Introduction
Research over the past two decades has highlighted that psychotic experiences (PEs) are far more prevalent in the population than psychotic disorders (Linscott and van Os, Reference Linscott and van Os2013). While approximately 5% of adults report PE phenomena (Linscott and van Os, Reference Linscott and van Os2013; McGrath et al., Reference Mcgrath, Saha, Al-Hamzawi, Alonso, Bromet, Bruffaerts, Caldas-De-Almeida, Chiu, De Jonge, Fayyad, Florescu, Gureje, Haro, Hu, Kovess-Masfety, Lepine, Lim, Mora, Navarro-Mateu, Ochoa, Sampson, Scott, Viana and Kessler2015; Maijer et al., Reference Maijer, Begemann, Palmen, Leucht and Sommer2018), the prevalence is higher in children and adolescents, with estimates ranging between 8% and 17% (Kelleher et al., Reference Kelleher, Connor, Clarke, Devlin, Harley and Cannon2012a; Maijer et al., Reference Maijer, Begemann, Palmen, Leucht and Sommer2018).
Initial research on the clinical significance of PEs focused on their association with future risk of psychotic disorders (Poulton et al., Reference Poulton, Caspi, Moffitt, Cannon, Murray and Harrington2000; Dominguez et al., Reference Dominguez, Wichers, Lieb, Wittchen and Van Os2011; Zammit et al., Reference Zammit, Kounali, Cannon, David, Gunnell, Heron, Jones, Lewis, Sullivan, Wolke and Lewis2013). Subsequent research found that individuals with PEs were also at risk of a range of non-psychotic disorders such as affective, anxiety and behavioural disorders (Kelleher et al., Reference Kelleher, Keeley, Corcoran, Lynch, Fitzpatrick, Devlin, Molloy, Roddy, Clarke, Harley, Arseneault, Wasserman, Carli, Sarchiapone, Hoven, Wasserman and Cannon2012b; Calkins et al., Reference Calkins, Moore, Merikangas, Burstein, Satterthwaite, Bilker, Ruparel, Chiavacci, Wolf, Mentch, Qiu, Connolly, Sleiman, Hakonarson, Gur and Gur2014; Jeppesen et al., Reference Jeppesen, Clemmensen, Munkholm, Rimvall, Rask, Jorgensen, Larsen, Petersen, Van Os and Skovgaard2015; McGrath et al., Reference Mcgrath, Saha, Al-Hamzawi, Andrade, Benjet, Bromet, Browne, De Almeida, Chiu, Demyttenaere, Fayyad, Florescu, De Girolamo, Gureje, Haro, Ten Have, Hu, Kovess-Masfety, Lim, Navarro-Mateu, Sampson, Posada-Villa, Kendler and Kessler2016b), with findings identifying an association between PEs and both concurrent (Calkins et al., Reference Calkins, Moore, Merikangas, Burstein, Satterthwaite, Bilker, Ruparel, Chiavacci, Wolf, Mentch, Qiu, Connolly, Sleiman, Hakonarson, Gur and Gur2014) and later mental disorders (Dhossche et al., Reference Dhossche, Ferdinand, Van Der Ende, Hofstra and Verhulst2002). An existing meta-analysis confirmed the association between PEs and both psychotic and non-psychotic disorders (Kaymaz et al., Reference Kaymaz, Drukker, Lieb, Wittchen, Werbeloff, Weiser, Lataster and Van Os2012). However, that analysis was primarily focused on adult samples and did not specifically examine the association among children and adolescents. Given that the prevalence of PEs in childhood is notably higher than adulthood (Kelleher et al., Reference Kelleher, Connor, Clarke, Devlin, Harley and Cannon2012a) and most individuals with lifetime PEs have their first onset by early adulthood (McGrath et al., Reference Mcgrath, Saha, Al-Hamzawi, Alonso, Andrade, Borges, Bromet, Oakley Browne, Bruffaerts, Caldas-De-Almeida and Fayyad2016a), clarifying if childhood and adolescent PEs are also associated with an increased risk of mental disorders (both psychotic and non-psychotic disorders) is therefore an important goal. With childhood and adolescent PEs being considered as an early pluripotent marker for subsequent psychiatric vulnerability (McGorry et al., Reference Mcgorry, Hartmann, Spooner and Nelson2018), it is also important to clarify any differences between cross-sectional and longitudinal relationships between PEs and mental disorders.
