Cheryl Corcoran

Diagnosis and treatment of a patient with both psychotic and obsessive-compulsive symptoms.

When a patient presents with both psychotic and obsessive-compulsive symptoms, the clinician is faced with a differential diagnosis that includes comorbid schizophrenia and obsessive-compulsive disorder (OCD), OCD with poor insight, and schizophrenia with antipsychotic-induced obsessive-compulsive symptoms. If the psychotic symptoms are subthresh-old or attenuated in form, the individual may have OCD and putative prodromal schizophrenia. The authors present a case to outline a strategy for differentiating among these possible diagnoses and for optimizing treatment.

Prediction of psychosis across protocols and risk cohorts using automated language analysis

Language and speech are the primary source of data for psychiatrists to diagnose and treat mental disorders. In psychosis, the very structure of language can be disturbed, including semantic coherence (e.g., derailment and tangentiality) and syntactic complexity (e.g., concreteness). Subtle disturbances in language are evident in schizophrenia even prior to first psychosis onset, during prodromal stages. Using computer‐based natural language processing analyses, we previously showed that, among English‐speaking clinical (e.g., ultra) high‐risk youths, baseline reduction in semantic coherence (the flow of meaning in speech) and in syntactic complexity could predict subsequent psychosis onset with high accuracy. Herein, we aimed to cross‐validate these automated linguistic analytic methods in a second larger risk cohort, also English‐speaking, and to discriminate speech in psychosis from normal speech. We identified an automated machine‐learning speech classifier – comprising decreased semantic coherence, greater variance in that coherence, and reduced usage of possessive pronouns – that had an 83% accuracy in predicting psychosis onset (intra‐protocol), a cross‐validated accuracy of 79% of psychosis onset prediction in the original risk cohort (cross‐protocol), and a 72% accuracy in discriminating the speech of recent‐onset psychosis patients from that of healthy individuals. The classifier was highly correlated with previously identified manual linguistic predictors. Our findings support the utility and validity of automated natural language processing methods to characterize disturbances in semantics and syntax across stages of psychotic disorder. The next steps will be to apply these methods in larger risk cohorts to further test reproducibility, also in languages other than English, and identify sources of variability. This technology has the potential to improve prediction of psychosis outcome among at‐risk youths and identify linguistic targets for remediation and preventive intervention. More broadly, automated linguistic analysis can be a powerful tool for diagnosis and treatment across neuropsychiatry.

Marijuana and adolescence: what can we learn from primates?

Inan article published concurrently with this editorial, Verrico and colleagues (1) report that D9-tetrahydrocannabinol (THC), the active ingredient of marijuana, administered to monkeys for 6 months during their adolescence, impairs development of spatial working memory. Normal development of spatial working memory was observed in adolescentmonkeys administered only saline. The spatial working memory task entailed monkeys 1) looking at a screen that had a square displayed at random in one of four corners, 2) touching the square, 3) looking at a “fixation” cross for an interval of 1, 4, 8, or 16 seconds, and 4) then pointing to the corner where the monkey remembered the square was before. The effect of chronic administration of THC was specific to spatial working memory in the monkeys, as the recall of a colored object independent of location was not affected by THC in terms of either accuracy or reaction time. The exposure to THC in adolescent monkeys was similar to that obtained by a human smoking one to two marijuana cigarettes, five times per week, for 6 months. Verrico and colleagues earlier found that acute administration of THC similarly impaired spatial working memory (2). These prior data are consistent with an extensive literature showing that THC “challenge” leads to transient working and other memory impairment across species, including rodents, humans, and nonhuman primates (3). But while the transient cognitive effects of challenge or acute administration of THC can be studied in humans (specifically adults) to observe direct effects of marijuana, random assignment to chronic administration of marijuana cannot be done in human adolescents, perhaps the most vulnerable group. A model of chronic cannabis exposure in monkeys offers the opportunity to observe the direct biological effects of the active component of marijuana on the developing adolescent primate brain and its cognitive function, independent of the confounds that exist in observational studies of teens. The effect of marijuana on the developing adolescent brain and its cognitive function is an especially salient public health issue in 2014. The prescription of marijuana for medical conditions is legal in nearly half of our 50 states. The sale of marijuana for recreational use is legal now in both Colorado and Washington, with other states weighing this option. The market for marijuana is large. For example, while in 2013 the state of Colorado collected

26. NOVEL APPROACHES TO PSYCHOSIS RISK: MOVEMENT, STRESS MODULATION, REWARD AND LANGUAGE

9million in tax revenue from “medical marijuana” dispensaries (4), the projected state tax revenue in 2014 in Colorado for retail marijuana is

33.1 DRIVERS OF STIGMA FOR THE CLINICAL HIGH-RISK STATE FOR PSYCHOSIS—IS STIGMA DUE TO SYMPTOMS OR THE AT-RISK IDENTIFICATION ITSELF?

