Purpose: Due to significant growth in the volume of information produced by cancer research, staying abreast of recent developments has become a challenging task. Artificial intelligence (AI) can learn, reason, and understand the enormous corpus of literature available to the scientific community. However, large-scale studies comparing the recommendations of AI and a multidisciplinary team board (MTB) in gastric cancer treatment have rarely been performed. Therefore, a retrospective real-world study was conducted to assess the level of concordance between AI and MTB treatment recommendations.
Methods: Treatment recommendations of Watson for Oncology (WFO) and an MTB were retrospectively analyzed 322 patients with gastric cancer from January 2015 to December 2018 and the degree of agreement between them was compared. The patients were divided into concordance and non-concordance groups and factors affecting the concordance rate were analyzed.
Results: The concordance rate between the AI and MTB was 86.96%. The concordance rates for each stage were 96.93% for stage I, 88.89% for stages II, 90.91% for stage III, and 45.83% for stage IV, respectively. In the multivariate analysis, age (p-value = 0.000), performance status (p-value = 0.003 for performance score 1; p-value = 0.007 for performance score 2; p-value = 0.000 for performance score 3), and stage IV (p-value = 0.017) had a significant effect on concordance between the MTB and WFO.
Conclusion: Factors affecting the concordance rate were age, performance status, and stage IV gastric cancer. To increase the validity of future medical AI systems for gastric cancer treatment, their supplementation with local guidelines and the ability to comprehensively understand individual patients is essential.
Keywords: Artificial intelligence; Concordance rate; Gastric cancer; Multidisciplinary team board; Watson for Oncology.
© 2023. The Author(s).