Introduction

Externalising disorders refer to a group of behavioural disorders appearing during childhood that represent the child’s negative conduct toward his external environment.1 In the research literature, externalising disorders include attention deficit/hyperactivity disorder (ADHD), oppositional/defiant disorder (ODD) and conduct disorder (CD).2 Evaluating the structure of these disorders usually demonstrates the connection between the tendency toward disinhibition and high-risk impulsive behaviours.3 These are the most prevalent reasons for referring children and adolescents to mental health centres, burdening families and individuals with heavy costs.4 Based on the existing studies, at least 5%–10% of school-aged children and adolescents have serious ongoing behavioural and emotional problems. A significant percentage of these problems can persist for years to come. The self-report by adolescents shows a high level of alcohol consumption, substance use, bullying and aggressiveness. These statistics are rather elevated even in industrialised countries.5 The systematic analysis of the Global Burden of Disease6 assessed the disability-adjusted life year and found the age-standardised rate for CD as 78.2 per 100 000 in 2005 and 79.3 per 100 000 in 2015. The proportion for ADHD was 8.6 in 2005 and 8.3 in 2015. For the abuse of drugs, alcohol, medicine, cocaine, amphetamine, hashish and other substances, the overall rate was reported as 597.7 in 2005 and 803.9 in 2015. The Global Burden of Disease Study7 also reported the age-standardised death rate for the abuse of drugs, alcohol, medicine, cocaine, amphetamine, hashish and other substances as 4.9 per 100 000 in 1990 and 5.6 per 100 000 in 2013.

Examining the prevalence of these disorders in clinical and non-clinical samples of children, adolescents and adults revealed a high prevalence of comorbidity.8 The prevalence of comorbidity was above 50% for ODD, CD, ADHD and substance use disorders. In addition to the high prevalence of comorbidity of these disorders, a high proportion of comorbidity was also reported between these and internalising disorders (ie, stress and depression). Epidemiological data showed that the concurrence of two or more disorders cannot be attributed to mere random factors.9 The comorbidity data have created a serious challenge for classification systems and aetiological hypotheses.10 Psychiatric disorder comorbidities, including externalising disorders, were rarely examined before the publication of DSM-III-R (diagnostic and statistical manual of mental disorders) in 1987. Previous versions of DSM used the diagnostic hierarchy, which usually meant comorbidities were ignored and not recorded. Eliminating the diagnostic hierarchy significantly increased the prevalence of comorbidities, and there was then a substantial rise in studies addressing the patterns of these diagnostic overlaps. The number of articles including the word ‘comorbidity’ went from 0 in 1985 to 243 in 1993 and over 1500 in 2013. This is an ongoing trend. The comorbidity and coexistence of two or more disorders have been the main challenge of the DSM diagnostic system and other disorder classification systems.11

Krueger suggested that comorbidity in psychiatric disorders is caused by fundamental psychopathological processes and focused his research on these main processes rather than their manifestations as separate disorders.12 By focusing on symptoms rather than disorders, bottom-up models or paradigms emerged as opposed to top-down paradigms. In bottom-up paradigms, the focus is on empirical data, from which the concept of psychopathology is derived. In contrast, top-down paradigms (such as DSM and ICD (International Statistical Classification of Diseases and Related Health Problems) diagnostic systems) prioritise the conceptualisation of the nature and structure of psychopathology. In bottom-up models, extensive factor analyses of psychiatric symptoms were performed to discover and explain the liabilities associated with the structure of psychopathology. The results of these extensive factor analyses have been varying.13