67 million, based on an estimated

O2.3. AUTOMATED ANALYSIS OF RECENT-ONSET AND PRODROMAL SCHIZOPHRENIA

578 million in sales (5). A recent Gallup poll shows that more than half of all Americans support the legalization of marijuana (6); therefore, it is likely that more states will join the trend of legalization and marijuana will become even more widely available. While marijuana can be sold legally only to adults in the United States, whether formedical or recreational purposes, it is nonetheless readily available to teens. The Even if performance of cognitive tasks does not seem to be affected in adolescent abusers, the brain’s reserve may already be compromised.

S21. EVENT-RELATED REPETITIVE TMS TO RIGHT POSTERIOR STS (BUT NOT OCCIPITAL FACE AREA) IN HEALTHY VOLUNTEERS (HV) BRIEFLY RECAPITULATES FACE EMOTION RECOGNITION (FER) DEFICITS OF SCHIZOPHRENIA (SZ)

Abstract Overall Abstract: Research on psychosis risk now encompasses novel and innovative approaches for understanding not only positive symptoms, but also impairment in sensorimotor function, stress regulation, reward learning, and language. These include the use of machine learning and cluster analysis with resting state functional connectivity analyses, in vivo measures of dopamine function in response to stress, computational modeling, and automated natural language processing analyses in collaboration with IBM. First, Vijay Mittal will describe subtypes of clinical risk, identifying a group with aggregated measures of sensorimotor dysfunction, developmental markers, negative symptoms and cognitive deficits, who have a discrete pattern of corticostriatal connectivity. Second, Romina Mizrahi will present her results from a study of dopamine response to stress in prefrontal cortex, using positron emission tomography, and correlations with cortisol release, across stages of illness, including schizophrenia and clinical risk, with healthy volunteers for comparison. Third, James Waltz will present data on the computational processes that may underlie both positive and negative symptoms, in respect to dopamine-based signals of salience. These include aberrant or erratic salience signaling, as well as a decreased ability to identify relevant salient stimuli, which could impair reward learning and motivation. His cohort includes individuals with psychosis, and those at clinical risk for it, as well as non-psychosis patient controls. Fourth, Cheryl Corcoran will describe the use of automated natural language processing (NLP), with machine learning (ML) to identify semantic and syntactic features that predict psychosis onset. She will show data on cross-validation of the classifier in a second risk cohort, and its correlation with demographics and manual linguistic features. Overall, there is an apparent norm of semantic coherence and syntactic complexity from which individuals with psychosis deviate, even prior to its onset. Finally, the discussant will review these data in the context of his experience and ongoing leadership in the field of psychosis risk research, leading audience discussion, and outlining a roadmap for future research in the field.

26.4 LANGUAGE DISTURBANCE AS A PREDICTOR OF PSYCHOSIS ONSET IN YOUTH AT ENHANCED CLINICAL RISK

Abstract Background The clinical high-risk state for psychosis syndrome (CHR) offers substantial potential benefits in terms of early identification and treatment for at-risk youth. Early treatment might lead to decreased symptoms, thus leading to reduced symptom-related stigma. However, stigma of the clinical high-risk state for psychosis designation might also initiate further stigma through the label of risk for psychosis. Identifying the effects of these sources of stigma is critical in order to best minimize stigma associated with CHR identification and to facilitate recovery. Methods Baseline stigma assessments were conducted with 170 clinical high risk state for psychosis individuals in a major, NIH-funded longitudinal study at Columbia University Medical Center, Harvard University Medical Center, and Maine Medical Center from 2012 to 2017. Labeling-related measures of stigma (e.g., “shame of being identified as at psychosis-risk”) adapted to the CHR group, and a parallel measure of symptom-related stigma (e.g., “shame of the symptoms associated with CHR”) were administered. These measures were examined in relation to outcomes of: a) self-esteem, b) quality of life, and c) social functioning, adjusting for sociodemographics and core CHR symptoms (e.g. attenuated psychotic symptoms). Results Results indicated that stigma related to symptoms was more strongly associated with all outcomes when compared with shame related to the risk-label. Stigma related to symptoms remained a significant predictor of self-esteem and quality of life even after accounting for stigma related to the risk-label and the effects of sociodemographics and CHR symptoms. Conversely, stigma related to the risk-label was no longer a significant predictor for outcomes after accounting for stigma related to symptoms. Discussion Overall, symptom-related stigma was a more salient correlate and was independently linked with self-esteem and quality of life even after accounting for the effects of stigma related to the risk-label. These results indicate that treating of symptoms through early identification and treatment may provide major benefit for CHR youth by also alleviating symptom-related stigma. These findings also indicate that CHR services should address stigma associated with symptoms immediately at first identification, as these have substantial effects on psychological and functional outcomes. These findings have implications for guiding implementation of specialized CHR services both in the United States and worldwide.

Impaired Motion Processing in Schizophrenia and the Attenuated Psychosis Syndrome: Etiological and Clinical Implications

Abstract Background Psychosis has significant effects on language, to the extent that its disturbance is one of the principal components of diagnosis and prognosis. In particular, two features of language seem to be prominently affected: discourse coherence, observed in patients as derailment (or tangentiality), and discourse richness, observed as poverty of speech. Using automated linguistic analysis on baseline interviews, we have shown in a previous study that it is possible to predict with high accuracy of conversion to psychosis (100%) among a cohort of clinical high-risk youth, by quantifying the subjects’ semantic coherence and syntactic complexity as proxies for derailment and poverty of speech, respectively.1 In the present study, we seek to explore to what extent the prodromal prediction model can discriminate recent-onset schizophrenia patients from matching controls, with the intent of understanding how the prodromal-onset transition is reflected in language. Methods Eighteen recent-onset schizophrenic patients and twelve matching controls had baseline interviews, using an open-ended protocol previously introduced.2 Using automated analysis, transcripts of interviews were evaluated for semantic and syntactic features predicting psychosis onset in an independent cohort. These features were then used to discriminate between patients and controls, applying the same classifier that predicted conversion to psychosis, namely converts laying outside the convex hull of non-converts. Additionally, we compared the discrimination power of this approach against alternative models including alternative linguistic features (e.g. metaphoricity3) and protocols (e.g. short prompts4). Results The convex hull of the controls subjects misclassifies only one of the patient samples, a result that amounts to 95% true positive rate. Surrogating by label randomization and accounting for false and positive negative rates results in a balanced accuracy of 80%, which is comparable to those obtained with alternative automated models, which range from 70% to 85%. Moreover, the present cohort is clearly separable from both converts and non-converts in the CHR cohort when projected in the feature space. Discussion The automated features optimized for prediction of psychotic onset convey highly significant information regarding the discrimination between recent-episode patients and controls. The directionality of effects, however, is not obviously derived from that observed in the prodromal-onset transition. This preliminary study provides the basis for a larger study to better understand language disturbances across these cohorts. References 1.Bedi, G., et al. npj Schizophrenia 1 (2015): 15030. 2. Ben-David, S., et al. Psychiatric Services 65.12 (2014): 1499–1501. 3. Gutiérrez, E.D., et al. Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. 2017. 4. Mota, N.B., et al. PloS one 7.4 (2012): e34928.

Taking care of the carers: support for families of persons with early psychosis

Abstract Background Profound FER deficits exist in Sz, causing social disability, though can be partly remediated with computer-based training. Neurostimulation might augment remediation if critical nodes were identified. We aimed to 1) briefly recapitulate FER deficits of Sz in HV using rTMS to rpSTS, 2) identify connectivity patterns of rpSTS regressed by FER, and 3) apply TMS to rpSTS with fMRI as readout. Methods 1) Nine healthy volunteers had rTMS (10 Hz; 500 msec; 110% RMT) to rpSTS or rOFA (counterbalanced; 10/10 system overlay with standard MRI) concurrent (1/3 trials) with stimuli (http://faces.mpdl.mpg.de/) for emotion or gender identification (button press). 14 Sz patients completed these tasks without TMS. 2) Whole-brain resting-connectivity analyses, seeded by rpSTS, was applied in 27 Sz and 35 HV who also completed the UPenn FER task. 3) BOLD fMRI was obtained in 4 HV pre- and post-TMS to rpSTS (1 Hz; 20 minutes). Results 1) In HV, rTMS to rpSTS only (not OFA) significantly slowed reaction time for FER only (not gender identification): overall F test for logRT (p=.001) with post-hoc rpSTS vs.OFA (p=.005) and rpSTS vs. non-stim trials (p=.004). rpSTS recapitulated slowed RT ad lower FER accuracy of Sz. 2) In both HV and Sz, rpSTS had significant resting connectivity with V1 (p= .00013), positively modulated by FER accuracy. 3) Analyses are ongoing. Discussion rpSTS is a critical node in the FER circuit with connectivity to primary visual cortex modulated by FER, whose disruption recapitulates FER deficits, making it a candidate target for remediatory neurostimulation.