Can preoperative CA19-9 and CEA levels predict the resectability of patients with pancreatic adenocarcinoma?

Authors


Professor Hong Joo Kim, Department of Internal Medicine, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine, 108, Pyung-Dong, Jongro-Ku Seoul, Korea. Email: hongjoo3.kim@samsung.com

Abstract

Background:  We aimed to explore the predictive ability of preoperative carbohydrate antigen 19-9 (CA19-9) and carcinoembryonic antigen (CEA) levels for assessing tumor resectability (R0 resection) in patients with pancreatic adenocarcinoma.

Methods:  The present study included 72 patients who had been treated surgically for potentially resectable pancreatic adenocarcinoma and 42 patients who had been treated surgically for palliation (bypass surgery) at our institution. Pancreatic adenocarcinoma was histologically confirmed by pathological examination of the resected specimen or, if unresected, by intraoperative biopsy.

Results:  For resectable disease, the mean and median values of CA19-9 were significantly lower than for R1/2 or unresectable disease. The best cut-off points for CEA, CA19-9, and tumor size to predict resectability were 2.47 ng/mL, 92.77 U/mL and 11.85 cm3, respectively. A CA19-9 ≥ 92.77 U/mL and both tumor markers no less than the cut-off levels predicted the possibility of R1/2 or unresectability with 90.6% and 88.6% accuracy, respectively. However, either tumor marker or both tumor markers less than the cut-off levels predicted the probability of R0 resection only with 27.1% and 40.6% accuracy, respectively. The independent contributing factors to resectability (R0 resection) by multivariate regression analysis were a CA 19-9 < 92.77 U/mL, a tumor size < 11.85 cm3, and a less advanced AJCC stage.

Conclusion:  The present study demonstrates that preoperative serum CA19-9 and CEA levels can be used for the prediction of resectability (R0 resection) in patients with pancreatic adenocarcinoma, which may enable a simple and cost-effective exclusion of such patients who are unlikely to benefit from surgery.

Introduction

Pancreatic adenocarcinoma is a devastating disease. Unfortunately, determining which patients have localized disease is not straightforward and, often, occult metastases are discovered during laparotomy. Hence, the curative resection of pancreatic adenocarcinoma can be carried out in only 10% of patients and resection margin-positive pancreatic tumors are associated with a poor prognosis.1–3

The only way to cure pancreatic adenocarcinoma is to remove the entire tumor with no residual disease (microscopic resection-margin negative).4 A preoperative assessment for the possibility of complete resection for patients with pancreatic adenocarcinoma is very important because precise estimation results in fewer unnecessary operations that do not afford survival benefit to the patients. Currently, the study of choice to stage pancreatic adenocarcinoma is computed tomography (CT). The accuracy of thin-cut, bolus-contrast, triple phase helical CT in predicting inoperability approaches 100%; however, the determination of resectability is only 75% to 80%.5–8 Therefore, approximately one-fourth of patients will have unresectable tumors discovered at the time of surgery and will have endured an unnecessary laparotomy. This issue is important because recovery from the unnecessary laparotomy further delays palliative systemic therapy.

The two most studied tumor markers that have been evaluated in the diagnosis and prognosis of patients with pancreatic adenocarcinoma are carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9).9–15 Little is known, however, about the association between the levels of these tumor markers and the existence of metastasis or locally advanced disease in patients with pancreatic adenocarcinoma.

The purpose of the present study was to determine the utility of these tumor markers for assessing the resectability (microscopic resection margin-negative) for patients with pancreatic adenocarcinoma.

Methods

This was a retrospective study involving 72 patients who had been treated surgically for potentially resectable primary pancreatic adenocarcinoma and 42 patients who had been treated with surgical palliation (bypass surgery) between June 2001 and June 2007 at an academic tertiary care referral center. Our study was approved by the Institutional Ethical Committee of Kangbuk Samsung Hospital in Seoul, Korea. All the patients' tumors were evaluated by preoperative bolus-contrast CT scan and were judged to be inoperable if the tumor showed extension to the superior mesenteric artery or celiac axis, and if there was evidence of distant metastasis on imaging.

