Abstract The association of early life adversities and psychosis symptoms is well documented in clinical populations; however, whether this relationship also extends into subclinical psychosis remains unclear. In particular, are early life adversities associated with increased levels of schizotypal personality traits in non-clinical samples? We conducted a systematic review and meta-analysis of associations between early life adversities and psychometrically defined schizotypal traits in non-clinical samples. The review followed PRISMA guidelines. The search using PubMed, Web of Science and EBSCO databases identified 1,609 articles in total. Twenty-five studies (N = 15,253 participants) met eligibility criteria for the review. An assessment of study quality showed that fewer than half of all studies were rated as methodologically robust. Meta-analyses showed that all forms of childhood abuse (emotional, physical and sexual) and neglect (emotional and physical) were significantly associated with psychometric schizotypy. The association of schizotypy traits with childhood emotional abuse (r = .33: 95%CI .30 to .37) was significantly larger than for all other form of abuse or neglect. Meta-regression analyses showed that the physical abuse-schizotypy relationship was stronger in samples with more women participants; and the sexual abuse-schizotypy relationship was stronger in younger samples. The current review identifies a dose-response relationship between all forms of abuse/neglect and schizotypy scores in non-clinical samples; however, a stronger association emerged for emotional abuse. More research is required to address the relationship of trauma types and specific symptom types. Future research should also address the under-representation of men.
Citation: Toutountzidis D, Gale TM, Irvine K, Sharma S, Laws KR (2022) Childhood trauma and schizotypy in non-clinical samples: A systematic review and meta-analysis. PLoS ONE 17(6): e0270494. https://doi.org/10.1371/journal.pone.0270494 Editor: Sarah Hope Lincoln, Case Western Reserve University, UNITED STATES Received: October 6, 2021; Accepted: June 10, 2022; Published: June 29, 2022 Copyright: © 2022 Toutountzidis et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All relevant data are within the manuscript (the relevant correlations and sample sizes are in the paper itself). As this is a systematic review and meta-analysis, all the data are derived from papers available in the public domain. Funding: Part of the review was funded by a PhD bursary from the University of Hertfordshire to DT. The funder did not have any involvement in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication. Competing interests: The authors have declared that no competing interests exist.
Introduction Since the start of the last century, psychotic symptomatology has been viewed as ranging from the extremely severe through to milder clinical and sub-clinical presentations [1, 2]. Both Bleuler and Kraepelin observed that sub-clinical psychosis symptoms exist prior to onset of schizophrenia and may even be observed in biological relatives of patients. The modern expression of this spectrum of experience is embodied in the continuum hypothesis of psychosis, which suggests that experiences such as hearing voices (auditory hallucinations) or delusional ideation occur in less distressing forms within the non-clinical population [3–5]. These experiences are not limited to the positive aspects of symptomatology (i.e., presence of odd feelings or behaviours), but extend also to negative symptoms (i.e., absence or lack of normal mental function) in non-clinical subjects [6]. Central to debates concerning the continuity of clinical, sub-clinical and healthy experiences has been the concept of schizotypy, which typically encompasses a broadly-defined phenotype of schizophrenia-like traits and behaviours that exist in the wider general population. The term “schizotypy” was first coined by Rado [7] and later adopted by Meehl [8], who viewed it as a form of latent personality organisation resulting from a specific biological/genetic predisposition. Later work by Claridge [9] clarified the distinction between models that view schizotypy as dimensional or, as in Meehl’s case, quasi-dimensional. Fully dimensional models would be those which suggest that schizotypy exists within both the healthy (and may even confer some advantages) and pathological range [10]. The continuum between healthy and pathological also encompasses other concepts, which have developed to describe psychosis-like experiences [PLEs; 11]. Schizotypy and PLEs refer to similar phenomena and both terms are frequently used interchangeably [12, 13] and reported as conceptually overlapping [14]. They might however, be distinguished in terms of the fact that PLEs are state-based and linked to