Key Takeaways
• Onc.AI’s Serial CTRS AI model outperformed traditional imaging benchmarks in predicting 12-month overall survival in NSCLC patients.
• The AI model was externally validated by GSK in a blinded retrospective analysis using data from the GARNET Phase I clinical trial.
• Serial CTRS uses standard CT imaging with no manual annotations and integrates seamlessly into existing oncology workflows.
• The technology is backed by an FDA breakthrough designation and broad real-world data from thousands of patients across global cancer centers.
Onc.AI, a digital health innovator specializing in artificial intelligence-driven oncology tools, presented new clinical findings at the 2025 American Association for Cancer Research (AACR) Annual Meeting that further solidify its flagship product, Serial CTRS, as a frontrunner in cancer prognosis technology.
The model was externally validated in collaboration with GSK, using patient data from the GARNET Phase I trial (NCT02715284) focused on non-small cell lung cancer (NSCLC).
Serial CTRS demonstrated significant improvements over traditional imaging methods in predicting long-term survival outcomes, emphasizing its potential to reshape early treatment response assessments in oncology.
Robust Validation in a High-Stakes Setting
Onc.AI’s Serial CTRS model was tested in Cohort E of the GARNET study, which included NSCLC patients treated with dostarlimab, GSK’s anti-PD-1 checkpoint inhibitor.
The analysis was performed by GSK in a blinded, independent, retrospective evaluation—an important measure to eliminate bias and ensure objectivity.
• Serial CTRS showed a hazard ratio (HR) of 2.91 (95% CI: 1.16–7.31) for distinguishing between patients with intermediate and high 12-month overall survival probabilities
• RECIST 1.1, the standard industry benchmark, had a HR of 1.34 (95% CI: 0.57–3.13)
• Tumor volume change assessments yielded a HR of 1.00 (95% CI: 0.43–2.34)
These results suggest that Serial CTRS can more accurately stratify patient risk, potentially improving trial efficiency and clinical decision-making.
Automation and Accessibility: A Seamless Clinical Fit
What sets Serial CTRS apart from conventional methods like RECIST 1.1 is its ability to analyze standard CT imaging without requiring manual annotations. This design allows it to be integrated into existing oncology workflows with minimal friction.
The model also maintained statistical significance after accounting for known prognostic variables such as:
-
Patient age
-
Baseline tumor volume
-
PD-L1 Tumor Proportion Score (TPS)
Such resilience further affirms the model’s value across diverse patient populations.
Industry Perspectives and Endorsements
“This important milestone builds on Serial CTRS’s FDA breakthrough designation. Our validation success, achieved with GSK, reflects the model’s strength derived from harmonizing diverse imaging data and a vast, heterogenous dataset spanning thousands of patients from hundreds of clinics across the U.S. and abroad.”— Akshay Nanduri, CEO of Onc.AI
“Onc.AI’s Serial CTRS holds transformative promise for pharmaceutical clinical development from Phase I through Phase III studies.”— Dr. George R. Simon, Vice President of Oncology, OhioHealth
These statements reflect broad confidence in the AI model’s utility from both industry leadership and clinical experts.
The Broader Impact: Redefining Trial Endpoints
In oncology drug development, endpoints like overall survival (OS) and progression-free survival (PFS) are often delayed and resource-intensive to measure. Serial CTRS offers a validated, non-invasive alternative that can deliver earlier insights.
• Enables faster decision-making in Phase I/II oncology trials
• Reduces dependency on long-term survival data
• Supports more efficient patient stratification and resource allocation
Its compatibility with widely used imaging techniques makes the model ideal for large-scale, multicenter studies, especially in global settings where annotation standards and workflows vary.
Powered by Diverse, Real-World Data
The reliability of any AI model depends heavily on the quality and diversity of its training data. According to Onc.AI, Serial CTRS was developed using:
• Imaging data from thousands of patients
• Sourced from dozens of healthcare systems
• Covering hundreds of clinics across the U.S. and international cancer centers
This breadth ensures the model’s applicability across different care environments, improving its chances of widespread clinical adoption and regulatory acceptance.
For more news and insights, visit AI News on our website.