Pancreatic adenocarcinoma was histologically confirmed by pathological examination of the resected specimen or, if unresected, by intraoperative biopsy. All other histological types, such as mucinous adenocarcinoma, intraductal papillary adenocarcinoma, acinar cell carcinoma, and endocrine tumors were excluded from the analysis. All patients were operatively staged by laparotomy based on the clinical staging system of the American Joint Committee on Cancer (AJCC) TNM classification.16 Tumor size measurements were determined by the three-dimensional volume of the largest diameter on the examination of gross pathological specimens for surgically resected cases, and the PACS program (Infinitt®, Seoul, Korea) assisted diameter measurement for the CT images obtained within 1 month before surgery for intraoperative biopsy cases.

Resectability was defined as a microscopic resection margin-negative R0 resection (R1; microscopically resection margin positive, R2; minimal macroscopic resection margin positive).

Laboratory results were reviewed for preoperative serum CEA, CA19-9, and total bilirubin levels drawn simultaneously no more than 2 weeks before surgery. These tumor markers were available in all 114 patients. As both CA19-9 and CEA undergo some degree of biliary excretion, the levels may have been artificially elevated because of the biliary obstruction caused by the tumor and, therefore, may not have accurately reflected the tumor volume. To adjust the effect of biliary obstruction on serum levels of CA19-9 and CEA, we adopted an adjusted CA19-9 or CEA to account for the degree of biliary obstruction. A threshold serum total bilirubin level of 2.0 mg/dL for adjusting the tumor markers was based on the reference range of total bilirubin levels reported by our clinical laboratory (0.3–1.3 mg/dL) and the reported alteration in the pharmacokinetics of various medications and their metabolites in the setting of hyperbilirubinemia caused by biliary obstruction.17,18 In various reports, the total bilirubin level has been shown to be a reasonable marker of altered biliary excretion, and a significant alteration has been shown to occur at levels greater than 1.5-fold the upper limit of normal, or approximately 2.0 mg/dL. Therefore, the adjusted tumor marker level (either CA19-9 or CEA) in patients with a total bilirubin level of 2 mg/dL or more (i.e. presumed altered biliary excretion) was determined by dividing the serum tumor marker level by the total bilirubin level. In patients with normal biliary excretion (i.e. total bilirubin level < 2.0 mg/dL), the actual tumor marker level was used.

Statistical analysis was initially carried out by means of analysis of variance for all three sample groups. We then used receiver–operating characteristic (ROC) curve analysis to determine the best cut-off values of CA19-9, CEA and tumor volume. Sensitivity, specificity, and positive and negative predictive values were calculated with each cut-off value of CA19-9, CEA and tumor volumes. A multivariate logistic regression model was used to determine the independent factors associated with resectability for these patients. A value of P < 0.05 was considered to indicate significance.

Results

Although 72 patients with pancreatic cancer appeared to have resectable tumors on the preoperative imaging studies, only 24 of them (33.3%) had completely resectable (R0) tumors on the postoperative pathological examination; the other 48 patients were identified as having unresectable (R1/2) tumors because of unexpected metastasis (total n = 16; liver metastasis, n = 5; peritoneal metastasis, n = 11) or locally advanced disease at the time of surgery (total n = 21; superior mesenteric artery invasion, n = 18; superior mesenteric and celiac axis invasion, n = 1; celiac axis invasion, n = 2) or residual tumor on the postoperative pathological examination (n = 11).

The age of the patients ranged from 41 to 88 years (mean ± standard deviation, 66.0 ± 10.2 years), and 60 (52.6%) of the patients were men. Sixty one of 114 (53.5%) enrolled patients had a smoking history of more than 5 pack-years. Thirteen of 24 (54.2%) patients who had R0 tumors and 48 of 90 (53.3%) patients who had R1/2 or unresectable tumors had smoked. Additionally, there was no significant difference in the amount of smoking between the two patient groups. In our study, there was no case of other malignancies which could elevate the serum levels of CEA and CA19-9, such as colorectal cancer, stomach cancer, lung cancer, ovarian cancer, intra- and extrahepatic cholangiocarcinoma and gallbladder cancer.

