AI Algorithm Demonstrates Ability to Predict the Efficacy of Sertraline in Patients With Depression

A new artificial intelligence (AI) tool is able to predict within a week whether an antidepressant will work in a patient with major depressive disorder (MDD), according to new research published in American Journal of Psychiatry. The study authors noted that with the help of the AI algorithms, they were able to determine whether or not the medication would work up to 8 weeks faster.¹

Image credit: zinkevych | stock.adobe.com

"This is important news for patients. Normally, it takes 6 to 8 weeks before it is known whether an antidepressant will work," said study author Liesbeth Reneman, MD, PhD, professor of neuroradiology at Amsterdam University Medical Center, in a press release.¹

The researchers analyzed whether the effect of one of the most prescribed antidepressants in the US and Europe, sertraline (Zoloft; Pfizer), could be predicted by the AI algorithm. In this double-blind, placebo-controlled, multisite, randomized trial, a total of 296 adult outpatients with unmedicated recurrent or chronic MDD. MR neuroimaging and clinical data were collected both prior to and after 1 week of treatment initiation. After 8 weeks, performance in predicting response and remission was collected and quantified using balanced accuracy (bAcc) and area under the receiver operating characteristic curve (AUROC) scores.^1,2

Of the 296 patients, 229 were included in the analysis, of which approximately 66% were female and the median age was 38 years. The results indicated that approximately 33% would respond to sertraline and the remaining patients would not. In addition, internal cross-validation performance when predicting the response to sertraline (bAcc = 68% [SD = 10], AUROC = .73 [SD = .03]) was notably stronger than chance. Further, external cross-validation on data from placebo non-responders (bAcc = 62%, AUROC = .66) and placebo non-responders who were switched to sertraline (bAcc = 65%, AUROC = .68) demonstrated differences which suggested a specificity for sertraline treatment compared with placebo treatment.^1,2

"With this method, we can already prevent [approximately 67%] of the number of 'erroneous' prescriptions of sertraline and thus offer better quality of care for the patient. Because the drug also has side effects," said Reneman in the press release.¹

Not only did the study results confirm that early sertraline treatment response can be predicted, but they also demonstrated that the AI models used are sertraline-specific compared with placebo and that the prediction of sertraline efficacy benefits from MRI data. Traditionally, patients are given an antidepressant and after 6 to 8 weeks—or sometimes several months—if the patient’s symptoms do not subside or improve, they are given another antidepressant. This process can be lengthy and continue until the patient finds an effective medication. According to the investigators, in 1 in 3 patients with MDD, there continues to be no improvement in symptoms even after multiple treatment steps, emphasizing an urgent need for a solution that allows for faster determination of antidepressants’ efficacy.^1,2

The study authors note that going forward, this method of prediction can better help personalize sertraline treatment to each individual patient while saving them time and cost on medications. However, there is currently no exact prediction tool. The study authors also state that future research will seek to improve the AI algorithm by providing it with additional information.^1,2

"The algorithm suggested that blood flow in the anterior cingulate cortex, the area of brain involved in emotion regulation, would be predictive of the efficacy of the drug. And at the second measurement, a week after the start, the severity of their symptoms turned out to be additionally predictive," said study author Eric Ruhé, psychiatrist at Radboudumc, in the press release.¹

References

1. Amsterdam University Medical Center. Artificial intelligence helps predict whether antidepressants will work in patients. News release. February 7, 2024. Accessed February 8, 2024. https://www.eurekalert.org/news-releases/1033410

2. Poirot, MG, Ruhé, HG, Mutsaerts, HMM, et al. Treatment Response Prediction in Major Depressive Disorder Using Multimodal MRI and Clinical Data: Secondary Analysis of a Randomized Clinical Trial. Am J Psychiatry. 2024:appi.ajp.20230206. doi:10.1176/appi.ajp.20230206