While antidepressant use continues to skyrocket across the United States, only a fraction of those using these drugs achieve remission of symptoms, even after trying several different medications. Even when remission is achieved, medications can lose their effectiveness over time. Patients struggle for months or years with ongoing symptoms and added side effects as their clinicians give their best guesses about what to try next.
Study uses computers to predict patient responses
A study published in the March 2016 issue of The Lancet Psychiatry described a new way to predict patient response to antidepressants using a pattern recognition computer algorithm based on data from major clinical trials. In this type of data analysis, the computer identifies patterns and interactions among variables, in this case depression symptoms, and produces outcome predictions for individual patients, rather than just averages for the whole group.
The authors examined data from two major clinical trials of antidepressants in adults. They looked at whether patients were receiving placebo, which antidepressant they were taking, and who went into remission after 12 weeks. They also took all of the depression symptom questions from the self-report questionnaires used in the clinical trials and determined which questions were most predictive of symptom remission. These questions and responses formulated the model used for the machine learning method of predicted analytics.
For patients taking selective serotonin reuptake inhibitors (SSRIs), accuracy of the study model was 59.6 to 64.6 percent, which was significantly higher than accuracy by chance alone, which was estimated at 51.3 percent. When the researchers tested clinician accuracy off the study model in a subset of patients, accuracy was lower than chance alone. They concluded that such computer-based statistical models might help predict patient responses to medications.
A computer algorithm such as the one described above can aid clinicians in finding the right therapy. Of course, computers cannot replace clinical expertise and experience, especially if individualized care is the goal. Also, such algorithms are based on the assumption that the future will be the same as the past. These models may also require expensive software and data management procedures.
Another way of predicting patient response to SSRIs is pharmacogenetic testing. Individual differences in the gene that determines how people metabolize medications can help predict how effective they will be. A recent study reported that 87 percent of patients who received pharmacogenetic testing had an improvement of depression and anxiety symptoms.
With so many people now taking antidepressants for off-label uses, it is important to not only monitor safety but efficacy as well. There is no sense taking medication that doesn’t work, with all the costs, risks and side effects associated with it. A better alternative is to turn to other therapeutic strategies that will improve symptoms sooner.
Sovereign Health of California treats individuals with mental illness, substance use disorders and dual diagnosis. We use multimodal diagnostic assessment and treatment to provide optimal long-term outcomes. Comprehensive treatment and ongoing aftercare provide the support patients need to recover from addiction and all of its consequences. To find out more about specialized programs at Sovereign Health, please call us at our 24/7 helpline.
About the author
Dana Connolly, Ph.D., is a senior staff writer for the Sovereign Health Group, where she translates current research into practical information. She earned her Ph.D. in research and theory development from New York University and has decades of experience in clinical care, medical research and health education. The Sovereign Health Group is a health information resource and Dr. Connolly helps to ensure excellence in our model. For more information and other inquiries about this article, contact the author at firstname.lastname@example.org.
Reach Out To Us Today! Most Private Insurance Accepted