The tumors involved the head (n = 74, 64.9%), body (n = 15, 13.2%) and tail (n = 25, 21.9%) of the pancreas. Although we have adopted an adjusted CA19-9 or CEA to account for the degree of bile-flow disorders, for the purpose of the exclusion of possible confounding effects of bile-flow obstruction in the serum levels of CA19-9 and CEA, further analyses by the location of tumor (especially for the location of pancreatic head) were carried out. In the present study, there were no significant differences in the results of the serum level of each tumor marker between the patients with pancreatic head cancer and the total enrolled patients of any tumor location. Thirty-five patients did not have distant metastasis or locally unresectable vessel invasion at the time of surgery and underwent curative resection (14, pylorus-preserving pancreaticoduodenectomy; 8, Whipple's operation; and 13, distal pancreatectomy) with regional lymph node dissection on the basis of tumor location. According to the baseline characteristics, age, gender, tumor location, and tumor cell differentiation were not related to resectability. However, AJCC staging and tumor size were significantly correlated with tumor resectability (Table 1). The median CEA and CA19-9 levels for all patients were 3.08 ng/mL and 122.25 U/mL, respectively. For the patients with resectable (R0) disease, the mean and median values of CA19-9 were significantly lower than for patients with R1/2 or unresectable disease (Table 2). The mean and median values of CEA for the patients with resectable (R0) disease showed a lower tendency than for patients with R1/2 or unresectable disease; however, it did not reach statistical significance (Table 2). Likewise, the existence of liver metastasis was associated with significantly higher CA 19-9 levels, but not with CEA levels (Table 2).

Table 1.  Baseline characteristics of patients with pancreatic adenocarcinoma
CharacteristicsTotalR0R1/2UnresectableP-value
  1. AJCC, American Joint Committee on Cancer; NS, not significant; SD, standard deviation.

No. patients (%)11424 (21.1)48 (42.1)42 (36.8) 
Age (mean ± SD, years)66.0 ± 10.265.6 ± 10.665.3 ± 9.867.1 ± 10.5NS
Male gender (%)60 (52.6)16 (66.7)23 (47.9)21 (50.0)NS
Tumor location (%)    NS
 Head74 (64.9)17 (70.8)34 (70.8)23 (54.8) 
 Body15 (13.2)2 (8.3)7 (14.6)6 (14.3) 
 Tail25 (21.9)5 (20.8)7 (14.6)13 (31.0) 
Tumor differentiation (%)    NS
 Well differentiated24 (21.1)4 (16.7)10 (20.8)10 (23.8) 
 Moderately differentiated57 (50.0)17 (70.8)19 (39.6)21 (50.0) 
 Poorly differentiated33 (28.9)3 (12.5)19 (39.6)11 (26.2) 
Tumor size (mean ± SD, cm3)29.3 ± 35.97.3 ± 6.225.4 ± 33.546.4 ± 40.4< 0.01
AJCC Stage (%)    < 0.01
 IA4 (3.5)4 (16.7)0 (0)0 (0) 
 IB2 (1.8)1 (4.2)1 (2.1)0 (0) 
 IIA32 (28.1)11 (45.8)17 (35.4)4 (9.5) 
 IIB23 (20.2)8 (33.3)12 (25.0)3 (7.1) 
 III20 (17.5)0 (0)8 (16.7)12 (28.6) 
 IV33 (28.9)0 (0)10 (20.8)23 (54.8) 
Table 2.  Comparison of preoperative CEA and CA19-9 levels with resectability and metastatic status in patients with pancreatic adenocarcinoma
VariableCategorynCEA (ng/mL)CA19-9 (U/mL)
MedianMean ± SDP-valueMedianMean ± SDP-value
  1. AJCC, American Joint Committee on Cancer; CA19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; SD, standard deviation.

