Childhood trauma and treatment resistance in first-episode psychosis: Investigating the role of premorbid adjustment and duration of untreated psychosis

Background: Early identification of treatment non-response in first-episode psychosis (FEP) is essential to outcome. Despite indications that exposure to childhood trauma (CT) can have adverse effects on illness severity, its impact on treatment non-response and the interplay with other pre-treatment characteristics is sparsely investigated. We use a lack of clinical recovery as an early indicator of treatment resistance to investigate the relationship between CT and treatment resistance status at one-year follow-up and the potential mediation of this effect by other pre-treatment characteristics. Methods: This prospective one-year follow-up study involved 141 participants recruited in their first year of treatment for a schizophrenia-spectrum disorder. We investigated clinical status, childhood trauma (CT), pre-morbid adjustment (PA), and duration of untreated psychosis (DUP) at baseline and clinical status at one-year follow-up. Ordinal regression analyses were conducted to investigate how PA and DUP affected the relationship between CT and one-year outcome in FEP. Results: 45 % of the FEP sample reported moderate to severe CT, with significantly higher levels of CT in the early treatment resistant group compared to participants with full or partial early recovery. Ordinal regression analysis showed that CT was a significant predictor of being in a more severe outcome group (OR = 4.59). There was a partial mediation effect of PA and a full mediation effect of DUP on the effect of CT on outcome group membership. Discussion: Our findings indicate that reducing treatment delays may mitigate the adverse effects of CT on clinical outcomes and support the inclusion of broad trauma assessment in FEP services.


Introduction
Psychotic disorders are heterogeneous in their initial presentation, course, and outcome (Friis et al., 2016), and multiple factors influence the treatment response (Demjaha et al., 2017).Up to one-third of patients with first-episode psychosis (FEP) do not respond to standard antipsychotic treatments and are considered treatment resistant (Howes et al., 2017;Howes et al., 2021).Treatment resistant patients experience more intense symptoms, higher levels of disability, higher risk of suicide, and lower quality of life (Howes et al., 2021;Kane et al., 1988;Mamo, 2007).The cost associated with treating treatment resistant patients constitutes 60-80 % of the overall cost of schizophrenia (Howes et al., 2021;Kennedy et al., 2014).Early identification of this patient group is essential, as adapted interventions, including the timely use of clozapine, have significant benefits (Qubad and Bittner, 2023).
To what extent different subtypes of CT are associated with different aspects of outcome is also sparsely investigated (Croft et al., 2019).There are suggestions that experiencing sexual abuse leads to a higher risk of psychosis than other forms of interpersonal trauma (Croft et al., 2019).Others have found links between sexual abuse and hallucinations (Grindey and Bradshaw, 2022), neglect and paranoia (Bentall et al., 2014), and between both sexual-and physical abuse and positive symptoms (Ajnakina et al., 2016).On the other hand, some argue that it is more likely that the number and severity of CT have a more significant impact on outcomes than the type (Croft et al., 2019;Gibson et al., 2016).
Being exposed to interpersonal CT has a fundamental impact on the stress-response system.It will also, for many, generate mistrust in others, thus disturbing social adjustment and limiting constructive helpseeking behavior.This suggests a relationship between CT and other pretreatment prognostic factors, but few studies have investigated the relationship between CT and PA or DUP.
In the TIPS cohort, significant associations between early interpersonal trauma, PA, and DUP have been suggested (Haahr et al., 2018).Hegelstad et al. (Hegelstad et al., 2021) found indications of a complex interplay between PA, CT, and time to first remission, where patients with poor PA exposed to CT had a higher risk of non-remission than other FEP patients at one-year follow-up.Ottesen et al. (Ottesen et al., 2021) found that FEP patients exposed to CT had more severe positive and depressive symptoms at follow-up than those unexposed.Patients who experienced physical or emotional abuse also used higher doses of antipsychotic medication after one year of treatment.In other cohorts, Smit et al., found that childhood physical neglect was associated with poorer general PA during early and late adolescence, and with poorer premorbid academic adjustment during childhood and early adolescence (Smit et al., 2021).Also, a study by Veru et al., (Veru et al., 2022) observed a significant association between CT and DUP, suggesting that DUP mediates some of the adverse effects of CT on the outcome, but this was not further explored.
In summary, suggestions exist for an interplay between CT, PA, DUP, and other risk factors for poor clinical outcomes in FEP.We have recently shown that adaptations of the criteria (Wold et al., 2023) for treatment resistance by the Treatment Response and Resistance in Psychosis Working Group (Howes et al., 2017) can identify FEP patients without symptomatic or functional remission after one year of treatment, thus serving as an early indicator of treatment resistance and that PA and DUP were significantly associated with meeting these adapted criteria (Wold et al., 2023).In the current study, we expand on these findings by investigating if CT contributes to the risk of early treatment resistance based on our proxy measure.No previous studies have investigated the role of CT in early treatment resistance and whether CT here interacts with other pre-treatment characteristics.

Aims
Our objective was to investigate associations between CT and early indications of treatment resistance in FEP at one-year follow-up and to what extent DUP and PA influence this relationship by investigating the associations between CT and CT subtypes, PA, DUP, and (the indicator of) early treatment resistance.

Setting and participants
The current study is part of the ongoing TOP (Thematically Organized Psychosis Research) study conducted at the Norwegian Centre for Mental Disorders Research (NORMENT).We included participants aged 18 to 65 years recruited within the first year of treatment from inpatient and outpatient psychiatric units at hospitals in the Oslo area between 2003 and 2019.The hospitals serve an area with a population of 660,000, about 88 % of the population of Oslo.Participants with DSM-IV schizophrenia, schizophreniform disorder, schizoaffective disorder, or psychotic disorder NOS (Legge et al., 2020) were included in this study.Exclusion criteria were severe brain injury, an estimated IQ below 70, and not speaking a Scandinavian language.Participants with one-year follow-up data who also had completed CT assessments were eligible for the current part of the study.CT assessment was first added to the study protocol in March 2007, and the 110 participants included after this date had a valid assessment.In addition, 31 participants recruited before May 2007 completed the CT assessment as participants in a substudy (Fig. 1).The total study sample for the current analyses thus comprised 141 FEP patients.
The study was conducted per the Helsinki Declaration of Ethics in medical research and approved by the Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate.We obtained written informed consent from all participants before the assessments.

Clinical assessments
Participants underwent comprehensive clinical assessments by clinical psychologists, medical doctors, or psychiatrists trained in the study protocol.The diagnosis was made by the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I), modules A-E (SCID I) (First, 2005).Diagnostic reliability was established through participation in a program developed at UCLA (Ventura et al., 1998), co-ratings of training videos, as well as regular diagnostic consensus meetings with a senior clinical researcher.The inter-rater reliability has been found satisfactory (Ringen et al., 2008), with an overall kappa score between 0.92 and 0.99 across different assessment teams (Høegh et al., 2020).Information about education, occupation, marital status, and a full illness history with treatment information was obtained through interviews, medical charts, blood samples, and adherence questionnaires.
Symptoms at baseline and one-year follow-up were measured with the Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987).The PANSS items were grouped according to the Wallwork five-factor model that consists of subsets of items constituting mean scores of positive, negative, disorganized, excited, and depressed symptoms (Wallwork et al., 2012).This model is more appropriate for assessing FEP populations than the original three-factor model (Wallwork et al., 2012;Langeveld et al., 2013;Ajnakina et al., 2018).

Assessment of childhood trauma (CT)
CT was rated using a Norwegian version of the Childhood Trauma Questionnaire-Short Form (CTQ-SF) (Bernstein et al., 2003).This is a retrospective self-report questionnaire with 28 items, yielding scores on five subscales of CT: emotional abuse, physical abuse, sexual abuse, physical neglect, and emotional neglect, as well as a total score.Each subscale is rated on a 5-point Likert scale with answers ranging from 1 (never true) to 5 (very often true) for five items (Bernstein et al., 1994).The CT data were analyzed as subscales and as the total CTQ score, with a minimum score of 25 and a maximum score of 125.For reporting prevalence, we dichotomized the CTQ scores based on cut-off scores for no/low and moderate/severe levels, as recommended by Bernstein and Fink (Bernstein and Fink, 1998): emotional abuse (13), physical abuse (10), sexual abuse (8), emotional neglect (15), physical neglect (10).The reliability and validity of the CTQ have been demonstrated previously (Bernstein et al., 2003;Bernstein et al., 1994;Kongerslev et al., 2019).

