Introduction
Childhood maltreatment encompasses any acts of commission or omission by a parent or other caregiver that result in harm, potential for harm, or threat of harm to a child, even if harm is not the intended result (Leeb, Paulozzi, Melanson, Simon, & Arias, Reference Leeb, Paulozzi, Melanson, Simon and Arias2008). It includes children's exposure to a range of adverse experiences including physical abuse, sexual abuse, emotional abuse, and neglect. Childhood adversity (CA) is an even broader term that also encompasses issues such as parental separation and divorce, domestic violence, and disruptions in caregiving. Some estimates of the frequency of CA worldwide suggest that about one-third of the general population is affected (Kessler et al., Reference Kessler, McLaughlin, Green, Gruber, Sampson, Zaslavsky and Williams2010), which poses a public health problem of great concern. In people with psychotic disorders, the problem is even greater, since almost 60% of this population experience moderate-to-severe maltreatment (Struck et al., Reference Struck, Krug, Yuksel, Stein, Schmitt, Meller and Brakemeier2020). A study on the global population-attributable fraction (PAF) of potentially modifiable risk factors for mental disorders concluded that the largest global PAF was 37.8% for CA and schizophrenia spectrum disorders (SSD) (Dragioti et al., Reference Dragioti, Radua, Solmi, Arango, Oliver, Cortese and Fusar-Poli2022). CA experiences are understood to be highly intercorrelated, and when factored together, robustly predict a broad range of psychopathological and physical problems across the life course (Felitti et al., Reference Felitti, Anda, Nordenberg, Williamson, Spitz, Edwards and Marks1998; Norman et al., Reference Norman, Byambaa, De, Butchart, Scott and Vos2012).
Despite increasing evidence pointing to the role of CA in the development, manifestations, and outcomes of psychotic disorders, its specific effects on the long-term outcome of psychotic disorders remain poorly understood. In recent years, two systematic reviews (Cotter, Kaess, & Yung, Reference Cotter, Kaess and Yung2015; Rodriguez et al., Reference Rodriguez, Aas, Vorontsova, Trotta, Gadelrab, Rooprai and Alameda2021) and three meta-analyses of mainly cross-sectional studies have addressed this question with variable results. While one meta-analysis concluded that the effect of childhood trauma on the evolution of first-episode psychosis (FEP) is unclear (Vila-Badia et al., Reference Vila-Badia, Del Cacho, Butjosa, Serra Arumí, Esteban Santjusto, Abella and Usall2022), others reported that childhood maltreatment was negatively associated with social functioning, but unrelated to independent living or occupational functioning (Fares-Otero et al., Reference Fares-Otero, Alameda, Pfaltz, Martinez-Aran, Schäfer and Vieta2023). The most comprehensive meta-analysis to date focusing on prospective assessment of outcomes provided evidence for an association between general childhood adversity, specific subtypes, and poor psychosocial functioning (Christy et al., Reference Christy, Cavero, Navajeeva, Murray-O'Shea, Rodriguez, Aas and Alameda2023). Notably, this meta-analysis included individual studies with a follow-up period of 1 month to 5 years; such a variable and short-term follow-up is of particular concern because the association between CA and functional outcome appears stronger over time (Rodriguez et al., Reference Rodriguez, Aas, Vorontsova, Trotta, Gadelrab, Rooprai and Alameda2021). Findings from FEP studies longitudinally examining the impact of adversity on functioning are also mixed. For instance, Alameda et al. (Reference Alameda, Ferrari, Baumann, Gholam-Rezaee, Do and Conus2015) showed long-lasting detrimental effects on functioning up to 3 years of follow-up, while neither Trotta et al. (Reference Trotta, Murray, David, Kolliakou, O'Connor, Di Forti and Fisher2016) nor Ajnakina et al. (Reference Ajnakina, Trotta, Forti, Stilo, Kolliakou, Gardner-Sood and Fisher2018) found such effects at 1-year and 5-year follow-ups, respectively. Furthermore, most previous studies on the topic did not consider important confounders such as socioeconomic status, familial risk of severe mental disorders, or obstetric complications, and none considered these confounders conjointly, which may partially explain the divergence in findings.
