Introduction

In the context of preventive psychiatry, clinical high risk for psychosis (CHR-P) criteria have been developed to identify individuals at increased and imminent risk of developing psychosis [1]. Individuals at CHR-P exhibit sub-threshold positive psychotic symptoms, impaired functioning, and low quality of life, with a 25% risk of developing a psychotic disorder within the following 3 years [2,3,4]. Preventive interventions are delivered to this clinical population with the aim of delaying, postponing, or, ideally, preventing the onset of full-blown psychotic disorders and reducing distress [5].

Notably, CHR-P individuals frequently exhibit comorbid mental disorders, including eating disorders (EDs), which have been found in 3% of this clinical population [6]. This co-occurrence is not surprising, as the co-existence of psychotic conditions and EDs or ED symptoms has been documented in several studies [7, 8]. For example, a systematic review of disordered eating in individuals with schizophrenia spectrum disorders revealed prevalence rates of 4.4–45% for binge eating symptoms, 27–60.6% for food addiction, 16.1–64% for food craving, and 4–30% for night eating syndrome among studies [9]. In addition, genome-wide association studies (GWAS) have identified shared genetic vulnerability in anorexia and schizophrenia [10]. Moreover, preliminary evidence suggests that, among individuals with schizophrenia, cases with premorbid EDs may be considered as a homogeneous subgroup [11].

However, most of the current evidence focuses on EDs in relation to full-blown psychotic disorders. In contrast, research on the co-occurrence of CHR-P and EDs is far less investigated. This is quite surprising since EDs and schizophrenia spectrum disorders may exhibit symptomatic and phenomenological overlap at or even before their onset, with clinical manifestations including delusional-like distortions of body image, a fragmented sense of body ownership, altered bodily experiences, delusional ideation focused on eating and food, and auditory hallucinations focused on body and eating [12].

Treating or, ideally, preventing EDs and ED-related symptoms in the CHR-P population could improve clinical outcomes for several reasons. First, EDs are potentially life-threatening mental disorders, which impact both general functioning and physical health [13, 14]. Second, EDs are associated with chronicity rates of 33% and 18% at less than 2 years and more than 10 years of follow-up, respectively, underscoring poor long-term courses of these disorders. Third, EDs may be associated with severe outcomes, including hospitalization (26% of cases) and mortality (0.4% of cases) [15]. Notably, addressing EDs in CHR-P individuals is also consistent with clinical guidelines and studies on CHR-P state, which emphasize the importance of addressing comorbid mental disorders in CHR-P individuals beyond psychosis prevention [6, 16,17,18,19,20,21].

Previous reviews have focused on comorbid mental disorders in CHR-P samples, including EDs, assessing their prevalence and impact on the transition to psychosis using a systematic review [22] and meta-analytic design [6]. However, no meta-analytic study has investigated predictors and clinical correlates of EDs in CHR-P youth, which could advance knowledge of the onset of EDs in CHR-P patients and suggest transdiagnostic preventive strategies. Moreover, the literature on clinical specificities of youth with both CHR-P and EDs/ED symptoms has never been appraised with a systematic review design.

The primary aim of this systematic review and meta-analysis is (a) to identify which clinical, socio-demographic, and study factors are predictive of ED diagnosis in CHR-P individuals. The secondary aim is to (b) characterize clinical presentations and outcomes in individuals with both eating pathology (i.e., a diagnosis of ED and ED-related symptoms) and CHR-P.

Methods

Search strategy and inclusion criteria

We conducted a Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020-compliant [23] systematic review and meta-analysis, searching PubMed, EBSCO/PsycINFO, and Web of Science for articles written in English and published between 1st January 2018 (i.e., the year of publication of the systematic review conducted by Albert et al. [22], which provided information on EDs in CHR-P subjects at the level of the single study) and 30th May 2023. The protocol of this study was registered on PROSPERO (CRD42023488792). Supplementary Materials S1 illustrates the PRISMA 2020 checklist. Supplementary Materials S2 show the search strategy, which was aimed to maximize the number of studies that included samples of CHR-P individuals. Two independent authors (GLB and MM) assessed articles at both title/abstract and full-text levels.