symptoms (typically positive), while schizotypy is trait-based. Schizotypal traits are often classified into three domains that correspond to the key symptom areas of schizophrenia: the positive domain typically incorporates traits that relate to anomalies of cognition (e.g., paranoid ideation, ideas of reference); the negative domain includes interpersonal, emotional and deficit traits (e.g., anhedonia, no close friends); and the disorganisation domain includes traits related to disruptions in the ability to organise and express thoughts and behaviour (e.g. odd behaviour and odd speech). The extent to which various conceptions of schizotypy are related to the risk of schizophrenia is a matter of debate [15]. Amongst many risk factors, adverse childhood experiences have been linked to the experiences of psychosis [16], which might include various forms of physical abuse, verbal abuse, sexual abuse, physical neglect, and emotional neglect. Adversities in childhood (occurring before the age of 18) are common, with approximately 40% of the general adult population reporting at least one type of adverse experience (e.g., parental mental illness, domestic violence, physical, emotional, and sexual abuse, neglect) before the age of 18 [17]. Much of what is known about the impact of childhood trauma on adult mental health is gleaned from studies of clinical populations. Less is known about the impact of early life events on people who do not go on to receive a clinical psychosis diagnosis. By examining non-clinical samples expressing high and low schizotypal traits scores, we might shed light on the differences in types and/or levels of trauma between clinical and non-clinical populations–and potentially increase understanding of the mechanisms through which trauma might lead to psychosis. One previous systematic review [18] of 25 studies reported that various types of childhood trauma were associated with schizotypy. The relationship of overall trauma with schizotypy revealed odds ratios across studies that ranged from 2.01 up to 4.15. The authors reported that associations exist for all trauma types, but concluded that it seemed stronger for emotional abuse. While they reported that trauma and schizotypy (especially positive traits) appeared to show a dose-response relationship, this relationship was based on a few individual studies rather than quantified and pooled across studies. Their review, which combined studies of both clinically-diagnosed schizotypal personality disorder (SPD) and those scoring high on self-rating schizotypy scales classified the majority of studies (14/25) as low in quality. Since the review by Velikonja and colleagues [18], research in the relationship between childhood trauma and schizotypy in non-clinical samples has grown considerably, with several new large studies published. The aim of the current study is to systematically review the available published literature reporting associations between childhood traumatic experiences and psychometric schizotypal traits, as well as PLEs, in non-clinical samples alone. In particular, we will focus on the relationship between specific childhood trauma including: physical, emotional, and sexual abuse, as well as neglect, and psychometric schizotypy in non-clinical samples. Furthermore, we will conduct the first meta-analysis to quantify the reported associations between trauma types and schizotypy. Hereafter, all references to emotional, physical and sexual abuse, as well as physical and emotional neglect are to abuse and neglect experienced in childhood and up to the age of 18.
Method Search strategy Literature searches were conducted in three major databases (PubMed, EBSCO, and Web of Science) using the following sets of search terms: Child* abuse OR physical abuse OR sexual abuse OR psychological abuse OR emotional abuse OR child* trauma* OR child* advers* OR maltreat* OR bully* OR bullied OR stress* events OR early trauma* OR emotional neglect OR physical neglect AND Schizoty* OR psychosis-like OR psychotic-like OR magical ideation OR suspiciousness OR delusional ideation OR odd belie* OR eccentric behavi* OR odd speech OR constricted affect OR unusual perceptual experiences OR ideas of reference OR paranoid ideation Publications between January 1980 and February 2021 were covered in all searches. The start-point was selected because the first investigations of the links between childhood trauma and psychosis outcomes appeared in the late 1980s, and because previous reviews focussing on clinical groups [e.g., 18, 19] failed to identify any relevant studies that took place prior to 1980. The Reference Manager EndNote X7 was used to manage the identified articles (n = 2144) and extract duplicates (n = 535). A total of 1609 articles were identified for title screening and 524 articles were found to be potentially