By modelling the hierarchical relationships between disorders and using the exploratory factor analysis for behavioural, emotional, social and thought problems in children and adolescents, Achenbach identified a bifactor model. The first factor included an external pole (involving aggressive and law-breaking behaviours) and an internal pole (involving intrapersonal disorders such as anxiety, depression and physical complaints without a medical reason). The second factor was unipolar and was presented as severe psychopathology that included problems such as strange behaviours, hallucinations, delusions, feeling harassed, confusion, fear of own impulses, compulsion and obsession.13 Inspired by the Achenbach model, Krueger et al suggested that the relationship between disorders reflects the two dimensions of internalisation and externalisation, which represent common genetic and environmental effects in disorders related to each spectrum.14 In adolescent samples, the internalisation–externalisation model of psychopathology has also been corroborated by other research. Caspi et al15 found evidence of a bifactor model that includes a general psychopathological factor and a second factor that includes the spectrum of internalisation and externalisation. In this context, individual symptoms are organised into distinct diagnoses. The extensive research by Caspi et al using adolescent samples supported this bifactor model. The symptoms are correlated in the bifactor model because all of these symptoms have general psychopathology in common as their feature. Overall, the mean score for the p factor indicates a general tendency towards psychopathology, and the mean score for each spectrum indicates a tendency towards problems associated with that spectrum. For example, in the externalisation spectrum, the probability of aggressive behaviours and behavioural problems is higher.14 The general psychopathology factor can explain the problems encountered in identifying the unique causes, consequences, biomarkers and treatment of individual disorders. This general factor emphasises the importance and usefulness of transdiagnostic treatment and comprehensive prevention approaches and can address multiple problems in a single framework.15 The presence of this general psychopathology indicates a positive correlation between these spectra.

Another model showed that the structure of psychological disorders can be summarised in three dimensions: (1) internal liability for depression and anxiety, (2) external liability for behavioural disorders and substance use disorders, and (3) liability for psychosis symptoms. This finding showed that in addition to the established dimensions of internal and external liabilities, there is a distinct third dimension that is characterised by confused thoughts.15 Of course, there is still some debate about the cause of thought disorder. For example, Caspi et al argued that thought disorder is not an independent dimension, rather it indicates the overall severity of general psychopathology.16

In addition to identifying the dimensions, numerous studies have examined the subfactors in each spectrum and have yielded disparate results. For example, some studies have not proposed any subfactors for externalising disorders, but internalising disorders have been divided into three subfactors, including distress, fear and eating.17 In some research, the externalising spectra have been modelled as an individual latent factor. Nevertheless, studies have shown that separate subfactors exist for this dimension, such as the bifactor model which distinguishes the subfactors of oppositional and defiant conduct from social norms.18

Studies investigating the latent structure of psychiatric disorders in children and adults have also shown that the underlying factors showing internalising and externalising liabilities in psychopathology explain comorbidity patterns in psychiatric disorders very well.19 Nonetheless, there is still controversy about the lower-level factors, but since there seems to be persistence in the latent structure of disorders between adolescence and adulthood, interventions aimed at the prevention and treatment of psychopathology in adolescents can reduce the risk of psychiatric disorders in adulthood.19

The new top-down models use the factor analysis method to construct dimensional scales similar to the scales based on the bottom-up models described earlier.13 The Hierarchical Taxonomy of Psychopathology (HiTOP) framework is used with the aim of integrating research on the experimental organisation of psychopathology to set up a comprehensive descriptive system. This hierarchical and dimensional classification system can solve problems related to the comorbidity of disorders, heterogeneity within the disorders, diagnostic instability and subthreshold symptoms.20 Compared with DSM diagnoses, the transdiagnostic dimensions of HiTOP explain heritability patterns, biological neural processes, functional disorders, environmental risk factors and variations in response to therapy. The HiTOP framework consists of five hierarchically organised levels. At the lowest levels of the hierarchy are signs and symptoms that combine to reflect the broader dimensions of psychopathology, from syndromes/disorders (such as the DSM diagnoses) to subfactors (such as fear and distress). The spectra (such as antagonism and detachment) and higher-level construct are two levels higher, which is a common factor in most forms of psychopathology.21

The external superspectrum consists of two spectra: (1) disinhibited externalising and (2) antagonistic externalising. The disinhibited externalising spectrum includes impulse-based tendencies regardless of potential consequences. Empirically, disinhibition is associated with socially forbidden behaviours that are psychologically aligned with the core of the structure, such as excessive use of psychedelic drugs with minimal regard for future consequences. The antagonistic externalising spectrum includes tendencies towards interpersonal antagonism and deliberately hurting other people with the least regard for their rights and feelings.20 Kotov et al22 suggested that disinhibition is above all associated with the issue of using drugs and includes substance use disorders, while conflict is prominent in aggressive disorders, which include CD, ODD, ADHD, intermittent explosive disorder and antisocial personality disorder.

