Usefulness of Pharmacogenomics Limited, But Can Be Part of Precision Psychiatry Future

When the David A. Mrazek Award for Psychiatric Pharmacogenomics was established in 2014, most thought it wouldn’t be long before award winners began to discuss tremendous breakthroughs in clinical care made possible via testing the metabolic genes of patients. Nearly a decade later, the promise has not yet been realized, but 2023 Mrazek Award recipient Jordan Smoller, M.D., Sc.D., said pharmacogenomics is still a cause worth pursuing. He presented his lecture on Tuesday at the Annual Meeting.

“It is a controversial topic,” said Smoller, the MGH Trustees Endowed Chair in Psychiatric Neuroscience and a professor of psychiatry at Harvard Medical School. He noted that both APA and the American Academy of Child and Adolescent Psychiatry recommend against using commercial pharmacogenomic tests in practice, and even the International Society of Psychiatric Genetics thinks only one test is valuable (a test for that may indicate an adverse reaction to carbamazepine or oxcarbazepine).

Yet, countless psychiatrists (including many in the audience who raised their hands) have been ordering tests for patients and stating that they helped make a better medication decision.

There have been several clinical trials comparing pharmacogenomic tests with usual decision-making for depression, and though the results have been inconsistent, there have been enough studies that researchers have been able to carry out meta-analyses of the combined data. As Smoller showed during his lecture, the combined data points to a modest, but real improvement in patient remission rates. Smoller cautioned that the pharmacogenomic studies aren’t blinded, which can skew results, and many are funded by industries with a vested interest. “But there is a signal there.”

Still, these tests have limitations, he continued. They typically offer an oversimplified “traffic light” readout of medication options (yes, no, or maybe) that don’t factor in a patient’s other medical conditions, other medications, or diet. And, while a test may eliminate a few medications from consideration, it does not indicate which medication is best.

“Pharmacogenomics, however, is just one corner of a bigger picture we call precision medicine,” Smoller said. “What if we combine genetic information with data from EHRs [electronic health records], biobanks, or mobile devices, and then use some amazing new tools like AI or language processing?”

Smoller then discussed some of his research team’s efforts at reaching that next frontier in psychiatric care, such as accurately predicting illness risk or developing targeted interventions for patients. One example is an AI model that analyzes both structured (prescription history) and unstructured (narrative notes) electronic health record (EHR) language to predict antidepressant response. The model is good enough to predict the class of antidepressant to which a patient will respond in about 74% of cases.

In a related project, Smoller’s team has also developed a program that combines clinical data with a patient self-report survey to stratify people into suicide risk categories. “There is multiple evidence that a majority of people who die by suicide see a doctor in the month preceding death; this is an opportunity to intervene,” he said. He has integrated his program into the EHR system at Harvard so it can offer real-time risk prediction as clinicians enter data during each patient visit.

Smoller thinks pharmacogenomics could be part of the same big data transition. “Right now, commercial tests are telling you about 0.01% of a person’s genome,” he said. “There are unexplored opportunities waiting for all these data we are not yet using.” ■