AI-Derived Electronic Tumor Marker (e19-9) Can Measure Treatment Response and Outcomes in CA19-9 Non-Producers with Pancreatic Ductal Adenocarcinoma

Background Clinicians use Cancer Antigen 19 − 9 (CA19-9) to assess treatment response and inform clinical decisions for patients with pancreatic ductal adenocarcinoma (PDAC). However, nearly 30% of patients with PDAC do not have an elevated CA19-9. Methods Using electronic health data, an artificial intelligence-based electronic tumor marker (e19-9) was created using common serum laboratory values to predict the expected CA19-9 level. The association between e19-9 and clinical outcomes was then measured in patients with PDAC who did not have an elevated CA19-9. Results The value of e19-9 was informative for predicting both completion of treatment and metastatic progression. Post-treatment e19-9 was independently associated with overall survival. A decline in e19-9 of at least 50% over treatment (AUC 0.79), and a post-treatment e19-9 of < 100 (AUC 0.84), mirror the clinical utility of CA19-9 for predicting the same outcomes. Conclusions The e19-9 correlates with important clinical outcomes among patients that do not have an elevated CA19-9 biomarker and has potential to guide clinical decisions.