relevant. Following abstract screening, 133 articles were identified for full-text screening and subsequently 25 articles were included in the present review (see S1 Appendix, for screened-excluded studies). All articles were independently screened by two authors (DT and KRL). Two studies [i.e., 20, 21] employed the same sample. Only findings from Gong, Wang (20) are presented in this review, as they provided a more detailed account than Liu, Gong (21) on the links between different schizotypy dimensions and various types of childhood trauma. Wigman, van Winkel [22] and Kramer, Simons [23] used largely overlapping samples, and thus we report only findings from the latter, based on their relevance to the current review, as they provided results between childhood adversity and three different schizotypal outcomes–instead of one result between a schizotypal persistent group and overall childhood trauma that provided by the former. Sheinbaum, Racioppi [24] and Gong, Wang [25] used subsamples from Sheinbaum, Kwapil [26] and Gong, Wang [20] respectively, and thus only results from the latter studies were included in the current review. Although not preregistered, the review followed PRISMA guidelines for reporting systematic reviews and meta-analyses. A PRISMA 2020 compliant flow diagram tool [27] was used to provide record numbers of each stage and reasons for reports exclusion in the full review stage (Fig 1) and the checklist documents if and where relevant information may be located in the paper (see S2 Appendix, for the PRISMA 2020 Checklist). PPT PowerPoint slide
PNG larger image
TIFF original image Download: Fig 1. PRISMA 2020 compliant flow diagram of each stage and details of excluded reports in full review. STs = Schizotypal traits, UHR = Ultra-high-risk. https://doi.org/10.1371/journal.pone.0270494.g001 Eligibility criteria Articles had to meet the following criteria to be included in the review: (a) be an original published research paper; (b) be written in English; (c) include a measure of childhood trauma for events before the age of 18 years (excluding peer victimisation; i.e., bullying; to assess for trauma where the perpetrator is an adult), (d) test for associations with schizotypy using any standardised measure of either single symptom/trait or multidimensional schizotypal personality traits in adults (aged over 18); and (e) include general population samples and not clinical psychosis or personality disorder cases (though not necessarily screened for other personal or family mental health problems). Screening and eligibility assessment were performed independently by two reviewers (DT and KRL); and any disagreements were resolved by consensus. Quality assessment tool Study quality was assessed using the quality assessment tool for childhood trauma and schizotypal traits research designed by Velikonja, Fisher [18]. Each study was rated on: sampling methods and sample sizes, selection and utilisation of standardised measures, and the assessment of confounding variables—scores from 0 to 2 were assigned on each item testing for the indicators. The maximum possible score for each study was 14 and the scorings (included in Table 3) were completed independently by two raters (DT and KRL), without disagreement. Corrected Covered Area (CCA) We calculated the corrected covered area [CCA; 28] to determine the degree of overlap for primary studies included in the current review and the one previous systematic review of Velikonja, Fisher [18]. Pieper, Antoine [28] suggest that CCA scores between 0% and 5% represent ‘slight overlap’, 6%–10% ‘moderate overlap’, 11%–15% ‘high overlap’, and scores greater than 15% are considered to represent ‘very high overlap’. CCA was calculated at 6.82% and so, represents slight-moderate overlap. This undoubtedly reflects the fact that 17 non-clinical studies have been published since Velikonja, Fisher [18] examined clinical and non-clinical samples. Meta-analysis of correlations The statistical analysis was conducted using Comprehensive Meta-Analysis (version 2.2). All relevant data for calculating effect sizes (correlation values, sample sizes) and moderator variables (age, percentage of women participants and year of publication) were extracted independently by two authors (DT and KRL). Effect sizes were calculated for the correlations between schizotypy scores and each of the following: emotional abuse, physical abuse, sexual abuse, emotional neglect and physical neglect. All correlation coefficients were transformed to Fisher’s z [29]. Synthesis of individual effect sizes to summary effect sizes was completed by conducting random effects meta-analyses. Results were then converted back from Fisher’s z to r for interpretation. Heterogeneity and variance among effect sizes of studies were assessed using the Q statistic and the I2 statistic.
