Aim: To explore whether psychological and psychopathological factors mediate the association between childhood adversity and psychosis-like symptoms in adolescence.

Background:

An increasing body of research has demonstrated an association between adverse childhood experiences and psychotic disorders (e.g., Bebbington et al., 2004; Fisher et al., 2010; Janssen et al., 2004; Shevlin et al., 2007a). As psychosis is considered to be a quantitative continuum from normality through attenuated psychotic symptoms to full clinical disorder (Chapman & Chapman, 1980; van Os et al., 1999), an underlying aetiological continuum is also assumed to exist (Johns & van Os, 2001; Myin-Germeys et al., 2003). Indeed, childhood adversity has also been linked to sub-clinical expressions of psychosis (e.g., Freeman & Fowler, 2009; Gracie et al., 2007; Lataster et al., 2006; Morgan et al., 2009; Nishida et al., 2008; Shevlin et al., 2007b; Thompson et al., 2009), including amongst the ALSPAC cohort (Schreier et al., 2009). These findings suggest that investigating samples with psychosis-like symptoms (PLIKS) may provide useful insights into clinical psychosis. However, little is known about the mechanisms underlying the adversity - psychosis association.

DIRECT: Traumatic events in childhood could directly increase the risk of developing PLIKS, perceptual aberrations or reality impairment (Allen et al., 1997; Honig et al., 1998). Indeed, genetic studies have demonstrated a role for environmental factors in psychosis proneness (Linney et al., 2003), and Cougnard et al. (2007) found that childhood abuse was associated with the initial appearance of PLIKS. These sub-clinical hallucinations or delusions could be considered to be traumatic reactions to severe adversity whose content is directly reminiscent of the adversity (Read et al., 2005). Such direct connections have been reported in between 3% - 50% of individuals studied (Gracie et al., 2007; Hardy et al., 2005; Read & Argyle, 1999). Therefore, a direct pathway between childhood adversity and PLIKS may be operating but perhaps only for a small minority of individuals.

INDIRECT: A range of indirect pathways have been suggested, including psychological mechanisms (low self-esteem, external locus of control, insecure attachment style) and the development of other mental health problems (depression, anxiety, PTSD). For instance, Gracie et al. (2007) showed that negative perceptions of the self and others partially mediated associations between lifetime trauma and PLIKS in a general population sample, but contradictory results were reported by Freeman and Fowler (2009). Insecure attachment styles are also prevalent amongst individuals who have experienced childhood abuse (Alexander et al., 1998) and patients with psychotic disorders (Dozier et al., 1991), especially those with a history of abuse (Tait et al., 2004), suggesting another possible indirect pathway. Furthermore, some studies have reported that the association between childhood abuse and psychosis is attenuated if depression is included as a confounder (Bebbington et al., 2004; Shevlin et al., 2007a), whilst Freeman and Fowler (2009) recently reported that anxiety mediated the association with paranoid delusions amongst abused individuals. Psychotic symptoms have also been hypothesised to emerge after the experience of post-traumatic stress disorder in those exposed to childhood adversity (PTSD; Braakman et al., 2009; Pepper & Agius, 2009). A range of other potential mechanisms may also be operating (e.g., stunted brain development, substance misuse, epigentic changes, gene x environment interactions, dysregulation of the HPA axis, hostile attributions, etc.) but these are beyond the scope of the proposed analysis.

Therefore, further exploration of the pathway(s) between childhood adversity and psychosis is required to test these different mechanisms. Moreover, the handful of studies that have explored this issue to date (e.g., Fowler & Freeman, 2009; Gracie et al., 2007; Whitfield et al., 2005) have all relied on cross-sectional samples, thus preventing temporal relationships from being accurately established.

Hypotheses:

(i) reported exposure to individual and multiple childhood adversities (peer victimisation, sexual abuse, domestic violence, physical abuse, parental loss, emotional abuse, harsh discipline and neglect) will be associated with separation anxiety, low self-esteem, an external locus of control, and higher levels of depression, anxiety and PTSD symptoms;

(ii) in turn these factors will predict greater endorsement of PLIKS at ~13 years; and

(iii) both psychological and psychopathological factors will explain a significant proportion of the variance in the pathway between the adverse childhood experiences and PLIKS with a direct pathway accounting for only a small proportion of the variance.

Analysis:

Single variables for each form of adversity will be constructed based on the presence of that adversity at any of the time points assessments were conducted versus absence across all time points. The mediation of the childhood adversity - PLIKS association by psychological and psychopathological factors will be investigated using a structural equation modelling framework in Mplus. The major advantage of this approach over path analysis is that measurement error arising from repeated assessments of the variables and missing data can be dealt with directly. Specifically two models will tested for each form of childhood adversity:

(i) full mediation of the relationship between childhood adversity and presence of suspected or definite PLIKS by separation anxiety, self-esteem, locus of control, depression, anxiety and PTSD symptoms; and

(ii) partial mediation such that a direct pathway from adversity to PLIKS is also present.

The weighted least-squares estimator will be used to analyse the models as some of the variables are likely to be non-normally distributed and it is robust to data missing at random. The two models will be compared using a chi-square test to determine whether including the direct pathway substantially improves the fully mediated model. The proportion of the variance accounted for by the direct and indirect pathways will also be reported to assess the degree of prediction. Potential confounders (age, gender, ethnicity, social class, IQ, family history of schizophrenia or depression and general family adversity) will be controlled for in the analysis.