Self-injury and externalizing pathology: a systematic literature review

Background During the last decade there is a growing scientific interest in nonsuicidal self-injury (NSSI). The aim of the current paper was to review systematically the literature with a special focus on the associations between self-injurious behaviours and externalizing psychopathology. An additional aim was to review terminology and measurements of self-injurious behaviour and the connection between self-injurious behaviours and suicide in the included publications. Methods A systematic literature search was conducted on 31st December 2016 in five databases (PubMed, OVID Medline, OVID PsycINFO, Scopus, Web of Science) with two categories of search terms (1. nonsuicidal self-injury, non-suicidal self-injury, NSSI, self-injurious behaviour, SIB, deliberate self-harm, DSH, self-injury; 2. externalizing disorder, attention deficit hyperactivity disorder, ADHD, conduct disorder, CD, oppositional defiant disorder, OD, ODD). Results Finally 35 papers were included. Eleven different terms were found for describing self-injurious behaviours and 20 methods for measuring it. NSSI has the clearest definition. All the examined externalizing psychopathologies had strong associations with self-injurious behaviours according to: higher prevalence rates in externalizing groups than in control groups, higher externalizing scores on the externalizing scales of questionnaires, higher symptom severity in self-injurious groups. Eight studies investigated the relationship between suicide and self-injurious behaviours and found high overlap between the two phenomena and similar risk factors. Conclusions Based on the current findings the association between externalizing psychopathology and self-injurious behaviours has been proven by the scientific literature. Similarly to other reviews on self-injurious behaviours the confusion in terminology and methodology was noticed. NSSI is suggested for use as a distinct term. Further studies should investigate the role of comorbid conditions in NSSI, especially when internalizing and externalizing pathologies are both presented.


Background
In the past decade self-injury became a hot topic in scientific literature. Nonsuicidal self-injury (NSSI) became an individual diagnosis in the 5th edition of the Diagnostic and Statistical Manual of Mental Disorders [1]; for the present it can be found only in Section III in the chapter 'Conditions for Further Study'. The definition for NSSI in DSM-5 follows the main instructions of the definition of the 'International Society for the Study of Self-injury' (ISSS), which was made in 2007 [2]. This definition emphasizes the deliberate, nonsuicidal purpose of the self-injurious act, which is not socially sanctioned. It underlines the importance to distinguish it from drug overdoses, culturally sanctioned behaviours (e.g. piercings), and repetitive, stereotypical forms among people with developmental disorders [2]. The DSM-5 suggests as a criterion of NSSI, that self-injurious acts should happen on at least 5 days in the past year. Moreover the DSM-5 underlines the non adaptive 'coping strategy' nature of NSSI: the individual who engages in NSSI must have the aim to get to a better emotional state after the action [1].
In the 4th edition of DSM [3], deliberated NSSI can only be found as a symptom of borderline personality disorder (BPD), however clinicians can meet a lot of patients with self-injury without BPD as well. Glenn and Klonsky [4] found in an adolescent psychiatric sample that the coexistence of NSSI and BPD is not more common than the coexistence of any other Axis I diagnoses and NSSI. Zetterqvist [5] found that in the population who met NSSI DSM-5 criteria there are more general psychopathology and impairment, than in the population who did not present NSSI behaviour or in the population who present NSSI behaviour, but did not meet the DSM-5 criteria.
Several risk factors of NSSI are described, such as prior history of NSSI, cluster b personality, hopelessness, female gender, depression, prior suicidal thought/behaviour, exposure to peer NSSI, eating disorder, abuse, etc. [6].
Many studies examined the association between NSSI and internalizing pathology; recently Bentley et al. [7] summarized these studies in their meta-analytic review: all the examined emotional disorders had an increased odds ratio for NSSI, except for bipolar disorder and social anxiety disorder. The association was the strongest with panic and post-traumatic stress disorder, otherwise there were no significant differences between the emotional disorders. Inconsistent methodologies are emphasized, such as the wide range of instruments used to assess NSSI and the multiple terms for self-injury, including NSSI.
