Background: Although CA 19-9 is most often used in pancreatic cancer as a diagnostic adjunct, or to follow response to treatment, its preoperative value has been reported to correlate with survival and recurrence. A corrected-CA 19-9 (c-CA19-9), obtained by dividing the CA 19-9 by total bilirubin, has been reported to improve this correlation. Our aim is to evaluate the predictive and prognostic value of CA 19-9 in a large single-institutional experience.
Methods: A retrospective review of all patients undergoing pancreatoduodenectomy from July 2001 through June 2007 at our institution was conducted. Preoperative serum CA 19-9 and total bilirubin levels were analyzed with histologic and survival data.
Results: Of 328 patients identified, 231 (58% male; 42% female) with a mean age of 66 (37-90) had both pre-operative serum CA 19-9 and total bilirubin levels and comprised our study group. Median follow-up was 2.1 years. All patients underwent pancreaticoduodenectomy for histologically confirmed pancreatic adenocarcinoma. Using receiver operator curves, neither CA 19-9 nor c-CA 19-9 demonstrated predictive value for lymph node status (c=0.55/0.56) or margin status (c=0.50/0.46). Tumor size and lymph node ratio very weakly correlated with CA 19-9 and c-CA 19-9 levels (Spearman correlation coefficients for tumor size: 0.26 and 0.28; for lymph node ratio: 0.17 and 0.16, respectively). Survival was not different at 1 year (73% vs. 68%), 3-years (33% vs 22%), or 5-years (27% vs 16%) for patients with CA 19-9 <=300 compared to those with values >300. Using corrected c-CA 19-9 with a cut-off of 50 also failed to demonstrate a significant difference in survival. Even at cutoffs of 500 and 100 for CA 19-9 and c-CA-19-9 respectively, there was no difference in survival compared to patients with lower levels (p>0.2). Patients with a CA 19-9 >300 or c-CA19-9>50 were at 1.30 and 1.29 times risk of death than those with CA 19-9 <300 or c-CA 19-9<50 respectively.
Conclusion: This large, single institution study demonstrates no histologic (lymph node or margin status) predictive value or prognostic value of CA 19-9 or c-CA19-9 in patients undergoing resection for pancreatic adenocarcinoma. These findings are in contrast to smaller previous studies that have suggested such a correlation. Our findings do not support broadening the use of CA 19-9 beyond aiding diagnosis and following therapeutic response.