Gender differences in the association between childhood physical and sexual abuse, social support and psychosis

Data were collected as part of the Aetiology and Ethnicity in Schizophrenia and Other Psychoses (AESOP) study, a multi-centre population-based incidence and case–control study of first episode psychosis. Cases were included in the study if they were between 16 and 64 years old, lived within specific catchment areas in south-east London and Nottingham, UK, had a first episode of affective or non-affective psychosis (International Classification of Diseases (ICD-10 codes F20–F29 and F30–F33) [17] between 1997 and 2000, and had no previous contact with secondary mental health services for psychosis. Exclusion criteria were evidence of organic psychosis, severe learning difficulties, and transient psychotic symptoms resulting from acute intoxication. Controls with no history of psychosis were randomly selected from the same geographical areas as the cases, and were also between the ages of 16–64 years. To exclude controls who may have experienced undiagnosed psychotic symptoms, all controls were screened using the Psychosis Screening Questionnaire (PSQ) [18].

The study was approved by the local research ethical committee at each study centre. Full details of the study methods have been described previously [19].

The Childhood Experience of Care and Abuse Questionnaire (CECA.Q) [20] was used to elicit information on experiences of childhood adversity before the age of 16. Specifically, we included in these analyses data on physical abuse from the main parent figures and sexual abuse by any person at least 5 years older than the victim. Screening questions relating to physical and sexual abuse were read out to all participants, and positive responses were followed up with more detailed questions. Researchers then used published guidelines to score the severity of the responses in a standardised manner [20]. The measure has satisfactory levels of concurrent validity and test–retest reliability within clinical and non-clinical populations [20, 21] as well as within patients with psychosis [22]. For data analysis, responses from the CECA.Q were dichotomised into severe, and no or non-severe experiences of sexual and physical abuse separately, using the most conservative published cut-points ( [20]; see [5] for further details). The group with none or non-severe instances of sexual and physical abuse is referred to as the ‘no abuse’ group within the categorical analyses presented in this paper.

The Significant Others Scale (SOS) [23] is a self-report questionnaire which measures perceived and ideal levels of practical and emotional support on a seven-point scale from up to seven significant people (e.g. father, partner). The SOS also elicits the number of significant others in participants’ social networks, as well as a discrepancy score (a measure of satisfaction) between ideal and perceived levels of emotional and practical support separately. The measure has been shown to have satisfactory concurrent and construct validity and test–retest reliability [24].

Participants completed the Medical Research Council (MRC) Socio-demographic Schedule [25] to collect data on age, gender, current employment status, education level, parental social class at participant’s birth, and ethnicity. This schedule was also used to collect data on the individual’s current living circumstances, current and long-term relationship status, number of confidants, and frequency of contact with friends and family. Finally, we used the Family Interview for Genetic Studies (FIGS) [26] to obtain data on history of mental health problems in parents (specifically psychosis, mania and depression).

Logistic regression was used to obtain the crude and adjusted odds ratios of the relationships between case–control status and both physical abuse and sexual abuse. This was done first with the sample unstratified, and then stratified by gender, to investigate the effects separately in men and women, with interaction terms fitted and likelihood ratio tests conducted to test the difference in odds ratios between men and women.

To assess whether social support modified the association between childhood adversity and psychosis, interaction terms were fitted to logistic models and likelihood ratio tests conducted to determine whether the interaction terms improved model fit. Subsequently, the same analyses were repeated, stratified by gender.

For all adjusted models, potential confounders included were sex (except in analyses stratified by gender), age (16–35 or 36–64 years), ethnicity (white British or other), study centre (London or Nottingham), education (no qualifications or any qualifications), current employment status [unemployed, economically inactive (i.e. students, house-persons), or employed], and any mental illness in parents. All analyses were carried out using STATA version 11 for Windows.

