AI integrated into molecular profiling workflows identified misclassified lung cancer cases, highlighting its potential role in improving diagnostic accuracy in clinical labs.
A study published in JAMA Network Open by Caris Life Sciences suggests that artificial intelligence (AI)-driven molecular profiling could play a meaningful role in catching cancer misdiagnoses that traditional diagnostic methods miss—potentially affecting thousands of patients annually in the US.
The research examined 3,958 lung cancer cases submitted to Caris with an existing diagnosis of squamous cell carcinoma (SCC). The company’s proprietary GPSai algorithm identified 123 of those cases as metastases originating from other primary sites, including cutaneous, urothelial, head and neck, and thymic cancers.
Of those reclassified cases, 88 patients—representing 71.5%—had guideline-preferred first-line systemic therapy change recommendations, meaning clinicians were alerted that a treatment adjustment could lead to better patient outcomes, according to a release from the company.
Extrapolating to the Broader Population
The potential scale of the problem extends beyond the study cohort. According to the Centers for Disease Control and Prevention (CDC), lung SCC accounts for roughly 21% of all lung cancer cases in the US. Applying the study’s misdiagnosis rate to CDC data, the company estimates that approximately 1,000 lung cancer cases each year in the US may represent potential misdiagnoses.
The clinical consequences of such errors can be significant, according to a release from the company. Treatment recommendations and prognostic expectations differ substantially across cancer types, meaning patients receiving care based on an incorrect primary diagnosis may not be receiving the most appropriate therapy for their condition, according to the release.
GPSai’s Growing Diagnostic Record
The JAMA Network Open study is part of a broader body of evidence the company says demonstrates the value of embedding AI into routine molecular profiling workflows. Since January 2024, Caris GPSai has overturned 3,857 diagnoses across the spectrum of cancer, according to the company.
“Caris GPSai has overturned 3,857 diagnoses across the spectrum of cancer since January of 2024,” says Matthew Oberley, MD, PhD, senior vice president, chief clinical officer, and pathologist-in-chief at Caris, in a release. “By integrating AI-driven tissue-of-origin predictions with comprehensive molecular profiling and pathology, we can give clinicians greater diagnostic confidence and ensure patients receive the most appropriate care.”
Integration With FDA-Cleared Assay
The GPSai algorithm is included as part of Caris’ MI Cancer Seek and MI Tumor Seek assays at no additional cost, according to the company. In November 2024, the Food and Drug Administration (FDA) cleared MI Cancer Seek—a tissue-based assay the company describes as the first and only simultaneous whole exome sequencing and whole transcriptome sequencing-based assay with FDA-approved companion diagnostic indications for molecular profiling of solid tumors.
By embedding GPSai into those assays, the company says providers gain an additional layer of diagnostic insight during routine profiling, without requiring a separate test or additional cost to the ordering clinician.
ID 310347417 © Valiantsin Suprunovich | Dreamstime.com