Discussion This systematic review identified 25 studies examining the association of childhood trauma and schizotypy in 15,253 non-clinical individuals. The current meta-analyses are the first to estimate the pooled effect size for the association between various forms of childhood abuse/neglect and schizotypy in subclinical samples and the emerging importance of this area of research is underscored by 17 of the 25 (68%) studies reviewed here being published since the sole previous systematic review by Velikonja, Fisher [18]. Schizotypy scores were significantly associated with all forms of abuse and neglect in a manner that is consistent with a dose-response interpretation. Almost all studies and their analyses report significant positive correlations between schizotypy and various forms of abuse and neglect (emotional abuse 11/11; physical abuse 10/13; sexual abuse 11/12; emotional neglect 7/7; and physical neglect 9/9). The current meta-analyses are however the first to quantify and compare the relative degree of association between these different forms of childhood trauma and schizotypy. Although schizotypy was significantly associated with all forms of abuse and neglect, only the associations with emotional abuse exceed the small range with an r = .33 (equivalent to an OR = 3.55). The association of schizotypal traits with emotional abuse was in the moderate range and crucially, significantly larger than the association for all other forms of abuse and neglect (where effect sizes did not differ significantly from each other). Our assessment indicated that the mean study quality ranking was marginally lower than the suggested cut-off for individual studies to be identified as methodologically robust [see 18]. Hence, most studies failed to meet the suggested cut-off. We also note that study quality has remained relatively low across time, with no signs of any recent increase. Despite this, study quality was not a significant moderator of the relationship between schizotypy and any of form of abuse or neglect. It is of course possible that the quality measure used fails to adequately capture some potentially relevant variability in study quality, although studies did show variability in quality scores (range 4–10 on a scale of 0–14). Certain specific areas of methodological weakness might however be identified. In particular, the vast majority of studies (21/25) failed to control for, assess or comment upon possible confounds–such as controlling for demographic variables, family history, current levels of anxiety or depression and so on. The studies were well-powered–taking the smallest effect size being 0.2 (for physical abuse) would require 150 participants to detect at 80% power. All studies included in the meta-analyses except two [62, 82] had sufficient sample size (with a median in excess of 400) to detect even the smallest effect size reported here. This power analysis would of course apply to detecting a simple correlation; however, most studies are assessing a range of associations between various types of trauma and various schizotypal traits and any power assessment needs to accommodate multiple testing. The current meta-analysis and effect sizes, hopefully provide some basis for more specific hypothesis testing in future studies. The stronger association between emotional abuse and schizotypy suggests a more consistent and pervasive link than appears to occur with either sexual or physical abuse. The latter forms of abuse differ from emotional abuse insofar as they might be linked more to individual abusive acts. Emotional abuse is also likely to accompany these other forms of abuse and neglect, and may heighten the effect of other abusive acts [130, 131]. It is also feasible that emotional abuse has a more pervasive influence partly because it might persist undetected for longer. Related to the potentially pervasive and interactive influence of emotional abuse, one limitation of current studies is that they rarely assess the independent contributions of specific abuse/neglect types and so effect size estimates may be non-independent. Two studies [63, 69] have however attempted to look at the independent impact of trauma types within a regression framework. Goodall, Rush [63] looked at emotional abuse, sexual abuse, physical abuse, emotional neglect and physical neglect as predictors of schizotypy scores (in 283 non-clinical participants) and found that only emotional abuse remained significant when all predictors were entered together. More recently, Toutountzidis, Gale [69] used a similar regression-based approach finding that schizotypy scores were significantly predicted by both emotional and physical abuse, although only emotional abuse remained a significant predictor when gender was entered into the model. Our meta-analyses identified substantial levels of heterogeneity for all effect sizes, with I2 values ranging from 68–87%. We explored this heterogeneity using meta-regression by focusing on the key participant variables of age and gender as well as study quality and year of publication. Velikonja, Fisher [18] speculated that exploring age and gender differences as possible mediators of the childhood trauma-schizotypy relationship might aid our understanding of the aetiology of psychotic symptoms and psychotic disorders. We found that age was a significant predictor of the sexual abuse-schizotypy association, with stronger associations emerging in younger samples. Age also significantly moderated the emotional neglect-schizotypy relationship, but associations were stronger in older samples. The opposing