Abstract
The value of DNA image cytometry in the differential diagnosis of benign and malignant epithelial lesions was tested analyzing 174 specimens from the biliary tract, the pancreatic ducts and the papilla of Vater. Patients and Methods: Cytologic specimens (n=45) or monolayer smears, prepared from three 50 μm thick sections by a cell separation technique (n=129), were stained according to Feulgen. The DNA content of 250 epithelial cells, chosen per random, was determined using a TV-image analysis system CM-1 (Hund, Wetzlar, Germany). The DNA content of 30 lymphocytes served as an internal standard for the normal diploid value in every individual case. Different DNA cytometric parameters and the mean nuclear area were calculated. Results: In comparison with the benign cases (n=120), the mean values of all DNA cytometric variables were markedly increased in the malignant (n=54) (p<0.001). However, a certain overlay of the data distribution had to be considered. Requiring a specificity of ≥95%, sensitivity values up to 87% (mean ploidy) could be obtained. An aneuploid DNA stemline was present in 76% of the tumor cases. The occurrence of 9cEEs were completely restricted to malignant specimens (specificity: 100%), but only present in 50% of the tumor cases. As an additional finding in some benign cases, DNA polyploidy up to 8c could be demonstrated (11/120, 9.2%). Conclusion: DNA image cytometry represents a relevant tool in the objective identification of malignant changes of specimens of the biliary tract, the pancreatic duct or the papilla of Vater, if suitable variables are used for DNA data interpretation. However, mainly due to the fact that several cases reveal only slight ploidy changes, a well-preserved DNA ploidy profile does not exclude malignant changes.
- Bile duct neoplasms
- pancreas neoplasms
- DNA ploidy
- computer-assisted
- cytometry
- image analysis
- differential diagnosis
The precise and correct interpretation of cytological specimens sampled during an endoscopic retrograde cholangiopancreaticography (ERCP) remains controversial up to now. Various types of epithelial cells, i.e. from the papilla of Vater, the bile ducts and the pancreatic ducts have to be considered. Furthermore, not only inflammatory changes but also poor cellular preservation by degeneration and the finding of epithelial activation make the interpretation of cytological changes more difficult.
Consequently, the sensitivity of cytology for the detection of malignant changes has been reported as comparably low with values between 27% and 60% for the biliary tree (1-6). As far as biliary strictures are involved, however, disruption by dilatation could enhance the results of bile cytology (5). In the undilated control group, 33.3% of the diagnoses were “positive for cancer ” or “suggestive” as compared to 75.5% in the dilated group (p=0.028). Respective rates for pancreatic carcinoma reached values between 40% and 65% (5-7). Only in a study on carcinoma of the papilla of Vater a high sensitivity rate of 91% could be obtained (6). Specificity values ranged between 96% and 100% indicating only a low risk of false positive tumor diagnosis (3, 7, 8).
To plan an adequate therapy, objective data are highly recommended based on tumor biologic or morphometric methods. In various tissues, malignant transformation has been predicted by DNA image cytometry, including precancerous squamous epithelial lesions of the oral (9), laryngeal (10), bronchial (11), vulvar (12) or cervical (13-14) mucosa or the skin (15), borderline lesions of the ovary (16), gastric dysplasia (17), hyperplastic endometrial lesions (18) or precancerous lesions of the endocervical glands (19). On lesions from the the bile ducts, the pancreatic ducts or the papilla of Vater only few data from the literature are available (20). Thus, as representative data on a possible relevance of the method concerned are still lacking, the aim of this study was to test the diagnostic significance of DNA image cytometry in these lesions more systematically.
Mean values and standard deviations of DNA image cytometric variables and the mean nuclear area for all 120 lesions.
Mean values and standard deviations of DNA image cytometric variables and the mean nuclear area for 68 lesions of the bile ducts and the gallbladder
Patients and Methods
Patient groups. 54 cases of carcinoma of the bile ducts or the gallbladder (n=20), of the pancreas (n=22) or of the papilla of Vater (n=12), were investigated as test group. One hundred and twenty normal or inflammatory specimens from the bile ducts or the gallbladder (n=48), the pancreatic duct or the pancreatic parenchyma (n=43) or the papilla of Vater (n=29) served as control group.
As cases of cytologic specimens from these sites were comparably rare in the files and the number of cells for an adequate investigation was partly too low, the 45 primary cytologic specimens from ERCPs (n=38) or from fine needle aspiration biopsies (n=7) were supplied by 129 cytologic smears, prepared from paraffin blocks by a cell separation technique.