Specifically, the aims of this systematic review and meta-analysis are (i) to assess the association between childhood and adolescent PEs and mental disorder (any mental disorder, any non-psychotic disorder and sub-categories of mental disorder) in non-help seeking individuals from the general population (2) to assess the effect of study design (cross-sectional or longitudinal design) on the relationship between childhood and adolescent PEs and mental disorders.
Method
Search strategy
A systematic review was conducted investigating all of the published literature (from inception to August 2017) pertaining to childhood and adolescent PEs (⩽18 years) and mental disorder in non-help-seeking community samples. Searches were carried out in August 2017 by CH using three electronic databases PUBMED, EMBASE and PsycINFO. A search strategy was devised with the assistance of a librarian. The search terms used were General population OR normal population OR normal healthy population OR healthy individuals OR community sample OR child and adolescent AND mental disorder OR psychiatric disorder OR psychopathology OR mental illness OR DSM* OR ICD* AND psychotic experience OR psychotic symptoms OR psychotic-like experiences OR psychotic-like symptoms OR auditory hallucinations OR hallucinat* OR delusion*. Additionally, the reference lists of all selected papers were searched for potential study inclusion.
Inclusion criteria
Only studies published in a peer reviewed journal and written in English were included in the review.
Sample
Only non-help-seeking samples of children and/or adolescents were used in this investigation. We included samples if the majority of participants were aged 18 years or younger at the first enquiry of PEs.
Exposure
For the purpose of this investigation, childhood and adolescent PEs were considered as exposure. Data on PEs reported by both questionnaire and interview format were included. Within the literature, PEs are reported either dichotomously (i.e. as the presence or absence of any PE phenomena) or by sub-types (e.g. auditory hallucinations or paranoia). All studies that reported PEs dichotomously were included in this investigation. Where PEs were not reported in this way, and only data on sub-types of PEs were reported, only studies that reported on auditory hallucinations or hallucinations were included. The decision to include these two categories of PE as valid outcomes for this investigation was based on evidence that endorsement of auditory hallucinations on questionnaires has demonstrated predictive validity for the presence of PEs when subsequently assessed by clinical interview (Kelleher et al., Reference Kelleher, Harley, Murtagh and Cannon2009; Laurens et al., Reference Laurens, Hobbs, Sunderland, Green and Mould2012; and Granö et al., Reference Granö, Kallionpää, Karjalainen, Roine, Ranta and Heinimaa2016). Where different ‘strengths’ or ‘levels’ of PEs were reported (e.g. ‘weak’ and ‘strong’ PEs or ‘definite’ and ‘possible’ PEs) only the strong or definite category was used to estimate the relationship with mental disorders (weak and possible PEs were combined with controls to improve community sample representation). If a study examined multiple reporting of PEs (e.g. participants were grouped into whether they had reported PEs never, once, twice or three times), these groups were combined (any PEs ever). Information on the criteria used for PEs in each selected study is given in Table 1.
Descriptive summary of the studies included
PE, psychotic experiences; K-SADS, Kiddie Schedule for Affective Disorders and Schizophrenia; DAWBA, Development and Well-Being Assessment; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders, fourth edition; YSR, Youth Self-Report Questionnaire; DIS-C, Diagnostic Interview Schedule for Children; DIS, Diagnostic Interview Schedule; CIDI, Composite International Diagnostic Interview; DSM-III-R, Diagnostic and Statistical Manual of Mental Disorders, revised third edition; SCL-90, Symptom Checklist-90; DIA-X/M-CIDI computerised version of the Munich-Composite International Diagnostic Interview; ICD-10, International Classification of Disease tenth edition; PLSI, Psychosis-Like Symptom Interview; and SCAN, Schedules for Clinical Assessment in Neuropsychiatry.