impact of age in these two types of trauma raises the possibility that additional variables are likely to be relevant. For example, age may impact how people process memories regarding maltreatment in childhood and this may depend upon the type of trauma experienced. It is also possible that temporal changes in social norms are relevant, such that earlier adverse events are not viewed as negatively as they might be now or the likelihood of such events may even have reduced across time. Turning to gender, the physical abuse-schizotypy link was larger in samples containing more women participants. Given the importance of gender in studies of trauma in clinical psychosis groups, surprising we identified only two non-clinical studies reporting associations separately for men and for women [61, 69]. Reviews have documented gender differences in the prevalence of different traumas reported in psychosis patients [132, 133]; however, the findings are not always consistent and crucially, we should not assume that trauma and schizotypy (or indeed psychosis) shows the same dose-response across men and women. For example, Toutountzidis, Gale (69) found that while men reported significantly more physical trauma experienced in childhood than women, it failed to correlate with schizotypy in men, but did in women. As noted above, the moderating role of gender was also borne out by our meta-regression analyses showing that the physical abuse-schizotypy link was stronger in studies with more women as participants. Turning to the role of age, the link between sexual abuse and schizotypy scores was larger in younger samples and several possible explanations exist. For example, it may be that sexual abuse differentially impacts schizotypy in those who are younger. Another possibility is that older samples are less willing to report sexual abuse (and/or that younger samples feel safer to disclose on this topic) and this the apparently stronger association in the younger is a confounded by willingness to reveal. The sampling is however somewhat skewed in terms of both age (with a median age of 25) and gender. The ratio of women to men participants in the studies reviewed here is around 2:1; and all studies except one [82] had a majority of women participants. The broadening of the focus onto older and more men as participants is something future studies should address. Although the overwhelming majority of studies report significant associations between early childhood trauma and schizotypy, a key unanswered question centres on whether any specific adversities are more reliably associated with any specific trait domains or symptoms. This review identifies considerable variation in the types of measures used, but crucially also at the level of analysis. Some have calculated a total score for early life adversities [70] while more recent studies have tended to separately assess different types of adversity (e.g., physical, emotional, sexual abuse and neglect: e.g., [69, 81, 86]). Similarly, some studies [64, 66, 67, 88] assessed several schizotypy dimensions (e.g., cognitive-perceptual, interpersonal, disorganised), some have broken these domains down further into multiple trait subscales (e.g., 8 subscales: [64]; 9 subscales: [69]), subscales and total scores [e.g., 88] while others have assessed total schizotypy scores only [e.g., 61]. Although variability in the level of analysis is likely to contribute to the high heterogeneity of effect sizes, the data did not permit looking at this in a sub-group analysis. Nonetheless, a simple vote-count suggests that certain relationships seem promising with some invariable associations emerging. The most consistent association being reported for childhood emotional abuse both with negative schizotypal traits (e.g., anhedonia, no close friends) (20/20 associations) and with total schizotypy scores (10/10); for physical/emotional neglect (6/6 studies) and the disorganised domain (e.g. odd behaviour and odd speech); and for total trauma scores (11/11 studies) with the positive domain (e.g., paranoid ideation, ideas of reference). It is notable that childhood sexual abuse showed inconsistent associations with the various schizotypy domains—being significant in 11/32 assessments for the positive domain; 3/23 for the negative domain; and 4/8 for disorganisation—but did show more reliable association with total schizotypy scores (10/14). We note however that vote-counts do not accommodate sample size differences across studies, statistical power, or the size of effects in studies and not an alternative to meta-analysis [134]. Unfortunately, it was not possible to derive sufficiently consistent levels and types of data to meta-analyse more specific associations between traumas and specific domains or even symptoms. Little work has thus far focused at the symptom-specific level of analysis. For example, in the case of voice-hearing, only one study [66] has thus far examined the relationship with trauma. Looking at a student sample, Cole and colleagues [66] found significant links between self-reported auditory hallucinations (LSHS) and all types of childhood traumas (sexual, physical, and emotional abuse, and neglect); and for self-reported delusional ideation (PDI-21) with all traumas apart from sexual abuse. This lack of attention is surprising given that so many studies of clinical samples with psychosis have reported links between sexual trauma and voice-hearing and delusions (for a recent meta-analysis, see [135]. The links between trauma and specific symptoms clearly merits further research in non-clinical samples.