The association between externalizing pathology and NSSI are understudied. Although Fox et al. [6] found that the odds ratios of 'misc externalizing symptoms' is higher (1.68) than 'misc internalizing symptoms' (1.37). According to our knowledge, there is only one review that investigated a specific externalizing disorder (attention deficit hyperactivity disorder -ADHD) and its association with self-injury, and found that ADHD is a risk factor of self-harm [8].
Therefore, our aim is to systematically review the studies that examined the association between self-injurious behaviour and externalizing psychopathology. Not only 'NSSI' was chosen as a search word for self-injurious behaviour, because the scientific literature is very confused about the terms [9]. Moreover, e.g. Brunner, Kaess et al. [10] noted this important aspect, often in clinical practice it is hard to distinguish the purpose of a selfinjurious act.
Before investigating the main aim of our study -1) prevalence rates, odds ratios or other associations between externalizing pathology and self-injuryit seemed to be important to review: 2) terminology and definitions of self-injurious behaviour, and 3) methods for measuring it. Finally, we examined: 4) associations between self-injury and suicide among the included publications.

Methods
A systematic literature search was made on 31th December 2016 in five computerised literature databases: PubMed, OVID Medline, OVID PsycINFO, Scopus, Web of Science. Search terms were: '(nonsuicidal selfinjury OR non-suicidal self-injury OR NSSI OR selfinjurious behaviour OR SIB OR deliberate self-harm OR DSH or self-injury) AND (externalizing disorder OR attention deficit hyperactivity disorder OR ADHD OR conduct disorder OR CD OR oppositional defiant disorder OR OD OR ODD)'. We found 902 papers in PubMed, 127 papers in OVID Medline, 120 papers in OVID PsycINFO, 116 papers in Web of Science and 16 papers in Scopus, making a total of 1281 studies including duplicates. We used EndNote X7 software for the systematization of the papers. After a duplicate search (both automatic by the software and manual by reading the authors and the titles again) we excluded 222 articles, so altogether we had 1059 individual papers (see the details in Fig. 1).
The flowchart of the inclusion/exclusion process can be seen in Fig. 2. Inclusion criteria were that studies investigating self-injury and externalizing pathology (a), are written in English (b), and reported original data (c). Exclusion criteria were studies: that were not written in English (b), were review/meta-analysis (c), were not about psychiatry/psychology (d), were not about selfinjury (e), were not about externalizing pathology correlation (f ).
A QUORUM flowchart of the selection process is summarized in Fig. 1. Publications included in the systematic review were selected in three steps: 1) from the 1059 individual papers, based on their titles 767 publications were excluded, 2) from the remaining 292, based on their abstracts 232 more were excluded 3) from the remaining 60 papers, 26 more were excluded after reading their full-texts. Finally, we added 1 paper by manual search. Therefore, a total of 35 papers are included in our systematic review (Table 1). First, each paper's title, abstract and full-text was investigated by the first author, which process was blinded reviewed by the second author. When there was a disagreement between the first and second authors whether a paper should be included or excluded to the systematic review, a common discussion and a final collective decision was made by all the three authors.
The methodology of this systematic descriptive review follows the PRISMA guidelines.

Terms and definitions for self-injurious behaviours in the included studies
Below, the different terms and their presence in the reviewed publications are displayed. It is important to clarify that in general the term 'self-injurious behaviours' is used for describing all of these acts, which are not the same as the specific term 'self-injurious behaviour (SIB)'. In 7 papers there was no clear distinction between the purpose of the investigated self-injurious act (whether it was with suicidal intent or not) [11][12][13][14][15][16][17], but in the remaining 28 papers the defined self-injurious behaviour was a nonsuicidal act .
There were 11 different terms for self-injurious behaviour in the investigated 35 papers.

Measurements of self-injurious behaviours in the included studies
Among the studies included in our review, diagnostic interviews, self-reported questionnaires and institutional records were used to measure self-injurious behaviours: altogether we found 20 different instruments among the investigated 35 papers.
In the remaining 7 publications, some kind of institutional (health care, prison) records or national databases were the sources of the data on self-injurious behaviours [12,14,16,24,25,27,37]. It is important to highlight the work of Kirkcaldy et al. [37]: although they based their examinations on the health records of German adolescents and youths, in the region where their study was conducted, the German Inventory Questionnairewhich contains questions about SIBis a regular part of the documentation.

