Email Icon Facebook icon Twitter Icon GooglePlus Icon Contact

User Top Menu

ICOSR 2017: Advances in Outcome Prediction for At-Risk Patients

24 Apr 2017

by Lesley McCollum

With sessions on biomarkers and risk prediction, early psychosis and people at high risk for schizophrenia were undoubtedly in the spotlight at the 2017 International Congress on Schizophrenia Research meeting in San Diego.

Philip McGuire of King’s College London opened his Sunday afternoon plenary talk by emphasizing the need to predict whether people will convert to psychosis and who will respond to treatment. Efforts have been quite successful at a group level for determining differences between people who do or do not develop psychosis, said McGuire. The clinical challenge is that decisions need to be made at an individual level. McGuire highlighted existing efforts at risk prediction, such as OASIS and the NAPLS risk calculator (see SRF related news). Multiple consortiums have now been developed to take on the challenge. EU-GEI, a multicenter project with partners across Europe, is assessing predictors in at-risk help-seeking people in order to create a deliverable tool that clinicians can actually use in practice. The National Institute of Mental Health-funded HARMONY initiative brings together several international large-scale studies of at-risk individuals, including PRONIA, NAPLS, PSYSCAN, and PNC, to coordinate data analysis in an effort to generate predictive tools that work across cultural groups.

Optimizing risk prediction

A number of different approaches are under investigation for individual risk prediction. McGuire presented new data on the NMDA receptor antibody, which is found in about 10 percent of ultra-high-risk patients. These patients have poorer cognitive functioning and more negative symptoms, and are less likely to go into remission. McGuire suggested identification of the NMDA receptor antibody—an easier and cheaper assessment than imaging measures—as one method for stratifying high-risk patients.

In the symposium on Tuesday morning, Jun Soo Kwon of Seoul National University in South Korea presented work implicating thalamocortical network abnormalities as a biomarker for psychosis risk. In addition to published findings revealing reduced thalamocortical connectivity in first-episode psychosis (FEP) patients and clinical high-risk (CHR) individuals (Cho et al., 2016), Kwon presented new data using diffusion kurtosis imaging to measure brain microstructure, which revealed thalamic structural alterations in 37 FEP and 76 CHR patients. The convergence of structural and functional connectivity data was encouraging, said Daniel Mathalon of the University of California, San Francisco, during a discussion he led following the symposium.

The symposium also covered a range of tools with promise for individual risk prediction. Larry Seidman of Harvard Medical School presented new data from the NAPLS2 project using neuropsychological data alone to generate four profiles of impairment related to different outcomes. Of 324 participants (including CHR and family members), 58 percent of the “very impaired” group converted to psychosis, and the group was heavily clustered with patients later receiving a diagnosis of schizophrenia.

In a similar vein, Dorien Nieman of the University of Amsterdam in the Netherlands presented data from the Dutch Prediction of Psychosis Study aimed to optimize a prediction model using a combination of premorbid adjustment score and EEG measures. The model distinguished CHR patients who would transition to psychosis by calculating a prognostic score for patients, stratifying each into three risk classes. Their initial study (Nieman et al., 2014) is currently being replicated as part of the HARMONY project.

In the final talk of the symposium, Nikolaos Koutsouleris of Ludwig-Maximilians-University in Germany and coordinator of the EU-funded PRONIA project also presented research aimed to optimize prediction by combining data measures. He demonstrated the powerful approach of machine learning to determine the best combination of data that can be collected in a clinic, presenting new data from a model incorporating structural and functional neuroimaging measures that could predict social functioning outcomes in CHR patients with high sensitivity and specificity. Findings from the PRONIA study are being validated as part of the HARMONY project to build a clinically useful predictive model that works across sites.

Personalizing treatment with prediction

In addition to predicting the onset of psychosis, the ability to predict treatment response could allow for earlier therapeutic intervention. In his plenary talk, McGuire highlighted a number of ongoing efforts to predict treatment response, including the EU-funded OPTIMISE—a study of 500 medication-naïve patients across Europe for predicting antipsychotic drug response in FEP. New data from this study using positron emission tomography in 26 FEP patients show higher dopamine levels in patients who responded to treatment compared to those who did not, and greater dopamine before treatment correlated with a greater improvement in positive symptoms.

The purpose of clinical risk prediction research is to figure out what to do for patients in the clinic, said Mathalon. A theme among the prediction techniques was stratification of patients to determine the best treatment course that suits their level of risk. However, in his talk, McGuire stated that there is currently no direction for matching risk predictions to treatment course. Stratification of patients and establishing treatment staging based on subgroups will hopefully provide a way forward for more personalized treatment.