A new study led by researchers at Emory University’s Winship Cancer Institute and the University of Pennsylvania’s Abramson Cancer Center shows how a first-of-its-kind platform using artificial intelligence (AI) can help clinicians and patients better understand individual patient care. It has been demonstrated that it may be useful in assessing health conditions and their extent. You may benefit from certain treatments being tested in clinical trials. This AI platform can help you make informed treatment decisions, understand the potential benefits of new treatments, and plan future care.
The study, published in Nature Medicine, was published by Ravi B., medical director of data and technology application sharing resources, associate professor of hematology, and board-certified medical oncologist at Emory University’s Winship Cancer Institute. – Led by Parikh, MD, MPP. Emory University School of Medicine’s Department of Medical Oncology is developing and integrating AI applications to improve care for cancer patients. Dr. Qi Long, professor of biostatistics and computer and information science, founder of the University of Pennsylvania Cancer Data Science Center, and associate director of quantitative data science at the University of Pennsylvania Abramson Cancer Center, is a co-investigator. -Senior author. The study’s lead author was Dr. Xavier Orcutt, a trainee in Parikh’s lab. Other study authors include Kang Chen, a doctoral student training in Long’s lab, and Ronak Mamtani, associate professor of medicine at the University of Pennsylvania.
Parikh and his fellow researchers have developed TrialTranslator, a machine learning framework for “translating” clinical trial results to real-world populations. By emulating 11 groundbreaking cancer clinical trials using real-world data, we are able to reproduce the results of real-world clinical trials and identify patient groups that respond well to clinical trial treatments and It is now possible to identify groups of patients who do not
“We hope this AI platform will provide a framework to help doctors and patients decide whether clinical trial results are applicable to individual patients,” Parikh says. “Additionally, this study may help researchers identify subgroups for which new treatments are not effective, potentially facilitating new clinical trials for those high-risk groups.”
“Our research demonstrates the tremendous potential of leveraging AI/ML to harness the power of rich and complex real-world data to best advance precision medicine.” Long he adds.
Limited generalizability of study results
Professor Parikh explains that clinical trials of potential new treatments are limited because less than 10% of cancer patients participate in clinical trials. This means that clinical trials are often not representative of all patients with the cancer. Even if clinical trials show that a new treatment strategy produces better results than standard treatments, “there are many patients for whom the new treatment will not work,” Parikh says.
Want more breaking news?
Subscribe to Technology Networks’ daily newsletter to get the latest science news delivered straight to your inbox every day.
Subscribe for free
“This framework and our open source calculator will enable patients and physicians to determine whether the results of Phase III clinical trials apply to individual cancer patients,” he said, adding, “This study “provides a platform for analyzing real-world generalizability.” Results of other randomized trials, including trials with negative results. ”
how did they conduct their analysis
Parikh et al. used Flatiron Health’s national database of electronic health records (EHRs) to review 11 landmark randomized controlled trials that investigated cancer drug regimens considered standard of care. studies that compared the effectiveness of different treatments). The four most prevalent advanced solid malignancies in the United States: advanced non-small cell lung cancer, metastatic breast cancer, metastatic prostate cancer, and metastatic colorectal cancer.
what they found
Their analysis found that patients with low- and intermediate-risk phenotypes, machine learning-based traits used to assess a patient’s underlying prognosis, were similar to those observed in the randomized control group. found that there was a survival benefit associated with treatment. ordeal. In contrast, patients with high-risk phenotypes showed significantly lower survival and treatment-related survival benefit compared to randomized controlled trials.
Their findings suggest that machine learning can identify groups of real-world patients to whom the results of randomized controlled trials are difficult to generalize. This means that “the prognosis of real-world patients is likely to be more heterogeneous than that of participants in randomized controlled trials,” the researchers added.
Why this is important
The research team concluded that the study “suggests that patient prognosis is a better predictor of survival and treatment response than eligibility criteria.” They recommend that prospective clinical trials “should consider more sophisticated methods of assessing patient prognosis at the time of entry, rather than relying solely on strict eligibility criteria.”
Additionally, they cite recommendations from the American Society of Clinical Oncology and Friends of Cancer Research, stating that “high-risk subpopulations in randomized controlled trials should be “Efforts should be made to improve group representation.” ”
Regarding the role of AI in research like this, Parikh said: “With proper oversight and evidence, we will soon see a growing tide of AI-based biomarkers that can analyze pathology, radiology, or electronic medical record information to help make predictions.” They determine whether a patient will respond or not respond to a treatment, whether the cancer can be diagnosed earlier, or whether a patient’s prognosis will improve. ”
reference: Orcutt X, Chen K, Mamtani R, Long Q, Parikh RB. Assess the generalizability of oncology test results to real patients using machine learning-based test emulation. Nat Med. 2025.Doi: 10.1038/s41591-024-03352-5
This article has been reprinted from the following material: Note: Materials may be edited for length and content. Please contact the citation source for details. You can access our press release publishing policy here.
