Personalized treatment for depression can soon become a reality, thanks to an artificial intelligence (AI) algorithm that accurately predicts antidepressant efficacy in specific patients.
A historical study of more than 300 patients with major depressive disorder (MDD) showed that a latent space machine learning algorithm adapted for resting electroencephalography (EEG) firmly predicted the patient's response to sertraline. The findings were generalizable in different study sites and EEG teams.
"We found that the use of the artificial intelligence algorithm can identify the EEG signature for patients who get good results with sertraline," study researcher Madhukar Trivedi, MD, a professor of psychiatry at the Southwestern Medical Center of the University, told Medscape Medical News. University of Texas at Dallas. .
"Interestingly, when we look for more, it became clear that patients with the same EEG signature do not get good results with placebo," he added.
The study was published online on February 10 in Nature Biotechnology.
Currently, major depression is defined using a variety of clinical criteria. As such, it encompasses a heterogeneous mixture of neurobiological phenotypes. Such heterogeneity may explain the modest superiority of antidepressant medication in relation to placebo.
While recent research suggests that resting EEG can help identify the predictive heterogeneity of treatment in depression, these studies have also been hampered by the lack of cross-validation and small sample sizes.
In addition, these studies have identified nonspecific predictors or have failed to generate generalizable neural signatures that are predictive at the level of each patient.
For these reasons, there is currently no solid neurobiological signature for an antidepressant-sensitive phenotype that can help identify which patients will benefit from antidepressant medications. However, Trivedi said, detailing such a firm would promote a neurobiological understanding of the response to treatment, with the potential for significant clinical implications.
"The idea behind this NIH-funded study was to develop biomarkers that can distinguish the results of the treatment between the drug and the placebo," he said. "To do so, we needed a randomized, placebo-controlled trial that had significant breadth in terms of biomarker evaluation and validation, and this study was specifically designed for this purpose in mind.
"There has not been a drug and placebo study that has analyzed this in patients with depression. So, in that sense, this was really a fundamental study," he explained.
To help address these challenges, the researchers developed a machine learning algorithm that they called Latent Space Regression of Scattered EEG (SELSER).
Using data from four separate studies, they first established the predictive signature of EEG at rest by training SELSER in data from 309 patients in the study Establishing moderators and bio-signatures of the antidepressant response in clinical care (EMBARC), a placebo-controlled and placebo-controlled study of neuroimaging, randomized clinical study of antidepressant efficacy.
The generalization of the antidepressant predictive firm was tested in a second independent sample of 72 depressed patients.
In a third independent sample of 24 depressed patients, the researchers assessed the convergent validity and neurobiological importance of the EEG signature predictive of treatment and at rest.
Finally, a fourth sample of 152 depressed patients was used to assess the generalizability of the results.
"Fantastic" result but validation is needed
These combined efforts were aimed at revealing a treatment-sensitive phenotype in depression, decoupling between medication and placebo response, establishing its mechanistic importance and providing initial evidence on the potential for treatment selection based on an EEG signature. in a state of rest.
The study showed that the improvement in patients' symptoms was strongly predicted by the algorithm. These predictions were specific for sertraline in relation to placebo.
When two samples of depression were generalized, the researchers also found that the algorithm reflected the ability to respond to general antidepressant medication and was differentially related to a result of repetitive transcranial magnetic stimulation (EMT) treatment.
"Although we only looked at sertraline," said Trivedi, "we also applied the signature to a sample of patients who had been treated with transcranial magnetic stimulation. And we found that the signature for TMS (response) is different from the signature for sertraline."
Interestingly, the antidepressant predictive signature identified by SELSER was also superior to that of conventional machine learning models or latent modeling methods, such as independent component analysis or principal component analysis.
This SELSER firm was also superior to a model trained only with clinical data, and was able to predict the result using resting EEG data acquired at a study site not included in the model training set.
The study also revealed evidence of convergent multimodal validity for the signature of antidepressant response by virtue of its correlation with the expression of a functional magnetic resonance signature based on tasks in one of the four data sets.
The signature strength of the resting state was also correlated with the prefrontal neuronal response capacity, as indicated by direct stimulation with single pulse TMS and EEG.
Given the ability of the algorithm to predict the outcome with sertraline and distinguish the response between sertraline and placebo at the individual patient level, researchers believe that SELSER may one day support personalized approaches to machine learning for the treatment of depression.
