Personalised medicine aims to assign a patient the optimal treatment, based on their predicted response; thereby moving from average group outcomes towards the prediction of outcomes for individual patients. The challenge of choosing a drug treatment for a particular patient with mental illness could be guided by an understanding of patient characteristics associated with better treatment outcomes. In the case of personalising treatment for severe mental illness, there is ongoing work focusing on genome-wide associations and biomarkers (for example, from functional magnetic resonance imaging and electroencephalography). However, these approaches have yet to bear fruit clinically; early attempts have been plagued by failures to replicate findings and small sample sizes. In the case of genomics, the low frequency of identified response-associated alleles means genetic testing alone is yet to be predictive in patients. Many of these approaches will also be challenging to deliver in routine practice, but any valid prediction arising from this work could be paired with the methods the aiMH lab is implementing.
Personalised medicine aims to assign a patient the optimal treatment, based on their predicted response; thereby moving from average group outcomes towards the prediction of outcomes for individual patients. The challenge of choosing a drug treatment for a particular patient with mental illness could be guided by an understanding of patient characteristics associated with better treatment outcomes. In the case of personalising treatment for severe mental illness, there is ongoing work focusing on genome-wide associations and biomarkers (for example, from functional magnetic resonance imaging and electroencephalography). However, these approaches have yet to bear fruit clinically; early attempts have been plagued by failures to replicate findings and small sample sizes. In the case of genomics, the low frequency of identified response-associated alleles means genetic testing alone is yet to be predictive in patients. Many of these approaches will also be challenging to deliver in routine practice, but any valid prediction arising from this work could be paired with the methods the aiMH lab is implementing.
Despite the rich data available, little attention has been paid to predictive analytics using other data sources such as electronic health registers, population-based registers or remote monitoring, which can provide real world outcomes and adverse event information for large numbers of patients with mental illness. Given the increasing number of variables available, flexible machine learning procedures may usefully complement hypothesis driven traditional regression methods, because of the ability to discover hidden risk factors and interactions, nonlinear, and higher-order effects, as well as to approximate intricate functions that are poorly represented by individual covariate or interaction terms. The aiMH Lab project - PrISM (Predicting pharmacological treatment Response In Severe Mental illness) aims to deliver a suite of prediction models for optimising treatment response and minimising risk of adverse effects at different illness stages in severe mental illness.
Publications
J. Hayes, D. Osborn, E. Francis, G. Ambler, L. Tomlinson, M. Boman, I. Wong, J. Geddes, C. Dalman, Glyn Lewis
BMC Medicine, 2021
Yue Wei, Vincent K. C. Yan, Wei Kang, I. Wong, D. J. Castle, Le Gao, C. Chui, K. Man, J. Hayes, W. Chang, E. Chan
JAMA network open, 2022
V. W. W. Ng, Le Gao, E. Chan, Ho Ming Edwin Lee, J. Hayes, D. Osborn, T. Rainer, K. Man, I. C. Wong
Psychological medicine, 2022
J. Hayes, S. Hardoon, J. Deighton, E. Viding, D. Osborn
Journal of Psychopharmacology, 2022