Assessment of duration of untreated psychosis (DUP)
DUP was established at baseline and measured in weeks from the onset of psychosis until the start of adequate treatment.The onset was defined as having a score equal to or above 4 on the Positive and Negative Syndrome Scale (PANSS) positive items P1 (delusions), P3 (hallucinatory behavior), P5 (grandiosity), P6 (Suspiciousness), and G9 (unusual thought content) for more than one week (Larsen et al., 2001).

Assessment of premorbid adjustment (PA)
Premorbid social and academic adjustment was measured at baseline with the Premorbid Adjustment Scale (PAS) (Cannon-Spoor et al., 1982).This is a clinician-rated 7-point scale that assesses social-and academic performance for different age groups.High PAS scores indicate poor adjustment.To avoid overlap with any potential prodromal phase before the first episode, we used the social and academic childhood subscales covering ages 0-11.

Functional assessment
The Global Functioning Scale (GFS) (Pedersen et al., 2007;Pedersen et al., 2018) was used to assess functioning.The GFS is a continuous measure where score 1 represents the worst imaginable level of impairment of psychosocial functioning, and a score of 100 represents the, hypothetically, most optimal level (Pedersen et al., 2018).Participants with a GFS value of ≥ 61 at the one-year follow-up were considered functionally recovered (Austin et al., 2013;Simonsen et al., 2017) since a score of 61 or above indicates that the participant can live independently, meets basic needs, holds employment, and maintains meaningful social relationships (Pedersen et al., 2018).For 17 participants, the precise GFS score at one-year follow-up was unavailable.We then utilized all accessible data to determine whether the GFS score for these participants was above 61.

Definition of one-year outcome groups
Because of delays in initiating and changing antipsychotic medications, few FEP participants had completed two adequate trials of antipsychotic medications in their first year of treatment, as required by consensus criteria from the Treatment Response and Resistance in Psychosis Working Group (TRRIP) (Howes et al., 2017).It is thus likely that not all treatment unresponsive FEP patients will be classified as treatment resistant by the TRRIP criteria at this time.In addition, the observation period for stability of remission was short.We therefore made adaptations of the TRRIP criteria and the consensus criteria for remission from the Remission in Schizophrenia Working Group (RSWG) (Andreasen et al., 2005) to create relevant outcome groups at one-year follow-up (See Table 1 & 2) (Wold et al., 2023).We here used a shorter time criterion (12 weeks) for stability of the current clinical status.In addition, we did not include the criterion of two adequate antipsychotic medication trials for treatment resistance.For further information about the background of these choices, see Wold et al., 2023(Wold et al., 2023).The three outcome groups were as follows: Early treatment resistance (as indicated by): not meeting RSWG remission criteria for at least 12 weeks at follow-up AND not regained functioning defined as a GFS score < 61.
Early clinical recovery (as indicated by): meeting RSWG criteria for remission AND regained functioning, defined as a GFS score ≥61.
Partial early clinical recovery: not meeting the two criteria above.