Outcome
For inclusion in this investigation, participants must have met criteria for a mental disorder in accordance with Diagnostic and Statistical Manual (DSM) or International Classification of Disease (ICD) standards (any edition). Diagnosed mental disorders were grouped into any mental disorder, any non-psychotic disorder and five specific categories of disorder: psychotic disorder, affective disorder (mania or depression), anxiety disorder (generalised anxiety, panic, obsessive-compulsive and phobias), behavioural disorder (conduct, oppositional defiant or attention-deficit/hyperactivity disorder) and substance use disorder (any).
Exclusion criteria
Sample
Help-seeking, high-risk samples were excluded from the investigation. This was done to increase the representativeness of the pooled estimate relative to the general population. Inclusion of help-seeking samples is likely to bias the pooled estimates. Additionally, case-control studies were excluded from the investigation. While pooled point estimates based on case-control studies are likely to be similar to cohort studies, the confidence intervals (CIs) for these estimates are narrower than for cohort studies as the number of individuals within the exposure group are inflated relative to the general population. For this reason, case-control studies were excluded.
Non-diagnostic assessment of psychopathology and temporality
Any study that did not use ICD or DSM diagnostic criteria to determine rates of mental disorder was excluded from the review and meta-analysis. Additionally, as PEs were considered our exposure of interest and mental disorder our outcome, for cross-sectional studies PEs and mental disorder had to be assessed contemporaneously and for longitudinal studies PEs had to precede the assessment of mental disorder. If the temporal relationship between PEs and mental disorder explicitly stated that mental disorder preceded PEs, the study was excluded.
Study selection and data extraction
Literature search was conducted by CH in August 2017 with studies reviewed by CH and RB. An assessment of study quality was conducted using the National Heart, Lung, and Blood Institute quality assessment tool for observational cohort and cross-sectional studies (see online Supplement A, https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools). The data extraction was conducted by two reviewers (CH and ND), with an initial reviewer consistency of 87.5%. Data extraction discrepancies between the two reviewers were resolved via joint discussion with a third reviewer (RB). We report the location of where the data was extracted from in each study in online Supplement B.
Metrics
Unadjusted odds ratios (ORs) were used when available. When ORs were not present but were calculable based on the information presented in the study (number of individuals were available), ORs were calculated. If the unadjusted ORs were not calculable or were not available within the text, adjusted ORs were used and the confounders are documented in Table 1. If the study presented alternative metrics (hazard ratio or risk ratio) relevant authors were contacted to request information for calculating ORs.
Disorder grouping and sub-categories
Mental disorders were grouped into two overall categories: (i) any mental disorder and (ii) any non-psychotic disorder. We also investigated five sub-categories of mental disorder: affective, anxiety, psychotic, behavioural and substance-use disorder. Affective disorders included both depressive disorders and mania. Anxiety disorders included generalised anxiety, separation anxiety, specific phobia, social phobia, panic, agoraphobia, obsessive compulsive disorder and post-traumatic stress disorder (PTSD). As the majority of studies reviewed use DSM-IV diagnostic criteria, PTSD was included under anxiety disorder classifications. Psychotic disorders included schizophrenia, schizophreniform disorder, non-affective psychosis and psychotic disorders not otherwise specified. Behavioural disorders included oppositional defiant, conduct and attention deficit/hyperactivity disorders.
When not explicitly reported in the text the ‘any mental disorder’ category and the ‘any non-psychotic mental disorder’ ORs were calculated by pooling the ORs of the disorders presented in the study. This was calculated by averaging the log odds for each disorder (i.e. logodds disorder X + the logodds disorder Y/No. of disorders) and averaging the standard error for each disorder (i.e. standard error of disorder X + standard error of disorder Y/No. of disorders). We used these metrics to calculate the OR and 95% CI for the ‘any mental disorder’ category for that study. We also used this approach if sub-categories of a disorder group were reported (i.e. reported on different types of anxiety disorders such as generalised anxiety disorder and panic disorder). This method reduced the likelihood of artificially narrowing the CIs.