ResectabilityR0242.343.35 ± 2.810.07649.66111.98 ± 156.230.034
R1/2484.9616.25 ± 34.35 233.031860.14 ± 3091.43 
unresectable423.988.82 ± 11.85 174.071559.83 ± 2985.07 
Liver metastasisYes298.0820.99 ± 37.230.064600.003188.09 ± 4089.710.004
No852.457.32 ± 15.99 80.14765.09 ± 1795.57 
Peritoneal metastasisYes99.6739.89 ± 51.490.104780.493967.94 ± 4703.700.113
No1052.668.30 ± 18.30 116.701159.77 ± 2448.94 
AJCC stageIA41.191.18 ± 0.030.02940.0540.05 ± 23.85< 0.001
IB27.477.47 ± 0.38 2005.282005.28 ± 2192.43 
IIA321.754.29 ± 9.17 41.46469.64 ± 1055.86 
IIB233.647.89 ± 9.45 87.77747.79 ± 2044.71 
III203.047.09 ± 9.08 135.83709.98 ± 1392.65 
IV338.0822.73 ± 40.02 600.003239.06 ± 4074.25 

ROC curve analysis predicting the values of CEA and CA19-9 for resectability (R0 resection) is shown in Figure 1. ROC analysis showed significant results for the preoperative serum levels of both CEA (AUC [area under curve] 0.631; 95% CI, 0.518–0.744; P = 0.049) and CA19-9 (AUC, 0.720; 95% CI, 0.619–0.821; P = 0.001) in predicting resectability (R0 resection). Likewise, tumor size (AUC, 0.771; 95% CI, 0.684–0.858; P < 0.001) was shown to be significant for predicting resectability (R0 resection). The best cut-off values for CEA, CA19-9 and tumor size to predict resectability (R0 resection) were 2.47 ng/mL (sensitivity, 62.0%; specificity, 60.0%), 92.77 U/mL (sensitivity, 67.8%; specificity, 75.0%) and 11.85 cm3 (sensitivity, 70.9%, specificity. 77.1%), respectively. Using these cut-off values, we tested the utility of each marker and the combination of the cut-off values for the prediction of resectability (R0 resection), which showed 90.6% and 88.6% negative predictive values when the CA19-9 was less than 92.77 U/mL and either tumor marker was less than the cut-off value, respectively (Table 3). These data also showed a 40.6% positive predictive value when both tumor markers were less than the cut-off values. These data indicated that both tumor markers greater than or equal to the cut-off levels or CA19-9 ≥ 92.77 U/mL predicted the possibility of R1/2 or unresectability with 88.6% and 90.6% accuracy, respectively. However, both tumor markers less than the cut-off levels (positive/positive) predicted the possibility of R0 resection with 40.6% accuracy (Table 3).

Figure 1.

Receiver–operating characteristic (ROC) analysis showed significant results for the preoperative serum levels of both carcinoembryonic antigen (CEA) (AUC [area under curve] 0.631; 95% CI, 0.518–0.744; P = 0.049) and carbohydrate antigen 19-9 (CA19-9) (AUC, 0.720; 95% CI, 0.619–0.821; P = 0.001) in predicting resectability (R0 resection). Likewise, tumor size (AUC, 0.771; 95% CI, 0.684–0.858; P < 0.001) was shown to be significant for predicting resectability (R0 resection). The best cut-off points for CEA, CA19-9 and tumor size to predict resectability (R0 resection) were 2.47 ng/mL (sensitivity, 62.0%; specificity, 60.0%), 92.77 U/mL (sensitivity, 67.8%; specificity, 75.0%) and 11.85 cm3 (sensitivity, 70.9%, specificity. 77.1%), respectively.

Table 3.  Assay parameters for serum CEA and CA19-9, and for the combination of the two in 114 patients with pancreatic adenocarcinoma
Assay parameterResectability (R0 resection)
CEA < 2.47 ng/mLCA19-9 < 92.77 U/mLEitherBoth
  1. CA19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; NPV, negative predictive value; PPV, positive predictive value.

No. patients51503832
Sensitivity (%)56.575.079.254.2
Specificity (%)57.864.443.378.9
PPV (%)25.536.027.140.6
NPV (%)83.990.688.686.6

To determine the independent contributing factors to the resectability (R0 resection), we carried out multivariate logistic regression analysis and showed that a CA 19-9 below the cut-off value (< 92.77 U/mL), a tumor size less than 11.85 cm3, and less advanced AJCC stage were independent predictors of resectability in patients with pancreatic adenocarcinoma (Table 4).