Prevalence of self-injurious behaviour
Epidemiological data about self-injurious behaviours in different groups can be found in 20 publications (see details below).
Five studies investigated general psychiatric populations (inpatients, outpatients or intensive home-based treatment population) [17,24,34,37,41]. Among them only one paper used NSSI as a term: Jacobson et al. [34], however they also used DSH to include suicidal behaviour, and they found 13% prevalence of NSSI and 17% of DSH. Feingold et al. [24] investigated 'self-injury' , and found its prevalence to be 32.4%. Kirkcaldy et al. [37] examined SIB with a prevalence rate of 59%. Preyde et al. [41] studied 'self-harm' and found 34% as its prevalence rate. 'Self-mutilation' was the term in the study of Zlotnick et al. [17], and the prevalence rate was 33.2%.
In the study of Carli et al. [19] there was an incarcerated population. They found the prevalence of 'self-mutilation' to be 17.0%.
In the remaining 8 studies, the aim of the investigation was to compare a group having a specific psychopathology with a healthy control group or with a group having another psychopathology.
Four publications examined attention deficit hyperactivity disorder (ADHD) as a group with specific psychopathology [12,18,30,31]. Among ADHD patients versus non-ADHD patients, there was a robust difference in the prevalence rates of self-injurious behaviours. Prevalence of self-injury was 50.6% in the combined type of ADHD, versus 28.9% in the inattentive type of ADHD, versus 19% in the non-ADHD control group [30]. Prevalence of self-injury among drug dependent patients with ADHD was 47.7%, versus 25.2% without ADHD [18]. Prevalence of DSH was 69% in an ADHD group and 32% in a control group [31]. In one study there was a significant difference in the incidence of DSP: it was 6.13/10,000 persons/years in the ADHD group and 1.36/ 10,000 persons/years in the control group [12].
Two studies examined intermittent explosive disorder (IED) as a specific psychopathology. McCloskey et al. [38] found the prevalence of SIB to be 16% in this group. Jenkins et al. [35] investigated participants with personality disorder with or without IED: they found the prevalence of NSSI to be 18%.
Only in one publication could data be found on the prevalence of self-mutilation among participants with conduct disorder (CD) [32]: they found its prevalence between 15.5 and 62.5%. It depends on the gender and alcohol dependence of participants, with the lowest prevalence being among non-alcohol-dependent boys, and the highest being among alcohol-dependent girls. However there was another study, which examined the prevalence of SIB among psychiatric offenders in a military institution with antisocial personality disorder [42]: the prevalence rate of SIB was 92.0%.

Correlations of self-injurious behaviours and externalizing psychopathology
Fifteen publications investigated the associations between self-injurious behaviours and ADHD [12,14,16,18,21,23,25,28,30,31,33,[41][42][43][44]. Four of them studied ADHD versus non-ADHD groups, and found robust differences in the prevalence or incidence rates of self-injurious behaviours (self-injury [18,30]; DSH [31]; DSP [12]). Higher symptom severity or higher scores on ADHD symptom scales or higher numbers of ADHD symptoms are associated with self-injurious behaviour in 6 papers: ADHD scores were higher in the SII group than in the control group or depressed group [21]; ADHD scores were higher in the self-harm group versus non-self-harm group [23]; and in the DSH group versus non-DSH group [33]; frequency of SIB was correlated with the number of ADHD symptoms [42]; those who engaged in self-harm in a residential treatment or intensive home-based treatment programme had higher ADHD symptom severity than those who did not engage in this behaviour [41]; and there was a significant association between symptom severity of ADHD and history of DSH behaviours [44]. Moreover, Fulwiler et al. [25] found the importance of childhood ADHD in adulthood self-mutilation among prisoners. Lam et al. [14] underline that the odds ratio of having comorbid attention deficit disorder (ADD) is 3 times higher in the self-harm group than in the non-selfharm group. Swanson et al. [43] found that both the inattentive and combined type of ADHD were associated with NSSI, however more severe forms and higher frequency of NSSI are more likely to be associated with the combined type. Young et al. [16] emphasize that the number of critical incidents (including self-injury) were higher among people with ADHD in an incarcerated population. According to the work of Guendelman et al. [28], maltreatment is a strong risk factor of both NSSI and any other self-harm acts (suicide attempts), and there is no significant difference in this risk effect between girls diagnosed with ADHD and comparison group.
Twelve publications examined the possible connection between conduct disorder (CD) or CD symptoms or antisocial personality disorder in adulthood and selfinjurious behaviours [11, 15, 20-22, 26, 32, 34, 36, 37, 40, 42]. Seven of them found significant positive associations: CD is a risk factor of adulthood DSP repetition [11]; both CD and oppositional defiant disorder (ODD) correlated with DSH, however there is no significant difference in the level of correlation between them [20]; there were higher CD scores in the SII group than in the control group or depressed group [21]; antisocial personality disorder is a risk factor of suicidal gestures (included self-injurious behaviour) [26]; prevalence of self-mutilation is high among CD patients, especially when they are girls and they are also alcohol-dependent [32]; higher initial levels of conduct problems are associated with NSSI [36]; and DSH is linked to violent behaviours (robbery, intimate partner violence, forced sex, cruelty to animals, use of a weapon) [15]. Two publications described a non-significant positive connection: CD symptoms are more frequent in a non-suicidal self-harm group than in the control group [22]; frequency of any disruptive behaviours was higher both in DSH and in NSSI groups than in the control group [34]. Nock et al. [40] found the presence of CD to be 49.4% in their 100% NSSI sample. Semiz et al. [42] found the prevalence of SIB to be 92% in a 100% antisocial personality disorder sample. However Kirkcaldy et al. [37] could not find any association between SIB and disruptive behaviour.
Four papers studied the link between oppositional defiant disorder (ODD) and self-injurious behaviours or SIB. Cerutti et al. [20] found, that ODD correlated with DSH but they did not find a significant difference with the correlation of CD and DSH. In the study of Crowell et al. [21], there are higher ODD scores in the SII group than in the depressed group or control group. Guertin et al. [29] found that ODD was more common in both the selfmutilator and suicide attempter group than in the suicide attempter only group. In Nock et al. [40], in the 100% NSSI sample the presence of ODD was 44.9%.
There was information about general externalizing symptoms in 5 publications. Crowell et al. [13] examined parasuicidal versus non-parasuicidal groups; in the parasuicidal group the T-scores of externalizing symptoms were significantly higher. Crowell et al. [21] investigated the SII group, control group and depressed group: Tscores of externalizing psychopathology were also significantly higher in the SII group than in the control or depressed groups. In the 100% NSSI sample of Nock et al. [40] the presence of any externalizing disorder was 62.9% while the presence of any internalizing disorder was 51.7%. Feingold et al. [24] did not study the direct link between externalizing pathology and self-injury, but they found a strong link between alcohol abuse and selfinjury, and also described the association between externalizing pathology and alcohol abuse. Gatta et al. [27] examined neuropsychiatric inpatients who engaged in NSSI and controls from high schools. They found significantly higher externalizing scores in the NSSI group than in the control group according to both self-report and parent-report.
In the remaining 3 studies, the authors examined traits of externalizing psychopathology: Carli et al. [19] found a significant difference in the prevalence of selfmutilation between high-impulsive and low-impulsive groups. You et al. [45] investigated the role of impulsecontrol problems in the frequency and severity of NSSI and found individuals with more impulse-control problems had more severe and frequent NSSI behaviour. Meza et al. [39] found the importance of response inhibition in the prediction of NSSI among individuals with or without ADHD.

Self-injury and suicide
In total 8 of the 35 articles assessed the relationship between self-injury and suicide [22,25,26,29,31,32,34,43]. To compare results by age groups, we will first overview 3 papers that studied adult samples [22,25,26] and then turn to the other 4 studies that assessed adolescence populations [29,31,32,34] and one longitudinal study that assessed young girls (M = 9.1 years) with 5 and 10 years follow-ups [43].
In their study on adult injecting drug users, Darke and Torok [22] assessed if the two phenomena (NSSI and suicidality) overlap. They found that 42.3% of those who reported NSSI also reported suicide attempt, and 39% of those who reported suicide attempt had also engaged in NSSI. In 83.3% of the cases with both suicidal and NSSI, the NSSI was present prior to suicide attempts.