Data analysis
All data analyses were conducted using Stata Version 15 (StataCorp, 2017).
Effects model
We used random effects models as we expected heterogeneity in the distribution of the ORs. Heterogeneity was statically measured using the I 2 metric. Heterogeneity was expected for many reasons including differences in: the temporal relationship between PEs and mental disorders (concurrent or subsequent), the methodology used to investigate PEs (questionnaire v. interview), the age of the participants, the manner of selection into each study and differences in diagnostic criteria. Model selection based on the heterogeneity estimates has been deemed inappropriate as the assumptions of the fixed and random models differ (Borenstein et al., Reference Borenstein, Hedges, Higgins and Rothstein2010).
Analysis 1
Firstly, we investigated the association between childhood and adolescent PEs and both any mental disorder and any non-psychotic disorder. Random effects pooled ORs are reported with estimates of heterogeneity across studies. Funnel plots (see online Supplements C and D) and the Egger's regression test for publication asymmetry (Egger et al., Reference Egger, Smith, Schneider and Minder1997) were examined and trim and fill (Duval and Tweedie, Reference Duval and Tweedie2000) adjustments were applied. Secondly, we stratified the results by study design to examine the effects design had on the relationship between PEs and mental disorder (any and any non-psychotic) separately. Finally, post-hoc meta-regressions were conducted on several variables that could explain the between-study variance in the relationship between PEs and any mental disorder (this was not conducted for any non-psychotic disorder as too few studies were available). The independent variables in these univariate regressions were PE assessment type (interview or questionnaire), study design (cross-sectional or longitudinal), population size and follow-up time (longitudinal studies only). Bubble plots and non-descriptive results of this analysis are presented in online Supplement E.
Analysis 2
In the second analysis we investigated the association between childhood and adolescent PEs with each sub-category of mental disorder: psychotic disorders, affective disorders, anxiety disorders, behavioural disorders and substance-use disorders. Similar to analysis 1, we report the random effects pooled ORs and heterogeneity based on I 2. Again, we stratify by study design to investigate the relationship between PEs and the sub-categories of each mental disorder in separate cross-sectional and longitudinal analyses. Finally, for longitudinal studies we report the population attributable fraction (PAF) where this was calculable (psychotic disorders only).
Results
Study selection
Based on the search terms, we extracted 3092 studies from the three databases. In total, 2877 remained after removing duplicates. The titles and abstracts of all 2877 were reviewed for relevance, which resulted in the identification of 186 for full text screening. Of those, 117 were excluded because the studies did not use a child or adolescent community sample. Based on the inclusion and exclusion criteria, 14 of the remaining 69 studies met criteria for inclusion in the review, 13 of which also met inclusion criteria for meta-analysis. The specific reasons for exclusion are given in the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) flow diagram (Fig. 1).
Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) flow diagram for study inclusion.