Strengths and limitations Although a major advantage of this review is the large number of participants and studies, the variety of measurement approaches meant that meta-analyses were limited to broad-brush approaches. Different studies of course have different aims and employ different measurement scales to assess both childhood trauma and schizotypy. In the case of measuring the latter, studies also vary considerably in whether they report total scores, domain (e.g., positive, negative, disorganised) subscale scores, or indeed, employ measures focussed specifically on symptoms (e.g., hallucinations or delusions). In general, studies assessing general population samples have not made a concerted effort to establish the parallels of the same trauma-symptom links highlighted in the clinical literature. For example, clinical studies have tended to emphasise the relationships of trauma types with specific symptoms such as delusions and hallucinations [135]. Future research would benefit from focusing on how childhood trauma might impact specific trait clusters/domains and symptoms in the healthy population. This would then make it more possible to look not just at more specific links, but to make more comparisons with the existing clinical literature. Turning to other measurement issues, ultimately the strength of reported relationships between trauma and schizotypy depend upon the reliabilities of the measures themselves. Although questions have arisen about the reliability of the self-reporting of childhood adversities in the context of clinical psychosis [e.g., 136], retrospective reports by psychosis patients are stable over time, not especially influenced by current psychopathology, consistent across different assessment instruments and correspond with clinical case notes [e.g., 137]. The most common measure of childhood trauma employed was the Childhood Trauma Questionnaire, [CTQ; 30, 31] being used in 15 of the 25 studies. A recent review [138] of 52 child abuse measurement instruments reported that the CTQ was the most thoroughly investigated and has the strongest levels of evidence overall and strong for: internal consistency, reliability, content validity, structural validity and convergent (hypothesis testing) validity. While we could find no evidence on the reliability of trauma reporting in schizotypy, recall is less likely to be affected by psychosis-like experiences or memory/cognitive deficits; although those with psychometrically defined schizotypy do show relatively subtle cognitive deficits on neuropsychological tests [139]; see also [140] for a recent meta-analysis. We also note that even when objective cognition is intact on testing, schizotypy has also been associated with subjective cognitive problems that may not be detectable on objective tests [141]. So, we cannot exclude the possibility that mild or subjective forms of cognitive impairment, including for memory and executive function, might impact memory recall and reliability in schizotypy. Six studies used a scale that tests for prodromal attenuated psychosis symptoms [65, 74, 81, 89, 110, 112]. Although these studies did not measure schizotypal traits per se, it is difficult to confidently assert whether they examine trait-like or state-like subclinical symptoms of psychosis for two key reasons. First, the cross-sectional nature of the studies does not permit a test of symptom/trait persistence. Second, all measures use similar questions to assess psychosis-like experiences in non-clinical samples (e.g., Have you had the sense that some person or force is around you, although you couldn’t see anyone?; both in PQ and SPQ). Third, only two studies using state measures were included in the meta-analyses [65, 74] and while these state measures revealed correlations that were a little smaller, they did not differ significantly from trait measure effect sizes. So, despite variability in the tests used and the level at which scoring occurs, the pattern of associations across studies and measures was robust. A clear limitation concerns the failure of most studies to meet the suggested cut-off for robust methodological quality; and that study quality has showed no signs of improving in recent years. While Velikonja, Fisher [18] noted study quality variability in their review, the 17 newer studies added here since that previous review do not evidence any recent increase in quality. We also note that individual quality items assessed here were not weighted and so, any simple comparisons across studies using summed total quality scores should be regarded with some caution [see 142–144]. In this context, we note that an area of specific methodological weakness has been the lack of control for potential confounding variables. For example, studies rarely screened for existing mental health problems, with only 4/25 (16%) screening for previous or current mental health disorders [65, 86, 88, 110] and one screened for possible neurological problems [71]. We also note the lack of diversity in sampling to date. Most studies have sampled