"Our findings advance the neurobiological understanding of antidepressant treatment through a computational model adapted to the EEG and provide a clinical pathway for the personalized treatment of depression," the authors write.
However, his work is far from over. Among the researchers' next steps is the development of an AI interface that can be widely integrated with EEGs across the country.
"Identifying this signature was fantastic, but you should also be able to validate it," said Trivedi. "And fortunately we were able to validate it in the three additional studies.
"The next question is whether it can be extended to other diseases."
Commenting on the findings of Medscape Medical News, Michele Ferrante, PhD, believes that soon there will be a time when algorithms like this are used to customize the treatment of depression.
"It is well known that there is no good biological evidence in psychiatry, but the promising computational tools, biomarkers and behavioral signatures to segregate patients according to the response to treatment are beginning to emerge for depression," said Ferrante, program head of Theory and Computing. Neuroscience Program at the National Institute of Mental Health in Bethesda, Maryland.
"I have no doubt that the accuracy in the ability to predict which patient will respond to each treatment will improve over time," added Ferrante, who was not involved in the current study.
However, he said, such approaches are not exempt from their possible drawbacks.
"The biggest challenge is to continually validate these computational tools as they continue to learn from more heterogeneous groups. Another challenge will be to ensure that these computational tools are well established, widely adopted, safe and regulated by the FDA as Software as a medical device," said.
The current algorithm should also undergo further tests, Ferrante said.
"It has been validated in an external data set," he said, "but now we need to conduct rigorous prospective clinical trials in which AI selectively assign patients to a treatment according to their biological signature, to see if these results are true. ".
"In the future, it would be important to implement computational models (that are) capable of assigning patients through the multiple treatments available for depression, including pharmaceuticals, psychosocial interventions and neuronal devices."
The study was directly and indirectly funded by the National Institute of Mental Health of the National Institutes of Health, the Stanford Neurosciences Institute, the Hersh Foundation, the National Key Research and Development Plan of China and the National Natural Science Foundation of China.
Trivedi disclosed financial relationships with (lifetime disclosure) Abbott Laboratories, Inc; Abdi Ibrahim; Akzo (Organon Pharmaceuticals Inc); Alkermes AstraZeneca; Axon Advisors; Bristol-Myers Squibb; Cephalon, Inc .; Brain CME Institute of Physicians; Concert Pharmaceuticals, Inc; Eli Lilly & Company; Evotec; Fabre-Kramer Pharmaceuticals, Inc; Forest pharmaceutical products; GlaxoSmithKline; Janssen Global Services, LLC; Janssen Pharmaceutica Products, LP; Johnson & Johnson PRD; Libby Lundbeck Meade Johnson; MedAvante; Medtronic; Merck Mitsubishi Tanabe Pharma Development America, Inc; Naurex Neuronetica; Otsuka Pharmaceuticals; Pamlab Parke-Davis Pharmaceuticals, Inc .; Pfizer Inc; PgxHealth; Phoenix Marketing Solutions; Rexahn Pharmaceuticals; Crest diagnosis; Roche Products Ltd; Sepracor; SHIRE Development; Mountain range; SK Life and Science; Sunovion; Takeda Tal Medical / Puretech Venture; Targacept; Transcript; VantagePoint; Vivus and Wyeth-Ayerst Laboratories. He has received grants / research support from the Agency for Healthcare Research and Quality; Cyberonics, Inc; National Alliance for Research in Schizophrenia and Depression; National Institute of Mental Health; and the National Institute on Drug Abuse.
Nat Biotechnol. Published online February 10, 2020. Summary
For more news from Medscape Psychiatry, join us on Facebook and Twitter
. (tagsToTranslate) Artificial intelligence (t) AI (t) AI (t) selective serotonin reuptake inhibitor (ssri) (t) selective serotonin reuptake inhibitor (t) ssri (t) inhibitor (selective serotonin reuptake) (t) functional magnetic resonance imaging (t) t) functional magnetic resonance imaging (t) fmri (t) major depression (t) major depressive disorder (t) unipolar depression (t) neurobiology (t) antidepressant (t) antidepressant drug (t) drug (antidepressant) (t) electroencephalogram (t) eeg (t) depression