The other two studies on adult samples focused on the differences between suicidal and NSSI behaviours [25,26].
In an adult prisoner sample, Fulwiler et al. [25] found differences between those committing suicide attempts and self-mutilation regarding their methods (more lethal methods in suicide attempt) and frequency (higher frequency in case of self-mutilation). The two groups also differed in psychiatric comorbidities: diagnosis of an affective disorder was present in 87% of the suicide attempt group compared to 19% in self-mutilators, while a syndrome of mixed anxiety/dysthymia patterns present from childhood or adolescence (56% in self-mutilation group) and history of childhood hyperactivity (75% vs 7%) characterized the self-mutilation group.
Garcia-Nieto et al. [26] also focused on the differences between self-injurers with and without intent to die. Their results pointed out the differences in the prevalence of certain personality disorders: histrionic personality disorder (HPD) (0% in neither gestures nor attempts; 46.7% in gestures; 1.6% in attempt, 12.5% in attempts + gestures group) and antisocial personality disorder (APD) (5.3% in neither gestures nor attempts; 6.7% in gestures; 4.8% in attempt, 1.2% in attempts + gestures group) as risk factors were associated with suicide gestures. Narcissistic personality disorder (NPD) (27% in suicide attempt group, 0% in all other groups) showed association with suicide attempts, and BPD (7% in neither gestures nor attempts; 33.3% in gestures, 9.7% in attempts, 43.8% in gesture + attempt group) seemed to be a risk factor for both suicide gestures and attempts.
In an adolescent community sample [31], those with ADHD reported more suicidal ideation (57% vs. 28%, P < 0.001) and DSH (69% vs. 32%, P < 0.001) than those who did not have the disorder. Female gender and depression/anxiety were also associated with both suicidal ideation and DSH, while behavioural disorder, substance use and certain family factors (e.g. living in nonintact family, living in family with financial problems) were associated with DSH but not with suicidal ideation and acts.
Jacobson et al. [34] studied an adolescent outpatients sample with no DSH (52%), NSSI only (13%), suicide attempt only (16%) and with both NSSI and suicide attempt (17%). They found that the only psychiatric diagnosis in which those engaging in NSSI differed from those who had not engaged in any type of DSH was BPD. When distinguishing between DSH groups, differences were found in the likeliness of comorbid depression, posttraumatic stress disorder (PTSD) and in the levels of suicidal ideation. Regarding MDD, with the NSSI only group as the comparison group and controlling for gender, the suicide attempt only group (odds ratio -OR =3.43; 95% CI = 1.17-10.00; p = 0.03) and the suicide attempt and NSSI group (OR = 3.55; 95% confidence interval -CI = 1.24-10.16; p = 0.02) were more likely to have depression. Regarding PTSD, when controlling for gender, participants in the suicide attempt only group (OR = 8.93; 95% CI = 1.04-76.92; p = 0.046) and the suicide attempt and NSSI group (OR = 10.09; 95% CI = 1.18-85.97; p = 0.035) were more likely to have PTSD diagnosis than those in the NSSI-only group. The NSSI-only group also reported significantly lower levels of suicidal ideation scores than the other DSH groups: Suicide attempt only (adjusted mean difference = −20.24, p < 0.001) and suicide attempt and NSSI group (adjusted mean difference = −25.92; p < 0.001). Meanwhile the two other DSH groups (suicide attempt only and suicide attempt and NSSI groups) did not significantly differ from each other regarding the rates of MDD, PTSD and the levels of suicidal ideation.
Ilomaki et al. [32] did not investigate the association between the two phenomena but they found that alcohol dependence was a common potential risk factor for both phenomena since it increased the risk for both suicide attempts 3.8-fold [CI95 = 1.06-13.44; p = 0.041] and self-mutilation 3.9-fold (CI95 = 1.09-13.76; p = 0.037) among adolescent girls with DSM-IV-diagnosed CD.

Discussion
In the current systematic review, 35 papers examining the correlation between externalizing psychopathology and self-injurious behaviours are summarized.