Included studies were: (a) original primary research studies written in English and published in peer-review journals, (b) evaluating CHR-P state with validated and reliable measures, (c) evaluating EDs and/or ED symptoms with validated and reliable measures/diagnostic criteria, (d) including individuals with both CHR-P and EDs/ED symptoms.

Excluded were: (a) grey literature (i.e., proceedings, books, and theses), (b) studies assessing EDs/ED-related symptoms in CHR-P patients only at follow-up, and (c) articles written in languages other-than-English. For each study aim, we retained the study with the highest number of participants when two or more reports displayed overlapping samples.

Established measures to assess CHR-P state were: Structured Interview for Prodromal/Psychosis-Risk Syndromes (SIPS) [24]; Comprehensive Assessment of At Risk Mental States (CAARMS) [25]; Schizophrenia Proneness Instrument, Adult Version (SPI-A) [26]; Schizophrenia Proneness Instrument, Child and Youth version (SPI-CY) [27]; Early Recognition Inventory based on IRAOS (ERIraos) [28]; Basel Screening Instrument für Psychosen (BSIP) [29]; Bonn Scale for the Assessment of Basic Symptoms (BSABS) [30].

Established measures/clinical criteria to assess EDs/ED symptoms included: the International Classification of Diseases (ICD) [31, 32], Diagnostic and Statistical Manual of Mental Disorders (DSM) [33,34,35,36], Mini International Neuropsychiatric Interview (MINI) [37], Structured Clinical Interview for DSM (SCID) [38, 39], Eating Disorder Inventory-3 (EDI-3) [40], Eating Attitudes Test [41], Eating Disorder Examination Questionnaire (EDE-Q) [42], SCOFF [43]. We contacted the study authors up to two times in case of potentially overlapping samples for each study aim or missing information.

Outcomes, data extraction, and quality assessment

The primary outcome was identifying the clinical, socio-demographic, and study factors with a significant effect on EDs in CHR-P subjects (aim a). The secondary outcome was to characterize clinical presentations and outcomes in individuals with both eating pathology (i.e., ED and ED-related symptoms) and CHR-P.

Two coders (GLB and ER) extracted data from the included articles. The following data were extracted from studies meeting our inclusion criteria: Digital Object Identifier (DOI); authors; year; country; study design; % males; mean age of the CHR-P sample; research/clinical center; measures to assess CHR-P/APS/EDs/ED symptoms; number of CHR-P individuals with comorbid ED; number of CHR-P individuals; effect sizes, including r, Standardized Mean Difference (SMD), rate ratio, along with their values and SD; mean intelligence quotient score in CHR-P individuals; mean functioning total score in CHR-P patients; depressive/anxiety symptoms/APS mean total score in CHR-P patients; childhood sexual abuse; appearance-related teasing victimization; presence/absence of control groups; self-disturbances; % CHR-P patients with any given ethnicity; % antipsychotics/antidepressants delivered at baseline to CHR-P patients; ED symptoms total score. We considered and coded as EDs each clinical diagnosis listed in the DSM/ICD among the Feeding and Eating Disorders/Feeding or Eating Disorders.

The quality of the studies included in this systematic review and meta-analysis was assessed by two independent judges (GLB and MM) using a modified version of the Newcastle–Ottawa scale (NOS). This version of the NOS was selected as it performs well in assessing both observational and randomized evidence and it has been widely adopted in meta-analytic studies evaluating comorbid mental health conditions in CHR-P individuals [6, 44]. This version of the NOS provides a maximum total score of 8, in which a higher score indicates a higher study quality.

In case of disagreements during the article screening, data extraction, or quality assessment phases, a structured resolution process was implemented. Initially, the two coders (GLB and MM for screening and quality assessment, GLB and ER for data extraction) reviewed the conflicting points independently and attempted to reach a consensus through discussion. If consensus could not be reached, the issue was reported to a third reviewer AT, who provided an independent evaluation. Final decisions were based on majority agreement or, when necessary, further consultation with additional team members (LM or TB).

Outcomes metrics, statistical analyses, and presentation of findings

For the primary outcome (a), predictors were estimated on the prevalence rate of EDs in CHR-P patients. As secondary outcome metrics, we initially considered the correlation (r) and CIs, standardized mean difference (SMD) or Odd Ratio and CIs. However, we did not perform meta-analytic estimates for the secondary outcome due to the insufficient number of studies (see results).

We conducted random-effect meta-analytic estimates for primary outcome (a). Before examining the impact of candidate predictors, we estimated the prevalence rate of EDs and CIs in CHR-P individuals. To calculate the pooled effect size, the metaprop() function [45] from the metapackage in R [46] was used, applying a logit transformation to proportions and the Generalized linear mixed-effects model (GLMM) through the metafor package [47]. I2 statistic [48] was computed to assess the heterogeneity among studies. We performed an influence analysis to assess the impact of individual studies on the overall effect size and heterogeneity. Identified outliers were removed, and the analysis was repeated to observe changes in heterogeneity. Publication bias was evaluated with Egger's test [49] and visual inspection of the funnel plot. We performed sensitivity analysis by excluding outliers and re-estimating the effect size and CIs.

A multi meta-regression model [45], testing the effect of each variable on the outcome parameter by controlling for all the other variables, was run to identify the effect of candidate predictors on the prevalence rate of EDs in individuals at CHR-P. Given the limited number of studies from certain countries, we grouped them to strengthen the statistical power into “America” (not limited to the United States) versus “other countries”, mirroring two distinct research networks [50]. Data analysis was conducted with R [51]. Regarding secondary aim (b), results are presented as a narrative synthesis.

Results

Study selection

The search returned 6752 reports. After removing duplicates, 3774 articles were retained and assessed at the title–abstract level (Fig. 1). Then, 1225 articles were examined at the full-text level, 1199 of which were excluded. Supplementary Materials S4 illustrates the reasons for exclusion. We included 26 articles in this systematic review and meta-analysis [52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77], which reported on a total of 2451 subjects, 2060 of which were included in the analysis, for study aim (a) and 589 subjects for secondary aim (b). Notably, 30 participants were included for both aim (a) and secondary aim (b). Table 1 illustrates the main characteristics of the studies. 17 reports were cross-sectional, 6 were longitudinal cohorts, and 3 were RCTs. 13 reports employed the SIPS to assess CHR-P, 9 employed the CAARMS, and 4 employed two or more tools. 19 reports employed diagnostic interviews to assess EDs and 6 employed clinical criteria. Finally, one report employed a self-report measure, and another report used a clinician-rated tool and a self-report measure to assess ED symptoms. The tools used are described in the Supplementary Materials S5. The quality of the studies ranged from 4.5 to 8, with a mean score of 6.38, indicating a good quality of the included reports.

Fig. 1
figure 1

PRISMA 2020 flow diagram

Full size image
Table 1 Characteristics of the included studies
Full size table

Predictors of EDs in CHR-P individuals: primary aim (a)

Due to the limited data in the retrieved reports regarding the predictors planned in the protocol (i.e., k < 10), we could not evaluate the impact of a range of factors with clinical value, ED symptoms or antipsychotic prescription. Thus, we considered only the following socio-demographic and study factors as candidate predictors: year of publication, country (i.e., America and other countries), study design (i.e., cross-sectional, longitudinal cohort, and RCT), measures to assess ED (i.e., diagnostic interview and clinical criteria), measure to assess CHR-P (i.e., CAARMS, SIPS and MIXED [i.e., two or more measures were used]), age, % males, and NOS score. Two studies [54, 55] were excluded from the analysis, due to missing data. We retrieved information about these factors for the remaining 22 studies.

