Accurate diagnosis of strictures involving the bile duct is essential to the planning of therapy and the choice of the right treatment option, such as surgical resection or endoscopic stenting. However, differentiation of malignant from benign ductal lesions at endoscopic retrograde cholangiopancreatography (ERCP) remains a challenge. Although cholangiographic features may be characteristic for malignant or benign disease, in many cases histological or cytological proof of the diagnosis is required to determine the optimal treatment for each individual patient. Histological and cytological tissue diagnoses may be obtained by several methods, including open biopsy, ultrasound or computed tomography guided fine needle aspiration or core biopsy, endoscopic forceps biopsy, endoscopic brush cytology, and bile aspiration cytology.^1–4

Brush cytology performed at ERCP has become the preferred initial method of pursuing tissue diagnosis in many patients with pancreatobiliary strictures.^5–9 The technique allows easy and convenient sampling and has a low complication rate.^3,9,10 The diagnostic specificity of biliary brush cytology is very high and few false positive diagnoses have been reported. The major limitation of the technique has been the relatively modest diagnostic sensitivity. The sensitivity rates reported in multiple studies are highly variable and range between 30% and 88%, with nearly 100% specificity.^3,4,9–12 In general, results of brush cytology for biliary strictures induced by pancreatic malignancies have proved to be inferior (on average 46%) to those observed for biliary malignancies (on average 68%).⁴

The success rate of brush cytology analyses is largely dependent on two factors: (1) the quality of the cytological material obtained at ERCP and (2) the expertise of the cytopathologist.

The quality of the cytological material is influenced by the processing technique, cellularity, cellular preservation, background, quantity of diagnostic cells, and cells not characteristic for biliary lesions such as duodenal mucosa.⁹ Many of these parameters are difficult to control and may require repeated brushings, alternative sampling approaches such as forceps biopsies,^13,14 or changes in sample processing techniques. The fraction of brush cytologies that were non-diagnostic due to reduced quality of the cytological sample has generally been described to be as low as 5%.^10–12 The study by Wight and colleagues¹⁵ highlights the paramount importance of the expertise of the cytopathologist. In this study, a review of 137 consecutive biliary brushings from 127 patients by two expert cytopathologists improved the sensitivity from 49.4% to 89%. Some of these problems arise from inconsistencies in the criteria used for classification of cells on cytological slides. While morphological criteria for benign and reactive changes of duct epithelial cells and for adenocarcinoma cells are well established and utilised in many studies,⁹ problems and inconsistencies mainly arise in the categorisation of lesions not fulfilling all criteria of malignancy. Classification of such cells has been termed to be a “cytological grey zone” by Selvaggi⁹ and includes categories such as atypical, dysplasia (low and high grade), and suspicious. The morphological criteria used for this classification show significant overlap in various studies,⁹ and some authors even include suspicious or atypical lesions in the calculation of sensitivity.^15–17 Thus comparison of the sensitivity and specificity rates obtained in various studies are hampered by the inconsistencies in the definition of cytological criteria in this “cytological grey zone”.

In this unsatisfactory situation, ancillary diagnostic modalities have been increasingly tested to improve the yield of brush cytologies. These include flow cytometry for DNA analysis of aneuploidy,^18,19 morphometry for assessing nuclear area, nuclear DNA content, chromatin distribution,^20,21 telomerase RNA,²² as well as CA-19-9 and CEA measurements¹⁹ in bile fluid. Many of these studies have shown some promising results but are either not widely available or provide only a limited advantage over cytology alone and have thus not led to a significant improvement.

The use of molecular techniques as adjunct to biliary brushing cytology has so far focused on the detection of K-ras codon 12 and p53 alterations. Immunohistochemistry analyses for p53 protein have yielded contradictory results^23–25 and are presently not used in the diagnostic routine. Analyses of K-ras codon 12 mutations in brush cytologies have also yielded unsatisfactory results. It has been shown that K-ras mutations are more frequently found in strictures induced by pancreatic cancers and are not, or less frequently, found in strictures induced by bile duct cancers. Van Laethem et al found a sensitivity for biliary disease of 24% in bile duct and pancreatic duct brushings compared with 81% for pancreatic diseases.²⁶ In the same study, however, K-ras mutations were found in 25% of patients with chronic pancreatitis, without evidence of malignancy even after a short period of follow up, thus reducing the specificity to 72% compared with 100% for cytology. This finding is supported by other studies reporting the presence of K-ras codon 12 mutations in chronic pancreatitis and in normal pancreas^27–29 without evidence of pancreatic cancer development during follow up. Thus at least for strictures induced by pancreatic diseases, analysis of K-ras mutations is of no additional value as their presence will not provide unequivocal proof for the malignant nature of the stricture. For biliary tract cancers the rates of K-ras mutations reported in the literature vary widely, ranging between 0% and 100%,³⁰ and the value of K-ras mutation detection in brush cytologies of biliary strictures appears even more conflicting. The location of the biliary tumour (proximal or distal bile duct, intrahepatic bile ducts, gall bladder), racial and geographic variation, as well as the methods used for mutation detection have been assumed to cause these differences in the incidence of K-ras codon 12 mutations. Furthermore, individual reports indicate that at least in patients with primary sclerosing cholangitis, K-ras mutations may as well be detected without evidence of carcinoma,³¹ which limits the use of K-ras mutation detection in patients with biliary strictures.

