Comprehensive Summary
The following study is a prospective, multicentre real-world evaluation of the clinical performance of a UK Conformity Assessed (UKCA) class Ila artificial intelligence as a medical device (AlaMD) compared with teledermatology assessments for triaging high-risk skin cancers under the NHS Urgent Suspected Cancer (USC) pathway. Across 12 NHS sites, 25,788 lesions were prospectively assessed by both the AlaMD and 81 teledermatologists between December 2023 and October 2024. Histopathologic confirmation was available for 18,744 lesions (72.7%). The study included 264 melanomas, 356 squamous cell carcinomas (SCCs), and 10 rare skin cancers. The AlaMD achieved significantly higher triage sensitivity for all high-risk cancers (98.6%, 95% Cl 97.3-99.2) compared with teledermatologists (95.9%, 95% Cl 94.0-97.2; P=0.004). Similar trends were observed for melanomas (98.1% vs 94.7%; P=0.04) and SCCs (99.2% vs 96.6%; P=0.02).
Outcomes and Implications
These findings demonstrate that AI-driven triage tools can perform at least as well as, and in some cases exceed, the sensitivity of teledermatology reviews for detecting high-risk skin cancers within real-world NHS workflows. Significantly, false negatives from either method were frequently corrected by the complementary review, highlighting the value of combining AI and clinician expertise in parallel. The study provides robust evidence and support benchmarking AlaMD sensitivity targets above 95% for melanoma and SCC detection, reinforcing AI’s potential as a reliable triage adjunct within national skin cancer pathways. Clinically, integrating such an AI system could improve diagnostic throughput, reduce time to cancer treatment, and standardize quality across dermatologic networks, while maintaining patient safety. Future studies should focus on prospective outcome tracking and cost-effectiveness analyses to guide large-scale NHS implementation of AI-augmented skin cancer triage.