Methods. After microscopic re-evaluation and microphotographic documentation, the primary cytologic specimens were rehydrated in graded ethanols up to 70% ethanol after removing the cover slip by xylene application.
On paraffin blocks under investigation, the region of interest was marked in order to prevent measurements on tissue areas beside the lesion. Then, monolayer smears were prepared from three 50 μm thick sections of this region by a cell separation technique (21, 22). Later, all slides were stained according to Feulgen as described in detail previously (23). Briefly, acid hydrolysis (4M HCl, 27.5°C, 55 minutes) was followed by a sixty minute incubation in Schiff's reagent at room temperature.
Mean values and standard deviations of DNA image cytometric variables and the mean nuclear area for 65 pancreatic lesions.
Mean values and standard deviations of DNA image cytometric variables and the mean nuclear area for 41 lesions of the papilla of Vater
Per specimen, the DNA content and the mean nuclear area of 250 epithelial cells was determined. For DNA quantification, per nucleus the integrated optical density and the nuclear area were measured at a wavelength of 570 nm (halfwidth of the interference filter: ±10 nm) using a TV-image analysis system CM-2 (Hund, Wetzlar, Germany) (24). The mean integrated optical density of 25-30 lymphocytes served as an internal standard for the normal diploid value in every individual case (error of the mean of the reference cell population: <3%). The results were documented in a DNA frequency histogram. From the distribution of the single values, the mean nuclear area and various DNA cytometric variables were calculated following the proposals of the recent ESACP consensus report (25), including the DNA stemline ploidy (26), the mean ploidy, the 2c deviation index (27), the 5c- and the 9c exceeding events (5cEE, 9cEE) (28, 29), the entropy (30) and the ploidy imbalance (18).
Statistics. For the comparison of the data sets of the three series of tumors, the Mann-Whitney U-test was used in order to analyze the data distribution. Multi-field tables were interpreted using the Pearson Chi2 test of independence. Statistical significance was accepted for p<0.05. Furthermore, for each variable, sensitivity and specificity values were calculated separately.
Sensitivity and specificity of DNA image cytometric variables and the mean nuclear area for the differential diagnosis of all 174 lesions at a specificity level ≥95% .
Sensitivity and specificity of DNA image cytometric variables and the mean nuclear area for the differential diagnosis of 68 bile duct lesions or gall bladder lesions at a specificity level ≥95% .
Results
Data distribution. The data distribution of the various DNA cytometric variables and of the mean nuclear area are presented in Table I. In comparison with the benign control cases, the mean values and the standard deviations were markedly increased in the group of the malignant lesions, most obviously for the 5cEE, the 2cDI and the ploidy imbalance. Also the mean nuclear area was clearly different between the two types of lesions. Similar results were obtained when the data were presented group-wise according to the localization of the lesions (Tables II, III, IV). For all variables, however, some overlay of the data distribution had to be considered. After stratifying the data in two groups according to the median value, multi-field tables revealed statistically significant differences in the Pearson Chi2-test as well for the DNA cytometric variables (p<0.001) and for the mean nuclear area (p=0.002).
Sensitivity and specificity of DNA image cytometric variables and the mean nuclear area for the differential diagnosis of 65 pancreatic lesions at a specificity level ≥95% .
Sensitivity and specificity of DNA image cytometric variables and the mean nuclear area for the differential diagnosis of 41 lesions of the papilla of Vater at a specificity level ≥95% .
DNA polyploidy in benign specimens. From the DNA data distribution it became obvious that some nuclei with DNA contents >5c (“5cEE”) were present in several benign lesions. As this indicator of DNA polyploidy (31) under benign conditions has not been reported yet in the tissue types included in this study, this aspect should be presented in more detail: Eleven benign tissues (9.2%, 11/120) contained a small number of nuclei with a DNA content >5c (5cEE=1, n=5; 5cEE=2, n=4; 5cEE=3, n=2), indicating polyploidy up to 8c. In pancreatic lesions (12.5%, 6/48) the rate of polyploid cases was higher than in bile tract lesions (8.3%, 4/43) or in lesions of the papilla of Vater (3.4%, 1/29). Thus, DNA polyploidy up to 8c represents a rare, but relevant finding which has to be taken into account if single cell interpretation should be applied. Nuclei with a DNA content > 9c (9cEE) were not present in the benign tissues at all (0%, 0/120).
Sensitivity and specifitity values for the stemline ploidy, considering 120 benign and 54 malignant lesions of the bile duct, the pancreatic duct an the papilla of Vater.
