Precision psychiatry envisions a future in which treatment is tailored using a detailed picture of each patient’s biology, behavior and environment. A psychiatrist might combine smartphone data on sleep and speech, blood tests, brain scans and neighborhood pollution levels, then compare those measurements with data from thousands of other people to identify the treatments most likely to help. The approach could make mental-health care more precise, but researchers and clinicians caution that many of its proposed tools remain unproven.
That vision is a world away from the experience of most people today. Despite major advances across medicine, psychiatry remains a frustratingly imprecise discipline. Doctors typically diagnose conditions and prescribe treatments based on symptoms and medical history. But mental-health diagnoses can be broad and vague: a treatment that transforms one person’s life may do little for someone else with the same diagnosis.
People with mental-health conditions therefore often repeat treatments and adjust dosages, searching largely through trial and error for an approach that works.
To improve this situation, psychiatrists want to use data-driven methods to divide broad diagnoses into more useful subgroups and tailor treatments to individuals. Some researchers are studying genes and brain circuits, while others are examining data on conversation, sleep and social situations. Researchers working on the Diagnostic and Statistical Manual of Mental Disorders (DSM), one of the field’s leading classification systems, are also considering how socioeconomic, cultural and environmental factors could be incorporated in the future.
“I think precision psychiatry is certainly part of the future,” says Tom Insel, a psychiatrist and former director of the National Institute of Mental Health (NIMH) who now co-directs several companies focused on mental health. “From the beginning, this patient needed this treatment and it was the best treatment.”
Some clinicians, however, say that a personalized approach is still unproven. Advances in psychiatric genetics, brain imaging and biomarker research have so far fallen short of expectations, while research continues to show how heterogeneous many mental illnesses are.
“There’s a model in psychiatry where a white knight comes over the hill and solves all our problems,” says Kenneth Kendler, a psychiatric geneticist at Virginia Commonwealth University in Richmond. “And we have been disappointed time and time again.”
More accurate classification
Personalized psychiatry may be a catchy phrase, but the underlying idea is not new.
For decades, psychiatry has sought to improve the accuracy of diagnosis and treatment. The third edition of the DSM, published in 1980, introduced specific diagnostic criteria for many symptoms. The aim was to improve reliability, so that a diagnosis such as major depression or schizophrenia would mean roughly the same thing to different psychiatrists.
But the DSM defines conditions largely through symptoms, and that approach creates difficulties. A person can receive multiple diagnoses because the same symptoms occur in several disorders. In addition, two people with the same diagnosis can have very different symptoms.
To meet the DSM criteria for major depressive disorder, for example, a person must experience at least five of nine symptoms during a two-week period. At least one must be depressed mood or a loss of interest or pleasure. The other symptoms include insomnia or excessive sleeping, weight gain or loss, feelings of guilt or worthlessness, and suicidal thoughts.
“Two people can get the same diagnosis even if they don’t share any of the necessary symptoms,” says Vikram Patel, a psychiatrist at Harvard Medical School in Boston, Massachusetts.
By the late 2000s, researchers had begun to question whether established disease categories were the best way to classify and treat people.
At the NIMH, Insel helped develop the Research Domain Criteria initiative, which seeks to study and measure processes such as reward, fear and cognition that cut across diagnostic categories. The approach encouraged researchers to examine shared behaviors and brain functions in different conditions. Although it has helped shift research towards the mechanisms underlying illness, it was not designed to replace symptom-based diagnostic systems or provide individuals with a reliable way to choose treatments.
Another way to improve accuracy is to measure symptoms and behavior more closely. A psychiatrist evaluating psychosis, for example, might listen to a patient’s account and assess qualities such as coherence, paranoia and grandiosity. This process draws on clinical expertise but also involves human subjectivity.
Insel envisions machine-learning systems listening to the same conversations and assigning measurements that are more objective and repeatable. “The paranoia score here is 6.1,” he suggests. “The coherence level is 2.2.”