Search yield
The search yielded 14 studies from 13 (n = 29 517) different community samples (Poulton et al., Reference Poulton, Caspi, Moffitt, Cannon, Murray and Harrington2000; Dhossche et al., Reference Dhossche, Ferdinand, Van Der Ende, Hofstra and Verhulst2002; Scott et al., Reference Scott, Martin, Bor, Sawyer, Clark and Mcgrath2009; McGrath et al., Reference Mcgrath, Welham, Scott, Varghese, Degenhardt, Hayatbakhsh, Alati, Williams, Bor and Najman2010; Dominguez et al., Reference Dominguez, Wichers, Lieb, Wittchen and Van Os2011; Kelleher et al., Reference Kelleher, Keeley, Corcoran, Lynch, Fitzpatrick, Devlin, Molloy, Roddy, Clarke, Harley, Arseneault, Wasserman, Carli, Sarchiapone, Hoven, Wasserman and Cannon2012b; Fisher et al., Reference Fisher, Caspi, Poulton, Meier, Houts, Harrington, Arseneault and Moffitt2013; Zammit et al., Reference Zammit, Kounali, Cannon, David, Gunnell, Heron, Jones, Lewis, Sullivan, Wolke and Lewis2013; Calkins et al., Reference Calkins, Moore, Merikangas, Burstein, Satterthwaite, Bilker, Ruparel, Chiavacci, Wolf, Mentch, Qiu, Connolly, Sleiman, Hakonarson, Gur and Gur2014; Adriaanse et al., Reference Adriaanse, Van Domburgh, Zwirs, Doreleijers and Veling2015; Jeppesen et al., Reference Jeppesen, Clemmensen, Munkholm, Rimvall, Rask, Jorgensen, Larsen, Petersen, Van Os and Skovgaard2015; Bechtold et al., Reference Bechtold, Hipwell, Lewis, Loeber and Pardini2016; Clemmensen et al., Reference Clemmensen, Van Os, Drukker, Munkholm, Rimvall, Vaever, Rask, Bartels-Velthuis, Skovgaard and Jeppesen2016; Cederlof et al., Reference Cederlöf, Kuja-Halkola, Larsson, Sjolander, Ostberg, Lundstrom, Kelleher and Lichtenstein2017). This included six cross-sectional studies from six different community samples and eight longitudinal reports from seven different community samples (average follow-up time of 10.5 years, range: 0.12–27 years). The characteristics of each study are presented in Table 1. Two community samples were represented in more than one investigation (the Copenhagen and the Dunedin cohorts). The two studies presenting data from the Copenhagen cohort examined different outcomes (see Table 1). We found an overlap between the outcomes in the two Dunedin cohort studies selected for review: Poulton et al. (Reference Poulton, Caspi, Moffitt, Cannon, Murray and Harrington2000) investigated the longitudinal relationship between PEs and any disorder, non-psychotic disorder, psychotic disorder, affective disorder and anxiety disorder while Fisher et al. (Reference Fisher, Caspi, Poulton, Meier, Houts, Harrington, Arseneault and Moffitt2013) investigated the longitudinal relationship between PEs and substance use disorder.
Prevalence of psychotic experiences
Data from 12 community samples (n = 29 365) were used to calculate the prevalence of PEs (prevalence estimates could not be calculated from Adriaanse et al., Reference Adriaanse, Van Domburgh, Zwirs, Doreleijers and Veling2015). The point prevalence (defined as the prevalence at time point 1 in all longitudinal studies) of children and adolescents reporting PEs was 9.83% (n = 2886). The prevalence in cross-sectional studies was 16.11% (k = 5; n = 1315) and the prevalence reported in longitudinal studies was 7.41% (k = 7; n = 1571).
There was a minor discrepancy in the prevalence of PEs between the methods of reporting. In questionnaire-based studies the pooled prevalence of PEs was 11.83% (n = 912; k = 5). In interview-based studies the pooled prevalence was 9.12% (n = 1974; k = 8).
Multiple reports of PEs
Two longitudinal studies reported PEs at multiple time points prior to the assessment of mental disorder (Dominguez et al., Reference Dominguez, Wichers, Lieb, Wittchen and Van Os2011; Bechtold et al., Reference Bechtold, Hipwell, Lewis, Loeber and Pardini2016). As so few studies (k = 2) reported on PEs at multiple time points, statistical analyses were not carried out to investigate the relationship between persistent PEs and mental disorders. Both of these studies investigated the relationship between PEs and subsequent psychotic disorder and both indicated a greater risk in those who repeatedly reported PEs.
