Dominic Kaddu-Mulindwa, Xianqing Mao, Caroline Duhem, Sigrid de Wilde, Stefan Rauh, Gilles Klein, Frank Jacob, Sophie Servais
Physicians working in primary care in Sweden consider GLP-1 RAs effective for treating obesity and believe these drugs will play an important role in obesity treatment in the future. These findings should be interpreted in the light of the low response rate and possible self-selection bias.
INTRODUCTION: Artificial intelligence (AI) is rapidly entering the field of medical oncology. Clinician literacy, ethical awareness, and governance structures are prerequisites for safe implementation, yet real-world data on adoption, training and ethical perceptions from small European healthcare systems remain scarce.
METHODS: We conducted an anonymous national cross-sectional online survey among all practicing physicians (medical oncologists, radiation oncologists, haematologist-oncologists) and oncology residents in Luxembourg (n = 42 eligible) between January and February 2026. The instrument covered AI knowledge, use, attitudes, ethical/legal perspectives, barriers and training needs. Proportions are reported with 95% Wilson score confidence intervals; analyses were descriptive.
RESULTS: In total 25 physicians responded (59.5%), 88% (95% CI 70.0-95.8) of whom had no formal AI training. All respondents had used large language models (LLMs), with 52% (33.5-70.0) reporting professional use for non-clinical tasks and 20% (8.9-39.1) for clinical decision support. Most participants (84%; 65.3-93.6) supported the use of AI as a clinical decision-support tool, yet 88% (70.0-95.8) of physicians perceived they would bear primarily legal responsibility for AI-related errors. The main barriers to implementation were lack of validated tools (68%; 48.4-82.8) and regulatory/legal uncertainty (64%; 44.5-79.8). Notably, 76% (56.6-88.5) of respondents reported encountering patients who brought AI-generated medical information to consultations.
DISCUSSION: AI adoption among oncologists in Luxembourg is already widespread, including emerging clinical use, despite limited formal training and unresolved medico-legal frameworks. This "implementation-governance gap" highlights the need for structured education, validated clinical tools, and regulatory clarity to ensure safe and ethically sound integration of AI into oncology practice.