The association of childhood adversity with outcome domains other than symptoms and functioning has not been examined using a very long-term follow-up of FEP. In two previous studies from our group, using a cohort of 243 FEP patients who were followed-up for a mean of 21 years, we found that CA predicted nonrecovery (Peralta et al., Reference Peralta, García de Jalón, Moreno-Izco, Peralta, Janda, Sánchez-Torres and Cuesta2022) and clinical staging (Peralta et al., Reference Peralta, de Jalón, Moreno-Izco, Peralta, Janda and Sánchez-Torres2023). Using the same population of psychotic disorders, the main goal of the present study was to examine the unique predictive ability of a continuous measure of CA on 27 outcome domains and subdomains, including symptoms, role functioning, quality of life, cognitive impairment, neurological dysfunction, and comorbidity, each rated at the final follow-up assessment. Secondary aims were to assess the dose–response relationship of an ordinal measure of CA with outcomes and to examine the prevalence and severity of CA across specific diagnoses of psychotic disorders.
Method
Study population
The study sample comprised participants from a longitudinal cohort study who were consecutively admitted for treatment of FEP (data collected from 1990 to 2008) and prospectively reassessed for several outcome domains (data collected from 2018 to 2021). Of the 510 patients assessed at baseline, 243 were successfully followed-up (46.4% of the initial sample and 57.3% of the survivors) and made the study sample. The followed and non-followed subjects did not differ in the main demographic and clinical variables (Peralta et al., Reference Peralta, García de Jalón, Moreno-Izco, Peralta, Janda, Sánchez-Torres and Cuesta2022). All participants and their legal representatives, if appropriate, signed written informed consent forms, and the local ethics committee granted ethical approval. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
A detailed description of the study methodology has been provided elsewhere (Peralta et al., Reference Peralta, García de Jalón, Moreno-Izco, Peralta, Janda, Sánchez-Torres and Cuesta2022, Reference Peralta, Moreno-Izco, Garcia de Jalon, Sanchez-Torres, Janda and Peralta2021). The main instrument for assessing background, clinical course, diagnosis, and some outcome variables was the Comprehensive Assessment of Symptoms and History (CASH) (Andreasen, Reference Andreasen1987, Reference Andreasen1992), and for variables used in the present study not included in the CASH, specific assessment instruments were employed (see below). The main sociodemographic and clinical characteristics of the participants are shown in online Supplementary Table S1.
Assessment of CA
We assessed CA using the Global Family Environment Scale (GFES) (Rey et al., Reference Rey, Singh, Hung, Dossetor, Newman, Plapp and Bird1997). The scale seeks retrospectively to quantify the adequacy of the environment in which a child was reared during the first 12 years of life and rates the worst 12 months. The construct measured by the GFES indexes psychosocial adversity, as measured by the Psychosocial Adversity Index (Rey et al., Reference Rey, Singh, Hung, Dossetor, Newman, Plapp and Bird1997). The GFES is a single-score measure with a structure similar to the Global Assessment of Functioning (GAF) scale, which is rated on a hypothetical continuum from 1 to 90 depending on the quality of the family environment the child was brought up with higher ratings indicating less adversity. For example, scores of 81–90 indicate an adequate family environment that is stable, secure, and nurturing for the child with consistent care, affection, and discipline; scores of 31–50 reflect poor family environment, persistent parental discord, hostile separation, exposure to more than one step parent, some abuse or neglect, and poor supervision; and scores of 1–10 reflect a very disturbed family environment, often resulting in the child being made a ward of the state, and evidence of severe abuse, or neglect, or severe deprivation are present.
The GFES has been validated in Spanish by our group and has been shown to have adequate reliability (intraclass correlation coefficient = 0.83) (Rey et al., Reference Rey, Peng, Morales-Blanquez, Widyawati, Peralta and Walter2000). The scale has shown good convergent validity with other rating scales that assess CA (Du Rocher Schudlich et al., Reference Du Rocher Schudlich, Youngstrom, Martinez, KogosYoungstrom, Scovil, Ross and Findling2015; Hawes et al., Reference Hawes, Lechowicz, Roach, Fisher, Doyle, Noble and Dadds2021; Marques-Feixa et al., Reference Marques-Feixa, Palma-Gudiel, Romero, Moya-Higueras, Rapado-Castro and Castro-Quintas2023). The senior authors (VP and MJC) scored the GFES at the FEP using a semi-structured interview focused on identifying signs of child vulnerability, adverse experiences, and family interactions. For all the patients, information was gathered through direct interviews with the parents and patient, which were conducted independently, and information from medical and social records. All these information sources were combined to rate the GFES, and in the case of contradictory or suspected unreliable information, as is often the case in highly disrupted families, we tried to contact other family members or significant others to clarify the ratings, which was the case in about 25% of the sample. In the present study, the GFES score system was reversed, with higher scores reflecting more adversity. According to the reversed GFES scoring system, scores of 1–10 indicate no CA exposure, scores of 11–30 indicate mild CA exposure, scores of 31–50 indicate moderate CA exposure, and scores of 51–90 indicate severe CA exposure.
Assessment of outcome domains
At follow-up and after a period of at least 6 months of clinical stability, outcome domains were assessed by field interviewers (EGJ and LM) using information from multiple sources, including direct interviews with the patients. The raters were blinded to the baseline information of the patients, including CA ratings. We assessed six major outcome domains: psychopathology, role functioning, self-rated quality of life, neurological impairment, cognitive performance, and comorbidity.
Psychopathology was rated using the Scale for the Assessment of Positive Symptoms (SAPS) and the Scale for the Assessment of Negative Symptoms (SANS), as included in the CASH, considering the severity of psychopathology over the last month. Role functioning was rated over the last year using the World Health Organization-Disability Assessment Schedule (WHO-DAS) (World Health Organization, 1988). Under the umbrella concept of ‘quality of life’, participants self-rated questionnaires on general quality of life ‒ the Schizophrenia Quality of Life Scale (SQLS) (Wilkinson et al., Reference Wilkinson, Hesdon, Wild, Cookson, Farina, Sharma and Jenkinson2000), perceived health status ‒ the EuroQol-5D (EQ-5D) (Brooks, Reference Brooks1996), personal recovery ‒ the Questionnaire of Personal Recovery (QPR-15) (Law, Neil, Dunn, & Morrison, Reference Law, Neil, Dunn and Morrison2014), and internalized stigma ‒ Internalized Stigma of Mental Illness (ISMI-9) scale (Hammer & Toland, Reference Hammer and Toland2017).
Neurological function was assessed using the Neurological Examination Scale (Buchanan & Heinrichs, Reference Buchanan and Heinrichs1989). Cognitive performance was assessed using the Screen for Cognitive Impairment in Psychiatry (SCIP) (Purdon, Reference Purdon2005). Due to lack of collaboration, 22 participants did not complete the SCIP and 18 participants did not complete the NES. Furthermore, we assessed social cognition in 165 patients who could cooperate using the Spanish version of the MATRICS Consensus Cognitive Battery (MCCB) (Nuechterlein & Green, Reference Nuechterlein and Green2006).
Lastly, comorbidity included number of metabolic syndrome criteria as assessed by the American Heart Association; National Heart, Lung, and Blood Institute (Grundy et al., Reference Grundy, Cleeman, Daniels, Donato, Eckel, Franklin and Costa2005), severity of drug abuse as rated with the Addiction Severity Index (ASI) (McLellan et al., Reference McLellan, Luborsky, Cacciola, Griffith, Evans, Barr and O'Brien1985), medical comorbidity as rated by the Elixhauser's Medical Comorbidity Index (EMCI) (Elixhauser, Steiner, Harris, & Coffey, Reference Elixhauser, Steiner, Harris and Coffey1998), number of lifetime suicide attempts, and number of comorbid DSM-5 non-psychotic mental disorders diagnoses at follow-up. Information to rate comorbidity was obtained using all available information, including personal interviews with the patients and significant others, such as clinical registers.
All these instruments have been validated for Spain, excepting the EMCI and the ISMI-9 scale, which were translated into Spanish by the authors.
Confounding variables
Sociodemographic confounders included sex, age at follow-up, length of follow-up, and parental socioeconomic status (p SES), all assessed with the CASH. Familial load of SSD was assessed in the participants' first-degree relatives using the Family History-Research Diagnostic Criteria (FH-RDC) (Andreasen, Endicott, Spitzer, & Winokur, Reference Andreasen, Endicott, Spitzer and Winokur1977) and we estimate the familial load score for SSD taking into account family size and age structure (Verdoux et al., Reference Verdoux, van Os, Sham, Jones, Gilvarry and Murray1996). Obstetric complications were assessed using the Lewis and Murray scale (Lewis, Owen, & Murray, Reference Lewis, Owen, Murray, Schulz and Tamminga1989). Family history and obstetric complications were documented using information from the patients' parents and clinical registers, and the mother was available in most cases to rate obstetric complications.
Statistics
We first examined the distribution of alternative CA ratings in the whole sample, and compared ratings across DSM-5 diagnoses by using Chi-squared or ANOVA followed by post-hoc tests.
To assess the relationships between CA and the outcomes, we fitted a series of univariate and multivariable linear regression models where each outcome was the dependent variable. Distributions of variables were evaluated for normality and transformed, as appropriate. The basic assumptions of the regression models were checked by examining residual plots, tolerance, and variance inflation factor. Hierarchical regression models were run to determine the independent contribution of CA exposure to predict outcomes. In the first step, we entered the six confounding variables as described above, and in the second step, we entered the childhood adversity score and examined the incremental fit (ΔR 2) over the covariates-only model. Standardized β values were provided to allow comparison between models, while ΔR 2 indexes the unique amount of variance explained by CA exposure in the outcome variance. Recently, Funder & Ozer (Reference Funder and Ozer2019) recommended that, within the context of assessing the longitudinal impact of a given exposure, the size of R 2 for small, medium, large, and very large effects are 0.01, 0.04, 09, and 0.16, respectively.
We examined the dose–response relationship between levels of childhood adversity (absence, mild, moderate, and severe) and continuous outcome variables by means of ANCOVA, controlling for the covariates and using polynomial contrast coefficients for linear trend testing. Each significant multivariate test was followed up with a post-hoc assessment of the significance between the levels of childhood adversity.
Because we examined the association of childhood adversity ratings with 27 different outcome domains or subdomains, we controlled the type I error rate using the Holm–Bonferroni method, where each outcome domain was considered as a family of hypotheses. All tests were two-tailed and the target α level was set at p < 0.05.
Results
Prevalence and severity of CA in the participants
There was no evidence of significant differences in the GFES score, levels of CA exposure, or a dichotomous rating of adversity exposure between participants and non-participants (online Supplementary Table S2).
Visual inspection of GFES scores revealed a semi-normal distribution (online Supplementary Fig.). The mean GFES score was 30.0 (s.d. = 21.8, range: 3–89), and the median score was 22 (interquartile range: 10–45). According to the predefined levels of severity in the GFES, 110 patients (45.3%) were unexposed to CA, 34 (14%) were exposed to mild adversity, 47 (19.3%) were exposed to moderate adversity, and 52 (21.4%) were exposed to severe adversity.
CA experiences were more prevalent and severe in patients with schizophrenia than in those with other psychotic disorders (p < 0.001) (Table 1). Levels of childhood adversity also differed significantly across psychotic disorders (χ2 = 58.9, df = 12, p < 0.001) (Fig. 1). CA experiences were more prevalent in patients with schizophrenia (74.3%), followed by those with schizoaffective disorder (53.8%), affective psychoses (32.7%), other psychotic disorders (31.6%), and brief psychotic disorders (25%). Severe CA exposure was highest in schizophrenia (37.2%) and lowest in brief psychotic disorder (no severe exposure).
