Development of a prediction system for KRAS mutation detection via circulating tumour DNA analysis among patients with unresectable pancreatic cancer

Background Pancreatic cancer (PC) is an aggressive malignancy with a poor prognosis and limited treatment options. Advances in molecular profiling and liquid biopsy, specifically the detection of circulating tumour DNA (ctDNA), offer new avenues for personalized therapy. KRAS mutations are present in approximately 63–70% of PC patients, but detection via liquid biopsy can be influenced by disease stage, metastasis site, and ctDNA concentration. The aim of this retrospective study was to develop a prediction model for KRAS mutation detection in unresectable PC patients using clinical variables. Methods We analysed 32 patients who underwent ctDNA testing from 2019 to 2024, utilizing FoundationOne® Liquid CDx and Guardant360® CDx panels. Multivariate analysis of the clinical factors was performed via logistic regression with stepwise selection to elucidate independent predictors of KRAS mutation detection. According to the independent predictors of successful KRAS detection, a nomogram was plotted Results Multivariate analysis revealed that liver metastasis, multiple metastatic sites, and disease progression were significant predictors of successful KRAS mutation detection. A nomogram and ROC curve demonstrated high predictive accuracy, with a sensitivity of 70%, specificity of 90.9%, and AUC of 0.83. Conclusions Our prediction system effectively stratified patients by the likelihood of KRAS mutation detection, offering a practical tool for selecting candidates for liquid biopsy. These findings underscore the importance of personalized approaches in PC management and suggest that patients without these key clinical factors may not benefit from ctDNA testing. Future studies should validate this model in larger cohorts.