Cases included in the analyses were around two times more likely to report severe childhood physical abuse compared with controls (OR 2.05, 95 % CI 1.22–3.46, p = 0.006), and this association held after adjusting for age, ethnicity, study centre, parental history of mental illness, employment status and education level (adj. OR 1.89, 95 % CI 1.05–3.39, p = 0.034). When stratified by gender, female psychosis cases were approximately three times more likely to report severe physical abuse before 16 years of age compared with women in the control group (OR 3.31, 95 % CI 1.56–7.03, p < 0.001), though there was no evidence of an association between physical abuse and psychosis in men (OR 1.22, 95 % CI 0.58–2.54, p = 0.597). A likelihood ratio test of the difference in odds ratios revealed an interaction between gender and physical abuse (χ ² = 3.55, p = 0.059). This was only slightly attenuated following adjustment for potential confounders (χ ² = 3.16, p = 0.076).

In contrast, reports of severe childhood sexual abuse were only marginally more likely in cases compared with controls (OR 1.44, 95 % CI 0.82–2.53, p = 0.199) and this did not reach conventional levels of statistical significance. However, when we repeated these analyses stratifying by gender, the results indicated that whilst there was no association between childhood sexual abuse and psychosis in men (OR 0.76, 95 % CI 0.26–2.24, p = 0.613), women in the cases group were twice as likely to report experience of severe sexual abuse before the age of 16 than female controls (OR 2.21, 95 % CI 1.11–4.41, p = 0.021). A likelihood ratio test of the difference in odds ratios between men and women showed an interaction with sexual abuse (χ ² = 2.78, p = 0.096), which remained robust following adjustment for potential confounders (χ ² = 2.90, p = 0.089).

There was evidence that the impact of severe physical abuse in childhood on odds of psychosis varied by number of significant others in adulthood (see Table 3). The adjusted odds ratio for those who reported abuse and had 5 or more significant others in their social networks was 0.99 (95 % CI 0.42–2.36), compared with 3.24 (95 % CI 1.42–7.38) for those who reported abuse and had fewer than 5 significant others (χ² = 3.90, p = 0.048). More tentatively, the impact of physical abuse on odds of psychosis varied by perceived practical support, in that the odds ratio for those who reported physical abuse and high practical support was 1.18 (95 % CI 0.45–3.08), compared with 2.43 (95 % CI 1.11–5.33) for those who reported abuse and did not receive practical support. However, the magnitude of this variation did not reach conventional levels of statistical significance (χ ² = 1.35, p = 0.245). In contrast, there was no evidence that the impact of physical abuse varied by current perceived emotional support.

In addition, and again tentatively, the impact of severe sexual abuse on odds of psychosis varied by perceived emotional support, whereby the odds ratio for those who reported sexual abuse and high emotional support was 0.93 (95 % CI 0.32–2.70), compared with 2.12 (95 % CI 0.86–5.20) for those who reported abuse and did not perceive themselves as having sufficient emotional support, though the magnitude of this variation did not reach conventional levels of statistical significance (χ ² = 1.37, p = 0.242). In contrast, there was no evidence that the impact of sexual abuse varied by current levels of perceived practical support or number of significant others.

When stratified by gender, there were no clear associations between childhood physical or sexual abuse, current social support and odds of psychosis in men (Table 4). In contrast, women were around four times more likely to be a case following childhood physical abuse compared with those who reported no exposure to abuse if they currently did not perceive themselves as having sufficient emotional support (OR 4.04, 95 % CI 1.47–11.09) or sufficient practical support (OR 4.90, 95 % CI 1.65–14.57), and were over six times more likely to be a case if they had fewer than five significant others (OR 6.14, 95 % CI 1.80–21.00). When adjusted for confounders, individuals exposed to abuse and who reported low perceived practical support and fewer significant others still had greater odds of psychosis compared with those who had experienced no abuse (adj. OR 4.33, 95 % CI 1.39–13.43 and adj. OR 6.73, 95 % CI 1.96–23.12, respectively). Moreover, a likelihood ratio test for the difference in odds ratios between those who had few significant others compared with those with five or more others demonstrated an interaction with physical abuse in women (χ ² = 3.55, p = 0.060). In contrast, there was no variation in odds of psychosis in women exposed to physical abuse according to levels of perceived emotional support in adulthood after adjusting for confounders (low emotional support: adj. OR 2.87, 95 % CI 0.98–8.37; high emotional support: adj. OR 3.56, 95 % CI 1.22–17.33).

