AI-Enabled Platform Improves Cancer Prevention Access in Unaffected Individuals
An online artificial intelligence (AI)-enabled platform was found to significantly improve completion, accuracy, and uptake of cancer prevention among unaffected individuals compared with in-clinic delivery, with high concordance to expert judgment, based on the results of an Indian prospective randomized study.1 These findings, which were presented at the inaugural European Society for Medical Oncology (ESMO) AI & Digital Oncology Congress (Abstract 175MO), appear to support the feasibility, validity, and scalability of the platform in low- and middle-income settings.
As presenting author Debapriya Mondal, MBBS, MD, DrNB, of Apollo Cancer Centre, Kolkata, India, explained, AI is increasingly applied in cancer diagnostics and treatment, yet its potential contributions to cancer prevention and hereditary risk assessment remain underexplored. He noted that limited screening uptake and persistent psychological barriers impede effective cancer prevention efforts in India.
“Digital solutions, like our tool, can bridge the gap in preventative oncology by adding accessible, scalable, and validated pathways aimed at population-level cancer prevention,” Dr. Mondal commented.
Study Details
Between June 1 and August 16, 2025, at four tertiary care hospitals in India, unaffected blood relatives of patients with cancer were invited to participate in the study. A total of 500 individuals were randomly assigned in a 1:1 ratio to either a traditional in-clinic preventative oncologic consultation or an online self-assessment using the online AI-enabled platform, OncoDefend (oncodefend.com).
OncoDefend is a Generative Pre-trained Transformer (GPT)-5 mini–powered platform informed by recommendations from the National Comprehensive Cancer Network, U.S. Preventive Services Task Force, American Cancer Society, National Cancer Grid of India, and the World Health Organization. It was built using a generative Application Programming Interface (API) with a React.js frontend, Node.js backend, and Amazon Web Services (AWS) infrastructure.
In the input layer, users provide consent before their information—including demographics, lifestyle details, habits, family and personal medical history, prior screenings, and comorbidities—is de-identified and processed. In the processing layer, data are supplied—via optimized prompts—to a generative model that produces a personalized cancer risk summary, a risk index, lifestyle recommendations, and guideline-based testing and screening suggestions. The output layer produces a final report that is then reviewed by a human oncologist for a simple agree-or-disagree assessment.
The investigators identified the assessment completion rate as the primary endpoint. Secondary endpoints included the concordance between AI-generated reports and expert assessments (from eight oncologists and two geneticists), as well as the proportion of participants who undertook at least one cancer risk–reducing action (ie, lifestyle action, genetic testing, screening, vaccination).
Key Findings
Assessment completion was found to be significantly higher among participants who were randomized to the online self-assessment vs in-clinic consultation (76% vs 15%; χ2 = 186.4; P < .0001; Cramér’s V = 0.61). A total of 45.8% and 5.2% of patients in these arms, respectively, initiated at least one preventative action (χ2 = 66.6; P < .0001; Cramér’s V = 0.37).
The AI system recommended germline testing for 63 participants, and human geneticists validated 61 of those recommendations (96.8%). Among the participants who initiated at least one preventive action after completing the AI-powered self-assessment, a total of 100 screening procedures, 50 germline tests, and 47 vaccinations were performed. The concordance between the AI-generated reports and oncologist review was 98.4%.
Dr. Mondal concluded, “Use of the online AI-enabled platform for cancer prevention and genetic risk assessment was feasible and significantly increased completion as well as uptake of preventative action compared with traditional in-clinic assessment. Concordance between [AI]-generated recommendations and oncologist review was high, and that included recommendation for germline testing, which confirms its clinical validity.”
DISCLOSURE: Dr. Mondal reported no conflicts of interest.
ASCO AI in Oncology is published by Conexiant under a license arrangement with the American Society of Clinical Oncology, Inc. (ASCO®). The ideas and opinions expressed in ASCO AI in Oncology do not necessarily reflect those of Conexiant or ASCO. For more information, see Policies.
Performance of a convolutional neural network in determining differentiation levels of cutaneous squamous cell carcinomas was on par with that of experienced dermatologists, according to the results of a recent study published in JAAD International.
“This type of cancer, which is a result of mutations of the most common cell type in the top layer of the skin, is strongly linked to accumulated [ultraviolet] radiation over time. It develops in sun-exposed areas, often on skin already showing signs of sun damage, with rough scaly patches, uneven pigmentation, and decreased elasticity,” stated lead researcher Sam Polesie, MD, PhD, Associate Professor of Dermatology and Venereology at the University of Gothenburg and Practicing Dermatologist at Sahlgrenska University Hospital, both in Gothenburg, Sweden.