Table 4.  Multivariate logistic regression analysis to predict resectability (R0 resection)
VariableCategoryOdds ratio95% Confidence intervalP-value
  1. AJCC, American Joint Committee on Cancer; CA19-9, carbohydrate antigen 19-9.

AJCC stageMore advanced6.063.77–15.630.002
Tumor size (cm3)< 11.85 cm32.951.59–19.610.022
CA19-9< 92.77 U/mL2.751.47–21.280.026

The 1-, 2-, 3- and 5-year cumulative survival rates were 73.0%, 20.9%, 15.7% and 15.7%, respectively, for patients with resectable (R0 resection) disease, 37.6%, 11.7%, 0% and 0%, respectively, for patients with R1/2 resection, and 3%, 0%, 0% and 0% for patients with unresectable disease, respectively (Fig. 2).

Figure 2.

Kaplan–Meier survival curves were drawn for patients with R0 resection, R1/2 resection and inoperable tumor. The 1-, 2-, 3- and 5-year cumulative survival rates were 73.0%, 20.9%, 15.7% and 15.7%, respectively, for patients with resectable (R0 resection) disease; 37.6%, 11.7%, 0% and 0%, respectively, for patients with R1/2 resection; and 3%, 0%, 0% and 0%, respectively, for patients with unresectable disease.

Discussion

Pancreatic adenocarcinoma is an aggressive tumor with a propensity for early dissemination to regional lymph nodes or distant metastatic sites (e.g. liver, peritoneum and lung). At the time of presentation, approximately 40% of patients have metastatic disease, whereas another 40% present with locally advanced disease. Of the remaining 20% believed to have tumors confined to the pancreas based on non-invasive imaging, up to one-half may have advanced, unresectable disease discovered at the time of operative exploration.19–25 As the failure to precisely define the extent of the tumor burden results in unnecessary surgical risk, a delay in appropriate systemic chemotherapy and an increase in health-care costs, the preoperative prediction of resectability is very important, but there is no precise modality for predicting complete resection of pancreatic adenocarcinoma.

In recent years, there have been considerable advances in the diagnostic approaches of pancreatic adenocarcinoma. CT is the most widely available and best-validated modality for imaging patients with pancreatic adenocarcinoma. Current CT imaging allows accurate evaluation of the local extent of pancreatic adenocarcinoma, but frequently misses small-volume hepatic or peritoneal metastases, which occur in a significant proportion of patients who undergo unnecessary surgical exploration for potentially unresectable disease.26 Although the technology for this imaging modality has improved tremendously, it still misses occult peritoneal or liver metastatic disease (< 1 cm) in 4% to 15% and occult vascular involvement in 4% to 19% of cases.23,24,27–29 The positive predictive value of CT for predicting resectability (45–79%) is low because the diagnostic criteria for diagnosing vascular invasion by the tumor favors specificity over sensitivity to avoid denying surgery to patients with potentially resectable tumors. Furthermore, the sensitivity of CT for small hepatic and peritoneal metastases is limited.30

CEA and CA9-9 are the most studied serum tumor markers that have been evaluated for diagnosis and prognosis in patients with pancreatic adenocarcinoma. However, little is known about the association between the levels of these markers and the existence of metastasis or locally advanced disease in patients with pancreatic adenocarcinoma. Recently, several researchers have shown the relationship between these markers and resectability in pancreatic adenocarcinoma.15,31–33 As for patients with an established diagnosis of pancreatic adenocarcinoma, markedly elevated levels of these tumor markers may reflect unresectability in those patients who were thought to have resectable disease on preoperative imaging. Fujioka et al.32 reported that combined preoperative CEA and CA19-9 levels are suitable for assessing expected curability and resectability in patients with pancreatic cancer. They used the definition of resectability as R0 resection (microscopic resection margin-negative) plus R1/2 resection (microscopic resection margin-positive or macroscopic resection margin-positive), but we strictly used R0 resection as a necessity for resectability because non-curative resection (R1/2) usually fails to achieve long-term survival. In our study, we have demonstrated such differences of median survival and 1-, 2-, 3-, and 5-year cumulative survival rates in the R0, R1/2, and unresectable groups (Fig. 2).

