Duration of Untreated Psychosis in First-Episode Psychosis is not Associated With Common Genetic Variants for Major Psychiatric Conditions: Results From the Multi-Center EU-GEI Study

Abstract Duration of untreated psychosis (DUP) is associated with clinical outcomes in people with a diagnosis of first-episode psychosis (FEP), but factors associated with length of DUP are still poorly understood. Aiming to obtain insights into the possible biological impact on DUP, we report genetic analyses of a large multi-center phenotypically well-defined sample encompassing individuals with a diagnosis of FEP recruited from 6 countries spanning 17 research sites, as part of the European Network of National Schizophrenia Networks Studying Gene-Environment Interactions (EU-GEI) study. Genetic propensity was measured using polygenic scores for schizophrenia (SZ-PGS), bipolar disorder (BD-PGS), major depressive disorder (MDD-PGS), and intelligence (IQ-PGS), which were calculated based on the results from the most recent genome-wide association meta-analyses. Following imputation for missing data and log transformation of DUP to handle skewedness, the association between DUP and polygenic scores (PGS), adjusting for important confounders, was investigated with multivariable linear regression models. The sample comprised 619 individuals with a diagnosis of FEP disorders with a median age at first contact of 29.0 years (interquartile range [IQR] = 22.0–38.0). The median length of DUP in the sample was 10.1 weeks (IQR = 3.8–30.8). One SD increases in SZ-PGS, BD-PGS, MDD-PGS or IQ-PGS were not significantly associated with the length of DUP. Our results suggest that genetic variation does not contribute to the DUP in patients with a diagnosis of FEP disorders.


Introduction
Despite historical pessimism about schizophrenia prognosis, 1 it has now been recognized that interventions at the onset of first-episode psychosis (FEP), which is an umbrella term used to refer to schizophrenia spectrum disorders or related psychotic disorders, can improve subsequent illness outcomes. 2,3 This recognition has led to development of early intervention services, which are founded on an assumption that duration of untreated psychosis (DUP), defined as the time from manifestation of first psychotic symptoms to initiation of adequate treatment, 4 influences treatment outcomes. [3][4][5] Despite the widespread introduction of early intervention services, however, individuals suffering with FEP still experience delays of approximately 1-2 years between onset of first psychotic symptoms and initiation of treatment, 6 prompting fears of serious consequences on patients' lives, including enduring deficits and disability. 6 It is, of course, possible that the relationship between DUP and psychosis outcomes may be a product of other factors 7,8 related to the organization of mental health system, treatment-seeking behaviors, quality of available treatment, 9 or poor premorbid functioning. Seen in this way, DUP may be a marker of the illness severity rather than a predictor of the illness itself.
Schizophrenia is a highly heritable disorder, with twin studies estimating its heritability to be more than 75%. 10 Genomic studies revealed that the genetic architecture of schizophrenia comprises multiple common risk alleles scattered across the whole genome. 11 Built on the results from the genomic studies, polygenic scores (PGS) analyses confirmed that schizophrenia is highly polygenic nature, 10,11 where its onset is influenced by many common genetic variants of small effects. [12][13][14] Further evidence highlighted that the impact of the combined effect of common genetic markers for schizophrenia, as measured with polygenic score for schizophrenia, extends beyond schizophrenia diagnosis. Indeed, a higher polygenic score for schizophrenia was shown to associate with more severe negative symptoms 15 ; whereas, longer DUP is also associated with more severe negative symptoms at first presentation. 16 Considering negative symptoms were linked to cognitive impairments and deficiencies in social and occupational domains in people with a diagnosis of schizophrenia, 2 all of which contribute to prolonged delay in seeking help, 17 it is feasible that length of DUP might be influenced by genetic factors. 8 However, this question has not been investigated.
Because the etiology of FEP is highly multifactorial, it is likely that other factors may have an important impact on the delay between onset of first psychotic symptoms and initiation of adequate treatment. Certainly, the length of DUP were shown to be influenced by reduced cognitive functioning or intelligence, 18 severity of depressive symptoms reported in patients with a diagnosis of FEP disorders 19 and bipolar disorders. 20,21 Similar to schizophrenia, PGS analyses showed that major depressive disorder, bipolar disorders, and cognition are highly polygenic in nature, [22][23][24] with an overlapping, though to varying degree, genetic underpinnings. For example, PGS that combined the additive effect of common genetic markers associated with bipolar disorders discriminated individuals with a diagnosis of schizophrenia 25 and major depressive disorder from healthy controls. 10,25 Although much uncertainty remains about their ultimate clinical utility, 26 PGS have the power to considerably advance our knowledge of the underlying nature of complex phenotypes. [27][28][29] Therefore, aiming to obtain insights into possible origins of DUP in people with a diagnosis of FEP disorders, we investigated associations between DUP and PGSs for schizophrenia, bipolar disorders, major depressive disorder, and cognition in a large multi-center phenotypically well-defined sample of individuals with a diagnosis of FEP disorders. Because the length of DUP in individuals with a diagnosis of schizophrenia spectrum disorders is reported to be considerably longer compared to other psychotic disorders, 17,30 we additionally investigated if our findings were applicable to all individuals with FEP disorders or were specific to patients with first-episode schizophrenia spectrum disorders. We hypothesized that there will be a positive association between polygenic propensity for schizophrenia, bipolar disorder, major depression, and intelligence with longer DUP in participants with a diagnosis of FEP disorders.

Sample
Participants were recruited and assessed as part of the incidence and first episode case-control study, conducted as part of the EUropean network of national schizophrenia networks investigating Gene-Environment Interactions (EU-GEI) study. 31 EUGEI study was designed to investigate risk factors for psychotic disorders between May 2010 and April 2015 in tightly defined catchment areas in 17 sites across 6 countries, which were the United Kingdom, The Netherlands, France, Spain, Italy, and Brazil. 32 The research sites within each country were purposefully selected to include a mix of urban and rural areas. 31,32 The inclusion criteria for FEP cases were: (1) presentation with a clinical diagnosis for a FEP as defined by International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) criteria 31 (codes F20-F33) within the timeframe of the study; (2) aged between 18 and 64 years (inclusive); and (3) resident within one of the 17 defined catchment area at the time of their first presentation to psychiatric services for psychosis. Because the construction of PGSs is dependent on the availability of the summary statistics from genome-wide association studies (GWASs), which are currently based on population of European descent, 33 for the purpose of the present study we limited participants to those who self-reported to be of European ethnicity. Exclusion criteria were: (1) a previous contact with specialist mental health services for psychotic symptoms outside of the study period at each site; (2) evidence of psychotic symptoms precipitated by an organic cause (ICD-10: F09); (3) transient psychotic symptoms resulting from acute intoxication (F1x.5); (4) severe learning disabilities, defined by an IQ less than 50 or diagnosis of intellectual disability (F70-F79); and (5) insufficient fluency of the primary language at each site to complete assessments. 31

Ethical Approval
All participants who agreed to take part in the casecontrol study provided informed, written consent following full explanation of the study. Ethical approval for the study was provided by relevant local research ethics committees in each of the study sites. 32

Assessments
Socio-Demographic Characteristics. Using the Medical Research Council Sociodemographic modified Questionnaire version, 34 data on socio-demographic characteristics, including gender and country of birth, at the time of the first contact with mental health services for psychosis were collated at each research site. Age at first contact was defined as the age at which a patient was in contact with mental health services for the first time due to their psychotic symptoms. Ethnicity was self-ascribed from the 16 35 was used to measure DUP, based on the assessment interview and mental health records, and defined in weeks as the difference between the date of the first positive psychotic symptom (hallucination, delusion or thought disorder-rated as 4 [moderate-severe] or higher on the Positive and Negative Syndrome Scale [PANSS]) 36 and the date of initiation of antipsychotic treatment. 37 The NOS scale provides a standardized and reliable way of recording early changes in psychosis and identifying relatively precise time points for measuring several durations in emerging psychosis. 35 The Operational Criteria Checklist (OPCRIT) 38 systems, whose reliability was assessed before and throughout the study (k = 0.7), was used by trained investigators to assess psychopathology in the first 4 weeks after the onset and generate research-based diagnoses based on ICD-10 diagnostic classification systems. 39 In the present study, diagnoses were grouped using ICD-10 codes into schizophrenia-spectrum disorders (F20-29), bipolar disorder (F30, F31), psychotic depression (F32, F33), and other psychosis.

Genetic Data
Samples were genotyped at the MRC Centre for Neuropsychiatric Genetics and Genomics in Cardiff (the United Kingdom) using a custom Illumina HumanCoreExome-24 BeadChip genotyping array covering 570038 genetic variants (Illumina Inc.). Quality Control. Quality control (QC) entailed removing samples based on call rate (<0.99), genotypephenotype mismatched information, suspected non-European ancestry, heterozygosity, and relatedness. Single-nucleotide polymorphism (SNPs) were excluded if the minor allele frequency was 5%, if more than 2% of genotype data were missing and if the Hardy-Weinberg Equilibrium P-value < 10 −6 ; nonautosomal markers were also removed. The baseline characteristics of participants who were genotyped or were not genotyped are provided in supplementary table 1. To account for any ancestry differences in genetic structures that could bias results, principal components analysis was conducted retaining top principal components (PCs). 40 Individuals of European ancestry were defined as having PC values within 6 SDs from the mean PC of the EUR in 1000G. Top 20 PCs were retained to adjust for possible population stratification in the association analyses. 40,41 Polygenic Scores. To calculate polygenic score for schizophrenia (SZ-PGS), bipolar disorder (BD-PGS), major depressive disorder (MDD-PGS) and intelligence (IQ-PGS), we used the summary statistics from the latest and largest genome-wide association studies 10,[22][23][24] utilizing PRSice 42 where quality-controlled SNPs were pruned using clumping procedure which allowed to obtain SNPs in linkage equilibrium with an r 2 < 0.25 within a 250 kb window. Each PGS was calculated using subsets of the total SNPs based on the P-value threshold of .05. The selected P value threshold of .05 for SNP inclusion was chosen based on evidence showing that it explains the most variance. 10,[22][23][24] To aid interpretability of the results, all PGSs were standardized to a mean of 0 (SD = 1).

Statistical Analysis
All analyses reported in the present study were performed using RStudio version 4.0.3. 43 Imputation of Missing Values. In the present study, unemployed (22.9% missing), DUP (15.3% missing), diagnosis (1.8% missing) and living alone (1.0% missing) variable had missing values (supplementary table 2). To avoid using an unrepresentative sample of complete cases that may result in incorrect risk predictions, 44,45 we conducted an imputation to handle the missing data. To impute the missing values, we employed missForest, 46 which is an iterative imputation method based on Random Forests. It handles continuous and categorical variables equally well and accommodates non-linear relation structures. 46 miss-Forest has been shown to outperform the well-known imputation methods, such as k-nearest neighbors and parametric multivariate imputation by chained equations. 46 To evaluate the quality of imputation, we estimated the imputation error Normalized Root Mean Squared Error (NRMSE) for continuous variables and proportion of falsely classified (PFC) for categorical variables. 46,47 A value close to 0 represents an excellent performance, and a value of 1 indicates poor performance. The imputation of the missing values yielded a minimal error (NRMSE = 0.08%; PFC = 0.13%), highlighting that the imputed values were very closely aligned with the observed values for both continuous and categorical variables. The distribution of the variables included in the analyses before and after the imputation are presented in supplementary table 3. Calculate Power and Predictive Accuracy of PGS. Using information on sample size (n), total number of independent markers in genotyping panel (m) and lower and upper P-values to select markers into polygenic score (p0, p0.5) we estimated the predictive accuracy (R 2 ) present in each PGS employed in the present study using Avengeme package implemented in R. 43 Consequently, using n = 619, and the number of SNPs included in PGS for schizophrenia (m = 26 281), bipolar disorder (m = 18 092), MDD (m = 19 508) and intelligence (m = 24 386), we estimated predictive accuracy for each PGS showing that SZ-PGS (R 2 = 0.134, P = 7.40 × 10 −23 ), BD-PGS (R 2 = 0.005, P = .044), MDD-PGS (R 2 = 0.036, P = 1.05 × 10 −6 ) and IQ-PGS (R 2 = 0.077, P = 4.24 × 10 −13 ) had sufficient, as indicated by significant P-values, predictive accuracy to be employed in the analyses. Regression Modeling. As the frequency distributions of DUP are severely skewed, DUP was normalized by taking the logarithm to base 10 (log 10 DUP) to allow the use of parametric regressions. Following log-transformation, log 10 DUP was normally distributed; distribution of DUP after normalization is presented in supplementary figure 1; the results from the correlations between log 10 DUP and each PGS are provided in supplementary table 5. For each PGS, 2 linear regression models were fitted to understand the role of covariates on the potential relationship of DUP with PGSs: Model 1: crude (unadjusted) model investigating an association between each PGS and DUP; Model 2: Model 1 plus adjusting for age at first contact with mental health services for psychosis, gender, genetic ancestry as measured with first 4 PCs, research sites and educational attainment. To measure prediction accuracy of each PGS, we utilized the incremental R 2 , which was calculated following the previously outlined steps. 48 Specifically, to calculate R 2 value for each model, we first regressed a phenotype on our set of controls without the PGS; we then re-ran the same regression but with the PGS included as a regressor. Sensitivity Analyses. To examine whether our findings were applicable to individuals with a diagnosis of FEP disorders or were specific to people with a diagnosis of first-episode schizophrenia spectrum disorders only, we repeated the analyses limiting them to those who received the diagnosis of schizophrenia spectrum disorders on the first contact with mental health services. We further investigated if the results would remain the same using unimputed (complete cases) variables. As this was an exploratory study, which does not strictly require adjustment for multiple comparisons, 49 we did not employ correction for multiple testing. All tests for analyses were 2-tailed; P-values ≤ .05 were considered statistically significant.

