Real-world analysis of treatment sequencing and late-stage liquid biopsy genomic profiling in advanced pancreatic cancer

Background Real-world evidence on treatment sequencing and the molecular features of terminal-stage pancreatic cancer is limited. We evaluated treatment sequences and late-stage liquid biopsy genomic profiles using a nationwide clinico-genomic database. Methods We retrospectively analyzed 16,782 patients with pancreatic ductal adenocarcinoma registered in the Center for Cancer Genomics and Advanced Therapeutics database in Japan. Overall survival was compared according to first-line regimen and first- to second-line sequence. Liquid biopsy data were analyzed in relation to the time before death, and a gradient-boosting survival model was developed to identify prognostic genomic features. Results As first-line therapy, gemcitabine plus nab-paclitaxel was associated with longer overall survival than FOLFIRINOX (hazard ratio 0.89, 95% confidence interval 0.84–0.93), despite comparable time to treatment failure. Gemcitabine plus nab-paclitaxel followed by FOLFIRINOX was associated with the most favorable survival, measured from either first-line (hazard ratio 0.84) or second-line initiation (hazard ratio 0.68). Liquid biopsies were performed closer to death than tissue-based tests (median 140 vs. 473 days). The KRAS variant allele frequency increased as death approached and was the most influential prognostic feature; a KRAS -only model performed comparably to the full model (concordance index 0.67 for both), and survival worsened with increasing KRAS variant allele frequency. Alterations in chromatin-remodeling and epigenetic regulatory genes, including SMARCA4 , ARID1A , KDM6A , and KMT2D , were associated with poor prognosis. Conclusion Treatment sequencing may influence survival in advanced pancreatic cancer, and the KRAS variant allele frequency in late-stage liquid biopsy is a robust prognostic biomarker.