Prescribing a medication or other therapy is as much art as science. Efforts to use blood-based tests for biomarkers and genetic tests to better match treatments to patients make sense. Psychiatrists are joining the chorus in calling for — and exploring — such tests. Yet this approach hasn’t paid off for psychiatry, and is elbowing aside other opportunities to help patients today.
No medication is perfect, and any doctor can tell you that certain treatments don’t work for certain patients. So-called precision medicine tests can help identify whether or not individuals will respond to a treatment.
In cancer, the genetic profile of some tumors can strongly influence the therapy chosen to fight it. For instance, a chemotherapy drug called cetuximab is effective against colorectal cancer when the tumor contains a normal K-ras gene, but isn’t effective against tumors with a mutated K-ras gene. In breast cancer, chemotherapies are often chosen based on the specific tumor subtype, which can be identified by genetic tests.
Despite the tremendous effort being poured into identifying biomarkers to help guide treatment for various psychiatric disorders, this work has yet to significantly improve clinical outcomes.
A big roadblock is that we haven’t yet found strong genetic mutations that can guide treatment for mental health issues. It’s not for lack of trying. In 2013, three large groups of investigators set out to analyze 1.2 million genetic variants in 2,256 patients with depression. They were hoping to find a genetic marker that reliably predicted whether a patient would get better with antidepressant medications.
They didn’t find one.
Continuing to look for genetic mutations linked to mental health certainly makes sense. But it’s already possible to match patients with effective treatments using other types of cheap and accessible information.
Analyzing an individual’s sociodemographic data, such as race or gender, and self-reported behavioral information, such as how poorly he or she sleeps, can often point the way to knowing how he or she will respond to a particular therapy. For example, depressed patients who were abused as children respond better to cognitive behavioral therapy than to an antidepressant called nefazodone.
Progress in refining treatment options hasn’t been limited to depression. In treating bulimia, an eating disorder, cognitive behavior therapy has been shown to be less effective among patients with poor social adjustment and low body mass index.
With ever-increasing computing power, researchers will likely be able to extract even more information from preexisting data. Several colleagues and I collected information on 25 variables from more than 4,000 patients who had participated in a depression clinical trial, including questions about sleep patterns, specific phobias, and psychiatric history. We used this information to train a computer to predict whether a patient would respond to an antidepressant called citalopram. Even though the algorithm did not include any biological data, it was successful in predicting treatment outcomes across two different clinical trials.
Based on those 25 variables, several colleagues and I have created a simple online questionnaire to help clinicians determine the best antidepressant treatment for their patients.
The machine learning approach can also work for other mental illnesses. Another group of researchers developed a computer algorithm to choose an antipsychotic medication for patients with a first episode of psychosis. It was accurate for patients in multiple countries throughout Europe.
Neuroscience-based approaches might also someday provide reliable and accurate information to guide treatment. Studies using functional magnetic resonance imaging or electrical recordings from the scalp of a specific brain region (the rostral anterior cingulate cortex) have found promising differences in brain activity that relate to antidepressant outcomes. A deeper biological understanding of psychiatric illness could lead to the development of better treatments.
Psychiatry has embraced other opportunities to innovate and personalize treatment pathways. Cognitive behavioral therapies for a number of disorders have been packaged into computer platforms — This Way Up is an example — and found to be effective both in terms of clinical outcomes and cost. Talkspace, a mental health startup in New York City, lets patients chat with licensed therapists online. In these regards, psychiatry deserves credit for seizing a relatively unique opportunity for delivering effective health care.
The aim of personalized medicine is to find the best treatment for every individual. For psychiatry to realize this ambition, it may have to temper its obsession with biological markers in favor of less glamorous approaches. If it does, we may be able to improve treatments for mental health far sooner than expected.
As a patient, I don’t care whether you test my blood, scan my brain, or ask if I get anxious in public places. I just want to get better.
Adam Chekroud is a PhD candidate in the department of psychology at Yale University and a cofounder of Spring Healthcare. He is also a co-inventor on a patent filed by Yale on an algorithm for selecting antidepressant treatments.