Sample Characteristics
The demographic characteristics of the analytic sample of FEP cases are presented in table 1. The sample comprised 619 (86.6% of N = 715) individuals of European ancestry for whom quality-controlled genome-wide genotyping and DUP were available. Those participants who were included in the study or excluded from the final cohort did not differ in terms of DUP, gender, marital status, employment, living arrangement, and diagnoses; however, the former group included participants who were younger (t (1112.5) = −2.31, P = .021) and had a lower educational attainment (x 2 (1) = 4.72, P = .030) compared to those who were included in the study (supplementary table 4

Length of DUP by European Countries and FEP Diagnoses
The median length of DUP in the whole sample was 10.  (supplementary table 6). When analyses were limited to participants with a diagnosis of first-episode schizophrenia spectrum, we did not find significant associations between each polygenic score and DUP in unadjusted and fully adjusted models (supplementary table 7).

Discussion
To our knowledge, this is the first study investigating the relationship of polygenic propensity for schizophrenia, bipolar disorder, major depressive disorder, and intelligence with duration of untreated psychosis.
Consistent with previous reports, 50-52 our findings showed that individuals with a diagnosis of FEP disorders had to endure a prolonged period coping with symptoms of psychosis without seeking appropriate treatments; though, this was heavily skewed with a smaller subset of participants experiencing over a year before first contact with mental health services. Similar to previous reports, 17,30 we observed that the median length of DUP in participants with schizophrenia spectrum disorders was significantly longer when compared to all other psychoses. These observed delays highlight that there is still a great need to improve recognition of the symptoms of FEP, including schizophrenia and pathways to care.
The neurodevelopmental theory of schizophrenia posits that genetic factors interfere with early brain development leading to the development of schizophrenia symptoms. 53 This, in combination of accumulated evidence for polygenicity of schizophrenia, 10 alluded to a possibility that the length of DUP might also be influenced by additive effect of multiple common genetic markers linked to schizophrenia. 8 However, this hypothesis was not confirmed by our findings. We further considered that high genetic predisposition for either bipolar disorder or major depression disorder might be associated with DUP in people with a diagnosis of FEP disorders. Once again, our findings were negative, as they were for PGS for intelligence. It may be argued that a limited power might have led to these non-significant results. Because the PGS employed in this study were built using the results from most recent and largest GWAS metaanalyses, our analyses were not constrained by our sample size. 43,54 Nonetheless, to ensure we captured the true polygenic contribution to DUP, we undertook calculations of power for each polygenic score, which revealed that there was considerable predicative power in each PGS to detect potential associations. Our results are further in line with a recent literature review highlighting that evidence of an association between DUP and brain structure in people with a diagnosis FEP disorders was minimal. 55 It is further argued that any relationships observed between untreated psychosis and psychosis illness course appears to be explained by lead-time bias. 7 Accordingly, those with a short DUP are in an earlier stage and therefore are likely to have better outcomes than those with a long DUP, who are in a later stage. 7 Cumulatively, our findings shed some doubt on the notion that genetic variation has substantial impact on DUP. 56 In light of these findings, a discussion of some alternative theories explaining the length of DUP is warranted. It has been suggested that the length of delay from first manifestation of psychotic symptoms to initiation of adequate treatment may be influenced by factors related to the organization of mental health system and process of referral to an appropriate service FEP. Reduced allocated resources for early intervention services 57 and limited availability of care 58 may also be important contributing factors to longer DUP. The lack of knowledge of what constitutes psychosis onset 5,37 and what help may be available for people affected by early psychosis and their families 59 were shown to be important factors influencing DUP. 5,37 The longer delays to seeking help for first-episode psychotic disorders were further linked fear of stigma. Therefore, DUP may be significantly reduced through educational and anti-stigmatizing campaign about the signs of early psychosis targeted at health care providers, public, and schools increasing the motivation to seek treatment. 58 Although evidences regarding successfulness of specific interventions in reducing DUP are still lacking and largely non replicated, 60 our findings should encourage the identification of potentially effective initiatives.

Methodological Strengths and Limitations
This is an extensive multi-site study of FEP with comprehensive data on a variety of environmental and genetic factors. The study included all incidence cases from welldefined catchment areas in 17 sites across 6 countries. As our analyses were focused on people with a diagnosis of FEP, the findings reported in the present study are less likely to be biased toward patients who experience multiple hospital admissions. 61,62 Given that our study was carried out in major urban and rural sites with heterogeneous populations suggest that the generalizability of our findings may extend to other centers with similar population profiles. Finally, because the calculation of PGS is based on well-powered GWASs, we did not require a large sample to test our hypotheses, which was further confirmed by estimated predictive power in each PGS employed in the analyses.
Nonetheless, important methodological considerations warrant a discussion. While it is likely most individuals who develop a psychotic disorder do present to services, at least in sites with well-developed public health systems, some who do not present will be missed and this may introduce selection biases. 31 Variations in referral procedures of patients with psychosis from primary to secondary mental health care settings and in the organization of secondary mental health care services across  Funders were not involved in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript and decision to submit the manuscript for publication. The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.