Validity of carcinoembryonic antigen and carbohydrate antigen 19-9 measurements in pancreatic cyst fluid with a serum-based immunoassay.

Imaging techniques, cytology, and biochemical analysis [including carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA 19-9) concentrations] of pancreatic cyst fluid obtained by endoscopic ultrasound-guided fineneedle aspiration are used to differentiate pancreatic cyst lesions, particularly those with malignant potential (1, 2 ). American College of Gastroenterology guidelines have outlined the published data relating to the diagnostic performance of pancreatic cyst fluid CEA and CA 19-9 (2 ) and concluded that cyst fluid CEA is the single most important factor in determining pancreatic cyst etiology. It is important, however, to appreciate that diagnostic cutoff values may be assay dependent. Many published reports do not describe the tumor marker assays used for analyzing pancreatic cyst fluid. In addition, the analytical validity of applying methods for serum tumor markers to the analysis of pancreatic cyst fluid does not appear to have been investigated in these studies. We investigated the analytical validity of the use of Roche E170 immunoassays for the measurement of CEA and CA 19-9 in pancreatic cyst fluid. We investigated pancreatic cyst fluid samples (n 21) obtained by ultrasound-guided fineneedle aspiration during routine investigation. Samples were stored at 20 °C and were thawed and vortex-mixed before analysis. CEA and CA 19-9 were measured with a Roche Modular Analytics E170 instrument. CEA and CA 19-9 were measured in a series of dilutions of each sample, which were performed manually with Elecsys Diluent Universal (Roche Diagnostics). The limits of quantification are 1–1000 g/L for the CEA assay and 2–1000 kU/L for the CA 19-9 assay. The in-house CV data (from serum analysis) are 4.4% at 3.21 g/L CEA, 3.0% at 37.1 g/L CEA, 7.5% at 12 kU/L CA 19-9, 2.8% at 20 kU/L CA 19-9, and 2.7% at 89 kU/L CA 19-9. In the case of CEA, 7 samples exhibited nonlinearity upon dilution (Table 1), with a progressive increase in the apparent CEA concentration as the dilution factor increased (3-fold in the most extreme case). In the case of CA 19-9, 7 samples exhibited nonlinearity upon dilution (Table 1), with an increase in the apparent CA 19-9 concentration as the dilution factor increased (60-fold in the most extreme case). In another sample (sample Q, Table 1), the CA 19-9 concentration was 452 kU/L when it was measured undiluted, but upon dilution the concentration increased to above the upper limit of detection (1000 kU/L). At a dilution of 1 volume in 100 000, the apparent CA 19-9 concentration was 1 160 000 kU/L. Of the 21 samples of pancreatic cyst fluid examined, 9 exhibited marked nonlinearity in the measurement of CEA and/or CA 19-9 upon dilution. There are a number of possible explanations for this finding. First, there may be differences in the antigen and immunoreactivity within cyst fluid compared with those of calibrators or serum. Although the CA 19-9 antigen in serum is associated with a mucin (3 ), it was originally discovered in gastrointestinal tumor cells as a ganglioside (4 ). The form present in cyst fluid is not known. CEA is known to display considerable heterogeneity, particularly in the extent and type of glycosylation (5 ). Second, matrix effects associated with cyst fluid, such as viscosity and the presence of nonvisible cellular debris, may contribute to the nonlinearity observed. None of the examined samples appeared viscous during manual dilutions, however, and there was no relationship between nonlinearity and the presence/absence of mucin in cytologic examinations. Third, the high-dose hook effect can lead to inaccurate results. The anomalies in CA 19-9 concentration noted upon dilution of sample Q are clearly due to a hook effect. In summary, this investigation has raised concerns about the performance of the Roche Elecsys assays for CEA and CA 19-9 in samples of pancreatic cyst fluid. Deviations from linearity on dilution are particularly marked for CA 19-9 measurement in certain samples. Other immunoassays for CEA and CA 19-9 may also be affected. These findings could have implications for the use of tumor marker analysis of cyst fluid in the investigation of pancreatic cysts. Caution should be exercised when measuring CEA and CA 19-9 concentrations in pancreatic cyst fluid with assays designed for the analysis of serum or plasma. Nonserum samples tested with such assays should always be subjected to additional quality-assessment measures, such as serial dilutions and spike recovery experiments. Further analytical validation is needed for assays used to measure CEA and CA 19-9 in pancreatic cyst fluid. The type of assay used and any investigation into its validity should be stated in future published studies that involve the measurement of these analytes in pancreatic cyst fluid.