Childhood adversity ratings across DSM-5 diagnosis of psychotic disorders
CA, Childhood Adversity; GFES, Global Family Environment Scale.
a Includes 42 patients with bipolar disorder and 10 patients with major depressive disorder.
b Includes nine patients with psychotic disorder not otherwise specified, six patients with schizophreniform disorder, and four patients with delusional disorder.
c 0 = no exposure; 1 = mild exposure; 2 = moderate exposure; 3 = severe exposure
Levels of childhood adversity exposure across diagnoses of psychotic disorders.
Note: SZ, schizophrenia (N = 113); SAD, schizoaffective disorder (N = 39); AFF, affective psychoses (N = 52); BPD, brief psychotic disorder (N = 20); OPD, other psychotic disorders (N = 19).
Associations between confounders and CA
Associations among cofounders are described in online Supplementary Table S3. Descriptive statistics for the confounders and their univariate associations with continuous and categorical ratings of CA are presented in Table 2. p SES, familial load of SSD, and obstetric complications were significantly related to continuous and dichotomic ratings of CA. Obstetric complications were the most contributing factor to CA, since they explained 13.3% and 10.3% of the variability in the continuous and categorical measures of CA, respectively.
Associations of confounder variables with continuous and categorical ratings of childhood adversity
GFES, Global Family Environment Score.
Univariate associations between CA and the outcomes
Univariate analyses showed widespread significant associations between the GFES scores and outcomes (Table 3). The GFES score was significantly related to all outcome variables except the EQ-5D score, EMCI score, and number of metabolic syndrome criteria. The variance explained in the outcomes by the GFES score ranged between 2.8% for NES motor coordination and 26.1% for the WHO-DAS total score.
Severity of childhood adversity (age 0–12 years) predicting outcomes in 243 patients with psychotic disorders
EQ-5D, EuroQol-5D; DSM, Diagnostic and Statistical Manual, 5th edition; ISMI, Internalized Stigma Scale; MCCB, MATRICS Consensus Cognitive Battery; NES, Neurological Examination Scale; QPR, Questionnaire of Personal Recovery; SAPS, Scale for the Assessment of Positive Symptoms; SANS, Scale for the Assessment of Negative Symptoms SCIP, Screen for Cognitive Impairment in Psychiatry; SQLS, Schizophrenia Quality of Life Scale; WHO-DAS, World Health Organization-Disability Assessment Schedule.
In bold are presented statistically significant associations after correcting for multiple comparisons.
Multivariate associations between CA and the outcomes
After adjusting for confounders, the association pattern between the GFES score and outcomes exhibited smaller effect sizes than that observed in unadjusted analysis; furthermore, the associations of the GFES score with motor coordination and number of suicide attempts were no longer significant after correcting for multiple testing (Table 3). Effect sizes of the associations decreased between 35% and 65% compared to those observed in the univariate analyses. For instance, while ΔR 2 for the association between the GFES score and SCIP total score in the univariate analysis was 0.109, in the multivariate analysis it was 0.040, which represents a 63.3% decrease. Large to very large effect sizes were observed for most role-functioning variables and negative symptoms (ΔR 2 between 0.105 and 0.181). Moderate effect sizes were observed, in that order, for positive symptoms, personal functioning, social cognition, impaired immediate verbal learning, internalized stigma, personal recovery, and SCIP total score (ΔR 2 between 0.040 and 0.066). All other significant associations had small effect sizes. The highest incremental variance over the covariates-only model was observed for familial functioning (18.1%) and social functioning (17.1%), such as the WHO-DAS total score (17.6%).
Dose–response relationship between levels of CA and the outcomes
Levels of CA exposure exhibited a dose–response association pattern with the outcomes (Table 4). A particularly strong dose–response relationship (p trend ⩽ 0.001) was observed for the following outcomes: all role functioning variables, positive and negative symptoms, immediate verbal learning, processing speed, SCIP total score, and NES total score. Other significant dose–response relationships (p trend > 0.001) mainly involved differences between no exposure and severe exposure.
Levels of childhood adversity exposure (age 0–12 years) predicting outcomes in 243 patients with psychotic disorders
See Table 3 for abbreviations.