In the context of this unsatisfactory situation, the study of Khalid and colleagues³² reported in this issue of Gut presents a novel approach for molecular analyses of brush cytologies of biliary strictures obtained during ERCP (see page 1860). As molecular indicators of malignancy, the authors used tumour suppressor gene linked microsatellite marker loss of heterozygosity (LOH) and K-ras codon 12 mutations. An interesting sampling procedure was used employing manual microdissection of normal and abnormal appearing cells on alcohol fixed Papanicolau stained cytology slides. DNA from two different sampling strategies was used in their study. One was termed “collective assembly” (CA) and involved the combination of separate aggregates of abnormal appearing cell clusters of one brush cytology sample to obtain a sufficient number of cells (approximately 1000) allowing direct molecular analyses. As this strategy was assumed to produce an averaging of mutational changes among the aggregated microdissected cells due to intratumoral heterogeneity of genetic alterations, a second approach was adopted. This approach involved the use of a whole genome amplification technique (WGA) of discrete clusters of 50–100 cells, which in theory should represent individual cytological lesions. However, the anticipated drawback of this technique is the introduction of artefacts and a bias due to the amplification step. As a normal control, the authors used either surgically resected non-neoplastic tissue from the same patient, where available, or normal appearing cellular material from the same brush cytology. To verify the validity of the results, the same molecular analyses were done with microdissected material obtained from surgically resected specimens from the same patients where available.

The most impressive result was that the molecular analyses of polymerase chain reaction amplified DNA from microdissected brush cytology cell clusters discriminated reactive from malignant cells with 100% sensitivity, specificity, and accuracy. Minor variations in chromosomal imbalances between the studied cytological samples and the corresponding surgically resected tissue were attributed to intratumoral mutational heterogeneity. Although the study is intriguing and delivers evidence for the potential value of molecular diagnostic tools for the differential diagnosis of gastrointestinal tumours such as pancreatobiliary malignancies, some issues must be kept in mind when assessing the data. Naturally, a study of 26 patients with mixed types of tumours and controls (six pancreatic cancers, 11 cholangiocarcinomas and nine not further defined benign biliary strictures) will always be preliminary and should be confirmed prospectively in a larger series of patient samples obtained in a blinded manner. The approach is based on the detection of abnormal looking cells on cytological slides. The authors define “abnormal looking cells” in inconclusive cases as cells fulfilling most, but not all, of the criteria for malignancy such as nuclear enlargement, pleomorphism, elevated N/C ratio, nuclear membrane integrity, and coarse chromatin. This classification varies from the categories used by many other cytologists which usually comprise categories such as suspicious, atypical, or dysplastic,⁹ reflecting the lack of standardisation of morphological criteria used to classify cells in this “cytological grey zone”. As in all cases a sufficient number of cells could be microdissected, there were obviously no non-diagnostic cases in the present study. Ten brush cytologies from biliary strictures were inconclusive in the cytological evaluation; only one was derived from a benign stricture. Eight of nine brush cytologies from benign strictures were negative for malignant or abnormal appearing cells. Thus the presence of abnormal looking cells in cytology alone was sufficient for the diagnosis of malignancy in 9/10 cases with inconclusive cytology and the presented molecular approach helped to exclude malignancy in one case. As mentioned above, some cytologists have used the presence of suspicious or atypical cells in brush cytologies for a classification of positive for malignancy.^15–17 In the absence of clear definitions and criteria for the classification of these types of cells, this interpretation appears controversial. The molecular approach presented in the study by Khalid and colleagues³² provides a powerful tool for standardisation of the morphological criteria used to categorise lesions in this “cytological grey zone”. It offers the possibility of identifying and describing cells that are found to be malignant in the molecular approach thus allowing an increase in the sensitivity of the cytological evaluation.

Unfortunately, this molecular approach offers no solution for non-diagnostic brush cytology samples as it requires identification of abnormal and normal cells. In this situation a technique not requiring identification of intact “abnormal looking” cells (for example, by using suspensions obtained from brush cytologies) would clearly be of great value.

A further interesting finding in the study by Khalid and colleagues³² is the observation that all cholangiocarcinomas in their series did not show K-ras-codon 12 mutations in contrast with the majority of pancreatic cancers. As mentioned above, data on the presence of K-ras mutations in biliary tract cancers are controversial and mutation rates reported in the literature show a high degree of variation.³⁰ Most studies however report that the frequency of K-ras mutations in biliary tract cancers is lower than in pancreatic ductal adenocarcinomas. Nevertheless, it appears that we cannot assume that K-ras mutations are absent in biliary tract cancers, and it even appears that mutations may arise in inflammatory disorders such as primary sclerosing cholangitis.³¹ Most likely the observation described in the current study is due to the low number of patients and we may expect to find a low but significant proportion of biliary tract cancers with K-ras mutations with an increasing number of patients. Thus the presence of a K-ras mutation in the brush cytology of a biliary stricture should not lead us to assume that the malignancy must be of pancreatic origin. What is more, as discussed above, the presence of K-ras mutations does not even indicate the presence of malignancy at all, as many patients with chronic pancreatitis show K-ras mutations.

To summarise, the study by Khalid and colleagues³² represents a significant advance as the presented LOH approach using microdissected cells provides a tool to verify the nature of cells in the “cytological grey zone” of brush cytologies of biliary strictures usually classified as suspicious, atypical, or dysplastic using different morphological criteria. It may thus help to standardise the criteria to define cells in the “cytological grey zone” and may allow an increase in the sensitivity of brush cytology by clearly identifying cells indicative of malignancy. So far this can only be done by follow up studies of patients or by histological analysis of surgically resected tumours.