Before assessing the impact of candidate predictors, we computed the meta-analytic prevalence of EDs in CHR-P individuals, yielding a proportion of 0.05 (95% C.l. 0.03–0.08, k = 22, n = 2060) (Fig. 2). The heterogeneity was high (84.7%). Influence analysis identified four outliers; after removing them, the pooled effect size was 0.03 (95% C.l. 0.02–0.04, k = 18, n = 1844), and heterogeneity was significantly reduced (I2 = 27.1%) (Supplementary Materials S6). The Egger’s test revealed the presence of publication bias (t = − 3.59, p < 0.01), as also shown by the asymmetry in the funnel plot (Supplementary Materials S7).

Fig. 2
figure 2

Meta-analytic prevalence of eating disorders in individuals at clinical high risk for psychosis

Full size image

The multi meta-regression model showed no significant effect of any predictors on the proportion of EDs in CHR-P individuals (R2 = 0.00%, p = 0.79) (Table 2).

Table 2 Multi meta-regression of candidate predictors of eating disorders in individuals at clinical high risk for psychosis
Full size table

Clinical characteristics of individuals with both CHR-P and EDs: secondary aim (b)

The included studies were too few and heterogeneous in populations and outcomes to perform meta-analytic comparisons for aim b. In Mensi et al. [69], after correcting for multiple comparisons, EDs severity in individuals with ED and with or without CHR-P was negatively associated with Body Mass Index (BMI), but was not significantly associated with global social functioning. Moreover, the authors found that individuals with both EDs and CHR-P exhibited more purging behavior, dysmorphophobia, ED symptoms, and higher scores in the composite subscales of EDI-3 [40] of Ineffectiveness Composite, Interpersonal Problems Composite, and Affective Problems Composite compared to individuals with EDs but without CHR-P. Moreover, individuals with both EDs and CHR-P received a higher amount of antipsychotic medication compared to individuals with only EDs.

In Hazan et al. [62], a large transdiagnostic sample recruited in enhanced primary care services with attenuated psychotic symptoms (APS; i.e., one of the CHR-P subgroups) exhibited more ED symptoms compared to clinical controls without APS at both baseline and 1-year follow-up. Moreover, CHR-P was a transdiagnostic marker of increased symptom severity and reduced response to treatment, denoting a high need for care. Nevertheless, despite the lower general functioning, individuals with APS showed a greater improvement in general functioning compared to the non-APS group. Notably, this was the only study in the current meta-analytic review focusing on ED-related symptoms, without a full-blown ED diagnosis.

In Sarac et al. [73], CHR-P individuals with EDs showed higher processing speed and greater white matter and brain volume compared to CHR-P individuals without EDs. The difference in speed processing was not maintained when adjusting for sex and socioeconomic status. The results were confirmed when controlling for white matter and brain volume. The groups did not show significant differences regarding positive and disorganized symptoms. For more details on the studies included for aim b, see Supplementary Materials S8.

Discussion

The current systematic review and meta-analysis aimed to identify potential predictors of EDs in CHR-P individuals and to characterize the clinical presentations and outcomes in individuals with both eating pathology (i.e., EDs and ED-related symptoms) and CHR-P. Two main findings emerged from the study. First, none of the tested variables predicted EDs in CHR-P youth (study aim a). Second, our findings provide preliminary insights into the clinical characteristics of individuals with both CHR-P and eating pathology.

5% of CHR-P individuals presented with EDs; however, no significant predictor of EDs was identified in individuals at CHR-P. This result partially aligns with a recent umbrella review focusing on individuals with EDs, which found no convincing evidence supporting the relationship between any risk factors and EDs and highly suggestive evidence only supporting the relationship between childhood sexual abuse and bulimia nervosa and between appearance-related teasing victimization and any ED [78]. This study also indicated that the ED field lags behind in terms of the evidence needed to implement preventive strategies for individuals with sub-threshold ED symptoms, recommending multi-centric cohort studies to identify modifiable risk factors. Of note, a potential example of such a project has been recently implemented in the CHR-P field, with an ongoing large-scale observational cohort study collecting and analyzing multimodal data to improve prognostic precision [79]. If translated to the ED field, an analogous project collecting multimodal data like clinical, environmental, cognitive, and neuroimaging data could increase precision in determining risk factors and prognosis and selecting appropriate preventive treatments for eating pathology. It should be noted that EDs and psychotic disorders may co-occur and the risk factors for both mental disorders should be considered, as suggested in a previous study focusing on multiple at-risk conditions [80]. Accordingly, research on prognostic outcomes in individuals with CHR-P and sub-threshold EDs could also benefit from transdiagnostic approaches, which focus on a “pluripotent” risk state for several mental disorders, encompassing trajectories of “homotypic progression” (e.g., a person with CHR-P developing psychosis) and “heterotypic progression” (e.g., a person with CHR-P developing anorexia nervosa) [81]. Although EDs are not among the most prevalent comorbid mental disorders in CHR-P samples, they warrant clinical attention as they may contribute to the complexity of the clinical picture. Growing international consensus highlights the importance of addressing the heterogeneity of the CHR-P population through clinical characterization and targeted interventions for specific subgroups [6, 19, 79]. Different tools were used to assess both CHR-P status and eating pathology across the included studies. Although the Positive SYmptoms and Diagnostic Criteria for the CAARMS Harmonized with the SIPS (PSYCHS) is under development [82], which harmonizes the main tools employed in the CHR-P field, at the time of our review, there was still notable heterogeneity in the assessment tools used.

Only three studies were included in the narrative synthesis, providing a clinical description of individuals with both CHR-P/APS and EDs/ED-related symptoms. One study including adolescents with EDs showed that CHR-P state was associated with increased overall mental health symptom severity and a greater need for antipsychotic treatments [69]. Moreover, individuals with both EDs and CHR-P were more frequently prescribed antipsychotic medications compared to individuals with only EDs, despite clinical guidelines discouraging their provision for at-risk syndromes (e.g., [19]). Hazan et al. [62] enrolled a transdiagnostic clinical sample of individuals seeking help in enhanced primary care services (i.e., headspace [83]), finding that APS was associated with more severe mental health symptoms, at both baseline and 1-year follow-up, and a poorer response to treatment. Timely treatment is also suggested in a case report focusing on an adolescent with both pica and CHR-P [84]. Although this study was excluded from our narrative synthesis due to the lack of standardized measures for EDs in the report, it suggests the importance of early assessment and treatment of both conditions to reduce symptoms and improve psychosocial functioning. Finally, regarding the differences between CHR-P individuals with or without EDs, Sarac et al. [73] showed better processing speed and greater brain volume in CHR-P individuals with EDs, but no greater symptom severity. Overall, our findings offer preliminary insights into the clinical characteristics of individuals with both CHR-P and EDs. However, the limited number of studies—two of which were cross-sectional/case–control and a longitudinal study with only 1-year follow-up—precludes drawing firm conclusions. Future transdiagnostic [80, 81, 85], prospective cohort studies are needed to explore long-term outcomes in individuals with both CHR-P and EDs, beyond diagnostic silos [20, 86].

Although the clinical population of patients with EDs is mainly composed of individuals assigned female at birth (i.e., 72% of cases; [15]), the current meta-analytic study found that the proportion of males had no effect on the prevalence of EDs in CHR-P individuals. Yet, further research is needed to explore the co-occurrence of EDs and CHR-P. In this regard, previous studies [69] suggested that both conditions may share a common “psychotic core”, characterized by a fragmented sense of subjectivity, often referred to as “anomalous self-experiences” or “self-disorders” [83]. In this vein, body-centered manifestations (such as interoceptive deficits), and bodily-directed symptoms (including body image concerns and the search for control or connection with a disembodied bodily dimension) experienced by EDs and CHR-P individuals, may represent expressions of this shared core issue. This aligns with Hilde Bruch’s [87] theoretical perspective which defined EDs as disorders of the self, characterized by a mind–body split and what has been termed as a disembodied self that concerns the feeling of dissociation from the given body, which feels "unreal" and remains unintegrated into a coherent experience of the self [88]. Future research comparing self-disorders in clinical groups with CHR-P (without EDs) and EDs (without CHR-P) is essential to clarify the extent of this potential phenomenological overlap.

