European Neuropsychopharmacology 47 (2021) 112–129 www.elsevier.com/locate/euroneuro Identifying clinical clusters with distinct trajectories in first-episode psychosis through an unsupervised machine learning technique Silvia Amoretti a,b,c,1, Norma Verdolini b,c,1, Gisela Mezquida a,b, Francisco Diego Rabelo-da-Ponte d, Manuel J Cuesta e,f, Laura Pina-Camacho b,g, Marta Gomez-Ramiro a, Concepción De-la-Cámara a,h, Ana González-Pinto b,i, Covadonga M. Díaz-Caneja b,g, Iluminada Corripio b,j, Eduard Vieta b,c,∗, Elena de la Serna b,k, Anna Mané b,l,m, Brisa Solé b,c, André F Carvalho n, Maria Serra c, Miguel Bernardo a,b a Barcelona Clinic Schizophrenia Unit, Hospital Clinic of Barcelona, August Pi I Sunyer Biomedical Research Institute (IDIBAPS), Neuroscience Institute, University of Barcelona, Spain b Biomedical Research Networking Center for Mental Health Network (CIBERSAM), Spain c Bipolar and Depressive Disorders Unit, Hospital Clinic, University of Barcelona, Institute of Neuroscience, IDIBAPS, CIBERSAM, 170 Villarroel st, 12-0, 08036, Barcelona, Catalonia, Spain d Laboratory of Molecular Psychiatry, Hospital de Clínicas de Porto Alegre, Federal University of Rio Grande do Sul, Brazil e Department of Psychiatry, Complejo Hospitalario de Navarra, Pamplona, Spain f Navarra Institute for Health Research (IdiSNA), Pamplona, Spain g Department of Child and Adolescent Psychiatry, Institute of Psychatry and Mental Health, Hospital General Universitario Gregorio Marañon, IiSGM, School of Medicine, Universidad Complutense, Madrid, Spain h Hospital Clínico Universitario and Instituto de Investigación Sanitaria (IIS), Department of Medicine and Psychiatry. Universidad de Zaragoza, Zaragoza, Spain i Department of Psychiatry, Araba University Hospital, Bioaraba Research Institute, Department of Neurociences, University of the Basque Country, Vitoria, Spain j Psychiatry Department, Institut d’Investigació Biomèdica-Sant Pau (IIB-SANT PAU), Hospital de la Santa Creu i Sant Pau; Universitat Autònoma de Barcelona (UAB), Barcelona, Spain ∗ Corresponding author at: Biomedical Research Networking Center for Mental Health Network (CIBERSAM), Spain E-mail address: EVIETA@clinic.cat (E. Vieta). 1The first two authors contributed equally to this work. https://doi.org/10.1016/j.euroneuro.2021.01.095 0924-977X/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/) European Neuropsychopharmacology 47 (2021) 112–129 k Department of Child and Adolescent Psychiatry and Psychology, Clínic Institute of Neurosciences, Hospital Clínic de Barcelona, 2017SGR881, University of Barcelona, IDIBAPS, Barcelona, Spain l Hospital del Mar Medical Research Institute (IMIM), Barcelona, Spain m Autonomous University of Barcelona (UAB), Barcelona, Spain n Department of Psychiatry, University of Toronto, Toronto, ON, Canada; Centre for Addiction and Mental Health (CAMH), Toronto, ON, Canada Received 6 November 2020; received in revised form 4 January 2021; accepted 18 January 2021 KEYWORDS Abstract First-episode The extreme variability in symptom presentation reveals that individuals diagnosed with a first- psychosis; episode psychosis (FEP) may encompass different sub-populations with potentially different Symptomatology; illness courses and, hence, different treatment needs. Previous studies have shown that so- Cognitive reserve; ciodemographic and family environment factors are associated with more unfavorable symp- Functioning; tom trajectories. The aim of this study was to examine the dimensional structure of symptoms Machine learning and to identify individuals’ trajectories at early stage of illness and potential risk factors as- sociated with poor outcomes at follow-up in non-affective FEP. One hundred and forty-four non-affective FEP patients were assessed at baseline and at 2-year follow-up. A Principal com- ponent analysis has been conducted to identify dimensions, then an unsupervised machine learning technique (fuzzy clustering) was performed to identify clinical subgroups of patients. Six symptom factors were extracted (positive, negative, depressive, anxiety, disorganization and somatic/cognitive). Three distinct clinical clusters were determined at baseline: mild; neg- ative and moderate; and positive and severe symptoms, and five at follow-up: minimal; mild; moderate; negative and depressive; and severe symptoms. Receiving a low-dose antipsychotic, having a more severe depressive symptomatology and a positive family history for psychiatric disorders were risk factors for poor recovery, whilst having a high cognitive reserve and better premorbid adjustment may confer a better prognosis. The current study provided a better un- derstanding of the heterogeneous profile of FEP. Early identification of patients who could likely present poor outcomes may be an initial step for the development of targeted interventions to improve illness trajectories and preserve psychosocial functioning. © 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/) complexes or dimensions, which can co-exist in the same in- dividual and cut across different types of psychoses (Peralta 1. Introduction and Cuesta, 2001). Thus, dimensions, unlike categories, are not mutually exclusive. Most dimensional studies observed The extreme variability in the presentation of symptom 4 or 5 dimensions within psychosis, with positive, negative, suggests that people diagnosed with first-episode psychosis disorganisation, affective and cognitive symptom domains (FEP) may encompass distinct sub-populations of patients most frequently reported (Potuzak et al., 2012). However, (Cocchi et al., 2014), with potentially different illness the number of dimensions identified has varied across courses and, hence, different treatment needs (Colasanti studies depending on the nature of the assessment tools et al., 2010). The identification of early clinical and sociode- used and the heterogeneity of included participants (Malla mographic features may be important in identifying subsets and Payne, 2005; Rodriguez-Jimenez et al., 2013;Compton of patients with similar characteristics, facilitating person- et al., 2014).This hampers the comparability of the studies alised treatment approaches (Compton et al., 2014). and their translation into consistent, uniform and replicable Two general models have been used so far to describe the data.Of note, a dimensional approach has been gradually varied clinical manifestations of psychosis, namely the cate- incorporated in recent diagnostic classifications. For ex- gorical and dimensional approaches. The categorical model ample, in the DSM-5 schizophrenia subtypes have been involves mutually exclusive classes, which are defined by excluded and a severity rating of core symptoms has been specific patterns of symptoms. Thus, it considers illness as included. Nonetheless, a mixed dimensional-categorical ap- being either present or absent. In line with this, the current proach to psychosis, and particularly FEP psychopathology, diagnostic systems, such as the Diagnostic and Statistical is still lacking in the currently available versions of both Manual of Mental Disorders (DSM-5) (American Psychiatric DSM and ICD (Vieta, 2016; Stein et al., 2020). Furthermore, Association, 2013) and the International Statistical Classi- clustering algorithms as fuzzy algorithm may be clinically fication of Diseases and Related Health Problems (ICD-10) useful for identifying a spectrum in subjects with FEP, since (World Health Organization, 2004) adopt a categorical it groups the subjects in a more natural way, not forcing a approach to classify psychotic disorders. On the other hand, subject to belong to one or the other group, unlike other the dimensional perspective proposes that symptoms of clustering methods (Sugranya and Shanthi, 2012; Wu et al., psychosis tend to cluster together within different symptom 2020). 113 S. Amoretti, N. Verdolini, G. Mezquida et al. The symptoms trajectories among patients with FEP are centres located throughout the Spanish territory from April 2009 to also heterogeneous (Abdin et al., 2017). Previous stud- April 2011. For the purpose of the present study, the final sample ies have shown that sociodemographic factors including consisted of 144 FEP patients fulfilling the following inclusion crite- younger age, male gender, education and clinical factors ria: 1) aged between 18 and 35 years at the time of first evaluation; such as having poor functioning at baseline or a longer dura- 2) presence of psychotic symptoms of less than twelve months’ du- ration; 3) ability to speak Spanish correctly and 4) signed informed tion of untreated psychosis (DUP) are associated with poor consent. Exclusion criteria were: 1) mental retardation according symptom trajectories (Austin et al., 2015; Verma et al., to DSM-IV criteria; 2) history of head trauma with loss of conscious- 2012). A meta-analysis revealed a significant association be- ness and 3) physical disease with mental repercussions. To ensure tween DUP and several outcomes, including positive and more homogeneous sample diagnoses, for the current study we only negative symptoms, depression/anxiety, social functioning included patients with non-affective FEP, as we considered affec- and global outcome (Perkins et al., 2005). Furthermore, tive FEP patients as a subgroup displaying several specific charac- poor premorbid adjustment and the early appearance of teristics in terms of clinical course, functional outcome and antipsy- negative symptoms are also predictive of subsequent worse chotic treatments. We considered only FEP patients that have been response to treatment (Malla and Payne, 2005; Rabinowitz diagnosed after 2 years of follow-up with a non-affective psychosis such as schizophrenia, schizophreniform, schizoaffective disorders et al., 2006). Numerous aspects related with the family en- and psychoses that are not otherwise specified, according to DSM- vironment, such as expressed emotions, have been shown IV-TR. In particular, only FEP patients with a complete psychopatho- to influence the progression of psychosis (Hinojosa-Marqués logical assessment at baseline and two-year follow-up have been et al., 2019). Thus, these environmental factors have also included in the present study. been considered potential factors for the risk, onset and This study was conducted in accordance with the ethical prin- progression of psychosis (Varese et al., 2012). A strong re- ciples of the Declaration of Helsinki and Good Clinical Practice lationship between cognitive performance and clinical im- and the Hospital Clinic Ethics and Research Board. All participants provement in FEP has also been described (Allott et al., provided written informed consent prior to their inclusion in the 2011; Rodriguez-Sanchez et al., 2013). More specifically, in study. the field of cognition-related factors, higher cognitive re- serve (CR) is associated with a later onset of psychosis and better recovery (Barnett et al., 2006). Thus, CR has been 2.2. Assessments considered as a positive moderator of the impact of pathol- ogy on clinical course, functional outcome and cognitive 2.2.1. Clinical and sociodemographic assessment performance (de la Serna et al., 2013; Amoretti et al., 2016, Clinical and sociodemographic data were systematically obtained 2018, 2020). for all participants and included: age, gender and parental so- Given their significant impact on both short and long-term cioeconomic status (SES). SES was determined using Holling- shead’s Two-Factor Index of Social Position (Hollingshead and outcomes such as psychosocial functioning and quality of Redlich, 1958); pharmacological treatment was measured by chlor- life, identifying both baseline and follow-up clinical phe- promazine equivalents (CPZ) based on international consensus notypes, sociodemographic and family factors in FEPs could (Gardner et al., 2010);and the DUP was calculated as the number provide relevant information that might allow the defini- of days between the first manifestations of psychotic symptoms un- tion of early personalised intervention strategies and tai- til the initiation of adequate treatment for psychosis. Drug misuse lored person-focused therapy. Consequently, the aims of this habits were also collected using a version of the European Addiction study were: Severity Index (EuropASI) (KokkeviandHartgers, 1995). Diagnoses at 2-year follow-up were determined with the Struc- 1. To examine the dimensional structure of psychopatholog- tured Clinical Interview for DSM (SCID-I-II) (First et al., 1997a, ical symptom presentation in FEP patients at different 1997b) according to DSM-IV criteria. Taking into consideration po- time points. tential changes across time and in order to ensure diagnostic sta- 2. To identify clinical subgroups of patients based on the bility, the diagnoses of the patients who completed the study were psychopathological dimensions at baseline and 2-year determined based on information gathered at the two-year follow- follow-up. up visit. 3. To search for differences between the different sub- A psychopathological assessment was carried out with the PANSS groups of patients in terms of sociodemographic, famil- (Kay et al., 1987), the Montgomery-Asberg Depression Rating Scale (MADRS) (Montgomery and Asberg, 1979) and the Clinical Global ial, clinical and psychosocial functioning. Impression Scale (CGI) (Guy, 1976). Higher scores indicate greater 4. To unravel the initial heterogeneity of psychopatholog- severity. ical presentation by identifying subgroups with distinct Traumatic Experiences in Psychiatric Outpatients Questionnaire longitudinal trajectories of disease. (TQ) (Davidson and Smith, 1990)was used to assess for traumatic experiences. In this study we used the TQ dichotomous clinical cut- off scores that differentiate between the presence or absence of 2. Experimental procedures significant abuse and neglect. 2.1. Sample 2.2.2. Family assessment The Family Environment was assessed by the Family Environment The sample of this study came from a multicentre, naturalistic and Scale (FES) (Moos and Moos, 1976). FES is a self-report instru- longitudinal project "Phenotype-genotype interaction. Application ment focused on the measurement and description of the inter- of a predictive model in first psychotic episodes” (PEPs Project) personal relationships among family members. FES evaluated fam- (Bernardo et al., 2013, 2019), all under the umbrella of the Span- ily emotional climate in different categories: Cohesion (for mu- ish Research Network on Mental Health (CIBERSAM) (Salagre et al., tual reliance); Expressivity (the extent to which family members 2019). A total of 335 patients with a FEP were recruited from 16 express their feelings directly); Conflicts (for open expression of 114 European Neuropsychopharmacology 47 (2021) 112–129 anger, aggressiveness and conflict); Independence (for an achieve- 2.3.2. Unsupervised machine learning technique ment orientated environment); Achievement Orientation (the ex- The machine learning analysis was performed using R software (ver- tent to which family members are independent in their decisions); sion R 3.6.3) and RStudio (version 3.5.3). We performed the fuzzy Intellectual–Cultural Orientation (for political, intellectual, cul- analysis clustering to identify clinical subgroups of subjects with tural interests); Active–Recreational Orientation (for participation FEP at baseline and 2-year follow-up based on the psychopatholog- in social activities); Moral–Religious Emphasis (for the importance ical dimensions detected in the PCA (Lefort-Nesmard et al., 2018), given to ethical and religious practices and values); Organisation using the R package “cluster” (version 2.1.0). The optimal number (for the organization in activities and responsibilities); and Control of clusters was determined by the Bayesian Information Criterion (the extent to which the family considers rules and established pro- (BIC) (Fraley and Raftery, 1998). We performed Dann and Gamma cedures). Indexes to verify the quality of clustering at baseline as well as The family history of psychiatric disorders was assessed using a at follow-up. Furthermore, we performed a Discriminant function semi-structured interview. Then, patients were classified into two analysis (DFA) using package ‘MASS’ (version 7.3–53) at baseline and categories: 1) those with a first-degree relative with a psychiatric at 2-year follow-up to confirm the clusters retained and to investi- disorder, and 2) those patients who did not have a first-degree fam- gate the predictive power of the clustering of each subject’s psy- ily member with a psychiatric disorder. chopathological dimensions to the clinical cluster (Jensen et al., 2015). 2.2.3. Functional assessment Overall, there is two clustering methods: hard and soft (fuzzy) The overall functional outcome was assessed by means of the Func- clustering. The former divided subjects into distinct subgroups, tioning Assessment Short Test (FAST) (Rosa et al., 2007). The FAST where each subject belongs exclusively to a single group, while the scale is used to evaluate 6 specific areas of functioning: autonomy, latter the subject may belong to more than one group, which may occupational functioning, cognitive functioning, financial issues, in- indicate a continuity or developmental aspect of a disease. In the terpersonal relationships, and leisure time, and has been validated fuzzy clustering, subjects at the boundaries between multiple clus- in FEP population (González-Ortega et al., 2010). Higher scores in- ters are not forced to belong entirely to one of the classes, but are dicate worse functioning. instead assigned degrees of association between 0 and 1, indicating The Premorbid Adjustment Scale (PAS) (Cannon-Spoor et al., their partial association. This type of grouping is much more natu- 1982) was applied retrospectively to assess premorbid adjustment. ral than other methods since there are situations in which a subject Higher scores indicate worse premorbid adjustment. may have mixed characteristics (Suganya and Shanthi, 2012). Furthermore, we performed a discriminant function analysis us- ing package ‘MASS’ (version 7.3–53) to confirm the clusters retained 2.2.4. Cognitive reserve assessment and to investigate the predictive power of the clustering of each To assess CR we have used the three most commonly proposed participant’s cognitive domain to the cognitive subgroup (Jensen proxy indicators of CR in psychiatry and particularly in FEP (Barnett et al., 2015). et al., 2006; de la Serna et al., 2013; Amoretti et al., 2016, 2018, 2020) which include premorbid intelligence quotient (IQ), 2.3.3. Statistical analysis education and lifetime participation in leisure, social and physical Descriptive analyses identifying differences between the clusters activities. Because patients already experienced a FEP when we and trajectories were conducted using chi-square for categorical evaluated CR, we could only estimate premorbid variables. The variables. Group differences were examined using unpaired t-tests estimated premorbid IQ was calculated with the vocabulary subtest for normally distributed variables, or using Mann–Whitney U tests of the Wechsler Adults Intelligence Scale (WAIS-III) as a measure re- for non-normal data. When comparing groups based on clinical, flecting crystallised intelligence. The vocabulary subtest seems to sociodemographic and functional variables, analysis of variance be stable during the progression of the disease and useful as a mea- (ANOVA) was used to show the overall differences between groups sure of estimated premorbid intelligence (de Oliveira et al., 2014). with post-hoc Turkey’s HSD test for multiple comparisons. To com- The second proxy, ‘education’ was assessed taking into account the pare the performance at baseline and two-year follow-up, a re- number of years of obligatory education that subjects had com- peated measures analysis of variance (ANOVA) was used. pleted as well as parents’ educational level and the subject’s life- Finally, we decided to identify trajectories of disease. Four- time school performance. Finally, lifetime participation in leisure, trajectory groups were described as “excellent prognosis”, “remit- social and physical activities was assessed by PAS scale (scholastic ting course”, “clinical worsening” and “chronic course”. Patients performance) and by asking about involvement in social activities, with an excellent prognosis trajectory were those who exhibited ability to perform physical activities and enjoyment of hobbies. mild symptoms at baseline and minimal or mild symptoms at follow- Higher scores correspond to better performance. A Principal up, whereas those with remitting course had moderate or severe Components Analysis (PCA) was performed to create a “Cognitive symptoms at baseline and minimal or mild symptoms at follow-up. reserve score” for each subject with the three main proxies. Patients with a clinical worsening trajectory exhibited mild symp- toms at baseline and negative/depressive or severe symptoms at follow-up. Finally, patients with a chronic course trajectory showed 2.3. Analysis moderate/severe symptoms at baseline and negative/depressive or severe symptoms at follow-up. 2.3.1. Pre-processing Data were analysed using SPSS, version 23. We considered A PCA was carried out with the aim of assembling items of PANSS p<0.05 as statistically significant. and MADRS in a series of coherent components that would be able to tackle psychopathological dimensions in patients with FEP at base- line and 2-year follow-up. Items with an unclear factor loading dis- 3. Results tribution (i.e., showing similar factor loadings in more than one component) or with low factor loadings (<0.45) were excluded from 3.1. Sociodemographic characteristics of the the components. Considering the high correlation among items, we sample used an oblique rotation (Loo, 1979). The oblique rotation method allows the extent to which the factors are actually correlated to A total of 144 patients with first-episode non-affective each other to be examined, and therefore renders it more realistic psychosis were included in the present study and were theoretically and empirically (Peralta and Cuesta, 2001). re-evaluated at 2-year follow-up. Sixty-seven percent were 115 S. Amoretti, N. Verdolini, G. Mezquida et al. male with a mean age of 25.26 ± 5.13 years (age at onset classified by DFA into 93.1% of the cases, demonstrat- of psychosis was 24.91± 5.14 years). The mean dose of ing the validity of the three clinical clusters. While at 2- antipsychotic medication (prescribed by the psychiatrist) at years follow-up, the DFA showed the presence of three baseline was equivalent to 619.72 ± 422.26 mg/day of CPZ significant discriminant functions, which explained 89.9%, (dose range 0 to 2400 mg/day) and 201.71 ± 279.98 mg/day 6.1%, 3.8% of the variance, respectively (Wilks’ λ=0.036, at follow-up (range, 0 to 1350 mg/day). The mean DUP χ 2(24)=455.39, p<0.001; Wilks’ λ=0.407, χ 2(15)=123.75, was determined as 102.26 ± 126.81 days (approximately 15 p<0.001, Wilks’ λ=0.689, χ 2(8)=51.24, p<0.001). The sub- weeks). After 2 years of follow-up, the severity of clinical jects were correctly classified by DFA into 94% of the cases, symptoms improved (values of partial eta squared range that is, it confirms the presence of the five clinical clusters. from 0.217 to 0.422, medium to large effect sizes), as When comparing the clinical and sociodemographic pro- did functional scores (η2 = 0.169) (see Supplementary files, patients with mild symptoms had lower DUP, lower Table 1). dose of antipsychotic, higher premorbid adjustment (PAS), lower illness severity, lower symptoms and greater psy- chosocial functioning (FAST) at baseline. There were no dif- 3.2. Dimensional structure of symptoms ferences between clusters in terms of age, gender, SES, age of onset, alcohol or cannabis use, family and history of trau- Concerning the PCA at baseline the Kaiser-Meyer-Olkin matic events/experiences (see Table 1). Patients with pos- (KMO) value, a measure of how suited the data is for factor itive and severe symptoms reported to have a higher con- analysis, was 0.844 and 0.876 at follow-up, exceeding the sume of alcohol (monthly alcohol consumption). recommended value of 0.6, and Bartlett’s Test of Spheric- At 2-year follow-up patients with minimal symptoms ity reached statistical significance, supporting the factora- had better premorbid adjustment, greater family cohesion bility of the correlation matrix. The six factors extracted and lower conflicts, more intellectual-cultural and active- at different time points are presented in Supplementary recreational orientation, lower illness severity, lower symp- Table 2 and 3 accordingly. Factors were defined as “Dis- toms and greater functioning. Sociodemographic and clini- organisation”, “Negative”, “Depressive”, “Anxiety”, “So- cal characteristics and psychosocial functioning among clus- matic/Cognitive” and “Positive”. When we compared the ters at 2-year follow-up are presented in Table 2. PCA at baseline and at two-year follow-up, the PCA dis- played the same components, showing the excellent sta- bility of these six components. However, the components 3.4. Trajectories of disease do not have exactly the same composition at both times of Four trajectory groups were determined as “excellent prog- the evaluations and the PCA item loadings differed between nosis” (32%), “remitting course” (37%), “clinical worsening” baseline and follow-up. (9%) and “chronic course” (22%) (see Fig. 2). A moderate symptoms group was excluded because it did not clearly dif- 3.3. Clinical subgroups of patients at baseline fer from either the minimal and mild symptoms group or the and follow-up negative and moderate, and severe symptoms group. The demographic and clinical characteristics of each trajectory In relation to clustering performance, the PAM algorithm and statistical comparisons between them are presented in achieved an average Silhouette width equal to 0.71, and Table 3. Dunn index equal to 0.75 at baseline. At 2-year follow-up, Twenty percent of those patients who exhibited mild the measures were 0.77, and 0.80, respectively. The Silhou- symptoms at baseline presented “negative and depres- ette indicates how well objects are clustered, ranging from sive” or “severe” symptoms at follow-up (clinical worsen- −1 (poorly) to 1 (well clustered). The Dunn index provides ing group). This group differed from those who remained the ratio between compactness within a cluster and sepa- with “minimal” or “mild” symptoms (excellent prognosis) ration between different clusters (Hassani and Seidl, 2017; on lower rates of CPZ (t = 2.01; p = 0.050), higher depres- Tomasini et al., 2017). sive symptoms (t=−2.05; 0.046) and positive family history The fuzzy algorithm through BIC method determined of psychiatric disorders (χ ²=7.60, p = 0.022). three distinct patient clusters at baseline: “mild symptoms” Sixty-two percent of patients with “moderate” or “se- (39%), “negative and moderate symptoms” (31%) and “pos- vere” symptoms at baseline presented a remitting course. itive and severe symptoms” (30%) (see Fig. 1a). In relation They showed better premorbid adjustment (t=−2.91; to clustering performance, it achieved a Dunn index equal p = 0.005) and higher CR (t = 2.13; 0.037) compared to to 0.79 and Gamma Index of 0.81. For these indexes, the those with a chronic course. closer to 1 the better the grouping (Hassani and Seidl, 2017; Finally, those patients who showed “minimal” or “mild” Tomasini et al., 2017). At follow-up, five distinct clusters symptoms (excellent prognosis + remitting course) had were found: “minimal” (34%), “mild” (26%), “moderate” better premorbid adjustment, higher family expression (13%), “negative and depressive” (13%) and “severe symp- and active-recreational orientation at baseline compared toms” (14%) (Fig. 1b). The clustering performance achieved to those who presented “negative and depressive” or “se- a Dunn index of 0.75 and Gamma Index of 0.78. vere” symptoms at follow-up (clinical worsening + chronic Additionally, the discriminant function analysis (DFA) ex- course). At two-year follow-up, they also showed greater hibited two discriminant functions, which explained 91.8% family cohesion, expression and active-recreational orien- and 8.2% of the variance, respectively (Wilks’ λ=0.14, tation, lower illness severity, lower symptoms and greater χ 2(12)=271.58, p<0.001; Wilks’ λ=0.72, χ 2(5)=43.96, functioning (sociodemographic and clinical characteristics p<0.001) at the baseline. The subjects were correctly of these groups are presented in Supplementary Table 4). 116 European Neuropsychopharmacology 47 (2021) 112–129 Fig. 1 Distribution of clinical subgroups of subjects with first-episode psychosis at baseline and at 2-years using fuzzy clustering algorithm A) Cluster distribution at the baseline. B) Cluster distribution at the 2-year follow-up. Each axis means a reduction of symptomatol- ogy domains and each point is a subject. Fig. 2 Trajectories of symptoms in first-episode psychosis. 117 Table 1 Baseline sociodemographic, clinical, functional and cognitive performance among clusters. Mild symptoms Negative and Positive and F P Post hoc (n = 56) [1] moderate symptoms severe symptoms (Tukey’s HSD) (n = 45) [2] (n = 43) [3] Sociodemographic variables Gender: Male N(%) 37 (66) 32 (71) 27 (63) χ²=0.70 0.705 – Age (M±SD) 25.00±5.14 25.16±5.33 25.70±4.98 0.24 0.790 – χ²=6.01 S. Amoretti, N. Verdolini, G. Mezquida et al. SES (%) High 14 (25) 10 (22) 8 (19) 0.815 – Medium-High 5 (9) 5 (11) 4 (9) Medium 15 (27) 7 (16) 12 (28) Medium-Low 15 (27) 17 (38) 16 (37) Low 6 (10) 6 (13) 3 (7) DUP 74.79±81.61 143.89±180.57 89.76±84.97 3.86 0.024 [1]<[2]∗ Age of onset 24.75±5.14 24.60±5.19 25.46±4.79 0.34 0.709 – CPZ 460.92±354.99 714.33±456.64 723.04±412.74 6.59 0.002 [1]<[2–3]∗ 118 Alcohol: Yes N (%) 30 (46) 25 (56) 30 (70) χ²=2.96 0.227 – Monthly alcohol 22.70±38.60 22.47±49.33 64.77±96.85 6.488 0.002 [1–2]<[3]∗ Cannabis: Yes N (%) 22 (39) 23 (51) 23 (54) χ²=2.37 0.306 – Monthly cannabis 35.88±62.73 62.96±125.56 27.05±41.05 2.262 0.108 – Trauma: Yes N (%) 34 (61) 24 (53) 27 (49) χ²=2.58 0.630 – PAS 38.96±20.01 52.28±24.42 45.86±19.33 4.52 0.013 [1]<[2]∗ Cognitive Reserve 74.62±12.63 72.94±10.07 77.32±11.29 1.46 0.235 – Family History: Yes N (%) 13 (23) 11 (24) 9 (21) χ²=0.16 0.924 Cohesion (C) 49.74±7.51 47.76±11.42 48.54±7.89 48.77±8.97 0.562 – Expressivity (EX) 50.83±9.66 49.44±10.10 47.67±10.06 49.47±9.93 0.322 – Conflicts (CON) 48.62±7.41 51.29±9.90 51.13±7.86 50.18±8.41 0.221 – Independence (IND) 49.57±9.16 51.02±9.38 48.18±9.15 49.61±9.22 0.389 – Achievement orientation (AO) 45.96±10.30 47.17±8.38 49.33±8.67 47.32±9.31 0.229 – Intellectual–cultural 46.92±9.67 49.05±10.30 46.51±11.14 47.46±10.29 0.487 – orientation (ICO) (continued on next page) Table 1 (continued) Active–recreational orientation 51.26±8.09 50.59±8.32 45.95±6.59 49.50±8.04 0.004 [1–2]>[3]∗ (ARO) Moral–religious emphasis (MRE) 46.26±7.88 45.10±7.54 45.67±8.30 45.73±7.86 0.776 – Organisation (ORG) 51.19±9.28 47.78±11.29 48.77±9.97 49.43±10.17 0.245 – Control (CTL) 47.94±8.85 48.02±9.52 49.69±5.98 48.48±8.32 0.561 – European Neuropsychopharmacology 47 (2021) 112–129 Clinical Variables PANSS positive 11.39±5.05 19.27±5.81 25.33±6.25 75.06 <0.001 [1]<[2–3]∗∗ , [2]<[3]∗∗ PANSS negative 13.39±5.40 23.56±5.53 20.77±6.91 40.13 <0.001 [1]<[2–3]∗ PANSS general 25.04±4.74 40.40±5.65 49.63±8.57 189.57 <0.001 [1]<[2–3]∗∗ , [2]<[3]∗∗ PANSS total 49.82±10.17 83.22±11.00 95.72±14.96 196.35 <0.001 [1]<[2–3]∗∗ , [2]<[3]∗∗ MADRS 6.05±5.39 10.91±6.58 20.65±8.50 56.62 <0.001 [1]<[2]∗ , [1–2]<[3]∗∗ 119 CGI 3.91±1.10 4.38±1.03 4.63±0.95 6.18 0.003 [1]<[3]∗ Psychosocial functioning Autonomy 2.96±2.97 3.91±3.37 6.07±3.25 4.20±3.42 <0.001 [1]<[3]∗∗ , [2]<[3]∗ Occupational 6.53±5.79 8.89±4.64 10.00±4.51 8.31±5.26 0.003 [1]<[3]∗ Cognitive 3.60±3.03 6.11±3.62 7.88±3.47 5.68±3.78 <0.001 [1]<[2]∗ , [1]<[3]∗∗ , [2]<[3]∗ Financial 0.76±1.57 1.30±1.64 2.49±1.84 1.45±1.82 <0.001 [1]<[3]∗∗ , [2]<[3]∗ Interpersonal relationships 4.07±4.19 7.73±4.84 9.86±4.01 6.96±4.97 <0.001 [1]<[2–3]∗ Leisure 1.58±1.63 2.43±1.72 2.81±1.94 2.22±1.82 0.002 [1]<[2–3]∗ FAST total 19.51±13.34 30.36±14.27 39.12±13.73 24.93 <0.001 [1]<[2–3]∗∗ , [2]<[3]∗ Abbreviations: HSD= Honestly Significant Difference; M=Mean; SD= Standard Deviation; SES=Socioeconomic status; DUP= Duration of untreated psychosis; CPZ= Chlorpromazine equiv- alents; PAS= Premorbid Adjustment Scale; PANSS= Positive and Negative Symptom Scale; MADRS= Montgomery-Asberg Depression Rating Scale; CGI=Clinical Global Impression Scale; FAST=Functioning Assessment Short Test. Significant differences (p<0.05) marked in bold. ∗ p < 0.05, ∗∗ p < 0.001. Table 2 Sociodemographic and clinical characteristics and psychosocial functioning among clusters at 2-year follow-up. Minimal Mild symptoms Moderate Negative and Severe F p Post hoc (Tukey’s symptoms (n = 37) [2] symptoms depressive symptoms HSD) (n = 49) [1] (n = 19) [3] symptoms (n = 20) [5] (n = 19) [4] Sociodemographic variables Gender: Male N(%) 34 (69) 25 (68) 12 (63) 13 (68) 12 (60) χ²=0.71 0.950 – Age (M±SD) 25.18±5.21 25.78±4.75 22.89±4.86 24.68±5.55 27.25±4.91 1.98 0.101 – SES (%) High 13 (27) 8 (22) 4 (21) 2 (11) 5 (25) χ²=17.76 0.603- Medium-High 4 (8) 4 (11) 3 (16) 2 (11) 1 (5) Medium 14 (29) 9 (24) 5 (26) 1 (5) 5 (25) Medium-Low 11 (22) 13 (35) 5 (26) 10 (53) 9 (45) Low 6 (12) 3 (8) 2 (11) 4 (21) 0 (0) S. Amoretti, N. Verdolini, G. Mezquida et al. Missing 1 (2) 0 (0) 0 (0) 0 (0) 0 (0) DUP 78.33±90.22 98.29±94.64 183.50±238.38 87.50±110.37 111.84±120.82 2.22 0.071 [1]<[3]∗ ANOVA not significant, Post-hoc test significant Age of onset 24.75±5.34 25.58±4.66 22.33±5.05 24.42±5.33 26.85±4.80 2.12 0.082 – CPZ 150.21±222.16 188.67±240.03 297.63±414.53 193.16±295.99 269.04±298.57 1.30 0.275 – Alcohol: Yes N (%) 29 (59) 24 (65) 10 (53) 13 (68) 9 (45) χ²=3.164 0.531 – Monthly alcohol 25.80±51.47 51.35±85.49 42.05±74.52 16±25.23 40±76.96 χ²=1.264 0.287 – 120 Alcohol FUP 29 (59) 24 (65) 10 (53) 12 (63) 8 (40) χ²=5.90 0.658 – Monthly alcohol FUP 25.73±41.73 22.24±36.02 34.42±89.84 32.16±50.34 12.40±30.31 χ²=0.630 0.642 – Cannabis: Yes N (%) 22 (45) 18 (49) 9 (47) 12 (63) 7 (35) χ²=3.271 0.513 – Monthly cannabis 27.33±48.82 39.97±63.44 37.11±56.89 41.11±55.86 85.05±176.27 χ²=1.738 0.145 – Cannabis FUP 8 (32) 5 (14) 5 (26) 4 (21) 3 (15) χ²=3.67 0.886 – Monthly cannabis FUP 8.76±26.52 11±34.23 2.74±7.16 23.68±49.57 11.95±32.67 χ²=1.136 0.342 – Trauma: Yes N (%) 25 (32) 22 (60) 14 (74) 9 (47) 9 (45) χ²=8.69 0.371 – PAS 40.53±19.69 39.64±18.34 53.89±25.49 54.63±22.13 48.84±24.20 2.97 0.022 Post hoc not significant Cognitive reserve 76.83±12.41 76.22±9.68 72.18±10.99 72.59±12.26 72.25±12.37 1.10 0.358 – Family History: Yes N (%) 11 (22) 6 (16) 3 (16) 4 (21) 9 (45) χ²=7.05 0.133 – Cohesion (C) 54.58±7.79 51.36±9.12 43.38±10.37 47.43±8.86 50.75±13.04 4.37 0.003 [1]>[3]∗ Expressivity (EX) 52.90±7.79 54.68±8.56 48.15±12.62 47.71±9.43 47.38±9.96 2.45 0.051 – Conflicts (CON) 47.38±5.17 47.32±4.62 54.69±10.21 48.29±7.41 46.00±4.04 4.13 0.004 [1–2]<[3]∗ , [3]<[5]∗ Independence (IND) 53.25±8.82 54.44±7.90 51.23±9.88 49.93±10.59 51.88±7.95 0.71 0.584 – Achievement orientation (AO) 48.58±9.28 44.80±8.20 43.69±8.62 44.93±9.07 43.38±13.80 1.27 0.288 – Intellectual–cultural orientation 53.48±11.29 49.52±11.94 41.92±10.36 44.57±9.07 51.00±13.96 3.41 0.012 [1]<[3]∗ (ICO) Active–recreational orientation 55.23±8.36 53.36±11.28 46.62±6.89 46.43±8.70 47.25±8.70 4.38 0.003 [1]<[3]∗ (ARO) (continued on next page) Table 2 (continued) Moral–religious emphasis (MRE) 46.00±8.23 47.80±10.83 44.38±5.58 47.21±5.41 47.75±9.16 0.45 0.769 – Organisation (ORG) 52.50±8.82 52.40±10.54 47.00±11.43 48.14±10.47 48.25±13.44 1.24 0.301 – Control (CTL) 47.93±8.10 46.48±8.30 50.85±12.75 52.29±9.34 43.50±9.61 1.73 0.150 – Clinical Variables PANSS positive 7.53±1.08 8.59±1.57 12.68±3.68 11.58±3.09 18.00±5.65 53.68 <0.001 [1–2]<[3–5]∗∗ , [3–4]<[5]∗∗ PANSS negative 9.39±3.07 12.03±3.90 17.53±5.40 21.32±4.70 21.95±3.52 58.54 <0.001 [1]<[2]∗ , [1–2]<[3–5]∗∗ , [3]<[4–5]∗ PANSS general 18.06±2.13 22.97±2.83 28.53±4.33 32.37±3.69 42.20±5.39 202.71 <0.001 [1]<[2–5]∗∗ , European Neuropsychopharmacology 47 (2021) 112–129 [2]<[3–5]∗∗ , [3–4]<[5]∗∗ , [3]<[4]∗ PANSS total 34.98±4.90 43.59±5.66 58.74±8.15 65.26±7.45 82.15±10.37 207.17 <0.001 [1]<[2–5]∗∗ , [2]<[3–5]∗∗ , [3–4]<[5]∗∗ , [3]<[4]∗ MADRS 1.16±1.60 4.59±3.90 7.53±5.27 10.95±4.54 16.45±6.76 56.23 <0.001 [1]<[2]∗ , [1]<[3–5]∗∗ , [2]<[4–5]∗∗ , [3]<[5]∗∗ , 121 [4]<[5]∗ CGI 1.80±1.08 2.30±1.08 3.53±1.12 3.79±0.86 4.15±1.18 26.77 <0.001 [1–2]<[3–5]∗∗ Psychosocial functioning Autonomy 1.29±2.08 2.08±2.11 4.61±2.85 4.95±3.22 5.32±2.16 17.21 <0.001 [1–2]<[3–5]∗∗ Occupational 3.46±4.83 4.08±4.55 8.17±5.03 7.32±4.16 10.11±3.33 10.56 <0.001 [1–2]<[3–4]∗ , [1–2]<[5]∗∗ Cognitive 1.33±2.36 3.24±3.28 5.28±3.25 6.84±2.67 6.68±2.91 20.73 <0.001 [1]<[2]∗ , [1]<[3]∗∗ , [1–2]<[4–5]∗∗ Financial 0.31±1.11 0.65±1.25 1.17±1.82 2.05±1.65 2.11±1.56 9.27 <0.001 [1]<[4–5]∗∗ , [2]<[4–5]∗ Interpersonal relationships 2.21±3.40 3.38±2.53 6.61±3.82 8.58±2.69 8.95±3.37 26.42 <0.001 [1]<[4–5]∗∗ , [2]<[4–5]∗ Leisure 1.06±1.71 1.41±1.26 2.44±1.25 2.84±1.68 3.16±1.34 10.61 <0.001 [1]<[3–5]∗∗ , [2]<[3]∗ , [2]<[3]∗ , [2]<[4–5]∗∗ FAST total 9.67±10.87 14.84±10.07 28.28±10.97 32.58±11.22 36.32±10.17 34.06 <0.001 [1]<[3]∗ , [1]<[4–5]∗∗ , [2]<[4–5]∗ Abbreviations: HSD= Honestly Significant Difference; M=Mean; SD= Standard Deviation; SES=Socioeconomic status; DUP= Duration of untreated psychosis; CPZ= Chlorpromazine equiv- alents; PAS= Premorbid Adjustment Scale; PANSS= Positive and Negative Symptom Scale; MADRS= Montgomery-Asberg Depression Rating Scale; CGI=Clinical Global Impression Scale; FAST=Functioning Assessment Short Test. Significant differences (p<0.05) marked in bold. ∗ p < 0.05, ∗∗ p < 0.001. Table 3 Sociodemographic and clinical characteristics and psychosocial functioning among clinical trajectories at baseline and 2-year follow-up. Excellent Remitting course Clinical worsening Chronic course F p Post hoc prognosis (n = 40) (n = 46) [2] (n = 11) [3] (n = 28) [4] (Tukey’s HSD) [1] Sociodemographic variables Gender: Male N(%) 29 (73) 30 (65) 5 (46) 20 (72) χ²=3.18 0.365 – Age (M±SD) 24.98±5.19 25.85±4.84 25.82±4.38 26.07±5.72 0.32 0.812 – SES (%) High 9 (23) 12 (26) 4 (36) 3 (11) χ²=17.80 0.273- Medium-High 4 (10) 4 (9) 1 (9) 2 (7) Medium 12 (30) 11 (24) 3 (27) 3 (11) Medium-Low 8 (20) 16 (35) 3 (27) 16 (57) Low 6 (15) 3 (7) 0 (0) 4 (14) Missing 1 (3) 0 (0) 0 (0) 0 (0) S. Amoretti, N. Verdolini, G. Mezquida et al. DUP 75.06±81.95 96.40±99.22 70.90±93.72 110.78±121.64 0.83 0.480 – Age of onset 24.67±5.39 25.49±4.76 25.82±4.64 25.61±5.41 0.29 0.832 – CPZ 482.78±313.31 720.44±408.12 281.39±199.58 710.31±397.94 6.62 <0.001 [1, 3]<[2]∗ , [3]<[4]∗ CPZ FUP 172.64±219.45 161.64±240.07 180.00±293.94 252.53±299.46 0.82 0.483 – Alcohol: Yes N (%) 22 (55) 31 (67) 7 (64) 15 (54) χ²=2.01 0.571 – Monthly alcohol FUP 20.65±30.13 50.83±88.09 48.09±14.50 29.54±62.76 χ²=1.692 0.172 – Alcohol FUP 24 (60) 29 (63) 7 (64) 13 (47) χ²=4.23 0.646 – 122 Monthly alcohol FUP 17.48±29.07 21.36±38.71 30.11±45.75 22.29±43.84 χ²=0.739 0.531 – Cannabis: Yes N (%) 17 (43) 23 (50) 4 (36) 15 (54) χ²=1.47 0.688 – Monthly cannabis 38.15±58.23 34.91±86.49 28.09±53.40 74.93±146.30 χ²=1.784 0.154 – Cannabis FUP 5 (13) 8 (17) 3 (27) 4 (14) χ²=3.30 0.770 – Monthly cannabis FUP 7.78±25.33 21.73±42.35 11.41±33.57 16.07±42.01 χ²=0.636 0.593 – Trauma: Yes N (%) 24 (60) 23 (50) 6 (55) 12 (43) χ²=4.58 0.598 – PAS 37.34±21.55 42.44±16.57 38.40±13.04 56.50±24.14 5.31 0.002 [1]<[4]∗ Cognitive reserve 75.56±12.87 77.47±9.58 73.92±12.87 71.73±12.01 1.31 0.274 – Family History: Yes N (%) 6 (15) 11 (24) 6 (55) 7 (25) χ²=7.42 0.060 – Cohesion (C) 50.08±6.81 49.50±8.10 49.20±8.83 46.59±11.42 0.94 0.425 – C FUP 52.49±7.74 54.33±9.16 52.50±10.95 47.19±10.14 2.27 0.087 – Expressivity (EX) 50.66±9.41 51.00±8.23 48.50±11.58 45.89±10.56 1.88 0.137 – EXP FUP 52.74±8.41 54.57±7.68 50.17±12.09 46.63±8.42 3.27 0.025 [2]>[4]∗ Conflicts (CON) 48.11±7.03 49.45±7.35 48.00±6.88 53.63±10.08 2.87 0.039 [1]<[4]∗ CON FUP 47.26±5.49 47.47±4.26 50.00±7.85 46.50±5.73 0.64 0.592 – Independence (IND) 49.66±8.59 50.40±8.65 49.60±9.45 49.52±9.30 0.08 0.973 – IND FUP 53.03±9.44 54.50±7.15 50.00±13.04 50.88±8.43 0.83 0.480 – Achievement orientation (AO) 47.39±9.09 48.52±8.62 41.00±15.28 48.33±8.87 1.78 0.154 – AO FUP 47.11±9.07 47.13±9.09 41.33±12.74 45.50±10.09 0.74 0.532 – Intellectual–cultural orientation (ICO) 46.82±10.19 49.60±10.64 48.90±8.69 44.70±10.09 1.38 0.252 – (continued on next page) Table 3 (continued) ICO FUP 50.06±10.74 54.17±12.36 51.83±10.07 45.06±11.36 2.28 0.085 – Active–recreational orientation (ARO) 52.03±7.78 49.50±8.52 49.80±8.22 45.48±6.42 3.71 0.014 [1]>[4]∗ ARO FUP 55.86±8.50 52.93±10.57 46.83±10.72 46.69±7.92 4.35 0.007 [1]>[4]∗ Moral–religious emphasis (MRE) 46.50±7.78 45.07±8.38 46.00±8.88 46.22±7.38 0.24 0.871 – MRE FUP 45.71±7.57 47.83±10.96 49.83±6.65 46.50±6.83 0.56 0.640 – Organisation (ORG) 51.87±8.80 48.76±10.18 50.70±10.99 48.37±11.16 0.89 0.447 – ORG FUP 52.14±9.37 52.83±9.66 47.33±10.73 48.50±11.85 1.04 0.380 – Control (CTL) 48.71±8.30 46.79±6.52 43.80±8.64 50.48±7.99 2.41 0.071 – CTL FUP 47.54±8.73 47.17±7.54 46.00±11.31 50.25±9.84 0.57 0.638 – Clinical Variables European Neuropsychopharmacology 47 (2021) 112–129 PANSS positive 10.73±4.52 23.39±6.97 12.00±5.83 21.86±6.70 37.97 <0.001 [1, 3]<[2, 4]∗∗ PANSS positive FUP 7.93±1.29 8.04±1.52 16.36±7.97 14.29±4.35 40.05 <0.001 [1–2]<[3–4]∗∗ PANSS negative 13.58±5.45 22.17±6.52 11.36±5.14 23.25±6.89 24.27 <0.001 [1, 3]<[2, 4]∗∗ PANSS negative FUP 10.35±3.63 10.67±3.81 20.36±3.98 22.14±4.09 75.80 <0.001 [1–2]<[3–4]∗∗ PANSS general 24.55±4.76 45.67±8.69 25.91±4.21 45.61±8.89 79.99 <0.001 [1, 3]<[2, 4]∗∗ PANSS general FUP 19.73±3.48 20.57±3.42 38.18±6.16 37.11±7.07 118.23 <0.001 [1–2]<[3–4]∗∗ PANSS total 48.85±10.47 91.24±14.38 49.27±9.07 90.71±14.99 105.45 <0.001 [1, 3]<[2, 4]∗∗ PANSS total FUP 38.00±6.84 39.28±6.68 74.91±15.88 73.54±11.03 139.58 <0.001 [1–2]<[3–4]∗∗ MADRS 5.25±5.28 15.43±9.89 9.00±5.74 16.14±8.48 15.28 <0.001 [1]<[2, 4]∗∗ 1.98±2.81 3.22±3.59 17.82±7.86 12.18±4.96 <0.001 [1–2]<[3–4]∗∗ , 123 MADRS FUP 67.10 [3]>[4]∗ CGI 3.93±1.07 4.59±1.13 3.64±1.21 4.32±0.82 4.10 0.008 [1, 3]<[2]∗ CGI FUP 1.98±1.12 2.04±1.09 4.45±0.69 3.79±1.10 30.75 <0.001 [1–2]<[2–4]∗∗ Psychosocial functioning Autonomy 3.10±3.23 5.04±3.62 2.27±2.41 5.22±3.25 4.47 0.005 [1]<[2]∗ Autonomy FUP 1.51±2.04 1.74±2.21 5.10±3.87 5.14±2.26 19.69 <0.001 [1–2]<[3–4]∗∗ Occupational 7.15±6.05 10.43±4.71 4.27±4.71 8.70±3.80 5.84 0.001 Occupational FUP 3.13±4.88 4.24±4.53 8.90±3.90 8.64±4.08 11.15 <0.001 [1–2]<[3–4]∗∗ Cognitive 3.90±3.24 7.13±3.77 3.00±2.00 7.37±3.05 11.27 <0.001 [1, 3]<[2]∗ Cognitive FUP 1.79±2.70 2.48±3.13 6.90±3.45 6.71±2.54 22.51 <0.001 [1–2]<[3–4]∗∗ Financial 0.67±1.51 2.04±2.04 1.00±1.84 2.07±1.41 5.75 0.001 [1, 3]<[2, 4]∗∗ Financial FUP 0.28±0.86 0.61±1.39 2.70±1.89 1.86±1.43 14.98 <0.001 [1–2]<[3–4]∗∗ Interpersonal relationships 3.97±4.18 8.80±4.53 3.27±3.47 9.33±4.51 14.30 <0.001 [1]<[2, 4]∗ Interpersonal relationships FUP 1.82±2.17 3.48±3.54 8.40±2.95 8.89±3.08 37.51 <0.001 [1–2]<[3–4]∗∗ Leisure 1.46±1.60 2.54±1.94 1.36±1.50 2.85±1.79 4.86 0.003 [1, 3]<[2, 4]∗∗ Leisure FUP 1.03±1.35 1.37±1.66 3.00±1.63 3.00±1.49 12.35 <0.001 [1–2]<[3–4]∗∗ FAST 20.26±13.83 36.00±14.88 15.18±10.78 35.56±13.49 14.57 <0.001 [1]<[2, 4]∗ FAST FUP 9.56±9.18 13.91±11.70 35.00±14.23 34.25±9.49 39.99 <0.001 [1–2]<[3–4]∗∗ Abbreviations: HSD= Honestly Significant Difference; M=Mean; SD= Standard Deviation; PC= Principal Component; SES=Socioeconomic status; DUP= Duration of untreated psychosis; CPZ= Chlorpromazine equivalents; PAS= Premorbid Adjustment Scale; FUP = Follow-up; PANSS= Positive and Negative Symptom Scale; MADRS= Montgomery-Asberg Depression Rating Scale; CGI=Clinical Global Impression Scale; FAST=Functioning Assessment Short Test. Significant differences (p<0.05) marked in bold. ∗ p < 0.05, ∗∗ p < 0.001. S. Amoretti, N. Verdolini, G. Mezquida et al. 4. Discussion with FEP to be identified. Three distinct patient clusters have been determined at baseline: the first cluster was As far as we are aware, this is the first study to use a characterised by mild symptoms (39%), a second cluster machine learning algorithms method to identify individual had negative and moderate symptoms (31%) and a third clinical trajectories of FEP patients and their risk factors presented positive and severe symptoms (30%). Patients based on their psychopathological presentation at baseline with negative symptoms in comparison with those present- and 2-year follow-up. Three main findings emerged from ing mild symptoms had higher DUP and doses of antipsy- the present study: 1) PCA of the PANSS and MADRS yielded chotics, worse premorbid adjustment, higher illness sever- six symptom factors (positive, negative, depressive, anxi- ity and lower psychosocial functioning. These results are in ety, disorganisation and somatic/cognitive) at baseline and line with previous studies and confirm that a poor premorbid 2-year follow-up; 2) There are three well defined clinical functioning is associated with worse psychosocial function- clusters at baseline (“mild”, “negative and moderate” and ing (White et al., 2009; Wunderink et al., 2020), and with “positive and severe”) and five at follow-up (“minimal”, the severity of clinical presentation, especially for patients “mild”, “moderate”, “negative and depressive” and “se- with a FEP presenting negative symptoms (Amoretti et al., vere symptoms”); 3) The risk factors contributing towards 2016; Mezquida et al., 2017). Previous reports have also sup- a clinical worsening in patients with non-affective FEP were ported the association between negative symptoms and DUP low doses of antipsychotics, high depressive symptomatol- (Perkins et al., 2005). At 2-year follow-up, five subtypes ogy and the presence of a positive family history whilst the of patients were derived: “minimal”, “mild”, “moderate”, protective factors for a remitting course were a high CR and “negative and depressive” and “severe symptoms”. Patients better premorbid adjustment. with minimal or mild symptoms had better premorbid ad- The six-factor model included 36 items at baseline (26 justment, higher family cohesion and lower conflicts as well PANSS items and 10 MADRS items) and 34 items at follow- as more intellectual-cultural and active-recreational orien- up (25 PANSS items and 9 MADRS items). The majority of tation, lower illness severity and greater functioning. An in- studies on the PANSS agreed that either four or five di- creasingly large number of studies have found an associa- mensions describe the psychosis construct, with positive, tion between negative family environment and poor prog- negative, disorganisation, and affective symptom dimen- nosis in FEP patients (Norman et al., 2005; Lee et al., 2013). sions most frequently reported (Potuzak et al., 2012) and Higher rates of negative family environmental styles are as- some studies reported an independent anxiety dimension sociated with an increased likelihood of relapse (Wuerker, (Cuesta et al., 2003; Allardyce et al., 2007). As depres- 1996; Butzlaff and Hooley, 1998). sive symptoms are common in patients with a FEP, with a In the present article, four trajectories were identified, reported prevalence from 7% to 75% in the different stud- namely “excellent prognosis”, “remitting course”, “clini- ies (Sim et al., 2004), our study also included one mea- cal worsening” and “chronic course”. According to previous sure of depression (MADRS). Previous investigations of the studies, a significant number of people achieving symptom MADRS structure have differed in research design (i.e. sam- remission and recovery were found (Van Os and Kapur, 2009; ples have included patients with a range of diagnoses) and Austin et al., 2015) and a small but significant group also analytic strategy, constraining comparisons across studies showed a deteriorating course (Rabinowitz et al., 2007). (Quilty et al., 2013). The number of factors identified in lit- Results indicate that the majority of participants showed erature ranged from one-unidimensional (Lobo et al., 2002; a general pattern of reduction of positive symptoms, in line Carmody et al., 2006); two (Rocca et al., 2002; Uher et al., with previous literature (Austin et al., 2015). However, pre- 2008); three (Suzuki et al., 2005; Ketharanathan et al., viously it has also been reported that negative symptoms 2015); and up to four dimensions (Williamson et al., 2006). typically showed less variation during the ten years after Almost all studies with three or four factors found dimen- the first episode (Austin et al., 2015). On the contrary, in sions characterised by sadness/dysphoria/anhedonia, anx- our study negative symptoms showed an important reduc- iety and vegetative/somatic symptoms. Thus, the factors tion. These results could perhaps be explained by the dif- identified in our study by the PCA of the PANSS and MADRS ficulty in detecting secondary negative symptoms caused scales are essentially the same as those described in the ma- by positive symptoms, depression, and/or extrapyramidal jority of other studies, namely positive, negative, disorgan- symptoms amongst other factors, which could be present ised, depressive, anxiety and somatic. However, the com- in our patients. These secondary symptoms represent a di- ponents do not have exactly the same composition at both agnostic challenge, particularly in the early stages after a times of the evaluations. The instability of the PCA could first episode, when positive symptomatology predominates express the role played by the phase of illness (Dollfus and (Mezquida et al., 2017). Petit, 1995). For example, at baseline the positive compo- Patients who presented a clinical worsening were char- nent included excitement, conceptual disorganization, sus- acterised by having lower CPZ and higher depressive symp- piciousness/persecution and lack of insight, but these items toms at baseline, and more presence of family history. De- appeared in disorganization component at follow-up. This pression is significant in the prediction of transition to FEP, results can be due to the fact that after a first-episode, poor clinical outcomes, quality of life, and suicide (Fusar- the positive symptoms tend to reduce, whereas follow-up Poli et al., 2014; Gardsjord et al., 2016). Similarly, a pos- studies have suggested that after antipsychotic medication itive psychiatric family history was found to be associated stabilization, negative symptoms tend to be stable and per- with worse outcome (Esterberg and Compton, 2012). Con- sistent (Ventura et al., 2015; Mezquida et al., 2017). firming previous studies, patients with a remitting course The six factors described above were used to conduct a showed better premorbid adjustment and higher CR com- cluster analysis that allowed clinical subtypes of patients pared to those with a chronic course (Mezquida et al., 2017; 124 European Neuropsychopharmacology 47 (2021) 112–129 Amoretti et al., 2016, 2018, 2020). Thus, these results sug- ondary negative symptoms (Potuzak et al., 2012; Mezquida gest that prescribing higher doses of antipsychotics as well et al., 2017), which is an unmet need of current research on as addressing depressive symptomatology can help to pre- FEP. Thirdly, the retrospective assessment of the premor- vent clinical worsening. Undoubtedly, there are risk fac- bid phase. While retrospective reports from patients about tors that cannot be changed, such as the presence of a their childhood and developmental adjustment may be sub- positive psychiatric family history. Nonetheless, more at- ject to recall bias, it has been shown that premorbid adjust- tention should be paid to these subgroup of FEP patients ment measured by PAS scale is related to important clinical as they already present a higher risk of worse outcomes. features of the illness (Rabinowitz et al., 2006: Austin et al., Since a high CR and better premorbid adjustment rep- 2015; Amoretti et al., 2020). It is important to note that resented the protective factors for a remitting course in this is a naturalistic, multicentric and longitudinal study and our study, targeting older, less-educated, poor patients and that there was no other way to collect this premorbid infor- their families seems necessary. Similarly, in the comparison mation. Finally, despite few patients were prescribed with between those patients who showed “minimal” or “mild” high doses of antipsychotics, according to a previous study symptoms versus those who presented “negative and de- by the same group assessing pharmacovigilance in the same pressive” or “severe” symptoms at follow-up, differences sample of FEP patients, a total of 87.2% of the patients were emerged in terms of premorbid adjustment, family expres- receiving adequate doses of antipsychotic, according to the sion and active–recreational orientation. There were no dif- recommended dose range in the specifications of the prod- ferences between the four trajectories in terms of alcohol uct for Spain for the patient age (Bioque et al., 2016). Due or cannabis use. Although cannabis use has been shown to to its naturalistic design and by the very nature of the sam- increase the risk of developing psychosis and is related to a ple included, some treatment prescription could be higher poor prognosis, our study did not find this association. This than dose recommended in prescribing guidelines. Nonethe- may be due to the fact that the age of initiation of cannabis less, these data provide a real-world point of view on the use, the duration of cannabis use and cannabis patterns may clinical prescription in FEP patients on a country basis. also play an important role (Forti et al., 2014) and were not In spite of its limitations, the strength of this study lies evaluated in this study. in the evaluation of FEP patients that represent quite an As a consequence, the findings of the present study high- ideal population for exploring the clinical expression of light that the symptom trajectories among patients with the core symptoms of psychoses. The inclusion of a well- FEP are heterogeneous (Abdin et al., 2017) and should not characterised and naturalistic sample of non-affective FEP, be over-simplified. Psychopathology at the beginning of the the longitudinal study design, the clinical assessment of episode must be taken into account to understand the tra- both symptoms and functioning, and the comprehensive and jectory of each person in an attempt to develop tailored sophisticated statistical assessment used are the strengths intervention and personalised treatment strategies. For in- of the present work, reporting clinically sound results that stance, different interventions should be applied to those can be implemented in daily clinical practice. patients presenting moderate symptoms at baseline and still In conclusion, the adoption of a dimensional approach in suffering the same symptomatology at follow-up than that FEP might represent an advantage in clinical practice by applied to those experiencing mild symptoms at baseline providing a more comprehensive description of the com- and ultimately presenting a worsening in the symptomatol- plexity of FEP psychopathology (Tonna et al., 2019). This ogy. Current staging models are based on the integration approach might also allow for an improvement in the accu- of clinical baseline information, such as psychopathological racy of targeting therapeutic interventions. Not only posi- presentation at the early stages of psychosis, with trajecto- tive but also negative and depressive symptoms are common ries of the disease. Defining each patient’s characteristics, in patients with FEP and are highly involved in their clini- severity, and prognosis in a more precise and individualised cal trajectory. Thus, this study highlights the importance of way might aid the clinical decision-making algorithm to de- addressing depressive and negative symptoms, cognitive re- cide which treatment is most effective at that specific stage serve and premorbid adjustment in the early stages of psy- on the basis of the patient’s characteristics (Salagre et al., chosis. 2018). Some limitations have to be taken into consideration be- fore translating these findings into clinical practice. Firstly, 6. Role of funding source a limitation of all CR studies evaluating psychiatric popula- tions is that there was no validated instrument to measure The funding sources were not involved in the study design, CR so criteria established and replicated in previous studies conduct, monitoring and preparation of the final database were followed. Indeed, the Cognitive Reserve Assessment and had no influence on the final data analysis of this report. Scale in Health (CRASH) (Amoretti et al., 2019) has been val- idated in 2019. It is the first tool to assess CR in patients with severe mental illness. A second limitation is that no spe- 5. Contributors cific scale was used to assess negative symptomatology, due to constraints associated with the PANSS scale (Blanchard SA and NV designed the study, managed the literature et al., 2011). Future studies making use of improved nega- searches and analyses, undertook the statistical analysis, tive symptom scales— such as the Brief Negative Symptom and wrote the first draft of the manuscript. GM, FDR, BS, Scale (BNSS) (Kirkpatrick et al., 2011) or the Clinical As- AC, MB and EV revised the first draft and added critical sessment Interview for Negative Symptoms (CAINS) (Kring comments to guide the redaction of the final manuscript. et al., 2013)—might differentiate between primary and sec- MJC, LPC, MGR, CC, AGP, CDM, IC, ES, MA and AM revise the 125 S. Amoretti, N. Verdolini, G. Mezquida et al. second draft of the article and provided critical comments S. Amoretti has been supported by a Sara Borrell to guide the redaction of the final manuscript. All the au- (CD20/00177), funded by Instituto de Salud Carlos III (ISCIII) thors within the PEPs Group recluted patients and healthy and co-funded by European Social Fund "Investing in your controls at their centers, provided the anonimous data and future”. revise the final manuscript. All authors approved the final The study has been supported by a BITRECS project con- manuscript. ceded to N. Verdolini. BITRECS project has received funding from the European Union’s Horizon 2020 research and inno- vation programme under the Marie Skłodowska-Curie grant Conflict of interest agreement No 754550 and from “La Caixa” Foundation (ID 100010434), under the agreement LCF/PR/GN18/50310006. E. Vieta has received research support from or served as E. Vieta thanks the support of the Spanish Ministry of Sci- consultant, adviser or speaker for AB-Biotics, Actavis, Al- ence, Innovation and Universities (PI18/00805) integrated lergan, Angelini, AstraZeneca, Bristol-Myers Squibb, Dainip- into the Plan Nacional de I + D + I y cofinanciado por el pon Sumitomo Pharma, Ferrer, Forest Research Institute, ISCIII-Subdirección General de Evaluación y el Fondo Eu- Gedeon Richter, Glaxo-Smith-Kline, Janssen, Lundbeck, Ot- ropeo de Desarrollo Regional (FEDER); the Instituto de Salud suka, Pfizer, Roche, Sanofi-Aventis, Servier, Shire, Sunovion, Carlos III; the CIBER of Mental Health (CIBERSAM); the Secre- Takeda, Telefónica, the Brain and Behaviour Foundation, taria d’Universitats i Recerca del Departament d’Economia the Spanish Ministry of Science and Innovation (CIBERSAM), i Coneixement (2017 SGR 1365) and the CERCA Programme the Seventh European Framework Programme (ENBREC), / Generalitat de Catalunya) and the CERCA Programme / and the Stanley Medical Research Institute, unrelated to the Generalitat de Catalunya. We would like to thank the De- present work. partament de Salut de la Generalitat de Catalunya for the M. Bernardo has been a consultant for, received PERIS grant SLT006/17/00357. grant/research support and honoraria from, and been on the Finally, we would also like to thank the authors of the speakers/advisory board of ABBiotics, Adamed, Angelini, PEPs group who participated in the development of this Casen Recordati, Janssen-Cilag, Lundbeck, Otsuka, Menar- manuscript, namely, Bioque M 1,2 , Pons MT 1 , Andreu- ini and Takeda. Bernabeu Á 2,7 , Ayora M 2,7 , Alonso-Solís A 2,10 , Grasa E 2,10 , C. De-la-Camara received financial support to attend sci- González-Ortega I 2,9,15 , Zorrilla I 2,9 , Saz P 8 , Santabárbara entific meetings from Janssen-Cilag, Almirall, Eli Lilly, Lund- J 2,16 , Nacher J 2,17 , Castro-Bleda MJ 18 , Toll A 2,12 , Martinez beck, Rovi, Esteve, Novartis, and Astrazeneca L 19 , Sanchez-Moreno J 2,3 , Anmella G 1 , Castro-Fornieles M. Gómez-Ramiro reports grants from INSTITUTO CAR- J 2,11 , Baeza I 2,11 , Contreras F 2,20 , Saiz-Masvidal C 20,21 , LOS III, grants from ADAMED, non-financial support from García Álvarez L 2,22,23 , Bobes-Bascarán T 2,23,24 , Gutiérrez M 2,25 JANSSEN–CILAG, non-financial support from LUNDBECK- , Segarra R 2,26 , Sanchez-Pastor L 27 , Rodriguez-Jimenez 2,27,28 OTSUKA, non-financial support from PFIZER, non-financial R , Usall J 29 , Butjosa A 29 , Sarró S 30 , Pomarol-Clotet E 30 support from ANGELINI, outside the submitted work. , Ibáñez A 2,31 , Sánchez-Torres AM 5,6 , Balanzá-Martinez V 2,32 R. Rodriguez-Jimenez has been a consultant for, spoken . in activities of, or received grants from: Instituto de Salud PEPs Group additional affiliations: 15 Carlos III, Fondo de Investigación Sanitaria (FIS), Centro de The National Distance Education University (UNED), Vi- Investigación Biomédica en Red de Salud Mental (CIBER- toria, Spain. 16 SAM), Madrid Regional Government (S2010/ BMD-2422 AGES; Department of Microbiology, Preventive Medicine and S2017/BMD-3740), JanssenCilag, Lundbeck, Otsuka, Pfizer, Public Health, Universidad de Zaragoza, Zaragoza, Spain. 17 Ferrer, Juste, Takeda, Exeltis, Casen-Recordati, Angelini. Neurobiology Unit, Program in Neurosciences and In- The rest of authors report no biomedical financial inter- terdisciplinary Research Structure for Biotechnology and ests or potential conflicts of interest. Biomedicine (BIOTECMED), Universitat de València, Biomed- ical Research Institute INCLIVA, Valencia, Spain. 18 VRAIN: Valencian Research Institute for Artificial Intelli- Acknowledgments gence, Universitat Politècnica de València, Valencia, Spain. 19 Hospital del Mar, Department of Psychiatry, Barcelona, We are extremely grateful to all participants. Spain. 20 This study is part of a coordinated-multicentre Project, Bellvitge Biomedical Research Institute IDIBELL, De- funded by the Ministerio de Economía y Competitividad partment of Psychiatry- Bellvitge University Hospital, Hos- (PI08/0208; PI11/00325; PI14/00612), Instituto de Salud pitalet de Llobregat- Barcelona, Spain. 21 Carlos III – Fondo Europeo de Desarrollo Regional. Unión University of Barcelona, Department of Clinical Europea. Una manera de hacer Europa, Centro de Investi- Sciences- School of Medicine, Barcelona, Spain. 22 gación Biomédica en Red de salud Mental, CIBERSAM, by the Department of Psychiatry, Universidad de Oviedo, CERCA Programme / Generalitat de Catalunya AND Secre- Oviedo, Spain. 23 taria d’Universitats i Recerca del Departament d’Economia Servicio de Salud del Principado de Asturias (SESPA) I Coneixement (2017SGR1355). Departament de Salut de la Oviedo, Instituto de Investigación Sanitaria del Principado Generalitat de Catalunya, en la convocatoria corresponent de Asturias (ISPA), Instituto Universitario de Neurociencias a l’any 2017 de concessió de subvencions del Pla Estratègic del Principado de Asturias (INEUROPA), Oviedo, Spain. 24 de Recerca i Innovació en Salut (PERIS) 2016–2020, modali- Department of Psycology, Universidad de Oviedo, tat Projectes de recerca orientats a l’atenció primària, amb Oviedo, Spain el codi d’expedient SLT006/17/00345. 126 European Neuropsychopharmacology 47 (2021) 112–129 25 Araba University Hospital, Bioaraba Research Institute, Fraguas, D., Corripio, I., Vieta, E., de la Serna, E., Morro, L., Vitoria, Spain. Garriga, M., Torrent, C., Cuesta, M.J., Bernardo, M., 2020. The 26 Cruces University Hospital, BioCruces Health Research impact of cognitive reserve, cognition and clinical symptoms Institute, University of the Basque Country (UPV/EHU), Viz- on psychosocial functioning in first-episode psychoses. Psychol. Med. 1–12. caya, Spain. 27 Austin, S.F., Mors, O., Budtz-Jorgensen, E., Secher, R.G., Instituto de Investigación Sanitaria Hospital 12 de Oc- Hjorthoj, C.R., Bertelsen, M., Jeppesen, P., Petersen, L., Tho- tubre (imas12), Madrid, Spain. rup, A., Nordentoft, M., 2015. Long-term trajectories of posi- 28 CogPsy Group, Universidad Complutense de Madrid tive and negative symptoms in first episode psychosis: a 10 year (UCM), Madrid, Spain. follow-up study in the OPUS cohort. Schizophr. Res. 168 (1–2), 29 Institut de Recerca Sant Joan de Déu, Esplugues del 84–91. Llobregat, Parc Sanitari Sant Joan de Déu, Sant Boi de Llo- Barnett, J., Salmond, C., Jones, P., Sahakian, B.J., 2006. Cognitive bregat, Barcelona, Spain. reserve in neuropsychiatry. Psychol. Med. 36, 1053–1064. 30 FIDMAG Germanes Hospitalàries Research Foundation, Bioque, M., Llerena, A., Cabrera, B., Mezquida, G., Lobo, A., Barcelona, Spain. González-Pinto, A., Díaz-Caneja, C.M., Corripio, I., Aguilar, E.J., 31 Bulbena, A., Castro-Fornieles, J., Vieta, E., Lafuente, A., Department of Psychiatry, Hospital Universitario Ramón Mas, S., Parellada, M., Saiz-Ruiz, J., Cuesta, M.J., Bernardo, M., y Cajal, IRYCIS, Universidad de Alcalá, Madrid, Spain. 2016. A Pharmacovigilance study in first episode of psychosis: 32 Department of Medicine, University of Valencia, Valen- psychopharmacological interventions and safety profiles in the cia, Spain. PEPs project. Int. J. Neuropsychopharmacol. 19 (4), pyv121. Bernardo, M., Bioque, M., Parellada, M., Saiz, J., Cuesta, M.J., Llerena, A., Sanjuán, J., Castro-Fornieles, J., Arango, C., Cabr- Supplementary materials era, B., 2013. Assessing clinical and functional outcomes in a gene-environment interaction study in first episode of psychosis Supplementary material associated with this article can be (PEPs). Rev. Psiquiatr. Salud Ment. 6, 4–16. Bernardo, M., Cabrera, B., Arango, C., Bioque, M., Castro– found, in the online version, at doi:10.1016/j.euroneuro. Fornieles, J., Cuesta, M.J., Lafuente, A., Parellada., M., 2021.01.095. Saiz-Ruiz, J., Vieta, E, 2019. One decade of the first episodes project (PEPs): advancing towards a precision psychiatry. Rev. Psiquiatr. Salud Ment. 12, 135–140. References Blanchard, J.J., Kring, A.M., Horan, W.P., Gur, R., 2011. Toward the Abdin, E., Chong, S.A., Vaingankar, J.A., Peh, C.X., Poon, L.Y., next generation of negative symptom assessments: the collabo- Rao, S., Verma, S., Subramaniam, M., 2017. Trajectories of pos- ration to advance negative symptom assessment in schizophre- itive, negative and general psychopathology symptoms in first nia. Schizophr. Bull. 37, 291–299. episode psychosis and their relationship with functioning over a Butzlaff, R.L., Hooley, J.M., 1998. Expressed emotion and psychi- 2-year follow-up period. PLoS One 12, e0187141. atric relapse: a meta-analysis. Arch. Gen. Psychiatry 55 (6), Allardyce, J., McCreadie, R.G., Morrison, G., van Os, J., 2007. Do 547–552. symptom dimensions or categorical diagnoses best discriminate Cannon-Spoor, H., Potkin, S., Wyatt, R., 1982. Measurement of pre- between known risk factors for psychosis? Soc. Psychiatry Psy- morbid adjustment in chronic schizophrenia. Schizophr. Bull. 8, chiatr. Epidemiol. 42, 429–437. 470–484. Allott, K., Liu, P., Proffitt, T.M., Killackey, E., 2011. Cognition Carmody, T.J., Rush, A.J., Bernstein, I., Warden, D., Bran- at illness onset as a predictor of later functional outcome in nan, S., Burnham, D., Woo, A., Trivedi, M.H., 2006. The Mont- early psychosis: systematic review and methodological critique. gomery Äsberg and the Hamilton ratings of depression: a com- Schizophr. Res. 125 (2–3), 221–235. parison of measures. Eur. Neuropsychopharmacol. 16 (8), American Psychiatric Association, 2013. Diagnostic and Statistical 601–611. Manual of Mental Disorders (DSM-5 R ), 5th edn. Arlington: Amer- Cocchi, A., Cerati, G., Lora, A., Meneghelli, A., Monzani, E., Per- ican Psychiatric Publishing, Arlington. cudani, M., Petrovich, L., Mirabella, F., Picardi, A., Preti, A., Amoretti, S., Bernardo, M., Bonnin, C.M., Bioque, M., Cabrera, B., 2014. Patients with first-episode psychosis are not a homoge- Mezquida, G., Solé, B., Vieta, E., Torrent, C., 2016. The im- neous population: implications for treatment. Clin. Pract. Epi- pact of cognitive reserve in the outcome of first-episode psy- demiol. Ment. Health 10, 1–8. choses: 2-year follow-up study. Eur. Neuropsychopharmacol. 26, Colasanti, A., Paletta, S., Moliterno, D., Mazzocchi, A., Mauri, M.C., 1638–1648. Altamura, A.C., 2010. Symptom dimensions as predictors of clin- Amoretti, S., Cabrera, B., Torrent, C., Mezquida, G., Lobo, A., ical outcome, duration of hospitalization and aggressive be- González-Pinto, A., Parellada, M., Corripio, I., Vieta, E., de la haviours in acutely hospitalized patients with psychotic exac- Serna, E., Butjosa, A., Contreras, F., Sarró, S., Penadés, R., erbation. Clin. Pract. Epidemiol. Ment. Health 6, 72–78. Sánchez-Torres, A.M., Cuesta, M., Bernardo, M., 2018. Cognitive Compton, M.T., Kelley, M.E., Ionescu, D.F., 2014. Subtyping reserve as an outcome predictor: first-episode affective versus first-episode non-affective psychosis using four early-course fea- non-affective psychosis. Acta Psychiatr. Scand. 138, 411–455. tures: potentially useful prognostic information at initial pre- Amoretti, S., Cabrera, B., Torrent, C., Bonnín, C.M., Mezquida, G., sentation. Early Interv. Psychiatry 8 (1), 50–58. Garriga, M., Jiménez, E., Martínez-Arán, A., Solé, B., Cuesta, M.J., Peralta, V., Gil, P., Artamendi, M., 2003. Psychopatho- Reinares, M., Varo, C., Penadés, R., Grande, I., Salagre, E., logical dimensions in first-episode psychoses. From the trunk to Parellada, E., Bioque, M., Garcia-Rizo, C., Meseguer, A., An- the branches and leaves. Eur. Arch. Psychiatry Clin. Neurosci. mella, G., Rosa, A.R., Contreras, F., Safont, G., Vieta, E., 253, 73–79. Bernardo, M., 2019. Cognitive reserve assessment scale in Davidson, J., Smith, R., 1990. Traumatic experiences in psychiatric health (CRASH): its validity and reliability. J Clin. Med. 28, 8. outpatients. J. Trauma Stress 3, 459–475. Amoretti, S., Rosa, A.R., Mezquida, G., Cabrera, B., Ribeiro, M., De la Serna, E., Andrés-Perpiñá, S., Puig, O., Baeza, I., Bombin, I., Molina, M., Bioque, M., Lobo, A., González-Pinto, A., Bartrés-Faz, D., Arango, C., Gonzalez-Pinto, A., Parellada, M., Mayoral, M., Graell, M., Otero, S., Guardia, J., Castro– 127 S. Amoretti, N. Verdolini, G. Mezquida et al. Fornieles, J., 2013. Cognitive reserve as a predictor of two year Kring, A.M., Gur, R.E., Blanchard, J.J., Horan, W.P., Reise, S.P., neuropsychological performance in early onset first-episode 2013. The clinical assessment interview for negative symptoms schizophrenia. Schizophr. Res. 143, 125–131. (CAINS): final development and validation. Am. J. Psychiatry de Oliveira, M.O., Nitrini, R., Yassuda, M.S., Brucki, S.M., 2014. Vo- 170, 165–172. cabulary is an appropriate measure of premorbid intelligence in Lee, G., Barrowclough, C., Lobban, F., 2013. Positive affect in the a sample with heterogeneous educational level in Brazil. Behav. family environment protects against relapse in first-episode psy- Neurol., 875960 2014. chosis. Soc. Psychiatry Psychiatr. Epidemiol. 49, 367–376. Esterberg, M., Compton, M., 2012. Family history of psychosis neg- Lefort-Besnard, J., Varoquaux, G., Derntl, B., Gruber, O., Ale- atively impacts age at onset, negative symptoms, and duration man, A., Jardri, R., Sommer, I., Thirion, B., Bzdok, D., 2018. of untreated illness and psychosis in first-episode psychosis pa- Patterns of schizophrenia symptoms: hidden structure in the tients. Psychiatry Res. 197 (1–2), 23–28. PANSS questionnaire. Transl Psychiatry 8 (1), 237. First, M., Gibbon, M., Spitzer, R., Williams, J., Benjamin, L., 1997a. Lobo, A., Chamorro, L., Luque, A., Dal-Ré, R., Badia, X., Baró, E., Structured Clinical Interview for DSM-IV Axis II Personality Disor- 2002. Validación de las versiones en español de la Mont- ders (SCID-II). American Psychiatric Publishing, Washington DC. gomery-Asberg depression rating scale y la Hamilton anxiety rat- First, M., Spitzer, R., Gibbon, M., Williams, J., 1997b. Structured ing scale para la evaluación de la depresión y de la ansiedad. Clinical Interview for DSM-IV Axis I disorders-Clinician (SCID-I). Med. Clin. (Barc) 118 (13), 493–499. American Psychiatric Press, Washington DC. Loo, R., 1979. The orthogonal rotation of factors in clinical re- Fraley, C., Raftery, A.E., 1998. How many clusters? Which clustering search: a critical note. J. Clin. Psychol. 