Research Information System
UHOD-ULUSLARARASI HEMATOLOJI-ONKOLOJI DERGISI, vol.35, no.1, pp.68-74, 2025 (SCI-Expanded, Scopus, TRDizin)
The role of artificial intelligence (AI) in oncology decision-making is rapidly expanding, yet its concordance with clinician recommendations and established guidelines remains unclear. This study evaluates the agreement between AI-generated adjuvant therapy recommendations for stage II colon cancer, clinician decisions, and international guidelines (NCCN, ESMO). We conducted a retrospective comparative analysis of 197 stage II colon cancer patients treated at Hacettepe University between 2014 and 2023. AI-generated recommendations (ChatGPT-4o) were compared with clinician decisions and NCCN/ESMO guidelines. Concordance rates were analyzed using Cohen's kappa and McNemar's test. Clinician adherence was highest with NCCN (89.8%) and ESMO (84.8%) guidelines. AI recommendations showed moderate agreement with clinician decisions (65.0%, kappa= 0.47). Statistically significant differences were observed between AI and clinical practice (p< 0.001), suggesting AI's more conservative approach. While AI demonstrates potential as a clinical decision-support tool, its moderate alignment with real-world decisions highlights the need for further refinement. Future improvements in AI interpretability, real-world validation, and clinician-AI collaboration are essential for its effective integration into oncology practice.