AI Prognostic Tool for NSCLC Earns FDA Breakthrough Designation
The FDA granted breakthrough status to Serial CTRS, an AI tool that improves survival predictions in non–small cell lung cancer.
- The FDA has granted breakthrough device designation to Serial CTRS.
- Serial CTRS is an artificial intelligence (AI)–based prognostic tool.
- The goal of the tool is to stratify patients with non–small cell lung cancer (NSCLC) into high- or low-risk mortality categories.
The FDA has granted breakthrough device designation to Serial CTRS, an AI-powered prognostic tool designed to classify patients with NSCLC into high- and low-risk mortality categories.1
This tool aims to enhance risk stratification and optimize treatment decision-making, allowing for more personalized, precise oncology care.
Developed by Onc.AI, Serial CTRS utilizes a deep-learning model to analyze diagnostic imaging scans and predict long-term survival outcomes in patients undergoing treatment. The model, which is part of Onc.AI’s broader pipeline of AI imaging solutions, is designed to assist oncologists in making more informed therapeutic decisions.
“We are honored to be awarded breakthrough device designation for our Serial CTRS AI model,” Akshay Nanduri, chief executive officer of Onc.AI, said in a press release. “Onc.AI aims to equip oncologists with vital, automated prognostic insights using routinely collected diagnostic imaging scans and ultimately improve treatment strategy and provide risk stratification throughout [the] journey [for a patient with cancer].”
Microscopic image of non-small cell lung cancer - Generated with Google Gemini AI
The AI model was trained using real-world datasets from patients with advanced NSCLC who were treated with immune checkpoint inhibitors.2 Researchers developed a multi-step pipeline incorporating image quality control, preprocessing, deep-learning feature extraction, and survival modeling based on serial CT scans.
The model was validated using additional real-world data from patients, with hazard ratios for overall survival (OS) compared with manual tumor volume segmentations and RECIST 1.1 response categories. Serial CTRS and tumor volume changes were categorized as high, medium, or low response, reinforcing its predictive accuracy.
At the 2024 SITC Annual Meeting, data from a multi-institutional study showed that Serial CTRS outperforms traditional assessment tools such as RECIST and tumor volume measurements in predicting OS among patients with NSCLC receiving immunotherapy.2,3 Serial CTRS on CT scans at baseline and at 3 months of follow-up was found to accurately predict long-term outcomes after only a few cycles of treatment.
The C-index for predicting OS was improved with Serial CTRS at 0.734 vs 0.631 with RECIST and 0.679 with tumor volume measurement changes. Additionally, in patients with stable disease, Serial CTRS showed a 12-month area under the receiver operating characteristic (AUROC) curve of 0.74, which outperformed tumor volume change (AUROC 0.62).
“As longstanding partners of Onc.AI, we are thrilled to see the application of Flatiron’s high-quality, curated real-world data in the development and validation of regulatory-grade AI models for clinical use,” said Jacqueline Law, vice president of Corporate Strategy at Flatiron Health, in the press release.1 “We look forward to supporting Onc.AI’s efforts collaborating with the FDA and achieving additional milestones together.”
“As part of our ongoing data and clinical collaboration with Onc.AI, we are excited to be evaluating Serial CTRS. Having been involved in product definition and evaluating results throughout the evolution of this product, I look forward to seeing this breakthrough technology enter the clinic and impact early phase trials and clinical development,” added Dwight Owen, MD, MS, associate professor of medicine and head of Thoracic Oncology at The James Cancer Center at Ohio State University, in the press release.
REFERENCES
Onc.AI awarded FDA breakthrough device designation for Serial CT Response Score deep learning model. News release. Onc.AI. February 6, 2025. Accessed February 10, 2025. https://tinyurl.com/3a2m2hrd
Sako C, Schmidt TG, Patel AA, et al. Deep learning serial CT response score predicts overall survival in advanced NSCLC treated with PD-(L)1 immune checkpoint inhibitors. J Immunother Cancer. 2024;12(suppl 2):1237. doi:10.1136/jitc-2024-SITC2024.1237
Onc.AI Serial Imaging Response Score outperforms traditional methods for early assessment of NSCLC immunotherapy outcomes. News release. Onc.AI. November 6, 2024. Accessed February 10, 2025. https://tinyurl.com/3rmsfnr8
Kim Evaluates New Regimens for EGFR+ Lung Cancer
January 20th 2025During a Community Case Forum event in partnership with the Medical Oncology Association of Southern California, Edward S. Kim, MD, MBA, discussed the FLAURA2 and MARIPOSA trials of newer regimens for EGFR-positive lung cancer.
Read More
Roundtable Roundup: Lung Cancer Molecular Testing and ALK-Targeted Treatment
January 18th 2025In separate, live virtual events, Vincent K. Lam, MD, and Chul Kim, MD, MPH, discuss molecular assays and treatment options for a patient with metastatic non–small cell lung cancer (NSCLC), with participants.
Read More