Similar associations, although weaker, were found in relation to childhood sexual abuse. Women who had experienced sexual abuse as a child and who currently had low perceived emotional support were approximately four times more likely to be a case (OR 4.41, 95 % CI 1.60–12.16) compared with those who experienced no abuse. This association held after adjusting for potential confounders (adj. OR 3.64, 95 % CI 1.25–10.61). A likelihood ratio test of the difference in odds ratios between those who had little emotional support compared with those with high support did not reach significance (χ ² = 1.71, p = 0.191). Moreover, women who had experienced sexual abuse as a child and who currently had few significant others were around three times more likely to be a case (OR 3.31, 95 % CI 1.11–9.87) compared with those who had experienced no abuse, and this association held after adjusting for potential confounders (adj. OR 3.55, 95 % CI 1.07–11.77). Lastly, women with a history of sexual abuse who perceived themselves as having high practical support were three times more likely to be a case compared with those who reported no abuse (adj. OR 3.14, 95 % CI 0.94–10.48). Formal tests for three-way interaction by gender, however, did not reveal any significant differences between men and women in the modifying effects of social support between childhood adversity and adult psychosis (see Online Resource 1).

There are several limitations which should be considered when interpreting the findings of this study. Firstly, we relied on retrospective reports of abuse, which have been criticised due to the potential of recall bias especially in those with severe mental illness [27]. However, reports of childhood abuse by individuals with psychosis have been shown to be stable over a seven-year time-period and independent of current symptoms [22], and prospective and retrospective reports of maltreatment have been equally associated with an elevated risk of psychopathology [28]. It is also important to bear in mind that longitudinal studies are not always feasible or justified, due to the large number of participants needed to generate enough cases of psychosis, and that questioning children about recent adverse events may cause substantial distress.

Secondly, cases with psychosis were not asked retrospectively about indices of social support, and therefore the data reflect size of social network and levels of support at the time of interview. Whilst it is possible that perceived social support and social network variables in individuals with psychosis may be consequentially influenced by symptomatology (e.g. those experiencing persecutory delusions or depressive symptoms may undervalue their social relationships) or contact with mental health services, there is limited evidence to suggest that such a “social network crisis” does exist directly following the onset of psychosis [29‐31]. Nevertheless, future research would benefit from investigating the resilience effects of social support earlier on in the developmental trajectory in those with and without a history of childhood adversity.

In a similar vein, the data collected was cross-sectional in nature, and thus causation cannot necessarily be implied. Nonetheless, evidence from a recent systematic review [16] and from an increasing body of more recently published research (e.g. [32‐35]) are suggestive of poor support and deficits in network functioning preceding overt psychotic symptoms. General population studies are a valuable tool for investigating the extended psychosis phenotype; a well-designed longitudinal study conducted in a sample of general population adolescents over 3 years tested the bidirectional association between social functioning and separate dimensions of subclinical positive psychotic experiences [36]. The authors reported a unidirectional relationship such that poor social functioning predicted the development of bizarre experiences and persecutory ideation over time, but not vice versa, thus raising the possibility that a lack of support is indeed contributory to the formation of subclinical psychotic experiences as well as to clinical disorder. There is, however, an evident need to explore antecedents to social isolation in individuals at risk of psychosis—amongst other factors, it could be a consequence of alterations in neurodevelopment in those at psychometric risk, or it could be partially attributable to stressful and traumatic events in childhood or adulthood which could lead to deterioration in social support.