Statistical analyses
We used the TSD (Tjenester for Sensitive Data) facilities at the University of Oslo for safe data storage.For statistical analyses, we used the Statistical Package for the Social Sciences (SPSS) Version 29.0 (Statistical Package for the Social Sciences (SPSS) for Windows ISSfW, Version 29.0, n. d.), STATA, version 18.0 (StataCorp, 2021) and the statistical programming language R (Team, 2021).Data were checked for normality, homogeneity of variance, and outliers.Non-normally distributed variables were log-transformed before inclusion in regression analyses.Some of the CTQ subscales were heavily skewed since many were not exposed to childhood trauma at all, while those who were often had been exposed to several types.Alpha was set at p < .05.
Group differences between the outcome groups in premorbid, sociodemographic, trauma-and clinical variables were examined with Chi-Square tests for categorical variables and ANOVAs with Tukey post hoc tests for continuous variables.Kruskal-Wallis tests were performed for non-parametric continuous variables.Group differences in the prevalence of moderate to severe CT were examined using Chi-Square tests.
Associations between CTQ total scores and CTQ sub-scores, PAS scores, DUP, and baseline symptomatology, were investigated using Pearson correlations.In the case of significant associations, multivariate relationships were further investigated using ordinal logistic regression analyses.
Ordinal logistic regression analysis (Proportional Odds model) was performed to investigate the influence of CT, DUP, and PA (as measured by the PAS) on the odds of meeting criteria for being in the different oneyear outcome groups (here treated as an ordinal assessment of levels of outcome severity).The PAS subscales were highly intercorrelated, and PAS social adjustment was chosen for the analyses because of the relevance of social aspects for the research question.Statistically significant variables that improved model fit (− 2 Log Likelihood, Akaike Information Criteria, and Bayesian Information Criteria) were kept for further analytic steps.We used a likelihood ratio test, a test of parallel lines, and a Brant test to test the proportional odds and parallel regression assumptions.The predictor variables were tested a priori to verify no violation of the multicollinearity assumptions.
We then examined whether PA and DUP mediated the relationship between the CTQ total score and outcome group membership at oneyear follow-up.The choice of approach was based on studies either using the product of coefficients with the logistic regression as recommended for mediation analyses of ordinal data (Liu et al., n.d.), or running mediation analyses of ordinal outcome on the Baron & Kenny's 4-step method (Baron and Kenny, 1986;Nguyen et al., 2023) and further using IBM SPSS® Hayes' PROCESS macro for determining the mediation effects (Hayes, 2018;Abdulla, 2019).Fig. 2 conceptualizes the mediation model.The ordinal regression model was used to establish that the effect might be mediated by investigating the role of CT in predicting outcome group membership at one-year follow-up (path c'), demonstrating that the independent variable (CT) is significantly affecting the dependent variable (Outcome).Further, an accumulative regression model was used to analyze the predicting role of CT on PA and DUP (path a1 and a2), and the ordinal regression model was used to investigate the predicting role of PA and DUP on outcome groups (path b1 and b2).We then tested the potential mediation using the ordinal outcome as a continuous variable using IBM SPSS® Hayes' PROCESS macro, model number 4 with multiple mediators (Nguyen et al., 2023;Hayes, 2018;Abdulla, 2019).

Results
The study sample comprised 141 participants, 83 (59 %) males.The mean age at baseline was 26.6 years (SD.7.5), and the mean years of education were 13.4 (SD 2.9).At baseline, 72 (51 %) had a diagnosis of schizophrenia, and the others met the criteria for a broad schizophrenia spectrum diagnosis.Median DUP was 30 weeks (range 0-1300).At oneyear follow-up, 62 (44 %) were in the early treatment resistance outcome group, 44 (31 %) were in the early clinical recovery group, and 35 (25 %) were in the partial early clinical recovery group.The demographic and clinical characteristics of the outcome groups are presented in Table 3.There were statistically significant differences between the early treatment resistance and early clinical recovery groups for almost all investigated premorbid and clinical variables, including CTQ, DUP, and PAS (Fig. 3) The only exceptions were gender, age, and age of onset.The partial early recovery group was in-between

Table 2
Adapted outcome criteria for identifying early treatment resistance and recovery (Wold et al., 2023).the early treatment resistance and early clinical recovery groups across all investigated areas, except for alcohol-and drug use (Table 3).For a description of diagnostic distribution and diagnostic stability, se tables S1-S3.

Recovery Partial Recovery
Treatment Resistance OUTCOME Fig. 2. Conceptual mediation model of the relationship between CT and outcome at one-year follow-up with PA and DUP as mediators.For calculations of effects, see Table 6.Paths a1 and a2 depicts whether variations in the level of the independent variable significantly account for variations in the presumed mediators (PA and DUP).Paths b1 and b2 depicts whether variations in the mediators (PA and DUP) significantly account for variations in the dependent variable (outcome).Path c' illustrates the direct effect.(29 %) emotional neglect, and 34 (24 %) physical neglect (Fig. 4).There were statistically significant associations between the different CTQ subscales, except sexual abuse and emotional neglect (Table 4).

CT and outcome groups at one-year follow-up
There was a statistically significant difference in total CTQ scores between the early clinical recovery and the early treatment resistance groups, with the early treatment resistance group reporting higher CTQ scores.Also, more participants in the early treatment resistance group (55 %) reported moderate to severe CT than in the early clinical recovery group (34 %).There were significant differences in all CTQ subscores between outcome groups, with significantly higher scores for the early treatment resistance group for all sub-scores, except the sexual abuse sub-scale (Table 3).For an overview of prevalence of moderate to severe childhood trauma subtypes per outcome groups, see Fig. S1.

The association between CT, pre-treatment-, and baseline clinical characteristics
The CTQ total score was statistically significantly associated with DUP, PA, and PANSS positive and depressive symptoms at baseline.We observed different association patterns for CT subtypes (Table 4).

PA and DUP as potential mediators of the impact of CT on outcome group membership
Ordinal regression analyses (Table 5) showed that CT had a statistically significant and independent contribution to the predicted outcome group membership (2LL = 168.552,x 2 = 8.92, df.= 1, p = .003).CT was a significant predictor (b = 1.52, s.e.= 0.53, p = .004) of the probability for a participant belonging to a more severe, compared to a more benign, outcome group (OR = 4.59).In other words, for each unit increase in total (log) CTQ score, the odds of a participant belonging to a more severe outcome group increased by a factor of 4.59.
After including PA in the model, the effect of CT weakened, and the effect was no longer statistically significant after also adding DUP (2LL = 257.273,x 2 = 28.08,df.= 3, p < .001),suggesting that PA and DUP mediated the effects of CT.The final model containing the complete set of three predictors had a 9.8 % improvement in the model fit relative to the intercept-only model containing only CT (McFadden pseudo R 2 = 0.98), considered consistent with a strong improvement in model fit (Tabachnick and Fidell, 2013).
We assessed the mediating role of PA and DUP on the relationship between CT and one-year outcome and found a significant indirect effect of the impact of CT on outcome group membership through PA (b = 0.074, t = 1.323).We also found an indirect effect of the impact of CT on outcome group membership through DUP (b = 0.193, t = 2.078).Hence, both PA and DUP mediated the relationship between CT and one-year outcome.Mediation summary is presented in Table 6.

Discussion
Forty-five % of the FEP sample reported moderate to severe CT, with emotional abuse, emotional neglect, and physical neglect as the most common subtypes.The early treatment resistance group reported more prevalent and severe CT than the early clinical recovery group as measured by the CTQ total score and CTQ subtypes, except sexual abuse.There were, as hypothesized, statistically significant associations between CT, DUP, and PA, as well as statistically significant associations between DUP, PA, and outcome group, suggesting an interplay between CT and other pre-treatment characteristics.Test of potential mediation The high prevalence of CT in FEP align with previous findings (Croft et al., 2019;Gibson et al., 2016).Our findings neither confirm nor disconfirm the possibility of different mechanistic effects of different types of traumas (Grindey and Bradshaw, 2022;Bentall et al., 2014), nor the possibility of dose-response relationships (Gibson et al., 2016).We, however, found a specifically strong association between emotional abuse and depressive symptoms (Pruessner et al., 2019).This may imply that CT subtypes could both directly affect the risk of early treatment resistance and may influence other aspects of symptom expression.
Previous studies of the association between CT and early outcomes of FEP are not equivocal (Ottesen et al., 2021;Trotta et al., 2016;Pruessner et al., 2021).The first years of treatment have the highest variability in individual clinical development.While psychotic-and affective symptoms show relatively prompt responses to interventions, negative cognitive symptoms change more slowly (Lyngstad et al., 2020;Flaaten et al., 2022).One potential reason for conflicting results in studies of the association between CT and early outcomes in FEP is, thus, the timing and type of outcome assessments.Retrospective studies indicate that treatment resistance is established during the first treatment period in 70-80 % of cases (Demjaha et al., 2017;Lally et al., 2016).Treatment resistance may thus be a stable phenotype that is discernible at this early stage and, accordingly, well suited for use in prediction studies.Our findings indicate that CT may be associated with a specific aspect of the outcome, i.e., early indications of treatment resistance, in line with previous findings from multi-episode patients (Hassan and De Luca, 2015).
Previous studies have also shown that both DUP and PA have strong independent effects on different aspects of outcome (Hegelstad et al., 2021;Veru et al., 2022;Lovretić et al., 2022;Fond et al., 2018), and it was recently shown that CT contributes to a longer DUP (Veru et al., 2022).There is a complex relationship between factors contributing to the risk of treatment resistance.To what extent these are accounted for in analyses may thus lead to apparent differences in study findings.A recent paper reporting on sex, gender, and outcome of FEP found a mediating effect of PA on the relationship between CT and clinical outcome (Luckhoff et al., 2023).To our knowledge, the extent to which DUP influences the associations between CT and outcome has not been explored previously.Since the measures of trauma are based on childhood reports, i.e., before the onset of psychosis, it is reasonable to conclude that DUP mediates the effects of CT on outcome and not vice versa.In addition, we do not have any theoretical or empirical grounds for hypotheses about unmeasured confounders that affect CT, DUP, and PA in the exposure-outcome analysis, exposure-mediator analysis, or mediator-outcome analysis.This may indicate that earlier intervention (shortening the DUP) may partly ameliorate the negative effects of CT on outcomes.
Due to the low incidence of first-episode psychosis, recruitment for this prospective cohort took place over several years.The participants in the current cohort had individual contact with their therapists and received antipsychotic medications supplemented by needs-based assertive outreach or short-term hospitalizations.No specialized trauma-based treatments for psychotic disorders were available in the recruitment period, and no participants received trauma-directed   (Hardy et al., 2024;Hardy et al., 2022).Based on the findings of the role of CT on outcomes, the next step could be to investigate whether these treatment approaches also have an effect on early psychosis outcomes.

