Prognostic tools in the management of clinical high risk for psychosis

Large consortia such as PSYCAN, PRONIA, and NAPLS have contributed to a fast-growing evidence base and increasing insights into the clinical, cognitive, and biochemical characteristics of patients suffering from psychosis. Prediction tools could considerably contribute to identifying patients at high risk for psychosis, but the development of these instruments has not been straightforward.

Although several neuroimaging studies have assessed brain abnormalities associated with the early stages of psychosis, this has not led to any specific tools to aid prediction of psychosis onset nor clinical outcome, according to Prof. Celso Arango (Hospital General Universitario Gregorio Marañón, Spain) [1]. These failures are due to several reasons, including the use of univariate analytical techniques and the lack of statistical power, external validation of potential biomarkers, and integration of non-imaging measures (e.g. genetic, clinical, or cognitive data) [2].

PSYSCAN is an international, longitudinal, multicentre study that implements machine learning to analyse imaging, clinical, cognitive, and biological data to facilitate the prediction of psychosis onset and outcome. Included were 702 participants (data cut-off: September 2020), consisting of 237 participants at clinical high risk for psychosis, 328 participants with a first episode of psychosis, and 137 controls. Clinical, cognitive, imaging, and peripheral data has been collected for all participants. The participants at clinical high risk of psychosis will be followed up for 24 months [1]. Based on the preliminary data, the objective of the study has changed from developing a tool which will help clinicians resolve key issues in managing patients with psychotic disorders to developing a tool that can be used in academic studies and clinical trials [2].

Dr Carrie Bearden (Semel Institute of Neuroscience and Human Behaviour, USA) presented data from the NAPLS study, which aims to refine algorithms to predict psychosis prospectively in at-risk youth and to assess biological markers that may improve clinical predictive algorithms. It further aims to determine whether neurobiological abnormalities are stable or progressive. Enrolled in this study were 743 subjects at high risk for psychosis. Follow-up data was available for 596 participants, of whom 84 developed a psychotic disorder. An individual risk prediction model was developed based on these patients, which is available online. This multivariate model achieved a concordance index of 0.71 [3].

Another study based on the phase 2 data from NAPLS evaluated the role of biomarkers and found that combining predictors with grey matter loss resulted in a higher psychosis predicting accuracy of 76% [4]. The polygenic risk score reflects the cumulative genome-wide impact of common genetic variation on a given phenotype into a single measure of genetic risk. It has been used in predicting other common diseases including breast cancer. When applied to evaluate the NAPLS participants, findings showed that participants with a higher polygenic risk score had an HR 1.47 (95% CI 0.91-2.37) of conversion to psychosis, which also occurred earlier than in other individuals. As the polygenic risk score was developed mainly with European patient data, it was noted that the score is more specific in European patients (non-Europeans HR 1.86 (95% CI 0.98-2.23)). By adding the polygenic risk score to the clinical prediction, the total predictive ability can be increased modestly. These clinical high-risk criteria are useful to elucidate predictors and mechanisms of the onset of psychosis, with prodromal criteria being among the most powerful risk factors for psychosis to date.

A multivariate combination necessary for perfect discrimination has not yet been obtained, but biological markers may improve the risk assessment, particularly grey matter loss and polygenic risk score. Challenges are sample size, the need for intervention, and heterogeneity of outcomes/risk profiles [1].

 

1. Arango C. What’s the role of prognostic tools in management of clinical high risk for psychosis? BS.04. ECNP Congress 2020.


2. Tognin S, et al. Schizophrenia Bulletin. 2020;46(2):432-441.


3. Cannon TD, et al. Study Am J Psychiatry. 2016 Oct 1;173(10):980-988.


4. Perkins DO, et al. Am J Psych. 2019;177(2):155-163.

Posted on 2 November, 20205 August, 2021