In bold are presented statistically significant associations after correcting for multiple comparisons.
* Also fit a quadratic function (p = 0.024).
Discussion
Main findings
The study's results can be summarized into five main findings. First, CA experiences were more prevalent and severe in patients with schizophrenia than in those with other psychotic disorders. The overall prevalence of adversity in psychotic disorders was 54.7%, and in schizophrenia it increased to 74.3%. Relatedly, 37.2% of patients with schizophrenia were exposed to severe CA, while it was <16% for other psychotic disorders. Second, p SES, familial load of SSD, and obstetric complications significantly impacted the ratings of adversity and its associations with outcomes, as shown by the finding that the effect sizes of the univariate associations between CA and outcomes decreased substantially when adjusted for these confounders. Third, CA uniquely predicted a wide range of poor outcomes with variable effect sizes, both between and within outcome domains. Overall, CA most impacted role functioning, followed, in that order, by symptoms, cognitive functioning, quality of life, neurological function, and comorbidity. Moderate to very large effect sizes were observed for CA predicting poor role functioning and symptoms, whereas small to moderate effect sizes were observed for the other outcome domains. Fourth, we found evidence of a dose-response relationship between levels of adversity and outcomes, with increasing severity of adversities exposure increasing the risk of poorer outcomes.
These findings add to the literature with evidence suggesting that CA exposure in psychosis patients leads, in a dose–response fashion, to widespread and long-lasting impairment in several outcome domains. Of particular relevance was the impact of CA on role functioning, since 17.6% of the variance in overall functioning was uniquely explained by CA. This is an outstanding finding given the long-time lag (of at least 36 years) between risk exposure and outcome assessment. Notwithstanding this, it should be noted that the variance explained by CA experiences in most outcome subdomains other than role functioning and negative symptoms was very small (i.e. around 5%).
A question that arises is whether our findings are specific to psychosis or whether they represent the non-specific effects of early adversity on a specific illness. Given the evidence of the extensive effects of CA on mental health outcomes irrespective of diagnostic categories (Norman et al., Reference Norman, Byambaa, De, Butchart, Scott and Vos2012; Teicher, Gordon, & Nemeroff, Reference Teicher, Gordon and Nemeroff2022), we believe that the second possibility is more plausible. Furthermore, given the higher prevalence and severity of CA in people with schizophrenia, most effects on outcomes appear to be mainly due to schizophrenia, a disorder with relevant neurodevelopmental risk factors and mechanisms (Murray, Bhavsar, Tripoli, & Howes, Reference Murray, Bhavsar, Tripoli and Howes2017). Children exposed to adversity are still in a sensitive period of neurological, cognitive, social, and emotional development, which helps to explain ‒ via the principle of multifinality (i.e. one risk factor may contribute to various outcomes) the widespread and enduring effects we observed. Indeed, CA appears to lead to neurodevelopmental disruption with substantial effects on brain structure and function, with broad and long-lasting consequences (Teicher, Samson, Anderson, & Ohashi, Reference Teicher, Samson, Anderson and Ohashi2016). The links of CA with self-perceived poor recovery and internalized stigma are also consistent with the social defeat hypothesis stating that early adversity leads to negative experiences of being excluded from the majority group (Selten, van der Ven, Rutten, & Cantor-Graae, Reference Selten, van der Ven, Rutten and Cantor-Graae2013).
Comparison with the literature
Comparison with other studies is difficult because of the differences in measures of CA and the lack of very long-term follow-up studies examining outcome domains. Notwithstanding this, some comparisons are in order. Regarding the prevalence of CA, it should be noted that our prevalence rate (54.7%) was similar to the almost 50% prevalence reported in studies of FEP in Spain (Alameda et al., Reference Alameda, Ferrari, Baumann, Gholam-Rezaee, Do and Conus2015; Arranz et al., Reference Arranz, Monferrer, Jose Algora, Cabezas, Sole, Vilella and Sanchez-Gistau2018; Barrigón et al., Reference Barrigón, Diaz, Gurpegui, Ferrin, Salcedo, Moreno-Granados and Ruiz-Veguilla2015). Additionally, we found that 62.9% of patients with schizophrenia experienced moderate to severe adversity, a figure similar to the 56.1% reported by Struck et al. (Reference Struck, Krug, Yuksel, Stein, Schmitt, Meller and Brakemeier2020) using the childhood trauma questionnaire. The similar prevalence rates of CA across studies and measures further adds to the validity of the GFES in assessing CA.
