AI15 The use of artificial intelligence as a medical device and teledermatology in the assessment of Merkel cell carcinoma: a National Health Service case series

Abstract This case series presents instances of detection of Merkel cell carcinoma (MCC) by both teledermatologists and a UK Conformity Assessed class IIa artificial intelligence as a medical device (AIaMD) in National Health Service (NHS) skin cancer pathways. MCC is a very rare type of skin neuroendocrine cancer. Incidence rates have risen markedly in recent years: the European age-standardized rate in England increased from 0.43 to 0.65 per 100 000 person-years between 2004 and 2018. MCC predominantly affects older adults, with a median diagnosis age of 81 years, and is more common in men than women (ratio 1.4 : 1). MCC is associated with the worst prognosis of all skin cancer subtypes, making timely diagnosis and treatment essential. However, only 40% of cases of MCC in 2018–19 were appropriately referred on the Urgent Suspected Cancer (USC) pathway in England. For this study, data were prospectively collected from consecutive patients presenting to 31 NHS pathways across England with suspicious skin lesions between April 2020 and October 2024. All lesions with a histology-confirmed diagnosis of MCC were included. Demographic data, clinical history and lesion characteristics were recorded for all patients. In total 124 626 patients with 187 441 lesions were assessed via AI-enabled skin cancer pathways over the study period. Overall, 12 patients were diagnosed with MCC (one patient had two primary MCC lesions), suggesting a population prevalence of 0.01% among patients presenting to skin cancer pathways. Of 13 lesions, six met the appropriate predetermined inclusion criteria for AIaMD assessment: all six were appropriately classified by AIaMD and routed to the USC pathway. The remaining seven MCC lesions were not suitable for AIaMD assessment due to either their size or the presence of ulceration. These seven lesions underwent conventional teledermatology evaluation and all seven were appropriately kept on the USC pathway by teledermatologists. The median age of patients diagnosed was 86 years (interquartile range 79–93) and 50% were male. Of these MCC lesions, six were located on the head, three on the arms, three on the legs and one on the shoulder. Based on over 124 500 patient assessments over a 4.5-year period, this case series of 13 MCC lesions suggests that both AIaMD and teledermatology review have high sensitivity for the detection of MCC.