Introduction

Non-suicidal self-injury (NSSI) refers to the deliberate and direct damage to one’s body tissue without suicidal intent [1]. Examples of NSSI include behaviors such as cutting, scratching, biting, or burning oneself. Adolescence and emerging adulthood are critical periods for NSSI, with approximately 20% of individuals reporting an onset before the age of 25 [2]. Reported gender differences in the prevalence of NSSI have been inconsistent; however, a recent meta-analysis found higher rates of NSSI in females compared with those in males, especially in clinical samples [3]. DSM-5 recently included the Non-Suicidal Self-Injury Disorder (NSSI-D) as a condition requiring further study [4]. NSSI-D requires NSSI on at least 5 days in the past year, significant distress due to NSSI, as well as several other conditions (e.g., psychological precipitant, intense urges, and the expectation of a favorable outcome following NSSI). Approximately 7% of adolescents and 1% of emerging adults meet diagnostic criteria for 12-month NSSI-D [5, 6].

While the definition of NSSI excludes eating disordered (ED) behaviors (i.e., binge-eating, purging, and restrictive eating), there is a strong relationship between NSSI and ED behaviors and syndromes [7]. Anorexia nervosa (AN) is an eating disorder characterized by a restriction of food intake that results in underweight, coupled with an irrational fear of gaining weight, and a disturbed body image [8]. Depending on whether AN patients engage in compensatory behaviors or not, a binge-eating/purging (AN-BP) or restrictive specifier (AN-R) is given. Bulimia nervosa (BN) occurs at a healthy or elevated weight and is characterized by binge eating, followed by compensatory behaviors to avoid weight gain [8]. In addition, DSM-5 added the binge eating disorder (BED) as a diagnostic entity that is characterized by distressing, recurrent episodes of binge eating in the absence of compensatory behaviors [4]. As is the case for NSSI, ED behaviors and disorders often start in adolescence and emerging adulthood [9, 10], with the weighted mean lifetime prevalence of any eating disorder being estimated around 8% for females (disorder-specific pooled estimates AN: 1.4%, BN: 1.9%, BED: 2.8%) and 2% for males (disorder-specific pooled estimates AN: 0.2%, BN: 0.6%, BED: 1.0%) [11].

Eating disorders are highly disabling, deadly, and costly psychiatric disorders [8]. Addressing why so many people diagnosed with ED behaviors self-injure was recently put forward as an important research topic by patients, caretakers, and clinicians [12]. To this end, the present report aims to summarize recent advances in the literature regarding (1) how often NSSI and ED behaviors co-occur and the severity of these behaviors when they co-occur, (2) the relationship between these behaviors over time, (3) reasons for why they might co-develop, (4) the potential psychosocial consequences of their co-occurrence, and (5) the implications for evidence-based practice. In doing so, we also identified remaining gaps of knowledge and pointed out several critical open questions for future research.

How Often Do NSSI and ED Behaviors Co-Occur?

In this section, we review studies that investigated the co-occurrence between NSSI and ED behaviors in individuals with a history of ED behaviors, NSSI, or both.

NSSI in Individuals Who Engage in ED Behaviors

In a recent meta-analysis, Cucchi and colleagues found that 27.3% of individuals diagnosed with an eating disorder report comorbid NSSI, with higher lifetime estimates of NSSI for patients diagnosed with BN (32.7%) compared with AN (21.8%) [13••]. For BED patients, lifetime prevalence rates of NSSI have been observed close to 20% [14]. There is compelling evidence that NSSI is more strongly associated with eating disorders of the binge-eating/purging subtype than the restrictive subtype [15]. Dzombak and colleagues found that more than half of adolescent inpatients with binge-eating/purging behaviors report NSSI in the month before treatment compared with 17% of patients who only engage in restrictive eating behaviors [16]. Other researchers also observed higher lifetime and past-month prevalence rates (median: 35% and 20%) of different NSSI behaviors among inpatient adolescents who engage in binge-eating/purging behaviors than those who report only restrictive eating (median: 17% and 7%) [17]. Similar findings have been observed in (emerging) adults with eating disorders [18, 19]. Among college students, however, restrictive eating has been associated with higher odds of NSSI beyond other ED behaviors [20]. NSSI should thus not be overlooked in individuals displaying only restrictive eating patterns as their susceptibility for NSSI may remain elevated compared with peers who do not report ED behaviors. Besides ED subtype, the treatment setting explains considerable variability in the presence of NSSI; more specifically within specialist/inpatient care facilities, estimates of NSSI are more than twice as high as non-specialist/outpatient settings [13••].

ED Behaviors in Individuals Who Engage in NSSI

Conversely, high rates of ED behaviors have been observed among individuals with a history of NSSI [21, 22, 23••, 24, 25]. In one study, Yiu and colleagues found that nearly four out of five individuals who engage in NSSI report at least one ED behavior within the past week, with binge-eating (57.6%) being the most prevalent and diuretic/laxative misuse (15.3%) being the least prevalent behavior [26]. In another study, 42% of young adults with a history of NSSI reported ED behaviors during a 2-week observation period [27••]. While rates of (sub)threshold eating disorders are estimated in the 8–25% range among individuals with a history of NSSI [23••, 24], future work is needed to provide a clearer picture of the prevalence of eating disorders among individuals with NSSI(−D).

Characteristics of NSSI/ED Severity in Individuals Who Engage in NSSI and ED Behaviors

Several researchers have evaluated the severity of NSSI and ED behaviors in individuals who engage in both behaviors. In a cross-sectional study, Sorgi and colleagues observed that the presence of binge-eating/purging behaviors among college students with a history of NSSI was associated with a greater variety of NSSI methods [28], which is, in turn, a reliable predictor of a more severe and chronic NSSI course [29,30,31]. Similar findings have been observed in clinical samples [32]. Conversely, a higher frequency of NSSI, a greater variety of NSSI methods, and a medical treatment history of NSSI have all been related to greater ED severity [23••, 26]. While these findings provide useful information about how the co-occurrence of NSSI/ED relates to the severity of NSSI and ED behaviors, this does not tell us how these behaviors, nor their characteristics, are related over time.

Do NSSI and ED Behaviors Influence Each Other’s Course?

The high co-occurrence between NSSI and ED raises questions about the directional nature of this comorbid relationship (i.e., NSSI increases risk of ED behaviors, ED behaviors increase risk of NSSI, or both behaviors reciprocally increase risk of each other). Yet, only a handful of longitudinal studies have investigated the temporal relationship between NSSI and ED behaviors. We first review what we learned from these studies and then highlight directions for future research.

The Temporal Relationship Between NSSI and Future ED Behaviors

In a cohort study, Wilkinson and colleagues found that the presence of repetitive NSSI, defined as two or more acts of NSSI in the past year, at age 14 predicted the onset of eating disorders by age 17 [33••]. No significant prospective association was found between a single act of NSSI and future eating disorders. In another prospective study, Riley and colleagues observed that the presence of lifetime NSSI at college entrance increased the risk of purging behaviors during the first year of college [34•]. Similar findings have previously been reported among female college students [35]. However, given that purging behaviors in both studies were assessed only in the past month, future work is needed to clarify whether NSSI is a risk factor for the onset and/or relapse of purging behaviors. Among individuals with a history of NSSI, Turner and colleagues investigated the temporal relationship with ED behaviors up to five times across 1 year, finding that an increase in NSSI frequency predicted greater ED severity 3 months later [23••].

The Temporal Relationship Between ED Behaviors and Future NSSI

In the study of Riley and colleagues [34•], college students with past-month purging behaviors were also at increased risk for an onset of NSSI during the first year of college. Similar findings have recently been reported for ED behaviors across separate cohort studies among adolescents and emerging adults [36, 37]. However, some scholars also found ED behaviors to be unrelated to future NSSI among adult women [38], although it should be noted that the prospective analyses of this study were plagued by a 60% dropout (n = 109). Finally, Turner and colleagues [33••] found that increases in ED symptoms also predicted more frequent NSSI at 3-month follow-up assessments. However, this relationship was moderated by emotion dysregulation, such that disordered eating was a risk factor of future NSSI among individuals scoring high on emotion dysregulation. Taken together, emerging evidence suggests that the NSSI-ED relationship may be bidirectional for some individuals.

What We Still Need to Learn About the Temporal NSSI-ED Relationship

First, it is surprising that we still know little about the developmental order in which NSSI and ED behaviors typically manifest [39]. In a cross-sectional study of patients with an eating disorder, the vast majority of those reporting comorbidities with NSSI indicated that they started to engage in NSSI after the onset of AN/BN, yet NSSI occurred for one in five before beginning ED behaviors [19]. Although this suggests that NSSI may often develop after the onset of eating disorders, longitudinal work is needed to determine the prevalence of different developmental patterns of NSSI and ED behaviors/disorders. Building upon these findings, it would be meaningful to clarify how long transitions within these patterns normatively take (weeks, months, or years) and what factors predict particular developmental patterns (i.e., that people develop one behavior before the other). Second, more work is needed teasing apart whether NSSI and ED behaviors also uniquely increase risk (i.e., onset, relapse, maintenance) of each other throughout adolescence and (emerging) adulthood. Third, apart from the developmental course, there is a lack of within-person studies that considered the temporal relationship between NSSI and ED behaviors in daily life [27••]. Providing greater clarity in each of these areas of investigation would not only advance our understanding of the temporal relationship between both behaviors but also provide meaningful information to help guide the deployment of targeted preventive interventions.

Why Do NSSI and ED Behaviors Co-Occur?

Over the past 5 years, relevant studies along two different research lines have been published that aid our understanding of why ED and NSSI co-occur so often within people. The first line of research focused on whether similar functions may underlie NSSI and ED behaviors, whereas the second line identified shared risk factors that may account for the high comorbidity between these behaviors.

Shared Functions and Motives

Theoretical models of NSSI and ED suggest that the functions of both behaviors may overlap [40]. The recent study of Muehlenkamp, Takakuni, and colleagues is important in this context [41•], as these authors compared a broad range of behavioral functions among people reporting only NSSI or only ED behavior. Findings revealed that similar intrapersonal (e.g., affect regulation, marking distress) and interpersonal functions (e.g., interpersonal boundaries, autonomy) may motivate NSSI and ED behaviors. However, functions of the intrapersonal domain were found to be more salient in motivating NSSI than ED behaviors, whereas the opposite was observed for functions of the interpersonal domain. Providing further support to the notion of functional equivalence, Turner and colleagues found that when asked about the reasons underlying NSSI or ED behavior in daily life, the majority of these instances (54.2–81.6% across behaviors) were reported to alleviate aversive feelings or thoughts [27•].

Beyond retrospective self-report, there is emerging evidence from ecological momentary assessment (EMA) that NSSI and ED behaviors are used to regulate both intrapersonal (i.e., mostly heightened negative affect) and interpersonal (e.g., perceived conflict and negative social appraisals) difficulties in daily life [42, 43]. Future prospective studies should clarify whether different functional profiles may explain why some individuals do, and others do not develop and maintain both harmful behaviors. Similarly, future work is needed to clarify the reinforcement processes and temporal unfolding of different functions in real-time. It may, for instance, be that interpersonal stressors lead to enhanced negative affect, which then more proximally increases the risk of NSSI and ED behaviors. Addressing both why and how different functional profiles relate to comorbid NSSI and ED engagement would provide important clues for the prevention and intervention of these behaviors.

Finally, a recent study of Fox and colleagues into underlying self-harming desires across NSSI and ED behaviors similarly revealed quantitative rather than qualitative differences [44••]. Specifically, individuals reported engaging in both types of behaviors with the intent of hurting oneself physically in the moment, but this desire appeared more central to NSSI than ED behaviors. Conversely, the desire to cause physical harm, in the long run, was also present across both behaviors, but this was most relevant in motivating restrictive eating behavior. In sum, these studies indicate that NSSI and ED behaviors can serve similar functions and motives to some degree, providing as such a first compelling reason for the high co-occurrence of these behaviors.

Shared Etiology of NSSI and ED Behaviors

The conceptual model of Claes and Muehlenkamp (Fig. 1), which was introduced in 2014 [7], postulates that NSSI and ED behaviors are multi-determined by a range of similar individual, social, and cultural risk factors that interact with each other within and across distal and proximal dimensions. Predisposing the development of NSSI and ED behaviors, the proposed model highlights shared distal factors on the individual (i.e., temperament and personality) and social level (i.e., family environment, traumatic experiences, and cultural pressures). Longitudinal studies confirm that dispositional factors related to emotionality (e.g., negative affectivity [45, 46]), disinhibition (e.g., negative urgency [47, 48]), and cognitive liabilities (e.g., trait rumination [49, 50], self-critical perfectionism [51]) can predispose the development of NSSI and ED behaviors. Cross-sectional studies also linked both behaviors to higher levels of subjective emotional reactivity [52•], anxiety sensitivity [53], and avoidance [54] as well as to lower levels of effortful control [55].

Fig. 1
figure 1

Conceptual NSSI-ED Risk Factor Model of Claes and Muehlenkamp (2014). Copyright © 2014 by Springer-Verlag. Adapted with permission from Non-suicidal Self-Injury and Eating Disorders: Dimensions of Self-Harm. In: Claes L., Muehlenkamp J. (Eds.) Non-Suicidal Self-Injury in Eating Disorders

Full size image

On the social level, longitudinal studies have implicated childhood-adolescent traumatic experiences [56, 57], dysfunctional attachment or parenting style [58,59,60,61], and peer victimization [36, 61, 62] in the development of NSSI and ED behaviors. Researchers also found associations between difficulties in parent-child relationships [63, 64], higher perceived sociocultural pressures (e.g., thin-ideal), and body-objectification and the presence of NSSI and ED behaviors [46, 65]. Taken together, the evidence to date suggests that individuals prone to aversive emotions and self-negative thoughts—as a result of distal intrapersonal and/or interpersonal vulnerabilities—while having lower control and a tendency to act impulsively when distressed are developmentally susceptible for NSSI and ED behaviors.

The postulated model also highlights a range of shared risk factors that more proximally trigger the initial development and contribute to the maintenance of NSSI and ED behaviors [7]. Providing support to this notion, poor emotion regulation has been implicated in the development and maintenance of NSSI and ED behaviors [29, 66,67,68]. Additionally, studies have shown an indirect path from childhood-adolescent traumatic experiences via emotion-dysregulation and dissociation to NSSI and ED behaviors [69, 70]. Increased self-criticism and decreased self-esteem have also been associated with NSSI and ED [71], and there is emerging evidence from longitudinal studies that heightened self-criticism increases risk of ED behaviors [72•] and predicts NSSI among ED patients [73•]. In a similar vein, while it is known that negative body attitudes precipitate and perpetuate ED disorders [46, 74], body dissatisfaction was recently found to predict NSSI among women beyond ED comorbidity [38]. Scholars also observed selection and socialization effects [75,76,77,78], such that young individuals whose friends contemplate or engage in NSSI and ED behaviors may be more likely to develop these behaviors themselves. Finally, research indicates that NSSI and ED behaviors are associated with a range of psychiatric symptoms and disorders [6, 79], with internalizing symptoms precipitating the onset of both behaviors [36, 49, 80]. While borderline personality disorder (BPD) has been associated with greater psychiatric comorbidity [81], studies recently linked problems with identity formation to the onset of both behaviors [82•].

Taking stock of the available research supports the model of Claes and Muehlenkamp [7] that lays out shared individual, social, and cultural risk factors as a framework (and second compelling reason) for understanding the high co-occurrence of NSSI and ED behaviors. Future prospective studies are now needed that consider NSSI and ED behaviors within the same cohorts to confirm and expand upon this model. Such studies would be ideally suited also to clarify unique factors that differentiate the development, maintenance, and course of both behaviors [83•].

What Are the Psychosocial Consequences of Comorbid NSSI and ED Behaviors?

While the section above illustrates that we have acquired substantial knowledge about risk factors underlying NSSI and ED behaviors, several researchers have also investigated the potential developmental consequences of engaging in these harmful behaviors. In what follows, we provide an update on this literature and point to open questions for future research.

NSSI and ED Behaviors: Double Trouble When It Comes to Suicide Risk?

Researchers have observed elevated rates of suicidal thoughts and behaviors among individuals who engage in NSSI and ED behaviors (especially binge-eating/purging subtypes) [6, 84], and recent meta-analyses found that both behaviors prospectively increase the risk of a future suicide attempt [85, 86]. Current theories suggest that people who engage in multiple self-damaging behaviors are at increased risk of suicide compared to those who engage in a single behavior [87], as repeated tissue damage and exposure to painful experiences might contribute to an enhanced capability for suicide. In a recent study, Brausch and Perkins observed that college students with a history of NSSI and ED behaviors (25%) were significantly more likely to have attempted suicide than those with only NSSI or ED behaviors (2–10% range) [88••]. Importantly, while the comorbid group did not report higher fearlessness of death, they did report lower fear about suicide than individuals in the single behavior groups. Individuals in the comorbid group also reported a higher suicide severity than the NSSI only group, while the ED only group reported the greatest suicide-related concerns.

In the study of Fox and colleagues mentioned above [44••], NSSI was associated with higher levels of suicidal thoughts than any ED behavior. Within ED subtypes, restrictive eating was most strongly associated with suicide- and death-related thoughts. In line with these findings, a recent study also found that restrictive eating (but not binge-eating or purging) was uniquely associated with suicide ideation among young individuals with eating disorders [89]. However, given the lack of direct comparisons over time, longitudinal studies are needed to clarify whether comorbid NSSI and ED behaviors, or particular behavioral patterns (e.g., NSSI and restrictive eating), result in a higher combined risk of future suicidality than either NSSI or ED behaviors alone. If so, the next question then becomes how the interplay between NSSI and ED behaviors in terms of future suicide risk can mathematically be best described (i.e., sub-additive, additive, or synergistic effects) and theoretically explained (i.e., what psychological mechanisms account for the higher susceptibility).

A Broader Developmental Perspective

Apart from the physical risk and societal burden associated with these behaviors, there is growing evidence that NSSI and ED behaviors may also negatively impact the psychosocial development of young people. Researchers have linked NSSI and/or ED behaviors with future risk of mental health problems (especially depression) [33••, 90], impaired family functioning [91, 92], decreased emotion regulation capabilities [67], identity issues [93, 94], lower self-esteem and quality of life [95, 96], stigma [97, 98], and academic failure [99, 100]. Hence, much could be learned from future investigations that adopt a broader developmental framework in which risk factors, NSSI and ED behaviors/disorders, and developmental tasks (e.g., identity formation, graduating, and finding work) might influence each other reciprocally throughout adolescence and (emerging) adulthood. Addressing this gap of knowledge could elucidate developmental cascades, and, again, make clear whether individuals who report comorbid engagement are more likely to experience adverse longer-term outcomes than peers who report either behavior alone. Providing greater clarity to these questions would aid scientific understanding and provide valuable information for prevention efforts and clinical risk assessment.

Evidence-Based Treatment of NSSI and ED Disorders

Several authors have examined best practice guidelines and interventions for the treatment of NSSI and ED disorders, which are first discussed. We then consider the added complexity of NSSI among eating disorder patients and reflect on the potential of emerging technological advances for the treatment and prevention of NSSI and ED behaviors.

Currently Best Practice Guidelines

For both behaviors, guidelines recommend outpatient services as the first line of treatment when there is no immediate risk to the individual. Concerning psychotherapies, cognitive-behavioral therapy-enhanced (CBT-E) is the first choice of treatment of all eating disorders [101,102,103]. As an exception to this rule, family-based therapy (FBT) should be considered in the treatment of AN in young individuals [103]. Some guidelines have also recommended FBT for the treatment of BN in younger patients [101]. For the treatment of BN and BED, interpersonal psychotherapy (IPT) has been recommended as an alternative to CBT-E [104]. While there is a paucity of randomized control trials (RCTs) investigating treatment efficacy for NSSI, promising effects have been observed for dialectical behavioral therapy (DBT), emotion regulation group therapy (ERGT), manual-assisted cognitive therapy (MACT), and dynamic deconstructive therapy [105, 106]. Concerning pharmacological interventions, second-generation antidepressants (i.e., selective serotonin reuptake inhibitors; SSRIs) can be considered for BN (in combination with psychotherapy) and BED [101, 102], whereas the central nervous stimulant lisdexamfetamine also appears moderately effective in reducing binge-eating [102]. Based on the literature, specific psychopharmacological therapy cannot be recommended for AN and NSSI.

Which Treatment to Use When ED Behaviors and NSSI Are Present Comorbidly?

Given the high comorbidity and observation that eating disorder patients who do—compared with those who do not—engage in NSSI present with a different, yet more complex, clinical profile that may affect the treatment response [14, 52•], it is problematic that RCTs typically do not include data on NSSI and suicidality. In the absence of clear guidelines, Marino and colleagues recently developed a decision-making model to aid clinicians in deciding which treatment (i.e., CBT-E or DBT) should be prioritized for adult clients with a complex eating disorder and comorbid conditions [107••]. The model directly refers to DBT when a client is diagnosed with BPD, functional assessment indicates that NSSI (or other harmful behaviors including ED behaviors) are engaged in for emotion regulation purposes, or when previous treatment attempts failed due to interfering behaviors. While future studies are needed to evaluate this novel decision-making model empirically, the clinical reality underlying this model illustrates the added complexity facing clinicians when treating clients with an ED who also self-injure.

At the same time, even though DBT also demonstrated effectiveness in reducing NSSI among adolescents with traits of BPD [108], not everyone needs and/or wants such specialized and intensive treatment. Limited resources, a lack of specialized therapists, and the finding that many (especially young people) do not find their way to conventional therapy [8, 109] call for an integrated stepped-care model in which the most effective, yet least resource-intensive intervention, is considered first [110].

Novel Technology-Based Interventions: Towards an Integrated Stepped-Care Approach

The recent proliferation of digital technology offers new opportunities for scientist-practitioners to better match treatments to the needs and preferences of individual patients. In 2011, the MACT was adapted for adolescents in the Cutting Down Program (CDP), a low-intensity psychotherapeutic intervention of 8–12 sessions [111]. The online version is currently underway [112], but a recent RCT showed that the face-to-face version was as effective in reducing NSSI and achieved faster recovery compared with an intensive treatment as usual [113•]. Several other accessible digital interventions are under development and have shown promising results that require replication in future RCTs [114]. Of note, a recently developed online low-intensity emotion-regulation individual therapy reduced both NSSI and comorbid harmful behaviors (including binge-eating and purging behaviors) [115•].

While online therapeutic programs have demonstrated effectiveness in reducing ED symptoms [116], researchers in the ED field are now evaluating the potential of EMA to improve clinical care by delivering Just-In-Time Adaptive Interventions (JITAIs) in daily life [117•]. To facilitate the clinical uptake of these promising developments, the RCTs should include data on NSSI and suicidality. Similarly, as emerging evidence demonstrates the feasibility of using technology to incorporate prevention into a stepped-care approach for these behaviors [e.g., 118], risk prediction models are warranted that consider multiple behaviors to develop an integrated approach in responding to NSSI and ED behaviors.

Conclusion

Collectively, the reviewed literature provides solid evidence of a robust cross-sectional association between NSSI and ED behaviors, and indicates a significant step forward in addressing the question of why so many people engage in both behaviors. However, much remains to be learned about the joint course of NSSI and ED behaviors over time. Longitudinal studies are now needed to provide novel data on the longer-term developmental and the shorter-term momentary relationship of these behaviors in daily life. While the former will be crucial in clarifying who is at high risk to develop comorbid engagement and experience its potentially negative consequences (e.g., increased suicide risk), the latter will be needed to determine when in the flow of ordinary life clients are at imminent risk and intervention is most indicated [117•].

Building upon these findings, intervention studies should target NSSI and ED behaviors simultaneously. Facilitating the uptake and making optimal use of the growing capacities of mobile technologies to facilitate the deployment of evidence-based interventions have become even more important than ever before, following recent abrupt changes in the delivery of clinical services due to COVID-19 restrictions [119]. While far from exhaustive, we hope that this review will be useful and helps to guide future work in addressing the comorbid NSSI-ED problem for the many individuals who struggle with these behaviors on a day-to-day basis.