exclusively from undergraduate students (16/25: 64%) and as already noted the ratio of women: men being approximately 2:1. The younger age (median 25) is perhaps understandable given the typical age of onset for psychosis-like symptoms–indeed, some studies have used this rationale for excluding participants outside of the 18–35 age range [e.g., 89]. The focus on more highly educated samples limits representativeness much more–certainly with comparison to the educational attainment of those with psychosis. In this context, we were also unable to assess whether ethnicity impacted effect sizes. Only one-in-three studies (9/25: [61, 62, 64, 66, 69, 71, 81, 89, 112]) provided any ethnicity breakdown. All apart from Alemany, Arias [71] were from the UK or USA and these identified a median proportion of white participants at 75%. The role of ethnicity is important here given evidence that the perceived discrimination in early life may be linked to later psychosis-like experiences [145–147]. This means that samples have often focused on largely young, highly-educated women and often lack information on ethnicity. Of those that we considered to consist of general population samples, some were clearly constrained by their sampling techniques. Amongst the remainder, Berenbaum, Valera (62) “sought respondents with unusual beliefs (e.g. belief in UFOs)” (p. 144); Fekih-Romdhane, Nsibi [82] assessed the unaffected siblings of those diagnosed with schizophrenia; Mongan, Shannon [112] recruited using the online paid participant pool from Mechanical Turk; Gaweda, Goritz [65] also from an online academic participant panel (called WiSO); Powers, Thomas [64] recruited from obstetric/gynaecological clinics; and Rössler, Hengartner [72] recruited a prospective study from amongst men conscripts and women on the electoral register in Zurich from mid 1970s. Finally, Toutountzidis, Gale [69] sampled the general public using an online survey and Startup [120] who used a university voluntary participant panel. We also note that only three studies screened out participants with any personal history of mental disorder [65, 86, 88], another for psychotic disorders [110] and one screened for a family history of suicide [86]. There were several studies that did not exclude family and personal history of mental health problems [e.g., 20, 63, 66, 67, 74] or whose sampling did not exclude those with self-declared mental health issues [69]. None screened and excluded for a family history of mental disorder. Indeed, family history was an inclusion criterion in one study i.e. the study of unaffected twins of those with schizophrenia [82]. Of course, the conclusion of the current systematic review and meta-analysis are prone to the limitations of correlational research generally. Aside from psychosis-like experiences occurring as a consequence of childhood adversity, the correlational nature of the research does mean that we should acknowledge alternative interpretations. In particular, trauma may result as a consequence of psychosis symptoms or in full-blown psychosis, as a result of involuntary treatment experiences [148, 149]. Despite this, given the subclinical levels of psychosis symptoms in adulthood, it seems less likely that trauma is induced by subclinical psychosis-like experiences earlier in childhood. Finally, as with psychosis, the links between early life trauma and schizotypy are undoubtedly complex and possible confounding factors cannot be eliminated. For example, it is known that higher levels of parental mental health problems are associated with various forms of abuse. Doidge, Higgins [150] assessed a range of possible child, parent and family risk factors for child maltreatment in a prospective 27-year population-based birth cohort of 2443 Australians. They found that parental mental health problems were associated with high intensity emotional abuse (OR = 2.06), low intensity emotional abuse (OR = 1.90), neglect (OR = 3.60), physical (OR = 2.31) and sexual abuse (OR = 3.33). It is also possible that children living with a parent who has mental health problems might view this as a form of trauma [see 151]. Future studies need to control for parental mental health status.
Conclusions This review shows that significant associations exist between childhood trauma and psychometrically defined schizotypy in the non-clinical population. While much research has focused on links between early life trauma and adult psychosis, including a possible causal role [19], the current findings show that trauma in childhood does not necessarily lead to the development of clinical presentations of psychosis. Such early life experiences may however impact the development and maintenance of potentially dysfunctional experiences and ways of thinking that remain sub-clinical. Despite the limitations of cross-sectional studies in terms of deriving causal inferences, studying childhood adversities that do not lead to clinical outcomes is as important as where they do–informing discussion around why some people are more ‘resilient’ than others.