Sensitivity and specifitity values for the mean ploidy, considering 120 benign and 54 malignant lesions of the bile duct, the pancreatic duct an the papilla of Vater.
Diagnostic data interpretation. The sensitivity of the parameters was calculated requiring a specificity of ≥95% (Table V, Figures 1 and 2). As DNA cytometric changes were only comparably slight in many cases, only one variable, the mean ploidy, revealed a very high sensitivity of 87% at a threshold of 2.35c. Most of the remaining variables showed sensitivity values between 65% and 76% . An aneuploid DNA stemline was present in 76% of the tumor cases. 9cEEs were infrequent also in the group of malignant tumors (sensitivity: 50%). The diagnostic value of the mean nuclear area was slightly lower than the DNA cytometric ones (sensitivity: 63%). Well comparable results were found when the different subgroups were analyzed separately (Tables VI, VII, VIII).
Discussion
Ultrasonography, computed tomography, cholangiography or endoscopy are very useful methods in the description and characterization of changes of the biliary tree, the pancreatic ducts and the papilla of Vater. However, the definite differentiation between malignant and benign lesions is restricted to a morphologic tissue diagnosis. As far as conventional diagnostic cytology should be used to confirm such a lesion as malignant, the limited sensitivity of this method on ERCP specimens has to be considered (2-6). Thus, it should be examined whether additional methods may contribute to the detection of malignancy in cytologically negative or suspicious cases. In this study, the value of DNA image cytometry was investigated, a method that has been proven to be of high diagnostic relevance in various tissue types (9-19).
On neoplastic lesions of the biliary tract, the pancreatic ducts or the papilla of Vater, however, respective studies have rarely been performed, mostly focusing on the as expected limited prognostic value of DNA image cytometry in these tumor entitites due to the poor overall survival (32-36). For diagnostic purposes, the interpretation of DNA image cytometric measurements has been performed in one study using so-called “histogram classes” and the 5cEE; considering 49 specimens, the combination of cytology and DNA image cytometry increased the sensitivity of the investigation to 88%, but the specificity decreased to 73% (20). The authors concluded that DNA image cytometry may help to select patients who should undergo a more rigorous evaluation for malignancy.
The present investigation applying various DNA cytometric variables on 120 cases underlines the high diagnostic value of modern DNA image cytometry. Requiring a specificity rate of ≥95%, a sensitivity rate of 76% -87% (Table V) may be obtained, using suitable thresholds for the identification of DNA aneuploidy. Following the results, in specimens from the biliary duct, the pancreatic ducts or the papilla of Vater the use of the DNA stemline ploidy or the mean ploidy may be especially recommended. Further, the finding of DNA values above the 9c level (9cEE) was completely restricted to carcinoma, being present in 50% (27/54) of the cases in small numbers (mean value for 9cEE: 3.02±5.77 [SD]). Thus, the occurrence of 9cEE is an infrequent, but important and hard marker in the characterization of a cell population as malignant. The presence of 5cEE in some benign specimens, although in only rare numbers (mean value: 0.15±0.54; maximal value: 3), underlines the capability of the epithelium to react with euploid polyploidization (31) up to 8c.
In conclusion, the results of this study justify the hypothesis that DNA single cell cytometry in lesions from the biliary or pancreatic ducts or from the papilla of Vater could be as relevant as it has been demonstrated for various precancerous conditions of squamous epithelium or glandular epithelium. It was possible to identify diagnostically helpful and hard markers of malignancy like DNA stemline aneuploidy, increased mean ploidy levels or the finding of 9cEEs. However, mainly due to the fact that several cases reveal only slight ploidy changes, a well-preserved DNA ploidy profile does not exclude malignant changes. In the future, the use of additional image cytometric tools should be tested more systematically, including morphometric variables. In a pilot study (37), a discriminant analysis of more than 30 parameters characterized significant differences between chronic pancreatitis and pancreatic adenocarcinoma (p<0.001) and between primary sclerosing cholangitis and cholangiocarcinoma (p<0.001) on a sensitivity and specificity level of 82% -85%, indicating a potential of this method as an adjuvant diagnostic technique in patients with pancreaticobiliary malignancy. Morphometric or cytometric approaches could also be combined with quantitative immunohistochemistry or with multiparameter fluorescence in situ hybridization (FISH) applications (38).
- Received September 5, 2008.
- Revision received January 19, 2009.
- Accepted February 2, 2009.
- Copyright© 2009 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved
