Search

It is widely hoped that statistical models can improve decision-making related to medical treatments. Because of the cost and scarcity of medical outcomes data, this hope is typically based on investigators observing a model's success in one or two datasets or clinical contexts. We scrutinized this optimism by examining how well a machine learning model performed across several independent clinical trials of antipsychotic medication for schizophrenia. Models predicted patient outcomes with high accuracy within the trial in which the model was developed but performed no better than chance when applied out-of-sample. Pooling data across trials to predict outcomes in the trial left out did not improve predictions. These results suggest that models predicting treatment outcomes in schizophrenia are highly context-dependent and may have limited generalizability.

Importance: Diverse models have been developed to predict psychosis in patients with clinical high-risk (CHR) states. Whether prediction can be improved by efficiently combining clinical and biological models and by broadening the risk spectrum to young patients with depressive syndromes remains unclear., Objectives: To evaluate whether psychosis transition can be predicted in patients with CHR or recent-onset depression (ROD) using multimodal machine learning that optimally integrates clinical and neurocognitive data, structural magnetic resonance imaging (sMRI), and polygenic risk scores (PRS) for schizophrenia; to assess models' geographic generalizability; to test and integrate clinicians' predictions; and to maximize clinical utility by building a sequential prognostic system., Design, Setting, and Participants: This multisite, longitudinal prognostic study performed in 7 academic early recognition services in 5 European countries followed up patients with CHR syndromes or ROD and healthy volunteers. The referred sample of 167 patients with CHR syndromes and 167 with ROD was recruited from February 1, 2014, to May 31, 2017, of whom 26 (23 with CHR syndromes and 3 with ROD) developed psychosis. Patients with 18-month follow-up (n = 246) were used for model training and leave-one-site-out cross-validation. The remaining 88 patients with nontransition served as the validation of model specificity. Three hundred thirty-four healthy volunteers provided a normative sample for prognostic signature evaluation. Three independent Swiss projects contributed a further 45 cases with psychosis transition and 600 with nontransition for the external validation of clinical-neurocognitive, sMRI-based, and combined models. Data were analyzed from January 1, 2019, to March 31, 2020., Main Outcomes and Measures: Accuracy and generalizability of prognostic systems., Results: A total of 668 individuals (334 patients and 334 controls) were included in the analysis (mean [SD] age, 25.1 [5.8] years; 354 [53.0%] female and 314 [47.0%] male). Clinicians attained a balanced accuracy of 73.2% by effectively ruling out (specificity, 84.9%) but ineffectively ruling in (sensitivity, 61.5%) psychosis transition. In contrast, algorithms showed high sensitivity (76.0%-88.0%) but low specificity (53.5%-66.8%). A cybernetic risk calculator combining all algorithmic and human components predicted psychosis with a balanced accuracy of 85.5% (sensitivity, 84.6%; specificity, 86.4%). In comparison, an optimal prognostic workflow produced a balanced accuracy of 85.9% (sensitivity, 84.6%; specificity, 87.3%) at a much lower diagnostic burden by sequentially integrating clinical-neurocognitive, expert-based, PRS-based, and sMRI-based risk estimates as needed for the given patient. Findings were supported by good external validation results., Conclusions and Relevance: These findings suggest that psychosis transition can be predicted in a broader risk spectrum by sequentially integrating algorithms' and clinicians' risk estimates. For clinical translation, the proposed workflow should undergo large-scale international validation.

At present, risk assessment for alcohol withdrawal syndrome relies on clinical judgment. Our aim was to develop accurate machine learning tools to predict alcohol withdrawal outcomes at the individual subject level using information easily attainable at patients' admission. An observational machine learning analysis using nested cross-validation and out-of-sample validation was applied to alcohol-dependent patients at two major detoxification wards (LMU, n = 389; TU, n = 805). 121 retrospectively derived clinical, blood-derived, and sociodemographic measures were used to predict 1) moderate to severe withdrawal defined by the alcohol withdrawal scale, 2) delirium tremens, and 3) withdrawal seizures. Mild and more severe withdrawal cases could be separated with significant, although highly variable accuracy in both samples (LMU, balanced accuracy [BAC] = 69.4%; TU, BAC = 55.9%). Poor outcome predictions were associated with higher cumulative clomethiazole doses during the withdrawal course. Delirium tremens was predicted in the TU cohort with BAC of 75%. No significant model predicting withdrawal seizures could be found. Our models were unique to each treatment site and thus did not generalize. For both treatment sites and withdrawal outcome different variable sets informed our models' decisions. Besides previously described variables (most notably, thrombocytopenia), we identified new predictors (history of blood pressure abnormalities, urine screening for benzodiazepines and educational attainment). In conclusion, machine learning approaches may facilitate generalizable, individualized predictions for alcohol withdrawal severity. Since predictive patterns highly vary for different outcomes of withdrawal severity and across treatment sites, prediction tools should not be recommended for clinical practice unless adequately validated in specific cohorts., Competing Interests: Conflict of interest All authors declare no conflict of interest., (Copyright © 2020 Elsevier Ltd. All rights reserved.)

Patients with schizophrenia (SCZ), as well as their unaffected siblings (SIB), show functional connectivity (FC) alterations during performance of tasks involving attention. As compared with SCZ, these alterations are present in SIB to a lesser extent and are more pronounced during high cognitive demand, thus possibly representing one of the pathways in which familial risk is translated into the SCZ phenotype. Our aim is to measure the separability of SCZ and SIB from healthy controls (HC) using attentional control-dependent FC patterns, and to test to which extent these patterns span a continuum of neurofunctional alterations between HC and SCZ. 65 SCZ with 65 age and gender-matched HC and 39 SIB with 39 matched HC underwent the Variable Attentional Control (VAC) task. Load-dependent connectivity matrices were generated according to correct responses in each VAC load. Classification performances of high, intermediate and low VAC load FC on HC-SCZ and HC-SIB cohorts were tested through machine learning techniques within a repeated nested cross-validation framework. HC-SCZ classification models were applied to the HC-SIB cohort, and vice-versa. A high load-related decreased FC pattern discriminated between HC and SCZ with 66.9% accuracy and with 57.7% accuracy between HC and SIB. A high load-related increased FC network separated SIB from HC (69.6% accuracy), but not SCZ from HC (48.5% accuracy). Our findings revealed signatures of attentional FC abnormalities shared by SCZ and SIB individuals. We also found evidence for potential, SIB-specific FC signature, which may point to compensatory neurofunctional mechanisms in persons at familial risk for schizophrenia.

Despite its initial promise, neuroimaging has not been widely translated into clinical psychiatry to assist in the prediction of diagnoses, prognoses, and optimal therapeutic strategies. Machine learning approaches may enhance the translational potential of neuroimaging because they specifically focus on overcoming biases by optimizing the generalizability of pipelines that measure complex brain patterns to predict targets at a single-subject level. This article introduces some fundamentals of a translational machine learning approach before selectively reviewing literature to-date. Promising initial results are then balanced by the description of limitations that should be considered in order to interpret existing research and maximize the possibility of future translation. Future directions are then presented in order to inspire further research and progress the field towards clinical translation., (Copyright © 2018. Published by Elsevier Inc.)

Identifying distinctive subtypes of schizophrenia could ultimately enhance diagnostic and prognostic accuracy. We aimed to uncover neuroanatomical subtypes of chronic schizophrenia patients to test whether stratification can enhance computer-aided discrimination of patients from control subjects. Unsupervised, data-driven clustering of structural MRI (sMRI) data was used to identify 2 subtypes of schizophrenia patients drawn from a US-based open science repository (n = 71) and we quantified classification improvements compared to controls (n = 74) using supervised machine learning. We externally validated the unsupervised and supervised learning models in a heterogeneous German validation sample (n = 316), and characterized symptom, cognition, and longitudinal symptom change signatures. Stratification improved classification accuracies from 68.5% to 73% (subgroup 1) and 78.8% (subgroup 2), respectively. Increased accuracy was also found when models were externally validated, and an average gain of 9% was found in supplementary analyses. The first subgroup was associated with cortical and subcortical volume reductions coupled with substantially longer illness duration, whereas the second subgroup was mainly characterized by cortical reductions, reduced illness duration, and comparatively less negative symptoms. Individuals within each subgroup could be identified using just 10 clinical questions at an accuracy of 81.2%, and differential cognitive and symptom course signatures were suggested in multivariate analyses. Our findings suggest that sMRI-based subtyping enhances the neuroanatomical discrimination of schizophrenia by identifying generalizable brain patterns that align with a clinical staging model of the disorder. These findings could be used to improve illness stratification for biomarker-based computer-aided diagnoses.

BackgroundTransition to psychosis is among the most adverse outcomes of clinical high-risk (CHR) syndromes encompassing ultra-high risk (UHR) and basic symptom states. Clinical risk calculators may facilitate an early and individualized interception of psychosis, but their real-world implementation requires thorough validation across diverse risk populations, including young patients with depressive syndromes.MethodsWe validated the previously described NAPLS-2 (North American Prodrome Longitudinal Study 2) calculator in 334 patients (26 with transition to psychosis) with CHR or recent-onset depression (ROD) drawn from the multisite European PRONIA (Personalised Prognostic Tools for Early Psychosis Management) study. Patients were categorized into three risk enrichment levels, ranging from UHR, over CHR, to a broad-risk population comprising patients with CHR or ROD (CHR|ROD). We assessed how risk enrichment and different predictive algorithms influenced prognostic performance using reciprocal external validation.ResultsAfter calibration, the NAPLS-2 model predicted psychosis with a balanced accuracy (BAC) (sensitivity, specificity) of 68% (73%, 63%) in the PRONIA-UHR cohort, 67% (74%, 60%) in the CHR cohort, and 70% (73%, 66%) in patients with CHR|ROD. Multiple model derivation in PRONIA-CHR|ROD and validation in NAPLS-2-UHR patients confirmed that broader risk definitions produced more accurate risk calculators (CHR|ROD-based vs. UHR-based performance: 67% [68%, 66%] vs. 58% [61%, 56%]). Support vector machines were superior in CHR|ROD (BAC = 71%), while ridge logistic regression and support vector machines performed similarly in CHR (BAC = 67%) and UHR cohorts (BAC = 65%). Attenuated psychotic symptoms predicted psychosis across risk levels, while younger age and reduced processing speed became increasingly relevant for broader risk cohorts.ConclusionsClinical-neurocognitive machine learning models operating in young patients with affective and CHR syndromes facilitate a more precise and generalizable prediction of psychosis. Future studies should investigate their therapeutic utility in large-scale clinical trials.

Ecological momentary assessment (EMA), a structured diary assessment technique, has shown feasibility to capture psychotic(-like) symptoms across different study groups. We investigated whether EMA combined with unsupervised machine learning can distinguish groups on the continuum of genetic risk toward psychotic illness and identify individuals with need for extended healthcare. Individuals with psychotic disorder (PD, N = 55), healthy individuals (HC, N = 25) and HC with first-degree relatives with psychosis (RE, N = 20) were assessed at two sites over 7 days using EMA. Cluster analysis determined subgroups based on similarities in longitudinal trajectories of psychotic symptom ratings in EMA, agnostic of study group assignment. Psychotic symptom ratings were calculated as average of items related to hallucinations and paranoid ideas. Prior to EMA we assessed symptoms using the Positive and Negative Syndrome Scale (PANSS) and the Community Assessment of Psychic Experience (CAPE) to characterize the EMA subgroups. We identified two clusters with distinct longitudinal EMA characteristics. Cluster 1 (NPD = 12, NRE = 1, NHC = 2) showed higher mean EMA symptom ratings as compared to cluster 2 (NPD = 43, NRE = 19, NHC = 23) (p < 0.001). Cluster 1 showed a higher burden on negative (p < 0.05) and positive (p < 0.05) psychotic symptoms in cross-sectional PANSS and CAPE ratings than cluster 2. Findings indicate a separation of PD with high symptom burden (cluster 1) from PD with healthy-like rating patterns grouping together with HC and RE (cluster 2). Individuals in cluster 1 might particularly profit from exchange with a clinician underlining the idea of EMA as clinical monitoring tool. [ABSTRACT FROM AUTHOR]

Psychotic disorders are frequently associated with decline in functioning and cognitive difficulties are observed in subjects at clinical high risk (CHR) for psychosis. In this work, we applied automatic approaches to neurocognitive and functioning measures, with the aim of investigating the link between global, social and occupational functioning, and cognition. 102 CHR subjects and 110 patients with recent onset depression (ROD) were recruited. Global assessment of functioning (GAF) related to symptoms (GAF-S) and disability (GAF-D). and global functioning social (GF-S) and role (GF-R), at baseline and of the previous month and year, and a set of neurocognitive measures, were used for classification and regression. Neurocognitive measures related to GF-R at baseline (r = 0.20, p = 0.004), GF-S at present (r = 0.14, p = 0.042) and of the past year (r = 0.19, p = 0.005), for GAF-F of the past month (r = 0.24, p < 0.001) and GAF-D of the past year (r = 0.28, p = 0.002). Classification reached values of balanced accuracy of 61% for GF-R and GAF–D. We found that neurocognition was related to psychosocial functioning. More specifically, a deficit in executive functions was associated to poor social and occupational functioning. [ABSTRACT FROM AUTHOR]

Background: Clinical high-risk states for psychosis (CHR) are associated with functional impairments and depressive disorders. A previous PRONIA study predicted social functioning in CHR and recent-onset depression (ROD) based on structural magnetic resonance imaging (sMRI) and clinical data. However, the combination of these domains did not lead to accurate role functioning prediction, calling for the investigation of additional risk dimensions. Role functioning may be more strongly associated with environmental adverse events than social functioning. Aims: We aimed to predict role functioning in CHR, ROD and transdiagnostically, by adding environmental adverse events-related variables to clinical and sMRI data domains within the PRONIA sample. Method: Baseline clinical, environmental and sMRI data collected in 92 CHR and 95 ROD samples were trained to predict lower versus higher follow-up role functioning, using support vector classification and mixed k-fold/leave-site-out cross-validation. We built separate predictions for each domain, created multimodal predictions and validated them in independent cohorts (74 CHR, 66 ROD). Results: Models combining clinical and environmental data predicted role outcome in discovery and replication samples of CHR (balanced accuracies: 65.4% and 67.7%, respectively), ROD (balanced accuracies: 58.9% and 62.5%, respectively), and transdiagnostically (balanced accuracies: 62.4% and 68.2%, respectively). The most reliable environmental features for role outcome prediction were adult environmental adjustment, childhood trauma in CHR and childhood environmental adjustment in ROD. Conclusions: Findings support the hypothesis that environmental variables inform role outcome prediction, highlight the existence of both transdiagnostic and syndrome-specific predictive environmental adverse events, and emphasise the importance of implementing real-world models by measuring multiple risk dimensions. [ABSTRACT FROM AUTHOR]

Schizotypy refers to a set of personality traits that bear resemblance, at subclinical level, to psychosis. Despite evidence of similarity at multiple levels of analysis, direct comparisons of schizotypy and clinical psychotic disorders are rare. Therefore, we used functional magnetic resonance imaging (fMRI) to examine the neural correlates and task-based functional connectivity (psychophysiological interactions; PPI) of smooth pursuit eye movements (SPEM) in patients with recent onset psychosis (ROP; n = 34), participants with high levels of negative (HNS; n = 46) or positive (HPS; n = 41) schizotypal traits, and low-schizotypy control participants (LS; n = 61) using machine-learning. Despite strong previous evidence that SPEM is a highly reliable marker of psychosis, patients and controls could not be significantly distinguished based on SPEM performance or blood oxygen level dependent (BOLD) signal during SPEM. Classification was, however, significant for the right frontal eye field (FEF) seed region in the PPI analyses but not for seed regions in other key areas of the SPEM network. Applying the right FEF classifier to the schizotypal samples yielded decision scores between the LS and ROP groups, suggesting similarities and dissimilarities of the HNS and HPS samples with the LS and ROP groups. The very small difference between groups is inconsistent with previous studies that showed significant differences between patients with ROP and controls in both SPEM performance and underlying neural mechanisms with large effect sizes. As the current study had sufficient power to detect such differences, other reasons are discussed. [ABSTRACT FROM AUTHOR]

Diagnostic heterogeneity within and across psychotic and affective disorders challenges accurate treatment selection, particularly in the early stages. Delineation of shared and distinct illness features at the phenotypic and brain levels may inform the development of more precise differential diagnostic tools. We aimed to identify prototypes of depression and psychosis to investigate their heterogeneity, with common, comorbid transdiagnostic symptoms. Analyzing clinical/neurocognitive and grey matter volume (GMV) data from the PRONIA database, we generated prototypic models of recent-onset depression (ROD) vs. recent-onset psychosis (ROP) by training support-vector machines to separate patients with ROD from patients with ROP, who were selected for absent comorbid features (pure groups). Then, models were applied to patients with comorbidity, ie, ROP with depressive symptoms (ROP+D) and ROD participants with sub-threshold psychosis-like features (ROD+P), to measure their positions within the affective-psychotic continuum. All models were independently validated in a replication sample. Comorbid patients were positioned between pure groups, with ROP+D patients being more frequently classified as ROD compared to pure ROP patients (clinical/neurocognitive model: χ 2 = 14.874; P <.001; GMV model: χ 2 = 4.933; P =.026). ROD+P patient classification did not differ from ROD (clinical/neurocognitive model: χ 2 = 1.956; P = 0.162; GMV model: χ 2 = 0.005; P =.943). Clinical/neurocognitive and neuroanatomical models demonstrated separability of prototypic depression from psychosis. The shift of comorbid patients toward the depression prototype, observed at the clinical and biological levels, suggests that psychosis with affective comorbidity aligns more strongly to depressive rather than psychotic disease processes. Future studies should assess how these quantitative measures of comorbidity predict outcomes and individual responses to stratified therapeutic interventions. [ABSTRACT FROM AUTHOR]

Depression frequently occurs in first-episode psychosis (FEP) and predicts longer-term negative outcomes. It is possible that this depression is seen primarily in a distinct subgroup, which if identified could allow targeted treatments. We hypothesize that patients with recent-onset psychosis (ROP) and comorbid depression would be identifiable by symptoms and neuroanatomical features similar to those seen in recent-onset depression (ROD). Data were extracted from the multisite PRONIA study: 154 ROP patients (FEP within 3 months of treatment onset), of whom 83 were depressed (ROP+D) and 71 who were not depressed (ROP−D), 146 ROD patients, and 265 healthy controls (HC). Analyses included a (1) principal component analysis that established the similar symptom structure of depression in ROD and ROP+D, (2) supervised machine learning (ML) classification with repeated nested cross-validation based on depressive symptoms separating ROD vs ROP+D, which achieved a balanced accuracy (BAC) of 51%, and (3) neuroanatomical ML-based classification, using regions of interest generated from ROD subjects, which identified BAC of 50% (no better than chance) for separation of ROP+D vs ROP−D. We conclude that depression at a symptom level is broadly similar with or without psychosis status in recent-onset disorders; however, this is not driven by a separable depressed subgroup in FEP. Depression may be intrinsic to early stages of psychotic disorder, and thus treating depression could produce widespread benefit. [ABSTRACT FROM AUTHOR]

• tDCS has modest and heterogeneous effects in major depressive disorder. • Identifying baseline individual predictors of response might enhance clinical outcomes. • XGBoost tree boosting machine-learning algorithms were employed in the ELECT-TDCS trial. • The algorithm achieved a balanced accuracy of 65% for tDCS; it was also therapy-specific. • TDCS response was predicted above-chance using only baseline phenotypic data. [ABSTRACT FROM AUTHOR]

Importance: Social and occupational impairments contribute to the burden of psychosis and depression. There is a need for risk stratification tools to inform personalized functional-disability preventive strategies for individuals in at-risk and early phases of these illnesses.Objective: To determine whether predictors associated with social and role functioning can be identified in patients in clinical high-risk (CHR) states for psychosis or with recent-onset depression (ROD) using clinical, imaging-based, and combined machine learning; assess the geographic, transdiagnostic, and prognostic generalizability of machine learning and compare it with human prognostication; and explore sequential prognosis encompassing clinical and combined machine learning.Design, Setting, and Participants: This multisite naturalistic study followed up patients in CHR states, with ROD, and with recent-onset psychosis, and healthy control participants for 18 months in 7 academic early-recognition services in 5 European countries. Participants were recruited between February 2014 and May 2016, and data were analyzed from April 2017 to January 2018.ain Outcomes and Measures: Performance and generalizability of prognostic models.Results: A total of 116 individuals in CHR states (mean [SD] age, 24.0 [5.1] years; 58 [50.0%] female) and 120 patients with ROD (mean [SD] age, 26.1 [6.1] years; 65 [54.2%] female) were followed up for a mean (SD) of 329 (142) days. Machine learning predicted the 1-year social-functioning outcomes with a balanced accuracy of 76.9% of patients in CHR states and 66.2% of patients with ROD using clinical baseline data. Balanced accuracy in models using structural neuroimaging was 76.2% in patients in CHR states and 65.0% in patients with ROD, and in combined models, it was 82.7% for CHR states and 70.3% for ROD. Lower functioning before study entry was a transdiagnostic predictor. Medial prefrontal and temporo-parieto-occipital gray matter volume (GMV) reductions and cerebellar and dorsolateral prefrontal GMV increments had predictive value in the CHR group; reduced mediotemporal and increased prefrontal-perisylvian GMV had predictive value in patients with ROD. Poor prognoses were associated with increased risk of psychotic, depressive, and anxiety disorders at follow-up in patients in the CHR state but not ones with ROD. Machine learning outperformed expert prognostication. Adding neuroimaging machine learning to clinical machine learning provided a 1.9-fold increase of prognostic certainty in uncertain cases of patients in CHR states, and a 10.5-fold increase of prognostic certainty for patients with ROD.Conclusions and Relevance: Precision medicine tools could augment effective therapeutic strategies aiming at the prevention of social functioning impairments in patients with CHR states or with ROD. [ABSTRACT FROM AUTHOR]