Meta-analysis
Analysis 1: PEs and any mental disorder and any non-psychotic disorder
Twelve of the 13 samples were used to investigate the relationship between child and adolescent PEs and mental disorder. Adriaanse et al. (Reference Adriaanse, Van Domburgh, Zwirs, Doreleijers and Veling2015) were not included in this analysis because PEs were measured as continuous variable in their study. We found that PEs were associated with a three-fold increased odds of any mental disorder (OR 3.08, CI 2.26–4.21, k = 12, see Fig. 2). When the investigation was narrowed to any non-psychotic disorder, those who report PEs had a 2.8-fold increase in the odds of meeting criteria for a mental disorder than their peers (OR 2.82, CI 1.86–4.28, k = 8, see Fig. 3). Visual assessment of the funnel plots and Egger's regression test did not suggest an asymmetry in the published literature for any mental disorder or any non-psychotic disorder (any mental disorder: t = 0.66, p = 0.526, see online Supplement C; and any non-psychotic disorder: t = 0.98, p = 0.364, see online Supplement D). Statistical adjustment for one potentially missing study using a trim and fill method somewhat adjusted the OR for any non-psychotic disorder in those with PEs (adjusted OR 2.51, CI 1.60–3.92, p < 0.001). Significant between-study heterogeneity was evident for the relationship between PEs and both any mental disorder (I 2 = 58.7, p = 0.005) and any non-psychotic disorder (I 2 = 54.5, p = 0.031).
Forest plot of the relationship between child and adolescent PEs and any mental disorder.
Forest plot of the relationship between child and adolescent PEs and any non-psychotic disorder.
Study design stratification
When investigated separately, we found a three-fold increase in odds of any mental disorder in children and adolescents who reported PEs among both cross-sectional (k = 5) and longitudinal studies (k = 7) (see Fig. 2). Between-study heterogeneity was evident across cross-sectional studies (I 2 = 71.4, p = 0.007) and was somewhat suggested across longitudinal studies (I 2 = 49.2, p = 0.067). When limited to non-psychotic disorders, both cross-sectional (k = 5) and longitudinal (k = 3) studies indicated increased odds of any non-psychotic disorder in children and adolescents reporting PEs (see Fig. 3). However, the number of studies available was limited in both design methods and there was significant heterogeneity across the cross-sectional studies (I 2 = 66.6, p = 0.018).
Meta-regression analysis
Given the significant between-study heterogeneity, we investigated whether a number of variables were likely to influence the relationship between PEs and any mental disorder. These included PE assessment type, study design type, the total population of the study and follow-up time in longitudinal studies. None of these variables had a significant effect on the relationship between PEs and any mental disorder (see online Supplement E).
Analysis 2. PEs and sub-categories of mental disorders
There was a significant association between PEs and all sub-categories of mental disorder (see Table 2 and online Supplements F and G). These results were particularly prominent for psychotic disorders (OR 3.96, CI 2.03–7.73), affective disorders (OR 3.83, CI 2.26–6.49, for depressive disorders only see online Supplement H) and substance use disorders (OR 3.41, CI 2.03–5.74). For example, analysis of data from the five longitudinal studies investigated the relationship between PEs and psychotic disorders found an approximate four-fold increased risk of psychotic disorders in those with PEs. The PAF was calculable from these five studies and indicated that childhood and adolescent PEs accounted for 23.2% of psychotic disorders. Heterogeneity was evident in the analysis investigating psychotic disorders and affective disorders.
The pooled ORs and the heterogeneity assessments of the relationship between childhood and adolescent PEs and each category of mental disorder (overall association and stratified by study design type)
I 2, percentage of heterogeneity; CI, confidence interval; N/A, non-applicable; –, no study available.
Bold values denote a probability of p < 0.05.
Study design stratification
In studies using a cross-sectional study design, children and adolescents reporting PEs had an increased risk of affective and behavioural disorders. There was significant heterogeneity in the investigation of affective disorders. In those using longitudinal study designs, childhood and adolescent PEs were associated with over a three-fold increased risk of substance use and psychotic disorders (Table 2). However very few studies (k = 2) investigated the longitudinal relationship between child and adolescent PEs and subsequent affective disorders or anxiety disorders. No study included in this investigation had examined the longitudinal relationship between PEs and subsequent behavioural disorders.