Consistent with the review on internalizing pathology and NSSI [7], we found wide-ranging terminology for self-injury: 11 different terms were used in 35 papers. Among these terms NSSI has the clearest definition, suggested by the 'International Society for the Study of Self-injury' (ISSS) in 2007 [2] and by DSM-5 [1]. DSH is the most unstable term, because it has many definitions, some authors use it only in a nonsuicidal meaning [20,31,33,44], while others use it as non-fatal but with both suicidal and nonsuicidal meanings [15,34]. The other issue is that self-poisoning was not included in the definition by most of the authors [15,20,31,33,44], but Jacobson et al. [34] use the term DSH both for physical self-damage and overdose cases as well.
Moreover, DSP appeared as an individual term for selfpoisoners [11,12]. SIB is mostly defined as both suicidal and non-suicidal self-injurious behaviour, but selfpoisoning is not included in the definition [37,38,40,42]. Summarizing the main issues of defining self-injurious behaviour are the following: whether the definition used should include self-poisoning or not, and whether it should be clarified as a suicidal or a nonsuicidal act. For example, Kapur et al. [52] questioned the concept of NSSI, because in some cases it is impossible to decide if the act was with suicidal intent or not, especially in an adolescent population. Nock [53] drew attention to this definition problem in his earlier review; while sometimes those who have engaged in self-injury cannot themselves clarify their intent, conceptually it is important to make this distinction. Brunner, Kaess et al. [10] made the distinction between self-poisoning and self-harming by focusing on the surface of the body.
The heterogeneity that is present in the terminology and definitions of self-injurious behaviours is also reflected in the diversity of instruments used to assess the phenomenon. The most frequently used instruments (LPC [13,21,34]; DSHI [20,35,44]) were present in no more than three plus three out of the 35 studies. Among the studies included in our review, diagnostic interviews, self-reported questionnaires, institutional (health care, prison) records and national databases were used to measure self-injury. In more than half of the papers (18/35) the method was an interview, or special questions as part of a longer interview [11,13,15,18,19,21,22,26,28,31,32,34,36,38,39,[41][42][43]. The instruments were based on the previously described different definitions of self-injury and therefore they may measure somewhat different phenomena. This may be one of the factors that make it difficult to compare the results of different studies, as we have already highlighted above. The prevalence rates of different self-injurious behaviours vary greatly both in clinical samples (13.0-59.0%) and in normal population based samples (2.9-41.9%). This variance can be explained partially by the definition and measurement heterogeneity, described above. However, using one term (e.g. NSSI) does not lead to more homogenous prevalence rates. In the specific psychopathology groups investigated, the diversity of prevalence rates could also be observed. Due to this heterogeneity, and with the aim of avoiding confusion, the original terms used by the authors of the papers are used in the 'Results' section, and the prevalence rates are mentioned in parallel. These various results underline the importance of standardized terminology and measurement for self-injurious behaviours, which would make the results of future studies comparable.
There were significantly higher prevalence rates in ADHD groups than in control groups [12,18,30,31]. Moreover, the papers found higher ADHD symptom severity in self-injurer groups [21,23,33,41,42,44]. The most severe and frequent forms of NSSI can be found in the combined types of ADHD versus the inattentive type and the control group [43], that strengthen the role of impulsivity in self-injurious behaviours, which is an important risk factor of selfinjurious behaviours. However, according to Hamza et al. [54], itis very complicated to measure, especially with self-reported questionnaires. Moreover, it seems that different aspects of impulsivity have different roles on self-injury. Mood-based impulsivity influences the initiation of self-injury, cognitive facets of impulsivity are related to the maintenance of self-injury and behavioural impulsivity is associated with self-injury under conditions of negative affect [55].
In addition to the finding that ADHD and selfinjurious behaviours are strongly associated, which is in agreement with a previous review on this topic [8], we found evidences for the association between selfinjurious behaviours and other externalizing psychopathology. There are high prevalence rates of self-injurious behaviours among patients with CD (15.5-62.5%, it depends on the gender and alcohol dependence of participants [32]. Semiz et al. [42] found the prevalence rate of SIB to be 92.0% among male offenders with antisocial personality disorder. Moreover, there were higher CD scores in the self-injurer group, than in the control group [21]; and in 6 further studies there are strong links between CD or disruptive symptoms and selfinjurious behaviours [11,15,20,26,32,36]. Only one paper from the 12 investigated could not find a link between CD and self-injury [37]. The relationship between self-injurious behaviours and ODD is underlined by 3 papers [20,21,29], and there is a connection with intermittent explosive disorder as well according to 3 studies [17,35,38]. General externalizing symptoms are also more frequent in self-injurer groups [13,21]. Nock et al. [40] found higher externalizing comorbidity than internalizing in self-injurer groups.
