ИИ-технологии выявляют подклинические базальноклеточные карциномы у пациентов из групп риска с точностью 94,4%

In patients at high risk for basal cell carcinomas (BCCs), AI-assisted line-field confocal optical coherence tomography (LC-OCT) was able to detect subclinical cancers at a high rate, according to a feasibility study published online in JAMA Dermatology. In the study, undertaken in a population with no clinical evidence of skin cancer, there was only one false positive, providing a positive predictive value of 94.4%, reported a multicenter team of German investigators led by Moritz Ronicke, MD, Department of Dermatology, Friedrich-Alexander University Erlangen-Nuremberg, Erlangen, Germany. “The findings suggest that systemic screening with AI-assisted LC-OCT may be a feasible approach for BCC detection at a preclinical clinical stage in populations at high risk for BCC,” Ronicke and his co-investigators concluded. This is a new capability, but the clinical value of detecting BCC at a subclinical level is still unclear. However, this is a first step for better characterizing the trajectory of BCC and studying earlier interventions, Ronicke told Medscape Medical News. Next Step in Early BCC Detection Is Justified “Our work justifies the next step, which is a prospective multicenter study following subclinical lesions at defined intervals, which would simultaneously establish the sensitivity of the approach and clarify the natural history,” he explained. In particular, prospective studies are expected to determine if “nodular BCCs arise from superficial precursors or whether some regress spontaneously,” Ronicke said. The 150 patients enrolled in the study had at least two risk factors for BCC, which included a personal history of BCC, a genetic syndrome associated with BCC, age 65 years or older, or chronic sun exposure (≥ 20 years occupational or ≥ 30 years leisure time exposure). More than 70% of those enrolled had a history of BCC. After excluding lesions suggestive of BCC, LC-OCT was used to screen inconspicuous facial skin — skin indistinguishable from normal. Integrated AI software was used to guide attention to regions of interest for potential subclinical BCC. If a lesion had a high probability score on repeated screens, it was evaluated by a physician using a systematic evaluation, including histology. The diagnosis of BCC was made independently of AI-assisted LC-OCT using criteria such as the presence of lobules with mille-feuille pattern or a bright rim surrounding the lobule. Subclinical BCCs Identified in 9.3% of Patients There were 17 subclinical BCCs identified in 14 (9.3%) patients. Most BCCs (76.5%) were superficial. Of the four remaining cases, three were nodular and one was infiltrative. The forehead was involved in six cases, the temple in five cases, the nose and cheeks in two cases each, and the jaw and periorbital area in one case each. In an accompanying editor’s note, Veronica Rotemberg, MD, PhD, who leads the Dermatology Imaging Informatics Group at Memorial Sloan Kettering Cancer Center in New York City, pointed out that this study identified “a previously uncharacterized category of disease: subclinical BCCs without visible correlates.” But she emphasized that clinical relevance is unknown. “Even in clinically visible BCCs, watchful waiting can be an appropriate strategy,” she wrote. Moreover, the study design did not permit any conclusions to be drawn about diagnostic performance. “Only lesions flagged by AI and judged suspicious by clinicians underwent biopsy, precluding calculation of sensitivity and specificity of the AI algorithm alone,” added Rotemberg, an associate editor at JAMA Dermatology. “Advances in detection,” she cautioned, “risk outpacing advances in clinical decision-making, raising the possibility of overdiagnosis and unnecessary intervention.” However, in an interview, she did not rule out the potential that very early BCC detection could lead to meaningful information. “Imaging modalities such as reflectance confocal microscopy have great potential to improve our understanding of biology. That would require high-quality observational studies, which are sorely needed in this space,” Rotemberg told Medscape Medical News. The fact that most of the BCCs identified in this study were superficial is potentially noteworthy, according to Andrew C. Krakowski, MD, chief of dermatology at St. Luke’s University Health Network in Easton, Pennsylvania. An author of a 2023 review paper on treating advanced BCC, he sees a great clinical potential if further studies prove the treatment of subclinical lesions improves outcomes. “Identifying basal cell carcinomas at the subclinical stage is a highly promising direction for clinical research because it shifts our treatment paradigm from surgical excision to noninvasive management,” Krakowski said in an interview. He pointed out that although most BCCs are curable with surgery, scarring is a risk. “Since this study found that the vast majority of these preclinical lesions are superficial, catching them early allows us to utilize tissue-sparing topical creams, cryotherapy, or light therapies, ultimately minimizing the cumulative surgical burden and aesthetic impairment for our high-risk patients,” said Krakowski, who was not involved with the study. Yet he, like the other experts, considers this study a first step toward a goal of showing that early detection and early treatment favorably affect clinical outcomes. Ronicke reported having financial relationships with AbbVie, Almirall, Fidia, Johnson & Johnson, Novartis, Pfizer, Recordati, UCB, and Urgo. Rotemberg reported having financial relationships with Atria Health and Research Institute, Inhabit, and Kaggle. Krakowski reported no potential conflicts of interest. Ted Bosworth, a career medical writer based in New York City, has been covering advances in clinical medicine, including dermatology, for several decades.