35, 762–765. method? Answers via model-based cluster analysis. Comput. J. Malla, A., Payne, J., 2005. First-episode psychosis: psychopathol- 41 (8), 578–588. ogy, quality of life, and functional outcome. Schizophr. Bull. 31 Fusar-Poli, P., Nelson, B., Valmaggia, L., Yung, A.R., McGuire, P.K., (3), 650–671. 2014. Comorbid depressive and anxiety disorders in 509 individ- Mezquida, G., Cabrera, B., Bioque, M., Amoretti, S., Lobo, A., uals with an at-risk mental state: impact on psychopathology González-Pinto, A., Espliego, A., Corripio, I., Vieta, E., Castro– and transition to psychosis. Schizophr. Bull. 40 (1), 120–131. Fornieles, J., Bergé, D., Escartí, M.J., Ibañez, Á., Penadés, R., Gardner, D., Murphy, A., O’Donnell, H., Centorrino, F., Sánchez-Torres, A.M., Bernardo, M., 2017. The course of nega- Baldessarini, R.J., 2010. International consensus study of tive symptoms in first-episode schizophrenia and its predictors: antipsychotic dosing. Am. J. Psychiatry 167, 686–693. a prospective two-year follow-up study. Schizophr. Res. 189, Gardsjord, E.S., Romm, K.L., Friis, S., Barder, H.E., Evensen, J., 84–90. Haahr, U., enVeldenHegelstad, W., Joa, I., Johannessen, O., Montgomery, S., Asberg, M., 1979. A new depression scale designed Langeveld, J., Larse, T.K., Opjordsmoen, S., Rund, B.R., Si- to be sensitive to change. Br. J. Psychiatry 134, 382–389. monsen, E., Vaglum, P., McGlashan, T., Melle, I., Røssberg, J.I., Moos, R.H., Moos, B.S., 1976. A typology of family social environ- 2016. Subjective quality of life in first-episode psychosis. A ten ments. Fam. Process. 15 (4), 357–371. year follow-up study. Schizophr. Res. 172 (1–3), 23–28. Norman, R., Malla, A., Manchanda, R., Harricharan, R., Takhar, J., González-Ortega, I., Rosa, A., Alberich, S., Barbeito, S., Vega, P., Northcott, S., 2005. Social support and three-year symptom and Echeburúa, E., Vieta, E., González-Pinto, A., 2010. Validation admission outcomes for first episode psychosis. Schizophr. Res. and use of the functioning assessment short test in first psy- 80, 227–234. chotic episodes. J. NervMent. Dis. 198 (11), 836–840. Peralta, V., Cuesta, M.J., 2001. How many and which are the psy- Guy, W., 1976. ECDEU assessment manual for psychopharmacol- chopathological dimensions in schizophrenia? Issues influencing ogy, revised. In: US Department of Health, Education, and Wel- their ascertainment. Schizophr. Res. 49 (3), 269–285. fare Publication (ADM). National Institute of Mental Health, Perkins, D.O., Gu, H., Boteva, K., Lieberman, J.A., 2005. Relation- Rockville, MD, pp. 76–338. ship between duration of untreated psychosis and outcome in Hassani, M., Seidl, T., 2017. Using internal evaluation measures to first-episode schizophrenia: a critical review and meta-analysis. validate the quality of diverse stream clustering algorithms. J. Am. J. Psychiatry 162, 1785–1804. Comput. Sci. 4, 171–183. Potuzak, M., Ravichandran, C., Lewandowski, K.E., Ongur, D., Hinojosa-Marqués, L., Domínguez-Martínez, T., Kwapil, T.R., Bar- Cohen, B.M., 2012. Categorical vs dimensional classifca- rantes-Vidal, N., 2019. Ecological validity of expressed emotion tions of psychotic disorders. Compreh. Psychiatry 53 (8), in early psychosis. Front. Psychiatry 10, 854. 1118–1129. Hollingshead, A., Redlich, F., 1958. Social class and mental illness: Quilty, L.C., Robinson, J.J., Rolland, J.-.P., Fruyt, F.D., Rouillon, F., a community study. Am. J. Public Health 97, 1756–1757. Bagby, R.M., 2013. The structure of the Montgomery-Åsberg de- Jensen, J.H., Støttrup, M.M., Nayberg, E., Knorr, U., Ullum, H., pression rating scale over the course of treatment for depres- Purdon, S.E., Kessing, L.V., Miskowiak, K.W., 2015. Optimising sion. Int. J. Methods Psychiatr. Res. 22 (3), 175–184. screening for cognitive dysfunction in bipolar disorder: valida- Rabinowitz, J., Harvey, P.D., Eerdekens, M., Davidson, M., 2006. tion and evaluation of objective and subjective tools. J. Affect. Premorbid functioning and treatment response in recent-onset Disord. 187, 10–19. schizophrenia. Br. J. Psychiatry 189, 31–35. Kay, S., Fiszbein, A., Opler, L., 1987. The positive and negative Rocca, P., Fonzo, V., Ravizza, L., Rocca, G., Scotta, M., Zanalda, E., syndrome scale (PANSS) for schizophrenia. Schizophr. Bull. 13, Bogetto, F., 2002. A comparison of paroxetine and amisulpride 261–276. in the treatment of dysthymic disorder. J. Affect. Disord. 70, Ketharanathan, T., Hanwella, R., Weerasundera, R., de Silva, V.A., 313–317. 2015. Diagnostic validity and factor analysis of Montgomery-As- Rodriguez-Jimenez, R., Bagney, A., Mezquita, L., Martinez-Gras, I., berg depression rating scale in Parkinson disease population. J. Sanchez-Morla, E.M., Mesa, N., Ibañez, M., Diez-Martin, J., Geriatr. Psychiatry Neurol. 29 (3), 115–119. Jimenez-Arriero, M., Lobo, A., Santos, J., Palomo, T., 2013. Cog- Kirkpatrick, B., Strauss, G.P., Nguyen, L., Fischer, B.A., nition and the five-factor model of the positive and negative Daniel, D.G., Cienfuegos, A., Marder, S.R., 2011. The brief syndrome scale in schizophrenia. Schizophr. Res. 143 (1), 77–83. negative symptom scale: psychometric properties. Schizophr. Rodríguez-Sánchez, J.M., Ayesa-Arriola, R., Pérez-Iglesias, R., Peri- Bull. 37, 300–305. añez, J.A., Martinez-Garcia, O., Gomez-Ruiz, E., Tabares-Seis- Kokkevi, A., Hartgers, C., 1995. European adaptation of a multidi- dedos, R., Crespo-Facorro, B., 2013. Course of cognitive deficits mensional assessment instrument for drug and alcohol depen- in first episode of non-affective psychosis: a 3-year follow-up dence. Eur. Addict. Res. 1, 208–210. study. Schizophr. Res. 150, 121–128. 128 European Neuropsychopharmacology 47 (2021) 112–129 Rosa, A.R., Sánchez-Moreno, J., Martínez-Aran, A., Salamero, M., Tonna, M., Ossola, P., Marchesi, C., Bettini, E., Lasalvia, A., Torrent, C., Reinares, M., Comes, M., Colom, F., Van Riel, W., Bonetto, C., Lenzi, J., Rucci, P., Iozzino, L., Cellini, M., Co- Ayuso-Mateos, J.L., Kapczinski, F., Vieta, E., 2007. Validity and macchio, C., Cristofalo, D., D’Agostino, A., de Girolamo, G., reliability of the functioning assessment short test (FAST) in De Santi, K., Ghigi, D., Leuci, E., Miceli, M., Meneghelli, A., Pi- bipolar disorder. Clin. Pract. Epidemiol. Ment. Health 7, 5. leggi, F., Scarone, S., Santonastaso, P., Torresani, S., Tosato, S., Salagre, E., Arango, C., Artigas, F., Ayuso-Mateos, J.L., Veronese, A., Fioritti, A., Ruggeri, M., Group, GET UP, 2019. Di- Bernardo, M., Castro-Fornieles, J., Bobes, J., Desco, M., mensional structure of first episode psychosis. Early Interv. Psy- Fañanás, L., González-Pinto, A., Haro, J.M., Leza, J.C., chiatry 13 (6), 1431–1438. Mckenna, P.J., Meana, J.J., Menchón, J.J., Micó, J.A., Uher, R., Farmer, A., Maier, W., Rietschel, M., 2008. Measuring de- Palomo, T., Pazos, Á., Pérez, V., Saiz-Ruiz, J., Sanjuán, J., pression: comparison and integration of three scales in the GEN- Tabarés-Seisdedos, R., Crespo-Facorro, B., Casas, M., Vilella, E., DEP study. Psychol. Med. 38, 289–300. Palao, D., Olivares, J.M., Rodriguez-Jimenez, R., Vieta, E., Van Os, J., Kapur, S., 2009. Schizophrenia. Lancet. 374, 635–645. 2019. CIBERSAM: ten years of collaborative translational re- Vares, F., Smeets, F., Drukker, M., Lieverse, R., Lataster, T., Viecht- search in mental disorders. Rev. Psiquiatr. Salud. Ment. 12 (1), bauer, W., Read, J., van Os, J., Bentall, R.P., 2012. Childhood 1–8. adversities increase the risk of psychosis: a meta-analysis of pa- Salagre, E., Dodd, S., Aedo, A., Rosa, A., Amoretti, S., Pinzon, J., tient-control, prospective- and cross-sectional cohort studies. Reinares, M., Berk, M., Kapczinski, F.P., Vieta, E., Grande, I., Schizophr. Bull. 38 (4), 661–671. 2018. Toward precision psychiatry in bipolar disorder: staging Verma, S., Poon, L.Y., Subramaniam, M., Abdin, E., Chong, S.A., 2.0. Front. Psychiatry 9, 641. 2012. The Singapore early psychosis intervention programme Sim, K., Mahendran, R., Siris, S.G., Heckers, S., Chong, S.A., 2004. (EPIP): a programme evaluation. Asian J. Psychiatr. 5 (1), 63–67. Subjective quality of life in first episode schizophrenia spec- Vieta, E., 2016. DSM-5.1. Acta Psychiatr. Scand. 134 (3), 187–188. trum disorders with comorbid depression. Psychiatry Res. 129, White, C., Stirling, J., Hopkins, R., Morris, J., Montague, L., Tan- 141–147. tam, D., Lewis, S., 2009. Predictors of 10-year outcome of first Stein, D.J., Szatmari, P., Gaebel, W., Berk, M., Vieta, E., Maj, M., episode psychosis. Psychol. Med. 39, 1447–1456. Anna de Vries, Y., Roest, A.M., de Jonge, P., Maercker, A., Williamson, D., Brown, E., Perlis, R.H., Ahl, J., Baker, R.W., To- Brewin, C.R., Pike, K.M., Grilo, C.M., Fineberg, N.A., Briken, P., hen, M., 2006. Clinical relevance of depressive symptom im- Cohen-Kettenis, P.T., Reed, G.M., 2020. Mental, behavioral and provement in bipolar I depressed patients. J. Affect. Disord. 92, neurodevelopmental disorders in the ICD-11: an international 261–266. perspective on key changes and controversies. BMC Med. 18 (1), Organization, World Health, 2004. ICD-10: International Statistical 21. Classification of Diseases and Related Health problems: Tenth Suganya, R., Shanthi, R., 2012. Fuzzy c-means algorithm-a review. Revision, 2nd ed. World Health Organization. IJSRP 2 (11), 1. Wu, C., Luedtke, A.R., Sadikova, E., Tsai, H., Liao, S., Liu, C., Suzuki, A., Aoshima, T., Fukasawa, T., Yoshida, K., Higuchi, H., Gau, S.S., VanderWeele, T.J., Kessler, R.C., 2020. Development Shimizu, T., Otani, K., 2005. A three-factor model of the MADRS and validation of a machine learning individualized treatment in major depressive disorder. Depress. Anxiety 21, 95–97. rule in first-episode schizophrenia. JAMA Netw. Open 3 (2), Tomasini, C., Borges, E.N., Machado, K., Emmendorfer, L., 2017. e1921660. A study on the relationship between internal and external va- Wunderink, L., van Bebber, J., Sytema, S., Boonstra, N., Mei- lidity indices applied to partitioning and density-based clus- jer, R.R., Wigman, J.T.W., 2020. Negative symptoms predict high tering algorithms. In: Proceedings of the 19th International relapse rates and both predict less favorable functional outcome Conference on Enterprise Information Systems, 1, pp. 89–98 in first episode psychosis, independent of treatment strategy. ICEIS 2017. Schizophr. Res. 216, 192–199. 129