How to Improve the Accuracy of Diagnosis of Malignant Biliary Strictures

Publisher Website
Google Scholar
Abstract
The optimal management of patients presenting with suspected malignant biliary obstruction often requires a tissue diagnosis. Although many such patients are managed without histologic confirmation of the tumor, such a confirmation would ideally permit more accurate decision-making with reference to total patient management, including potential radiotherapy or chemotherapy. Endoscopic retrograde cholangiopancreatography (ERCP) supplies an unique opportunity to provide biliary decompression, and simultaneously to obtain a histologic or cytologic specimen from a patient with an unexplained biliary or pancreatic stricture. It would be preferable to have one technique for the pancreatobiliary tree with a cancer detection rate similar to that associated with the biopsy of upper gastrointestinal and colonic neoplasms; however, this goal has not been attained. Biliary brush cytology is the most commonly applied technique at ERCP, and also the most extensively studied. Numerous recent series of intraductal brush cytology report a wide range of sensitivity from 30 % to 69 % [ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ]. However, many of these studies used only small samples. In this issue, Glasbrenner et al. [ 10 ] report on a large prospective series of patients, and the diagnostic accuracy of brush cytology in biliary strictures of different etiologies and locations. From March 1996 to November 1997, at their institution, 86 prospective patients with obstructive jaundice were evaluated by ERCP. A final diagnosis (benign or malignant) was established in 78 patients, with brushings being obtained successfully in all patients. Malignancies (confirmed by histology, surgery, autopsy or clinical follow-up) were identified in 57 patients (31 pancreatic cancer, 20 cholangiocarcinoma, 4 metastatic adenocarcinoma, 1 hepatoma, 1 ampullary carcinoma). When cytology samples with dysplastic or malignant cells were regarded as malignant, the overall results for cancer detection were as follows: sensitivity 56.1 % (32/57); specificity 90.5 % (19/21); positive predictive value 94.1 % (32/34), negative predictive value 43.2 % (19/44), and diagnostic accuracy 65.4 % (51/78). The diagnostic yield of brush cytology was greater with proximal stenoses, probably reflecting the increased sensitivity in patients with cholangiocarcinoma (80%; 16/20) as compared to pancreatic cancer (35.5 %; 11/31). Specificity was less than 100 %, as two patients with dysplasia identified by brush cytology were found to have chronic pancreatitis, with malignancy excluded at surgery. The authors concluded that brush cytology was helpful in differentiating between benign and malignant biliary strictures, especially in suspected cholangiocarcinoma. Dysplastic cells might occur in the absence of malignancy, and therefore should be interpreted cautiously. Glasbrenner et al. [ 10 ] have presented a large series on biliary brush cytology. While they confirm that the technical success rate is high, the cancer detection rate (56 % in their study) is still discouraging. Their results reflect the general trend seen in other studies evaluating brush cytology, where higher sensitivities are seen for cholangiocarcinoma than for pancreatic or metastatic carcinoma [ 1 ] [ 11 ] [ 12 ]. This finding is presumably related to the fact that metastatic or pancreatic tumors that initially compress the bile duct extrinsically are unlikely to be diagnosed by brushing the epithelial surface during the earlier stages of the disease. In contrast, a primary bile duct tumor originating from the ductal epithelium would appear to be more easily diagnosed by cytology [ 13 ]. Thus, series with a higher percentage of primary bile duct cancers and ampullary cancers report the highest sensitivities for brush cytology. Brushing the pancreatic duct may increase the diagnostic yield in patients with pancreatic cancer [ 11 ]. However, pancreatic cancers often disrupt the duct, and prevent passage of the brush through the tumor in more than 25 % of patients. In an attempt to improve the sensitivity of cancer detection, other methods have been employed more recently. Howell et al. originated the use of the ERCP endoscopic fine-needle aspiration (FNA) technique, and reported a sensitivity, from biliary samplings, of 62 % in affected patients, including 53 % of those with pancreatic cancer and 80 % of those with cholangiocarcinoma [ 14 ]. These results were superior to brush cytology (cancer detection rate of 8.3 % in this series). Endobiliary forceps biopsy allows examination of tissue specimens below the bile duct epithelium. The results of four prospective series [ 1 ] [ 2 ] [ 3 ] [ 15 ] have been encouraging, with a cancer detection rate, among 275 patients, of 65 %. Stents placed for palliation can be removed and sent for cytologic evaluation of adherent cells, as exfoliated malignant cells may become trapped in the biofilm and sludge of an occluded endoprosthesis. A few small studies have evaluated this approach, with a wide range of sensitivities (36 - 79 %) [ 16 ] [ 17 ]. Unfortunately, this method is impractical, because diagnosis is delayed until the stent is removed. Aspiration of bile or pancreatic juice is the easiest method of obtaining tissue when evaluating biliary strictures for malignancy. These results have also been largely disappointing, with cancer sensitivities in the 6 - 32 % range [ 1 ] [ 7 ] [ 16 ] [ 18 ]. The desmoplastic nature of certain tumors, or failure of the neoplasm to invade the ductal epithelium, are probably responsible for these results. It has been suggested that endoscopic manipulation of the stricture (dilation [ 19 ]; irrigation with saline [ 20 ]), prior to bile collection, may increase tumor exfoliation, making more malignant cells available for diagnosis. Prospective comparative trials, however, have not demonstrated any increased cancer detection rate [ 1 ] [ 7 ] [ 16 ]. Preliminary data from our institution suggests that...

This publication has 0 references indexed in Scilit: