|Home | About | Journals | Submit | Contact Us | Français|
To evaluate the diagnostic accuracy of intraoperative frozen sections diagnosis of liver lesions thought to be malignant tumours.
285 frozen sections of liver from 173 patients were reviewed. The examinations were done between 1998 and 2004.
Final histological diagnosis was divided into positive (32%) and negative (68%) for malignancy. In four cases (2%), diagnosis was deferred to paraffin section. There was one false positive and two false negative diagnoses. Sensitivity was 96.9% and specificity was 99.1%, and the overall accuracy to determine the lesions was 95%. The cases were further analysed to ascertain the nature of diagnostic difficulties, which comprised pathological misinterpretation, sampling error, and technical imperfections. Biliary hamartoma was the most common entity that was confused with malignant tumours in frozen sections.
The data are in accordance with those of similar studies in other sites, and confirm that the frozen section is an accurate and reliable method for intraoperative diagnosis of suspected liver lesions.
Core or fine needle biopsy of the liver is less invasive than open exploration, associated with lower morbidity and mortality, and can be applied to patients who are too ill to undergo surgery. However, it has been reported that needle core biopsy in malignant liver lesions is associated with a high risk of needle track tumour seeding.1,2,3 Intraoperative biopsy and frozen section diagnosis of focal hepatic lesions has an important role in diagnosis and it is often carried out in specific circumstances—for example, hepatic nodules suspected of being metastatic deposits—especially in cases where further surgical intervention may be needed, such as metastatic carcinoid and single metastatic lesions.4,5,6 Sometimes during laparotomy or laparoscopy done for any reason, the surgeon may unexpectedly find a liver mass and sample it for a frozen section diagnosis. Anther indication for frozen sections these days is liver biopsy at the time of allograft procurement during liver transplantation.
The main purpose of frozen section of a liver mass lesion is to differentiate malignant from benign lesions, which is important in determining the immediate consequences for the patient. Although differentiating between benign and malignant liver tumours is not usually a major problem, difficulties may arise when the biopsy samples are small or the quality of frozen section is poor, or when it is necessary to distinguish between reactive processes and benign or malignant tumours of the same cell type.7 Therefore, it is important to know the most definitive criteria for making a diagnosis and to understand the pitfalls inherent in frozen section biopsies, so that the most appropriate diagnostic sample can be obtained.
The most important differential diagnosis of nodular lesions in the liver are as follows: metastatic tumours including carcinomas of the stomach, gall bladder, pancreas, colon, lung, breast, and genitourinary tract, followed by lymphomas and sarcomas8; and primary lesions such as bile duct adenoma, bile duct hamartoma, focal fatty change, liver cell adenoma, haemangioma, focal nodular hyperplasia, large regenerative nodule, partial nodular transformation, inflammatory pseudotumour, and primary malignant tumours of the liver.9,10
Frozen section has long been used as a diagnostic tool for intraoperative histopathological determination of various surgical conditions and has acceptable accuracy for clinical use in different tissues including gynaecological, pancreatic, breast, and head and neck cancers and others.11,12,13 However, there is little evidence of the value of frozen section as a diagnostic tool in focal hepatic lesions. In this study, frozen section of liver lesions was evaluated to determine its accuracy and reliability, and to determine causes of error and any pitfalls that may arise.
We identified 285 consecutive hepatic frozen sections from the archives of the department of histopathology, University Hospitals of Leicester, in a seven year period from 1 January 1998 to 31 December 2004. Data such as the age, sex of the patient, the histological diagnoses, and the number of biopsies per patient were recorded. All frozen and corresponding paraffin sections were reviewed and the cases were evaluated according to matching of diagnosis between frozen and permanent paraffin sections, and the histological diagnosis in each case. In addition, the quality of frozen section was assessed and causes of poor quality evaluated.
Descriptive statistics were used to define the accuracy of frozen section in the cases enrolled in the study by comparing the results of frozen section with those of standard paraffin sections. Diagnostic indices were determined, including sensitivity, specificity, positive and negative predictive values, and false positive and false negative rates. All analyses were carried out using SPSS for Windows, version 10.0
In all, we reviewed 285 liver biopsies that had been resected and evaluated by frozen section from 173 patients in our hospitals during the study period. The number of biopsies per patient varied from 1 to 3 (median 1). This study included 110 male and 63 female patients. The median age of the patients was 61 years (range 12 to 90). The final diagnoses of all cases are shown in table 11.
The diagnoses of four cases could not be stated with certainty from frozen section. These were reported as suspicious and a definitive diagnosis was deferred. In all cases, paraffin sections showed malignant neoplasms. One case was metastatic diffuse gastric carcinoma and the others were metastatic cholangiocarcinoma (1) and adenocarcinoma (2). We regarded all four deferred cases as inconclusive, as the definite pathological diagnosis of malignancy was made on the paraffin sections. In these four cases, frozen section played no role in the decision about the further management of the patients. We therefore excluded them from the statistical analysis in evaluating the diagnostic value of frozen section. The overall accuracy of the remaining 169 cases is shown in table 22.. The sensitivity of the test in distinguishing malignant from benign lesions was 96%, the specificity was 99%, and the overall accuracy of the test was 95%. There was no association between number of biopsies per case and benign or malignant diagnoses (p>0.05).