Comparison with other studies regarding outcomes is of little value since as noted in the introduction, most previous studies were cross-sectional and, if prospective, the follow-up period was too short and variable to provide insightful comparisons. As a matter of fact, systematic reviews and meta-analyses provide evidence for a link between CA and cognition (Vargas & Mittal, Reference Vargas and Mittal2018), functioning (Christy et al., Reference Christy, Cavero, Navajeeva, Murray-O'Shea, Rodriguez, Aas and Alameda2023), and symptoms (Alameda et al., Reference Alameda, Christy, Rodriguez, Salazar de Pablo, Thrush, Shen and Murray2021; Bailey et al., Reference Bailey, Alvarez-Jimenez, Garcia-Sanchez, Hulbert, Barlow and Bendall2018), but with overall small effect sizes and mixed evidence for subdomains thereof. In line with the literature (McKay et al., Reference McKay, Kilmartin, Meagher, Cannon, Healy and Clarke2022), we found evidence of CA predicting psychiatric comorbidity, such as number of comorbid diagnoses, drug abuse, and suicidality. However, contrary to expectations, we found that CA did not predict medical comorbidity or metabolic syndrome features. We are not aware of previous studies prospectively addressing this question in psychotic disorders, although in line with our findings, a systematic review of 20 cross-sectional studies did not find an association between CA and metabolic syndrome in participants with severe mental illness (Balaji & Sankaranarayanan, Reference Balaji and Sankaranarayanan2023). There is a lack of studies examining the long-term effects of CA on quality-of-life measures and neurological dysfunction, thus our findings regarding these outcomes need to be confirmed by other authors.
Measurement issues
Single-item measures of psychological experiences are often viewed as being psychometrically suspected. To quantify childhood adversity, we used the GFES, which consists of a single-score measure that retrospectively rates CA within the family. Thus, the GFES has important limitations derived from (1) using a single global rating of CA, (2) excluding adversities outside the family context, (3) lack of rating of specific adversities, and (4) the retrospective assessment of childhood events. Indeed, the GFES has been criticized for being a global measure and having the potential cost of losing specificity because adversities that influence one sort of outcome may not necessarily be the same as those that affect others (Rutter, Reference Rutter1997). A further limitation of the GFES is that it has been rarely used in research, which limits comparison with other studies.
That said, it is important to note that the GFES is a reliable and valid retrospective measure indexing cumulative psychosocial adversities of the child (Rey et al., Reference Rey, Singh, Hung, Dossetor, Newman, Plapp and Bird1997). Moreover, the GFES has several advantages over most other CA scales. First, it includes rearing practices, such as parental conflicts and inconsistencies about discipline or expectations (i.e. child's overprotection), which may negatively impact psychological development, and are rarely considered in other rating scales. Second, because adversities tend to cluster together, the scale accounts for their interactive effects on the global score. Third, the GFES does not merely rate adversity experiences, but also an adequate and nurturing family environment for the child, which can buffer the effect of adversities outside the family (Garner & Yogman, Reference Garner and Yogman2021). Lastly, the GFES assesses early adversities (under age 12), which conveys higher power to predict poor outcomes than late adversities (Alameda et al., Reference Alameda, Ferrari, Baumann, Gholam-Rezaee, Do and Conus2015).