Research on the validity of the model is still underway. Assuming the existence of a dimensional hierarchy model, it is expected that the prevalence of comorbidities in the subspectra would be higher and less correlated with the other spectra. Also, considering the general factor of psychopathology, a positive correlation is expected between the prevalence of comorbidities in the externalising subspectrum with the other spectra. Despite the sensitivity of the adolescence period and its importance in the formation and continuation of subsequent pathologies, we see fewer studies in this area about the prevalence of these disorders and their structure at this age. Therefore, the purpose of this systematic review is to examine the prevalence of comorbidities of externalising disorders in adolescence.

To the best of our knowledge, there was no relevant systematic review examining the prevalence of comorbidities of externalising disorders. Nevertheless, some systematic reviews have examined the comorbidity of one disorder from the externalising spectrum. A systematic review of the comorbidity of electronic cigarette smoking and mental health problems examined 40 articles including adolescents and young adult groups by focusing on three spectra of internalising disorders (depression and anxiety, suicide and eating disorders), externalising disorders (ADHD and CD) and transdiagnostic concepts (impulsivity and perceived stress). This review study showed that electronic cigarette smoking is associated with an increased prevalence of mental health problems, especially in adolescents. The time frame of the search was limited and the reviewed studies were only cross-sectional.23

Another systematic review study also examined the prevalence of ODD comorbidity in children and adolescents.24 Their results showed that although ODD occurs in all age groups, its comorbidity rates differ in the special age group of children and adolescents, and also boys and girls have different comorbidity patterns. The studies reviewed in the research were limited to English-language studies and limited sources were used for the search. Dullur et al25 conducted a systematic review to examine comorbidity rates between ADHD and gaming disorder. They demonstrated a relationship between these two disorders by reviewing 29 observational studies, and the relationship between this disorder and ADHD was greater than the other subclusters. Their sources were limited and no meta-analysis was carried out by them either. In another systematic review and meta-analysis, Nazar et al26 examined the comorbidity of ADHD and eating disorders. They reviewed 16 articles (12 studies on eating disorders in people with ADHD and 4 articles on ADHD in people with eating disorders) and found that the prevalence of eating disorders is increasing in people with ADHD (3.82). This finding indicates that people with ADHD have a relatively high risk of developing an eating disorder. In a meta-analysis and meta-regression analysis, the prevalence of comorbidity of ADHD and substance use disorders was examined in 29 English-language studies, and the results showed that the prevalence of comorbidity was 23.1. The study sources were limited in that article as well, and no systematic review was performed by its researchers. In another meta-analysis, Tung et al27 examined the comorbidity pattern in girls with ADHD. In that study, the comorbidity of internalising and externalising disorders was studied and the review of 18 articles showed that the comorbidity of this disorder with other disorders of the externalising spectrum (ODD: 5.6, CD: 9.4) was higher than with disorders of the internalising spectrum (anxiety: 3.2, depression: 4.2). On the same subject, in a systematic review protocol and meta-analysis, Catalá-López et al28 examined the prevalence and comorbidities of ADHD. Their study included cross-sectional observational studies of the general population and the school environment. In the first stage, the purpose of the study was to examine the prevalence of ADHD. In the second stage, it examined the prevalence of any physical or psychological problems associated with ADHD. The heterogeneity causes were also evaluated.

Among previous relevant systematic reviews and meta-analysis articles, only two articles have used both the systematic review and meta-analysis approaches. Most of these articles had language limitations and most had not used grey literature. In a few studies, the selection process, quality assessment and data extraction were all duplicated. Also, in none of the studies were all three of these steps duplicated simultaneously. Moreover, limited search terms had been used in all the studies. Compared with these previous studies and based on the prior principle, a systematic review, meta-analysis and database search were carried out in this study without any language restrictions and with consideration for grey literature resources to examine the prevalence of comorbidity of externalising disorders (ADHD, CD, substance use disorders and ODD) in adolescents.