Among all the cases, three had frozen section diagnoses incompatible with the permanent sections, with two false positive and one false negative diagnoses. In one case, frozen section was reported as showing a focus of metastatic adenocarcinoma. However, permanent sections showed Von Meyenberg complex and a bile duct adenoma, the bile duct adenoma having been present on the frozen section and misinterpreted as malignancy. The other case was a well differentiated adenocarcinoma with abundant desmoplastic stromal response which was misinterpreted by frozen section as showing no evidence of malignancy. On review, we disagreed with the initial report of the frozen section and attributed these errors to misinterpretation or overdiagnosis. In the false negative case, frozen section was reported as showing no evidence of malignancy but permanent section showed a focus of adenocarcinoma. On review, we agreed with the initial report—the frozen section was benign, the permanent section containing adenocarcinoma, and attributed this to sampling error, as the frozen tissue received was not processed entirely and the additional tissue that was put on for permanent section revealed the focus of carcinoma.
We also reviewed minor discordances where, for example, two cases were reported as bile duct hamartoma in frozen section, but the lesion was not found in permanent section, suggesting it has been cut through. A further case was reported as bile duct adenoma in frozen section but permanent section showed Von Meyenberg complex.
Having already identified sampling error and pathological misinterpretation as sources of error, we reviewed the technical quality of the sections. Poor quality sections were found in 18 cases. We found that the main causes of poor quality are as follow: presence of calcification (3), extensively necrotic tissue (2), thick section (2) and processing and staining problems (11).
Most patients with a focal hepatic lesion, especially those with a history of previous malignancy, will undergo extensive preoperative radiodiagnostic tests, such as ultrasonography, computed tomography, and magnetic resonance imaging. These techniques usually offer an accurate clinical diagnosis14,15; however, in certain circumstances, such as in patients amenable to resection or when the surgeon unexpectedly finds a liver mass, frozen section diagnosis is mandatory for the proper management of the patient. The accuracy and diagnostic value of frozen sections has been evaluated in tumours of different tissues. However, previous studies of the accuracy of frozen section in the diagnosis of focal hepatic masses are lacking. Thus in the present study we aimed to evaluate the value of frozen section in the histopathological diagnosis of surgically resected liver lesions and to determine causes of error and any pitfalls that might occur.
Overall, malignant tumours were found in 32% of cases and benign tumours in 9%, other benign lesions in 26%, and no evidence of malignancy in 33%. Regarding diagnostic accuracy of frozen section in determining malignancy status, after excluding the four deferred cases in which the definite diagnosis was made on paraffin sections, we found that the overall accuracy of the test was 95% and the negative predictive value was 98%. These results are similar to that of previous studies of frozen sections in other tissues.11,12,13,16 Errors were one false negative diagnosis because of sampling error and two false positive diagnoses. Among these three error cases, one had major implications for the management of the patient, in whom the surgeon had to undertake another operation to remove the resectable tumour which turned out to be malignant. Although, we regarded the cause of error in the two false positive cases as an overdiagnosis error, as evidenced by the permanent sections, pathological misinterpretation has been reported as a well recognised pitfall in the distinction between bile duct adenoma and metastatic adenocarcinoma.17,18 Our results are consistent with other studies such as that of Prey et al, who undertook an extensive review of frozen sections of 4057 tissue blocks from various organs and found an accuracy of 91.5%. Almost all of the misdiagnoses were because of sampling errors and misinterpretation.19
As there were some pitfalls making the diagnosis less than perfect, we explored the influence of the various factors involved. We found that the major pitfall in frozen sections occurred in the distinction between Von Meyenberg complex or bile duct hamartoma and well differentiated adenocarcinoma, which is a well recognised problem in the diagnosis of focal hepatic lesions,20 and in cases of metastatic carcinoma that show abundant desmoplastic stromal response with a paucity of malignant cells. We noted that certain types of metastatic carcinomas in the liver are difficult to diagnose, especially in frozen section, such as metastatic signet ring carcinoma of the stomach and lobular carcinoma of the breast. Thus previous knowledge of the primary tumour of the patient, if any, has a substantial impact on the interpretation of frozen sections of liver masses. We also found that bile duct epithelium may be lost during processing of frozen sections, and in some cases of bile duct hamartoma, for example, the bile ducts may appear as cystic spaces or holes without discernable lining or content in the frozen section.
With regard to the quality of the frozen sections, although we found approximately 8% to be of poor quality, this did not interfere significantly with the diagnosis, and in most cases a correct diagnosis was made regardless of the quality of the section. The number of biopsies was also explored as an influencing factor. Although we did not find an association between number of biopsies and a benign or malignant diagnosis, we noticed that the yield of malignant diagnoses increased with increasing numbers of biopsies. For example, in some cases, malignancy was diagnosed on the third biopsy and one case it was diagnosed using nine blocks. In agreement with this observation, Tangjitgamol et al did not find an association between the numbers of frozen sections of ovarian masses and malignant diagnosis.11 Other studies found that underdiagnoses were caused by sampling error and recommended that a greater number of frozen sections should be taken.21,22
In conclusion, frozen section appears to be an accurate and reliable test for intraoperative diagnosis of focal liver lesions in general. However, some problems persist and histopathologists should remain vigilant.