While many researchers of childhood adversity strongly advocate the use of multi-item scales, several authors have recently challenged this idea by signifying that single-item measures can have acceptable psychometric properties and are a potentially viable alternative to multi-item scales for construct measurement purposes (Allen, Iliescu, & Greiff, Reference Allen, Iliescu and Greiff2022; Vargas & Mittal, Reference Vargas and Mittal2018). As a matter of fact, experiences of childhood adversity are highly interrelated (Dong et al., Reference Dong, Anda, Felitti, Dube, Williamson, Thompson and Giles2004) and compiling a single additive cumulative ‘risk’ score can increase the power of the design (Evans, Li, & Whipple, Reference Evans, Li and Whipple2013), so it makes sense to use a global measure of adversity.
Implications
Our findings suggest that interventions aimed at reducing exposure to CA are important for reducing adult morbidity of psychotic disorders. Most articles on this topic typically conclude with a call for increased funding for public health and policy interventions, with a broader perspective on health that extends beyond the medical system. However, primary prevention has proven difficult and will ultimately require societal changes that improve the quality of family and household environments during childhood. A review of programs for young children and families facing adversity concluded that there was mixed evidence on their effectiveness, and that more effective interventions are particularly needed in the prenatal period and first three years after birth for the most disadvantaged children and families (Shonkoff, Reference Shonkoff2016). Secondary prevention of the effects of CA will first require increased recognition of their occurrence and, second, an effective understanding of both personal and familial resilience abilities to cope with these experiences. Lastly, tertiary prevention in the case of psychosis will require routine clinical inquiry into CA experiences within FEP programs. This has potentially important treatment implications for identifying individuals at elevated risk of poor outcomes. While there is currently no clear guidance on how to target childhood trauma in the early treatment of psychosis, there are several promising lines of inquiry, including desensitization and reprocessing therapy and trauma-focused cognitive-behavioral therapy (Hardy et al., Reference Hardy, Keen, van den Berg, Varese, Longden, Ward and Brand2024).
Strengths and limitations
We used a representative sample of patients with first-admission psychosis, who were assessed for CA at FEP and re-assessed after a mean of 21 years of follow-up for 27 outcome domains and subdomains. By adjusting for relevant confounders, we showed that CA experiences had unique and widespread impact across outcomes with variable effect sizes. We additionally found that the identified effects had a dose–response relationship, and that CA exposure was more prevalent and severe in schizophrenia than in other psychotic disorders. This allowed the demonstration of a temporal relationship between CA exposure and outcomes lending confidence to causal inference.
The specific limitations of using the GFES to assess CA have already been discussed; however, some other limitations of our study should be noted. First, loss to follow-up is inevitable over a 21-year period, and we had a 46.4% attrition rate of living subjects. Nevertheless, previous research simulating the effects of selective dropout in longitudinal studies found that it may not reduce the validity of predicting outcomes (Wolke et al., Reference Wolke, Waylen, Samara, Steer, Goodman, Ford and Lamberts2009), and we did not find evidence of selection bias regarding baseline variables, including rates and severity of CA. Second, we relied on retrospective reports of CA, which are subject to mis-reporting due to recall bias. More specifically, CA experiences were ascertained during the FEP, where might be a reporting bias for more severe presentations. Nevertheless, evidence suggests that retrospective reports have moderate-to-good consistency over time, and may be a reliable and valid indicator of early life adversities, thus having a worthwhile place in research (Hardt & Rutter, Reference Hardt and Rutter2004), particularly when assessed using multi-information sources (Vargas & Mittal, Reference Vargas and Mittal2018). We used a semi-structured interview to retrospectively rate CA using multiple information sources, which minimizes recall bias and provides additional important contextual details of adversity exposure (Gayer-Anderson et al., Reference Gayer-Anderson, Reininghaus, Paetzold, Hubbard, Beards, Mondelli and Morgan2020). Finally, it is possible that different types of CA impact the outcomes differently. In this sense, our study can be viewed as a general approach to this issue that should be further examined using more specific measures of CA, as well as timing and duration of adversity.
Acknowledgement
We are extremely grateful to all subjects who took part in this study.
Funding statement
The study was funded by the Spanish Ministry of Economy, Industry and Competitiveness (grant PI16/02148 and 19/01698) and the Regional Government of Navarra (grant 31/17 and 41/18).
Competing interest
The authors declare no conflicts of interest
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291724001223
