subjected to some criticism, including heartbeats underestimation, low correlation between actual and reported heartbeats, increased heartbeat detection at slower heart rate and with shorter time intervals [95]. However, most of these concerns have been strongly disputed in recent literature [96, 97], and the Schandry task remains the most widely used tool for assessing IA and is often considered the gold standard in interoception research [50]. Fourth, the evaluation of IA was based on behavioral assessments and self-reported measures of interoception. Finally, the reliability of the retrospective assessment of CT experiences, as measured by the CTQ in adulthood, may be affected by uncontrolled recall bias, including minimizations in reports of CT. However, the CTQ is widely adopted and considered to be a reliable psychometric instrument for evaluating early adverse experiences [43]. To our knowledge, this is the first study that investigated the combined influence of phasic alterations in HRV and CT on IA, targeting a selected cohort of individuals diagnosed with AN-R within a year of onset. The findings support current models indicating that CT may affect the whole mind-brain-body relationship [98]. Accordingly, interventions grounded in an embodiment-informed framework—aimed at enhancing patients' ability to perceive and recognize their bodily sensations and emotions—may represent a promising approach for treating individuals with EDs who have experienced childhood maltreatment.

Abbreviations

AN

Anorexia nervosa

AN-R

Anorexia nervosa restricting subtype

CT

Childhood trauma

CTE

Childhood traumatic experiences

HCs

Healthy controls

HRV

Heart rate variability

IA

Interoceptive accuracy

Author contributions

Conceptualization, LM; Data Curation, LM, CS, MA; Formal Analysis, LM, CS; Methodology, LM, CS, MA, GM, MDS, VG, GS; Supervision, GM, MDS, VG, GDK, GS, Writing – original draft, LM, CS, EC, GM, MDS; Writing – review & editing, GDK, DJ, VDG, FC. All authors have visualized and approved the final version of the manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study received approval from the UCSC Institutional Review Board (PROTOCOL ID: 2014). All participants provided written informed consent to be enrolled in the study.

Consent for publication

Not applicable

Competing Interests

The authors declare no competing interests.