Determining the CA19-9 concentration that best predicts the presence of CT-occult unresectable features in patients with pancreatic cancer : a population-based analysis

Background: Serum CA19-9 concentration may be useful in triaging patients with pancreatic cancer for more intensive staging investigations. Our aim was to identify the CA19-9 cut-point with the greatest accuracy for detecting unresectable features not identified by CT scan, and to examine the performance of this and other cut-points in predicting the outcome of staging laparoscopy (SL). Methods: Patients with pancreatic cancer were drawn from two state-wide cancer registries between 2009 and 2011. We used classification and regression tree (CART) analysis to identify the CA19-9 cut-point which best predicted the presence of imaging-occult unresectable features, and compared its performance with that of a number of alternative cut-points. We then used logistic regression to test the association between CA19-9 concentration and detection of unresectable features in patients who underwent SL. Results: From the CART analysis, the optimal CA19-9 cut-point was 440 U/mL. CA19-9 ≥ 150 U/mL had a similar Youden Index, but greater sensitivity (69% versus 47%). This remained true for those who had obstructive jaundice at the time of CA19-9 sampling. CA19-9 concentration greater than or equal to 110 U/mL, 150 U/mL and 200 U/mL was associated with significantly greater odds of unresectable features being detected during SL. Conclusion: Elevated serum CA19-9 concentration is a valid marker for CT-occult unresectable features; the most clinically appropriate cut-point appears to be ≥ 150 U/mL irrespective of the presence of jaundice. Clinical trials which evaluate the value of CA19-9 in the staging algorithm for pancreatic cancer are needed before it is routinely used in clinical practice.