Discussion
This is, to the best of our knowledge, the first systematic review of studies looking at risk of mental disorders in non-help-seeking children and adolescents who report PEs. Childhood and adolescent PEs were associated with increased odds of psychotic and affective, anxiety, behavioural and substance use disorders.
The prevalence of PEs in included studies was 9.3%, which is in keeping with meta-analytic estimates (Kelleher et al., Reference Kelleher, Connor, Clarke, Devlin, Harley and Cannon2012a; Maijer et al., Reference Maijer, Begemann, Palmen, Leucht and Sommer2018). It also is in keeping with the observation that PEs are more prevalent in early life than in adulthood (Linscott and van Os, Reference Linscott and van Os2013). Childhood and adolescent PEs were associated with a three-fold increased odds of having any mental disorder or any non-psychotic disorder in both cross-sectional and longitudinal studies. Roughly a quarter of psychotic disorders were attributable to PEs in childhood or adolescence.
These results align with the suggestion that childhood PEs are an early stage pluripotent psychiatric marker (McGorry et al., Reference Mcgorry, Hartmann, Spooner and Nelson2018) and may therefore be considered as an early trans-diagnostic marker for vulnerability to subsequent mental disorder. This theory is empirically supported by follow-up research using a sub-sample of the Philadelphia cohort, which found that those with persistent PEs had increased rates of psychotic, affective and behavioural disorders while those with transient PEs had an increased rate of subsequent depressive disorders (Calkins et al., Reference Calkins, Moore, Satterthwaite, Wolf, Turetsky, Roalf, Merikangas, Ruparel, Kohler, Gur and Gur2017). All three of their PE groups also had higher treatment history prevalence than controls at follow-up. Similarly, Fisher et al. (Reference Fisher, Caspi, Poulton, Meier, Houts, Harrington, Arseneault and Moffitt2013) found that, by age 38, 93.3% of those who reported PEs in childhood had met criteria for a mental disorder at some stage of their life. These results suggest that PEs may be a useful marker for identifying those at risk of subsequent mental disorder. Our own research has highlighted that childhood PEs are not just a marker for subsequent risk but including PEs in assessments actually improves the prediction of subsequent psychopathology over and above the effects of a history of mental disorder, childhood functioning and traumatic experiences (Healy et al., Reference Healy, Gordon, Coughlan, Clarke, Kelleher and Cannon2018). While the meta-analysis results from this investigation support the theory that those who report childhood and adolescent PEs have an increased risk of subsequent mental disorders (any and any non-psychotic disorder), there were very few longitudinal studies investigating the relationships between PEs and specific categories of non-psychotic disorders (Poulton et al., Reference Poulton, Caspi, Moffitt, Cannon, Murray and Harrington2000; Dhossche et al., Reference Dhossche, Ferdinand, Van Der Ende, Hofstra and Verhulst2002; Fisher et al., Reference Fisher, Caspi, Poulton, Meier, Houts, Harrington, Arseneault and Moffitt2013; Cederlof et al., Reference Cederlöf, Kuja-Halkola, Larsson, Sjolander, Ostberg, Lundstrom, Kelleher and Lichtenstein2017). More research, specifically targeting the relationship between childhood PEs and subsequent non-psychotic disorder is therefore warranted.
In addition to the longitudinal outcomes, results from cross-sectional study design provided converging evidence suggesting that those who report childhood and adolescent PEs are more likely to have a co-occurring non-psychotic disorder. The synthesis of the cross-sectional literature provides evidence that childhood and adolescent PEs are a potential feature of non-psychotic disorders. Research over the last two decades has challenged the homotypic nature of these phenomena, as those who report PEs have increased rates of a variety of different disorders (Calkins et al., Reference Calkins, Moore, Merikangas, Burstein, Satterthwaite, Bilker, Ruparel, Chiavacci, Wolf, Mentch, Qiu, Connolly, Sleiman, Hakonarson, Gur and Gur2014). However, similar to longitudinal research, the number of studies investigating the relationship between these phenomena and mental disorders is still relatively limited and subsequent research is necessary to fully examine the prevalence of each sub-category of mental disorder and PEs.