Background: Functional deficits associated with the Clinical High Risk (CHR) status very often lead to inability to attend school, unemployment, as well as social isolation, thus calling for predictors of individual functional outcomes which may facilitate the identification of people requiring care irrespective of transition to psychosis. Studies have revealed that a pattern of cortical and subcortical gray matter volumes (GMV) anomalies measured at baseline in CHR individuals could predict their functional abilities at follow up. Furthermore, literature is consistent in revealing the crucial role of several environmental adverse events in increasing the risk of developing either transition to psychosis, or a worse overall personal functioning. Therefore, the aim of this study is to employ machine learning to test the individual and combined ability of baseline GMV data and of history of environmental adverse events in predicting good vs. poor social and occupational outcome in CHR individuals at follow up. Methods: 92 CHR individuals recruited from the 7 discovery PRONIA sites were included in this project. Social and occupational impairment at follow up (9–12 months) were respectively measured through the Global Functioning: Social (GF:S) and Role (GF:R) scale, and CHR with a follow up rating of 7 or below were labeled as having a poor functional outcome. This way, we could separate our cohort in 52 poor outcome CHR and 40 good outcome CHR. GMV data were preprocessed following published procedures which allowed also to correct for site effects. The environmental classifier was built based on Childhood Trauma Questionnaire, Bullying Scale, and Premorbid Adjustment Scale (childhood, early adolescence, late adolescence and adulthood) scores. Raw scores have been normalized according to the psychometric properties of the healthy samples used for validating these questionnaires and scale, in order to obtain individual scores of deviation from the normative occurrence of adverse environmental events. GMV and environmental-based predictive models were independently trained and tested within a leave-site-out cross validation framework using a Support Vector Machine algorithm (LIBSVM) through the NeuroMiner software, and their predictions were subsequently combined through stacked generalization procedures. Results: Our GMV-based model could predict follow up social outcome with 67.4% Balanced Accuracy (BAC) and significance (p=0.01), while it could not predict occupational outcome (46.6% BAC). On the other hand, our environmental-based model could discriminate both poor vs. good social and occupational outcomes at follow up with, respectively, 71% and 66.4% BACs, and significance (both p=0.0001). Specifically, the most reliable features in the environmental classifier were scores reflecting deviations from the normative values in childhood trauma and adult premorbid adjustment, for social outcome prediction, and in bullying experiences and late adolescence premorbid adjustment, for occupational outcome prediction. Only for social outcome prediction, stacked models outperformed individual classifiers’ predictions (74.3% BAC, p=0.0001). Discussion: Environmental features seem to be more accurate than GMV in predicting both social and occupational outcomes in CHR. Interestingly, the predictions of follow up social and occupational outcomes rely on different patterns of occurrence of specific environmental adverse events, thus providing novel insights about how environmental adjustment disabilities, bullying and traumatic premorbid experiences may impact on different bad outcomes associated with the CHR status. [ABSTRACT FROM AUTHOR]

Background: Evidence exists that cannabis consumption is associated with the development of psychosis. Further, continued cannabis use in individuals with recent onset psychosis (ROP) increases the risk for rehospitalization, high symptom severity and low general functioning. Clear inter-individual differences in the vulnerability to the harmful effects of the drug have been pointed out. These findings emphasize the importance of investigating the inter-individual variability in the role of cannabis use in ROP and to understand how cannabis use relates to subclinical conditions that predate the full-blown disease in clinical high-risk (CHR). Specific symptoms have been linked with continued cannabis consume, still research is lacking on how different factors contribute together to an elevated risk of cannabis relapse. Multivariate techniques have the capacity to extract complex patterns from high dimensional data and apply generalized rules to unseen cases. The aim of the study is therefore to assess the predictability of cannabis relapse in ROP and CHR by applying machine learning to clinical and environmental data. Methods: All participants were recruited within the multi-site, longitudinal PRONIA study (www.pronia.eu). 112 individuals (58 ROP and 54 CHR) from 8 different European research centres reported lifetime cannabis consume at baseline and were abstinent for at least 4 weeks. We defined cannabis relapse as any cannabis consume between baseline and 9 months follow-up reported by the individual. To predict cannabis relapse, we trained a random forest algorithm implemented in the mlr package, R version 3.5.2. on 183 baseline variables including clinical symptoms, general functioning, demographics and consume patterns within a repeated-nested cross-validation framework. The data underwent pre-processing through pruning of non-informative variables and median-imputation for missing values. The number of trees was set to 500, while the number of nodes, sample fraction and mtry were optimized. All hyperparameters were tuned with the model-based optimization implemented in the mlrMBO R package. Results: After 9 months 50 individuals (48 % ROP, 52 % CHR) have relapsed on cannabis use. Relapse was over all timepoints associated with more severe psychotic symptoms measured by PANSS positive and PANSS general (p<0.05) and a significant interaction between positive symptoms and time of measurement (p<0.05). Our random forest classifier could predict cannabis relapse with a balanced accuracy, sensitivity, and specificity of, respectively, 66.5 %, 66.0 % and 67.0 %. The most predictive variables were a higher cumulative frequency of cannabis consumption in the last 3 months, worse general functioning in the last month, higher density of place of living, younger age and a shorter interval time since the last consumption. Discussion: Our results using a state-of-the-art machine learning approach suggest that the multivariate signature of baseline demographic and clinical data could predict follow up cannabis relapse above chance level in CHR and ROP. Our findings revealing that cannabis relapse is associated with more severe symptoms is in line with previous literature and emphasizes the need for targeted treatment towards abstinence from cannabis. The information of demographic and clinical patterns might be useful in order to specifically address therapeutic strategies in individuals at higher risk for relapse. This might include special programs for younger patients and taking model in an independent sample. [ABSTRACT FROM AUTHOR]

Background It is important to decompose the clinical heterogeneity of affective and non-affective psychoses into transdiagnostic subgroups that disentangle prognostic outcomes and inform aetiological understanding. Research thus far has predominantly focused on symptoms, but there is a need to consider a wider range of clinical variables with prognostic and genetic validation. We aimed to detect psychosis subgroups using a novel unsupervised machine learning method taken from cancer genomics and used on a clinical battery of demographics, medical history, symptoms, quality of life, general functioning, and cognition. We validated the subgroups by examining their: a) illness course over 1.5-years; and b) schizophrenia polygenic risk scores (SCZ-PRS). Methods Ongoing multicenter, naturalistic, longitudinal study beginning in 2011 with three follow-up time-points collected at 6-month intervals across 18 sites. Convenience and referred sample of 765 cases (n=411 at follow-up) collected from secondary and tertiary care centers with DSM-IV diagnoses of schizophrenia, bipolar affective (I/II), schizoaffective, schizophreniform, and brief psychotic disorder (45% female; mean age (yrs; SD) = 42.7(12.9)). A clinical battery of 188 variables assessing the domains listed in the objectives was decomposed using sparse non-negative matrix factorization consensus clustering. Subgroups were validated with mixed models and supervised machine learning analyses of illness course. SCZ-PRS was compared between the subgroups using analysis of variance and supervised machine learning. Results Five subgroups were found (permutation-based p<0.001), four of which were transdiagnostic and labelled as later-onset affective psychosis, suicidal psychosis, depressive psychosis, and male, high functioning psychosis. The fifth group was labelled severe psychosis subgroup and predominately consisted of individuals with schizophrenia diagnoses. Subgroup differences in illness courses were found for psychosis symptoms (F(4,406)=9.6, p<0.001), depression symptoms (F(4,406)=3.9, p=0.004), general functioning (F(4,406)=5.8, p<0.001), and quality of life (F(4,406)=3.1, p=0.02). The later onset affective and the high functioning psychosis subgroups demonstrated similar levels of functioning and stable, linear illness courses. The depressive and severe psychosis subgroups exhibited the lowest functioning and quadratic illness courses. Exploratory SCZ-PRS analyses did not reveal differences, but specific tests indicated higher SCZ-PRS in the severe psychosis subgroup across all thresholds. Discussion Five subgroups were detected with distinctive illness courses that crossed traditional nosological boundaries. The study demonstrates the importance of data-driven approaches and highlights the critical importance of illness trajectories to future psychosis taxonomies. [ABSTRACT FROM AUTHOR]

Background: Childhood trauma (CT) is associated with an increased risk for psychiatric disorders like major depression and psychosis. However, the pathophysiological relationship between CT, psychiatric disease and structural brain alterations is still unknown. Methods: PRONIA (‘Personalized Prognostic Tools for Early Psychosis Mangement’) is a prospective collaboration project funded by the European Union under the 7th Framework Programme (grant agreement n° 602152). Considering a broad set of variables (sMRI, rsMRI, DTI, psychopathological, life event related and sociobiographic data, neurocognition, genomics and other blood derived parameters) as well as advanced statistical methods, PRONIA aims at developing an innovative multivariate prognostic tool enabling an individualized prediction of illness trajectories and outcome. Seven clinical centers in five European countries and in Australia participate in the evaluation of three clinical groups (subjects clinically at high risk of developing a psychosis (CHR), patients with a recent onset psychosis (ROP) and patients with a recent onset depression (ROD)) as well as healthy controls (HC). To investigate the high-dimensional patterns of CT experience, measured by the childhood trauma questionnaire (CTQ), in HC and our three patient groups (PAT) (n=643), we used a Support Vector Machine (SVM). Furthermore, we tested whether patient-specific CT exposure is associated with structural brain changes by VBM analyses. Results: We found that patients and HC could be separated very well by their CTQ pattern, whereas the different patient groups showed no specific CTQ pattern. Furthermore, an association with extensive grey matter changes suggests an impact on brain maturation which may put individuals at increased risk for mental disease. Discussion: We have demonstrated in this large multi-center cohort that adverse experiences in childhood contribute transdiagnostically to the riskr for developing a psychiatric disease. The observed association between CTQ scores and structural changes suggests an impact of adverse childhood experiences on brain development. Resulting alterations may add to a neurobiological vulnerability for depression and psychosis. A role of both features for other mental disorders could be assumed and warrants further investigation. [ABSTRACT FROM AUTHOR]