Strengths and limits

One strength of our study lies in the use of three non-overlapping search engines (PubMed, EBSCO/PsycINFO, and Web of Science), which enhanced our ability to retrieve a broader range of studies. This allowed us to include a relatively large number of studies—especially considering that we focused on mental health conditions (i.e., EDs) that have been neglected in the CHR-P field. Second, we performed both quantitative (i.e., meta-analysis and multiple meta-regression) and qualitative (i.e., narrative) syntheses to provide a comprehensive clinical description of individuals with both CHR-P and EDs/ED symptoms.

This study has also several limitations that should be considered. First, the number of studies included for aim b was minimal (k = 3). Thus, our results, especially the narrative synthesis, should be considered exploratory rather than conclusive. Nevertheless, we were able to identify research gaps in the field, proposing areas for future research. Second, we could not examine the impact of all predictors planned in the a priori protocol, as most of the included studies reported minimal information about our variables of interest. For example, variables such as eating symptoms and antipsychotic prescriptions were excluded, as there were too few studies consistently reporting on them. As a consequence, we were unable to assess the potential impact of these variables on the prevalence of EDs in CHR-P individuals. Third, our meta-analysis of the proportion of EDs in CHR-P individuals showed high heterogeneity, which was not explained by the multiple meta-regression model. This should not be strictly viewed as a study limitation, as high heterogeneity is common in this field [6]. Moreover, the tested predictors included almost exclusively socio-demographic variables. Thus, unexplained heterogeneity should be expected. Fourth, this study did not provide evidence about the predictors of specific EDs, like anorexia nervosa or bulimia nervosa, as we did not retrieve enough studies to run meta-regression analyses focusing on these conditions. Moreover, the studies included in the narrative synthesis did not provide evidence about the clinical specificities of any specific ED. Fifth, we excluded studies due to missing data, which can lead to selection bias and reduce the generalizability of the results, potentially resulting in the loss of contextual diversity from underrepresented populations or settings [89]. Sixth, potential publication bias was indicated by Egger's test and funnel plot asymmetry. However, influence analysis showed that outliers did not significantly affect the results. In addition, there is no clear rationale to expect a systematic bias toward over- or underreporting EDs in CHR-P populations, suggesting the asymmetry is more likely due to statistical variability. To further reduce potential bias, future studies should incorporate grey literature, such as theses and conference proceedings, for more comprehensive analyses. Finally, we did not include articles on psychotic-like experiences, as we focused exclusively on CHR-P/APS, as operationalized by narrowly defined clinical criteria.

Conclusion

This is the first systematic review and meta-analysis focusing on the clinical specificities of individuals with both CHR-P and EDs. The results highlight the need for future research to clarify the relationship between these conditions and underscore the importance of prospective, multi-centric cohort studies to explore potential predictors and clinical outcomes. Cooperation between multidisciplinary teams specialized in EDs and CHR-P may be crucial for ensuring appropriate referrals, continuity of care, and tailored intervention strategies. Future studies with a transdiagnostic approach are warranted to confirm our findings, which should be considered preliminary, rather than conclusive.

What is already known on this subject?

The co-existence of psychotic conditions and eating disorders (EDs) or ED-related symptoms has been documented in several studies. In contrast, research on the co-occurrence of clinical high risk for psychosis (CHR-P) and EDs is far less investigated. Research on predictors and clinical correlates of EDs in CHR-P individuals could advance knowledge of the onset of EDs in CHR-P patients and suggest transdiagnostic preventive strategies.

What this study adds?

No factor was found to significantly impact the proportion of EDs in individuals at CHR-P. This study characterized individuals with both CHR-P and ED/ED-related symptoms; however, the limited number of included studies is insufficient to support definitive conclusions. Transdiagnostic, prospective, cohort studies are needed to explore long-term outcomes in individuals with both CHR-P and EDs, beyond diagnostic silos.