Heterogeneity
Most analyses revealed heterogeneity in the effects reported across studies. This was expected, given the variability between the studies in design characteristic such as PEs assessment type and follow-up time. These characteristics may affect the relationship between PEs and mental disorder. To investigate this, we ran four univariate meta-regressions (online Supplement E). None of the variables we investigated had an effect on the relationship between PEs and any mental disorder. It is possible that other study or sample characteristics could have influenced this relationship. Such characteristics might include participant demographic characteristics or cross-study cultural differences, differences in diagnostic instrument or the interactive effects of a number of features. Additionally, the number of studies available precluded an investigation of how study characteristics affect the relationship between PEs and specific mental disorders.
Strengths and limitations
Strengths of the current study include investigation of both longitudinal and cross-sectional studies. A limitation is that some of the cross-sectional studies asked about lifetime (not current or recent) PEs. Interestingly, however, previous research has shown that, even when asked about lifetime experiences, most young people who report PEs have experienced these symptoms within the last year (Kelleher et al., Reference Kelleher, Connor, Clarke, Devlin, Harley and Cannon2012a). The studies examined were restricted to published reports within peer-reviewed journals adding to the credibility of the findings; however, this also leaves open the possibility of publication bias. However, visual assessment of funnel plots and statistical assessment using Egger's regression test for the main analysis suggests that there is minimal asymmetry in the overall investigation. It was noted that there are a number of studies that examine the relationship between PEs and psychopathology using non-diagnostic questionnaires, such as the Strengths and Difficulties Questionnaire (Goodman, Reference Goodman2001). While this investigation was restricted to the relationship between childhood and adolescent PEs and clinically defined mental disorder, we acknowledge that there is body of literature using these methods (Laurens et al., Reference Laurens, West, Murray and Hodgins2008; Bartels-Velthuis et al., Reference Bartels-Velthuis, Jenner, Van De Willige, Van Os and Wiersma2010; Dolphin et al., Reference Dolphin, Dooley and Fitzgerald2015; Bartels-Velthuis et al., Reference Bartels-Velthuis, Wigman, Jenner, Bruggeman and Van Os2016). The majority of these studies have indicated an increased risk of internalising and externalising behavioural problems in those who report PEs. This, coupled with the results of the current study, suggests converging evidence across assessment tools in the relationship between childhood and adolescent PEs and psychopathology. However this remains to be confirmed. Only two of the studies in this investigation had examined PEs at multiple time points (Dominguez et al., Reference Dominguez, Wichers, Lieb, Wittchen and Van Os2011; Bechtold et al., Reference Bechtold, Hipwell, Lewis, Loeber and Pardini2016). This limited our ability to meaningfully assess the relationship between persistent PEs and mental disorder. It has been reported, using non-diagnostic questionnaires, that children with persistent PEs have an elevated risk of internalising and externalising problems relative to transient PEs and healthy participants (Downs et al., Reference Downs, Cullen, Barragan and Laurens2013). Future research should investigate the relationship between persistent PEs and common mental disorder using diagnostic clinical assessments.
Conclusion
Children who report PEs are at increased risk of psychotic, affective, anxiety, behavioural and substance use disorders. Further research is necessary to understand why some young people with PEs go on to develop psychotic disorders while other young people with PEs go on to develop, for example, affective disorders (or, indeed, some young people with PEs do not develop any mental disorder at all).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291719000485
Author ORCIDs
Colm Healy, 0000-0001-7974-1861
Acknowledgements
This research was funded by a European Research Council Consolidator Award to M Cannon (Grant Code 724809 iHEAR) and a Health Research Board Award (HRA-PHR-2015-1323).
Conflict of interest
None.
