Novel System Outperforms Existing Prognostication Models for Squamous Cell Carcinoma
Artificial intelligence-derived risk score prognostication system demonstrated superior sensitivity across all outcomes
By Dermsquared Editorial Team | June 20, 2025
WEDNESDAY, June 18, 2025 (HealthDay News) -- A novel system outperforms current prognostication systems for predicting poor outcomes in cutaneous squamous cell carcinoma (cSCC), according to research published online June 11 in JAMA Dermatology.
Neil K. Jairath, M.D., from the New York University Grossman School of Medicine in New York City, and colleagues examined whether a customized generative pretrained transformer model could be used to develop a novel class-based risk stratification system that would outperform the current standard. A systematic review of the literature was conducted that assessed risk factors for poor outcomes in cSCC to build the retrieval augmented generation (RAG) knowledge base. A novel class-based risk stratification system was developed using the RAG-enabled generative pretrained transformer (GPT) model that assigned point values for risk factors, culminating in a GPT-based prognostication system called the artificial intelligence-derived risk score (AIRIS). Performance of the system was validated on a combined prospective and retrospective cohort of 2,379 primary cSCC tumors.
The researchers found the AIRIS prognostication system showed superior sensitivity across all outcomes (locoregional recurrence [LR], 49.1 percent; nodal metastasis [NM], 73.7 percent; distant metastasis [DM], 82.5 percent; and disease-specific death [DSD], 72.2 percent), and it had the highest area under the receiver operating characteristic curve values (0.69, 0.81, 0.85, and 0.80 for LR, NM, DM, and DSD, respectively); compared with the Brigham and Women's Hospital and American Joint Committee on Cancer Staging Manual, eighth edition, discriminative ability was significantly enhanced.
"Leveraging large language models in developing risk stratification systems represents a promising avenue for advancing patient care and delivering insights into trends not yet capitalized on," the authors write.
Two authors disclosed ties to the biopharmaceutical industry; one author holds a patent for methods and materials for assessing and treating cSCC.