Background: Childhood maltreatment (CM) is a major psychiatric risk factor and leads to long-lasting physical and mental health implications throughout the affected individual’s lifespan. Nonetheless, the neuroanatomical correlates of CM and their specific clinical impact remain elusive. This might be attributed to the complex, multidimensional nature of CM as well as to the restrictions of traditional analysis pipelines using nosological grouping, univariate analysis and region-of-interest approaches. To overcome these issues, we present a novel transdiagnostic and naturalistic machine learning approach towards a better and more comprehensive understanding of the clinical and neuroanatomical complexity of CM. Methods: We acquired our dataset from the multi-center European PRONIA cohort (www.pronia.eu). Specifically, we selected 649 male and female individuals, comprising young, minimally medicated patients with clinical high-risk states for psychosis as well as recent-onset of depression or psychosis and healthy volunteers. As part of our analysis approach, we created a new Matlab Toolbox, which performs multivariate Sparse Partial Least Squares Analysis in a robust machine learning framework. We employed this algorithm to detect multi-layered associations between combinations of items from the Childhood Trauma Questionnaire (CTQ) and grey matter volume (GMV) and assessed their generalizability via nested cross-validation. The clinical relevance of these CM signatures was assessed by correlating them to a wide range of clinical measurements, including current functioning (GAF, GF), depressivity (BDI), quality of life (WHOQOL-BREF) and personality traits (NEO-FFI). Results: Overall, we detected three distinct signatures of sexual, physical and emotional maltreatment. The first signature consisted of an agedependent sexual abuse pattern and a corresponding GMV pattern along the prefronto-thalamo-cerebellar axis. The second signature yielded a sexdependent physical and sexual abuse pattern with a corresponding GMV pattern in parietal, occipital and subcortical regions. The third signature was a global emotional trauma signature, independent of age or sex, and projected to a brain structural pattern in sensory and limbic brain regions. Regarding the clinical impact of these signatures, the emotional trauma signature was most strongly associated with massively impaired state- and trait-level characteristics. Both on a phenomenological and on a brain structural level, the emotional trauma pattern was significantly correlated with lower levels of functioning, higher depression scores, decreased quality of life and maladaptive personality traits. Discussion: Our findings deliver multimodal, data-driven evidence for a differential impact of sexual, physical and emotional trauma on brain structure and clinical state- and trait-level phenotypes. They also highlight the multidimensional nature of CM, which consists of multiple layers of highly complex trauma-brain patterns. In broader terms, our study emphasizes the potential of machine learning approaches in generating novel insights into long-standing psychiatric topics. [ABSTRACT FROM AUTHOR]

Childhood adversities and trauma (CAT) are associated with poor functional outcome. However, the influence of the single CAT aspects on the risk of a poor functional outcome within different mental disorders has not been investigated so far. Our aims were (i) to predict individual functional outcome based on CAT (ii) to examine whether the prediction power differs within different diagnostic groups (clinical high-risk for psychosis (CHR), psychosis, affective disorders, anxiety disorders) (iii) to compare the specific patterns of CAT experiences, influencing functional outcomes in these groups. Clinical data of 707 patients (mean age: 25.09 years (SD = 5.6), 65.5% male) of the Cologne Early Recognition and Intervention Center were assessed with the Trauma And Distress Scale. Functional outcome was estimated by the Social and Occupational Functioning Assessment Scale and school educational attainment. Using machine learning, we generated individualized models to predict functional outcome and to identify specific CAT patterns. Across the entire sample, the best prediction for the functional outcome achieved a balanced accuracy (BAC) of 0.6. After splitting into the single diagnostic groups, an improvement with best results in the psychosis group (BAC = 0.70) was observed. Considering specific CAT patterns, the most predictive items depicted a positive and caring environment – or the absence of these, a positive self-image and experiences of bullying. Our results indicated that CAT was differentially associated with functional outcome in the various mental disorders. Thus, the importance of mediating variables, that might explain the interindividual differences in the vulnerability to CAT, like resilience factors, appeared to be crucial. • Investigation of patterns of childhood trauma using machine-learning techniques. • Large cohort of patients (n = 707) with different psychiatric diagnoses. • Childhood trauma was differentially associated with functional outcome. • Best results were achieved in the psychosis group (balanced accuracy = 0.70). [ABSTRACT FROM AUTHOR]