Predicting Long-Term Psychiatric Outcomes in Schizophrenia: New Research from MRC LMS
In a groundbreaking collaboration between the MRC London Institute of Medical Sciences (LMS), King’s College London, the Universities of Cambridge, Birmingham, Bristol, and others, researchers have achieved a significant milestone. For the first time, they have demonstrated that commonly recorded demographics and biomarkers can serve as predictors of long-term psychiatric outcomes in schizophrenia.
Schizophrenia, a complex mental condition characterized by symptoms of psychosis and difficulty distinguishing reality, affects approximately 1 in 100 individuals. Alarmingly, about a quarter of those diagnosed will develop a treatment-resistant form of the disorder (TRS). Currently, there is no effective way to predict whether someone will develop TRS from their initial experience of psychosis. This is a critical gap because evidence suggests that clozapine, the only licensed treatment for TRS, is more effective when administered early. However, in clinical practice, there are often significant delays before considering clozapine as a treatment option.
Previous research teams have explored methods to predict the risk of developing TRS. However, many existing approaches rely on data that becomes available only after the illness has progressed (e.g., hospitalizations or response to treatment). Additionally, these methods often require specialized technology and a dedicated team for result analysis, making them less practical in a clinical setting.
In a recent publication in collaboration with the MRC LMS Psychiatric Imaging Group, a team has developed an algorithm called Model fOr cloZApine tReaTment (MOZART). This algorithm can predict TRS using routinely collected data such as age, sex, and standard blood test results. What sets MOZART apart from other methods is its ability to predict treatment resistance from the very first experience of psychosis. This has the potential to assist in determining whether an individual would benefit from clozapine treatment. However, further testing and validation of the model are necessary before it can be implemented in a clinical setting.
Dr. Emanuele Osimo, lead author and psychiatrist in the MRC LMS Psychiatric Imaging Group and the MRC LMS Computational Regulatory Genomics group, expressed the significance of their work, stating, “We have dedicated considerable efforts to develop a risk prediction model that could be used clinically. MOZART performs well at low-risk thresholds, allowing for low-risk interventions such as close monitoring. It also demonstrates good calibration, both internally and externally, indicating its potential application to new samples beyond the ones used in its development.”
The initial development of the MOZART model involved data from 785 patients in Cambridgeshire and Birmingham, encompassing factors like age, sex, ethnicity, and routine blood markers. The model was subsequently externally validated using data from over 1000 patients in South London. The model’s performance was assessed based on discrimination (ability to distinguish future TRS cases from non-cases) and calibration (comparison of predicted risk with observed cases). These crucial evaluations ensure the model’s usefulness and its potential for application in external samples.
In the future, MOZART’s predictions could personalize treatment plans for individuals with schizophrenia and enable effective monitoring of TRS development. The research team aims to expand their model by including a larger patient population and diverse settings for further external validation. Additionally, they will continue collaborating with the LMS Computational Regulatory Genomics group to explore additional variables, such as genetic markers, that may serve as signals for the development of TRS.