By Rich Haridy
An international team of researchers has used machine learning to produce a novel blood test that can predict a young person’s risk of developing a psychotic disorder, such as schizophrenia, years before the condition develops.
Only around a quarter of young people who display mild, transitory psychotic symptoms at an early age ultimately go on to develop a serious psychotic disorder. Schizophrenia, for example, is generally not clinically diagnosed until a person reaches their twenties. However, the condition is known to present a number of signs and symptoms than can precede the full-blown psychotic episodes often needed for clinical diagnosis.
This early pre-clinical phase of a psychotic disorder is often referred to as the prodromal stage. In the case of schizophrenia, prodromal symptoms appear in nearly three quarters of patients up to five years before the first episode of psychosis occurs.
David Cotter, a molecular psychiatrist from the Royal College of Surgeons in Ireland and senior author on the new study, suggests early detection of those most at risk of developing psychotic disorders is vital for administering preventative treatments.
“Ideally, we would like to prevent psychotic disorders, but that requires being able to accurately identify who is most at risk,” says Cotter.
The new study first looked at blood samples from a number of 12-year-olds classified as at a clinically high-risk of psychosis. Over recent years several tools have been developed to identify adolescent subjects at the highest risk of developing psychosis.
The 12-year-old subjects were followed until around the age of 18, so the researchers were able to differentiate blood samples between those who went on to suffer a psychotic episode and those who didn’t. Using machine learning, the researchers homed in on a unique pattern of proteins that distinguished those who ultimately went on to develop a psychotic disorder.
Ten particular proteins were identified as most predictive, and the test was subsequently validated in a separate dataset. Using the most accurate protein pattern, the researchers were able to correctly determine which high-risk subjects would go on to develop a psychotic disorder by the age of 18 with a 93-percent accuracy.
The test was less accurate in predicting those high-risk 12-year-olds that did not go on to develop a psychosis by the age of 18. However, considering only between 16 and 35 percent of young people considered at clinical high risk ultimately transition to a full psychotic disorder, even this low level of accuracy could be useful in stratifying those younger patients more likely to develop psychosis.
“Our research has shown that, with help from machine learning, analysis of protein levels in blood samples can predict who is at truly at risk and could possibly benefit from preventive treatments,” says Cotter. “We now need to study these markers in other people at high risk of psychosis to confirm these findings.”
Another compelling insight offered by this new study is the finding that many of these protein markers predicting psychosis are linked with inflammatory processes. There is a small, but burgeoning, body of study finding links between psychosis and autoimmune conditions, suggesting systemic inflammation can influence a number of psychiatric illnesses.
The new research was published in the journal JAMA Psychiatry.