Finally, whilst we carried out a formal test of the differences in the protective effects of social support between men and women (Online Resource 1), this did not result in any statistically significant findings. Nevertheless, the effects are such that they merit reporting given that the results are in the hypothesised direction and that the sample size is such that it would be difficult to detect a significant effect. However, we acknowledge that caution is warranted when interpreting the results, and future research would benefit from the use of larger samples in order to allow firmer conclusions to be drawn.

Previous research has shown that women benefit more from the buffering effects of social support than men, in particular in reducing the risk of experiencing symptoms of depression [15, 37], yet the reason why remains unclear. This current study provided tentative support for such a gender difference in the protective effects of social support following childhood maltreatment in psychosis. It is evident that women benefit more from social support than men, yet the reason why remains very unclear. There are several stages within the support process in which individuals may face barriers to receiving the full positive impact of social support [38]; these include recognising a need for help, deciding what aspect of support may be most beneficial in the situation, and whether there are others who can and are willing to provide the support they need. Then support must be asked for, or if it is freely offered, then the individual has to decide whether to accept it. It is possible that men and women differ in one or more of these stages thus leading to findings of differential health benefits of social support between genders.

Understanding the differences between genders in the positive effects of social support is also impeded by a lack of understanding of how support buffers stress. Several potential mechanisms have been proposed. Firstly, tangible aid or information may change the nature of stressful events [39]. However, research has shown that simply perceiving that one has more available support predicts better adjustment to stress [8, 38], which may act by altering the appraisal of the stressful event [40]. In addition, social support may alter the individual’s self-esteem or their perceived control over their environment, which have both been directly associated with a better adaptation to stressful events [41, 42]. Men and women have been found to differ in the perceived stressfulness of events and their mastery over stressful events [43], and thus men may be less likely to seek support from others close to them.

It has also been proposed that actively seeking help or receiving help results in greater ego-costs for men than women [44] due to traditional gender roles which reinforce independence and low emotional disclosure in men [45]. Divulging the need for help in men, more so than women, may also precipitate another potential cost of decreased perception of self-efficacy or control. Finally, having a larger social network may change the number of resources available to address the stressful event, or increase the likelihood of normative health behaviours [46] or access to normalising explanations for anomalous experiences and abnormal beliefs [47]. This last theory may explain the findings found in this study population, which demonstrated that size of social network was most strongly associated with resilience to psychosis following abuse, especially in women.

These mechanisms are all of potential importance in understanding the apparent relationship between childhood adversity and risk of psychosis. At a critical developmental period, childhood physical and sexual abuse can create a lasting cognitive vulnerability in the form of negative schematic beliefs, resulting in low self-esteem, which in adulthood could lead to an inability to deal with stress and increase vulnerability to psychosis [48]. In addition, and in line with the stress-vulnerability model of psychosis, previous exposure to child abuse may lead to stress sensitisation whereby the individual is found to be more sensitive and perceive higher levels of stress following adverse events [49]. In short, initial research suggests that low self-esteem and negative schemata, and stress sensitisation contribute to the development of psychosis, and may be a long-term consequence of having experienced adverse childhood events. Given that social support in adulthood has been shown to modify each of these variables to promote emotional well-being and to decelerate decompensation to other negative mental health outcomes, it is apparent that exploring the protective effects of social support in the relationship between childhood adversity and psychosis, and the mediating mechanisms, is an important and valuable next step.

Lastly, it would be interesting to investigate further the effects of levels of support and size of networks that exist around the time of abuse, as well as different sources of support, in mitigating the effects of childhood adversity on development of psychosis. As most instances of childhood maltreatment (except sexual abuse) are likely to occur within the family, the individual may have a more negative perception of family support, than perception of friend support, and thus the former may not have a substantial effect in protecting against the development of psychosis. Indeed, in female adult victims of child abuse, perceived friend support, but not family support, has been found to act as a buffer against PTSD [9] and symptoms of depression in adulthood [37]. Differential effects of sources of support, and of support closer to the time of adversity, have yet to be explored in psychosis.