In a more recent report, Zhang et al.33 reported that the preoperative serum CA19-9 level is a useful marker for evaluating the resectability of pancreatic adenocarcinoma. They obtained higher cutoff levels of CA19-9 (> 353.15 U/mL) than in our study (> 92.77 U/mL) and reported 93.1% and 78.5% sensitivity and specificity, respectively. They also reported positive and negative predictive values as 84.38% and 90%, respectively. Compared to our results of 75.0%, 64.4%, 36.0%, and 90.6% for sensitivity, specificity, and positive and negative predictive values, respectively, they showed much higher sensitivity, specificity, and positive predictive values and comparable negative predictive value. On the other hand, the report by Fujioka et al.32 have obtained the best cut-off level of CA19-9 for resectability in patients with pancreatic adenocarcinoma as 157 U/mL and shown that the sensitivity, specificity, and positive and negative predictive values of preoperative CA19-9 to predict the resectability for these patients were 76%, 46%, 57% and 71%, respectively. The report of Fujioka et al.32 showed more similar results to our study findings than the report of Zhang et al.33 However, the specificity and negative predictive values of our study showed more favorable results. As Fujioka et al.32 and Zhang et al.33 did not use the adjusted CA19-9 level, the best cut-off value of CA19-9 in their studies could be higher than our adjusted CA19-9 level.

Our study demonstrated that a CA19-9 level no less than the cut-off value and both tumor markers no less than the cut-off levels predicted the possibility of R1/2 or unresectable operation with 90.6% and 88.6% accuracy, even with a preoperative CT demonstrating only a localized and resectable tumor. It provides a considerable predictive value compared to conventional imaging studies such as bolus contrast, helical CT scan.

So then, how can we assess and manage the patients in whom preoperative CT demonstrates only a localized and resectable tumor, but whose tumor markers show higher levels than our cut-off levels? We recommend further preoperative evaluations such as positron emission and computed tomography (PET-CT) and/or diagnostic laparoscopy for detecting occult metastases to decrease the number of unnecessary laparotomies. However, patients in whom preoperative CT demonstrates only a localized and resectable tumor and with tumor markers showing lower levels than our cut-off levels, the possibility of occult distant metastases may be quite low, and the progression to surgical resection will be a step in the right direction.

Interestingly, our data showed that for the patients with resectable (R0) disease, the mean and median values of CA19-9 were significantly lower than for the patients with R1/2 or unresectable disease, but the mean and median values of CEA for the patients with resectable (R0) disease was shown to have a lower tendency than those for patients with R1/2 or unresectable disease; however, it did not reach statistical significance. A similar trend has been reported in the report by Fujioka et al.;32 however, they have found significant differences between the CEA levels of R0, R1/2 and unresectable groups.

As with other diagnostic modalities, the utility of CA19-9 has several confounding limitations. Patients who are negative for the Lewis blood group antigen (a-, b-), approximately 4% to 15% of the general population, do not synthesize CA 19-9. We did not test for Lewis antigen status in our study. Only one-half of cancers less than 2 cm are associated with an elevated CA 19-9.34,35 In addition, false-positive elevations in CA 19-9 exist in benign conditions, such as in patients with extrahepatic biliary obstruction caused by pancreatitis and choledocholithiasis.36,37 This is believed to be due to hepatic insufficiency to degrade and secrete CA19-9.38 Several studies have shown that the association of elevated levels of CA19-9 with the diagnosis of pancreatic cancer is significantly obscured in the face of obstructive jaundice, and the cut-off value should be adjusted for hyperbilirubinemia.36,39 We have attempted to adjust for CA19-9 and CEA values in reference to total bilirubin level in our study.

In conclusion, preoperative serum CA19-9 level can be independently used for the prediction of resectability (R0 resection) in patients with pancreatic adenocarcinoma, which may enable simple and cost-effective exclusion of patients who are unlikely to benefit from surgery.

Acknowledgment

This study was financially supported by a grant from the IN-SUNG Foundation for Medical Research.

Ancillary