Although the aim of this review was not specifically to explore the relationship between NSSI and suicidality, 8 out of the 35 articles contained results regarding this topic. Although the two phenomena might overlap [22,34] and might have shared clinical characteristics such as BPD [26], depression and anxiety [31] or alcohol dependence [32], most of these articles focused on factors distinguishing between suicide and NSSI. Those with and without suicidal intent might show differences in the frequency and methods of their self-injuring, comorbid symptoms and personality disorders [25,26]. Some results show that while externalizing symptoms are associated with NSSI [25,43], internalizing symptoms might play a bigger role in suicidality [25,34,43]; while others found higher numbers of both internalizing and externalizing symptoms among those who reported both suicidal behaviour and NSSI [29,31].

Limitations
The main limitation of our review is the heterogeneity of both the terminology and instruments, as we mentioned above, which makes it hard to compare the results of the included papers. In most of the cases, the distinction that the act was with or without suicidal intent was clear, however, among these 35 papers there were 7 papers, where this distinction was not clarified [11][12][13][14][15][16][17]. The fact that similar terms for self-injurious behaviour were also measured with multiple instruments in different studies also complicates the understanding of the topic.
A further limitation is that we may be missing some relevant papers on this topic. Firstly, because we included only articles that are written in English. Secondly, we have chosen the standard method and searched for publications in large indexed literature databases. This led to 14 papers on the topic of ADHD and self-injury, however, Allely [8] used a less conservative search method and checked 'Google Scholar' as well and found 15 articles on the topic of ADHD and NSSI, which is a sub-topic of our review topic. There were small differences in our keywords as well, that could also be an explanation for the differences in our findings. They used 'self-poisoning' as a search word as well. Finally there were 5 overlapping publications [25,30,31,33,42], but they found 9 publications that we did not (5 of them fit our inclusion criteria as well, but they had not appeared in our search) and we found 3 articles in their time window, that they had not [18,21,41]. The remaining 5 publications we found miss their time window [12,14,23,43,44]. In spite of all these differences in the papers, Allely [8] also concluded that ADHD and self-injury have a strong association, as we did, so these results support each other. Finally, it is important to underline that our aim was not to focus only on ADHD but on the whole spectrum of externalizing disorders.
Another methodological limitation could be that our work is a narrative review. A meta-analytic review, or a full meta-analysis would be more useful to compare the results of these very heterogenous papers. To handle this limitation and make the comparisons easier, sample sizes, design of the original studies and statistical results are presented in Table 1.
Another limitation could be that since our focus in this review was to investigate the association between NSSI and externalizing pathology, the role of externalizing problems in the relationship between NSSI and suicidality might be overrepresented, while other important factors might remain hidden. A review focusing directly on the NSSI-suicidality relationship, including two of the articles discussed in this paper [29,34] was just recently published [56]. According to their conclusion, the two phenomena are correlated and share similar vulnerabilities and risk factors, moreover, NSSI itself is found to be a risk factor for suicidal behaviour.

Conclusions
In conclusion, reviewing these 35 papers on self-injury and externalizing psychopathology underlines the confusion in terminology on self-injurious behaviours, which could lead to difficulty in comparing studies on the topic. Based on our findings, we suggest the use of the term NSSI in future studies, as NSSI has the most precise definition. Furthermore, it became an individual diagnostic category in DSM-5 [1].
Summarizing the results of these papersalthough the connection between internalizing pathology and selfinjurious behaviours are well-studied [7] there is a strong association between self-injurious behaviours and externalizing pathology and individual externalizing disorders as well. Based on this knowledge, it seems to be important to investigate the presence of parallel internalizing and externalizing pathology and its connection to NSSI, because it may help to identify the most endangered populations.