John Torous, a psychiatrist at Beth Israel Deaconess Medical Center in Boston and Harvard Medical School, is working on a related approach known as digital phenotyping. It uses information from smartphones and wearable sensors to study behavior. The term emerged about a decade ago, when researchers proposed that sensors already built into smartphones could continuously measure behaviors such as movement and communication and potentially aid diagnosis.1
A small study subsequently found a link between activity levels and depressive or manic symptoms in people with bipolar disorder.2 Another reported an association between location data and symptom severity in schizophrenia.3
The challenge is turning these measurements into clinically useful information. Torous says medical training generally does not teach clinicians how to interpret such data, and researchers have not yet established standards or protocols for using it. “We are still in the early stages in that sense,” he says.
Artificial intelligence could help by analyzing multiple data streams and packaging them into information that clinicians can use, he suggests.
The approach also raises privacy concerns. Data on a person’s location, movement, sleep and communication can be highly sensitive and could be used to identify them. In addition, a systematic review published this year found wide variation among digital-phenotyping studies in the devices and analytical methods they use.4 That variation could limit the reproducibility of the findings and the technology’s clinical use.
Biological markers
Looking more deeply into biology could offer another way to improve psychiatric accuracy. As genetics and neuroscience advance, some researchers hope their findings will divide a wide range of psychiatric syndromes into distinct diseases with different biological mechanisms and, ultimately, different treatments.
That has not happened yet, and some researchers argue that it may be a futile effort. Biological insights helped scientists distinguish between two types of diabetes about a century ago, but “psychiatry had never approached this,” Kendler says. Mental-health conditions can be “highly heterogeneous.”
Even so, genetics has produced one of psychiatry’s more promising potential biomarkers. By analyzing the genomes of thousands of people with schizophrenia in genome-wide association studies, researchers have identified genetic variants that are associated with risk.
For many health conditions, the small effects of numerous genetic variants combine to increase susceptibility. Geneticists refer to this combined estimate as a polygenic risk score. People in the highest part of the polygenic-risk distribution for schizophrenia may have about a 10-fold increase in risk, Kendler says, although the size of the increase varies between studies. Published estimates, for example, have placed the risk for people in the top 1% at about six times that of everyone else.5 But the scores alone cannot predict who will develop symptoms.
“Are we at that point now where we can take that score and immediately change what we’re doing diagnostically? No,” Kendler says.
In five or 10 years, Kendler says, a score that distinguishes between schizophrenia and bipolar disorder could be useful in difficult cases where the distinction affects treatment.
Researchers are also investigating hormone and metabolite levels as potential psychiatric biomarkers. But Kendler says such measurements “almost always” turn out to be nonspecific. It is difficult to translate average biological differences between groups into tests that can determine what is wrong with a particular person or which treatment that person should receive.
Brain imaging presents a similar problem. Magnetic resonance imaging (MRI) studies have reported many associations between brain characteristics and psychiatric symptoms, but numerous findings have been difficult to reproduce.
A 2022 analysis of about 50,000 participants found that associations between behavioral and mental-health traits and differences in brain structure and function were generally much smaller than earlier studies had suggested.6
Helen Mayberg, a neurologist at the Icahn School of Medicine at Mount Sinai in New York City, says brain scans can still provide clinically useful information when researchers ask sufficiently specific questions.
Mayberg has spent decades mapping brain circuits involved in depression, focusing on a region of the cortex called the subcallosal cingulate. It lies beneath the corpus callosum, a dense bundle of fibers connecting the two sides of the brain.
In one study, researchers used functional MRI to scan the brains of 122 people with depression.7 The participants were then randomly assigned to one of two treatment groups. One received 12 weeks of cognitive behavioral therapy, while the other received antidepressants for the same period.
The researchers found that connections between the subcallosal cingulate and three other brain regions differed according to treatment outcome. People who responded to psychotherapy and those who responded to drug therapy generally showed different connectivity patterns from people who had the opposite response.
In other words, the connectivity pattern could serve as a biomarker for assigning treatment. “The scans done at baseline can actually help make that decision,” Mayberg says.
There is still a considerable distance between potential biomarkers and routine clinical tests. Mayberg says she was told many times that “no one would do a scan to determine how to treat a human being” because it would be too expensive and time-consuming.
Source: www.nature.com