Strengths and limitations
This study has several strengths, including a well-characterized sample of FEP participants in a prospective naturalistic longitudinal design.The participants were recruited from all treatment units in a large catchment area without pre-selection and assessed by trained personnel, ensuring that the study sample was representative.
Using CTQ, we relied on participants' recollection of exposure to CT, which can be limited by recall bias.Several validation studies of the CTQ exist (Georgieva et al., 2021), and it has been shown to have adequate reliability and sensitivity (Lardinois et al., 2011;Stowkowy et al., 2016;Edwards et al., 2001).
The somewhat limited sample size could affect statistical power.

Conclusion
Forty-five % of the participants in the current FEP sample reported experiences of CT, with emotional abuse, emotional neglect, and physical neglect as the most prevalent subtypes.CT was more prevalent in the early treatment resistance group, indicating that CT may contribute to early treatment resistance.We also identified a potential mediating effect of DUP and PA on the relationship between CT and outcome at one-year follow-up.Our findings are consistent with the hypothesis that CT could have a causal influence on the outcome of FEP.If so, identifying modifiable mediators could inform tertiary prevention-and promote early treatment strategies.Our finding also underlines the need for trauma assessment in FEP to avoid diagnostic unclarity that may increase treatment delays and the need to incorporate adapted trauma treatments into FEP treatment when relevant.
Abbreviations: PANSS, Positive and Negative Syndrome Scale; GFS, Global functioning scale.

Fig. 3 .Fig. 4 .
Fig. 3. Levels of Childhood Trauma measured by total CTQ scores, DUP in weeks, and Premorbid social adjustment measured by the PAS across outcome groups.

Table 3
Demographic, pre-treatment, baseline, and one-year clinical characteristics in total sample and between outcome groups.

Table 4
Correlations between CT, CT subscales, premorbid and baseline characteristics to determine significance for inclusion in mediation analysis.

Table 5
Ordinal regression analysis (PO model) showing variables that had statistically significant independent contributions to the predicted outcome group membership at one-year follow-up.

Table 6
Mediational analysis of the impact of CT on outcome group membership, mediated by PA and DUP. .There are recent reports of promising results for several different trauma-focused psychosocial interventions for patients with psychosis Abbreviations: CT, Childhood Trauma; CI, confidence interval; PA, premorbid adjustment; DUP, duration of untreated psychosis.K.F.Wold et al.interventions