Preoperative CA 19-9 kinetics as a prognostic variable in radiographically resectable pancreatic adenocarcinoma

Research output: Contribution to journalArticlepeer-review

23 Scopus citations


Background and Objectives: Serial levels of CA 19-9 are correlated with treatment response and survival; however, little is known about CA 19-9 kinetics in the absence of therapy. We hypothesize that preoperative CA 19-9 kinetics predict rate of resectability as well as survival. Methods: Retrospective review of 72 patients with radiographically resectable pancreatic adenocarcinoma with two pre-operative CA 19-9 levels prior to planned pancreaticoduodenectomy. Primary outcome measures were resectability and overall survival. Results: Forty-seven out of 72 patients (65%) had resectable disease. Unresectable patients had higher absolute change in CA 19-9 than patients with resectable disease (97 U/ml vs. 34 U/ml) as well as higher rate of change (4 U/ml/day vs. 1 U/ml/day). Receiver operating characteristic curves identified predictive thresholds for absolute (≥50 U/ml) and rate of CA 19-9 change (≥1 U/ml/day) that accurately identified unresectable patients. Survival analysis revealed that a change in CA 19-9 <50 U/ml and a rate of change <1 U/ml/day predicted improved survival (P=0.04, P=0.02); however, for patients with resectable disease, CA 19-9 changes did not predict survival. Conclusions: Preoperative kinetics of CA 19-9 predict resectable disease for pancreatic cancer. These variables also predict overall survival; however, these do not predict survival for those with resectable disease.

Original languageEnglish (US)
Pages (from-to)293-298
Number of pages6
JournalJournal of Surgical Oncology
Issue number3
StatePublished - Mar 1 2015


  • CA19-9
  • Pancreas cancer
  • Tumor marker

ASJC Scopus subject areas

  • Surgery
  • Oncology


Dive into the research topics of 'Preoperative CA 19-9 kinetics as a prognostic variable in radiographically resectable pancreatic adenocarcinoma'. Together they form a unique fingerprint.

Cite this