Fine-needle aspiration biopsy (FNAB) of pancreatic lesions has become popular to establish a tissue diagnosis before chemotherapy and/or surgery. The diagnostic sensitivity and specificity of FNAB have improved as a result of several articles regarding cytologic criteria for pancreatic adenocarcinoma. However, false-negative and “suspicious for malignancy” rates remain relatively high, in part because of the underdiagnosis of well differentiated adenocarcinoma (WDA). Existing cytologic criteria do not specifically focus on WDA. In this study, the authors attempt to add to, redefine, and test cytologic criteria for WDA of the pancreas in FNAB specimens.
The authors retrospectively reviewed the specimens of 291 consecutive computed tomography-guided FNABs of pancreatic lesions performed at the study institution between 1995 and 1999. The original cytologic diagnoses were confirmed by cell blocks (131), surgical tissue (84), or clinical follow-up. The FNAB specimens were evaluated for the presence of 10 cytologic criteria: 1) anisonucleosis, 2) nuclear membrane irregularity, 3) nuclear crowding/overlapping/three-dimensionality, 4) nuclear enlargement (if there are more than two red blood cells), 5) gap versus confluent cell spacing, 6) hyperchromasia, 7) macronucleoli, 8) mitosis, 9) chromatin clearing, and 10) necrosis.
The original cytologic diagnoses were nondiagnostic in 24 cases, benign in 27, suspicious for malignancy in 15, and malignant in 225. Among the 225 malignant lesions, 74 cases were diagnosed as WDA. Cytologic criteria 1–4 were observed in 92–99% of WDA cases, whereas criteria 5–10 were present in only 7–38% of WDA cases. Six of 15 suspicious and 4 of 27 negative cases were of low cellularity but retrospectively met the cytologic criteria 1–4 for WDA (the diagnosis was confirmed on clinical follow-up).
Pancreatic malignant tumors are the fifth leading cause of cancer-related death in the United States1 and the incidence of these tumors continues to rise. The survival rate of patients with these tumors is extremely poor, with an overall 5-year survival rate of less than 5%.2 Fine-needle aspiration biopsy (FNAB) has become popular to establish a tissue diagnosis before chemotherapy and/or surgery. The diagnostic sensitivity and specificity of FNABs have improved as a result of several articles establishing cytologic criteria for pancreatic carcinoma.3–8 However, false-negative rates and categorization as “suspicious for malignancy” remain relatively high, in part because of the underdiagnosis of well differentiated adenocarcinoma (WDA). The exisiting cytologic criteria do not specifically focus on WDA. In this study, we add to, redefine, and test cytologic criteria in a large series of WDA cases.
MATERIALS AND METHODS
We identified 326 consecutive cases in which computed tomography (CT)-guided FNAB of pancreatic lesions was performed at the University of Texas M. D. Anderson Cancer Center between January 1995 and September 1999. Thirty-five cases were excluded from this study because the slides were not available for review. The remaining 291 cases constituted the study group. One hundred fifty-one patients were male and 140 female. Ages ranged from 24 to 94 years, with a mean of 62 years. All patients presented with clinical and radiologic evidence of possible malignancy.
The FNAB procedures had been performed as follows. Under CT scan guidance, two to four FNABs were performed using sterile 22-gauge needles. Aspirates were air dried for Diff-Quik stain (American Scientific Products, McGraw Park, IL) and fixed in alcohol for modified rapid Pap stain. The slides were reviewed retrospectively using 10 diagnostic cytologic criteria: 1) anisonucleosis (variation in nuclear size greater than four times within a same epithelial group), 2) nuclear membrane irregularity, 3) nuclear crowding/overlapping/three-dimensionality, 4) nuclear enlargement (if more than two red blood cells), 5) gap versus confluent cell spacing, 6) hyperchromasia, 7) macronucleoli, 8) mitosis, 9) chromatin clearing, and 10) necrosis.
Diagnoses were confirmed using surgical specimens in 84 cases, cell blocks prepared from a needle rinse solution in 131 cases, and by clinical follow-up in the remaining cases. The original cytologic diagnoses were nondiagnostic in 24 cases, benign in 27, suspicious for malignancy in 15, and malignant in 225 (Table 1). Among the 225 malignant lesions, 214 cases were primary malignant tumors of the pancreas, 74 of which were classified as WDA. The most prevalent criteria for diagnosing WDA were anisonucleosis (97% of cases), nuclear membrane irregularity (97% of cases), nuclear crowding/overlapping/three-dimensionality (92% of cases), and nuclear enlargement (99% of cases). The criteria of gap versus confluent cell spacing, hyperchromasia, macronucleoli, mitosis, chromatin clearing, and necrosis were present in 38%, 36%, 14%, 22%, 14%, and 7% of WDA cases, respectively. The cytologic features of retrospectively confirmed WDA cases (originally diagnosed as WDA, suspicious, or benign) and benign lesions are summarized in Table 2. Figures 1 and 2 demonstrate the cytologic features of a representative benign lesion and a WDA, respectively. Gap versus confluent cell spacing means a distinct open space formation within a confluent sheet of epithelial cells (Fig. 2E), which differs from nuclear crowding/overlapping/three-dimensionality (Fig. 2A). There were four false-negative cases (Fig. 3). Each case met a substantial number of the cytologic criteria for WDA. Six suspicious cases also retrospectively met the criteria for WDA (all of which were malignant on clinical follow-up).
Table 2. Comparison of Cytologic Features of Confirmed Pancreatic WDAs and Benign Lesions
WDA (n = 74) (%)
Suspicious (n = 6) (%)
False negative (n = 4) (%)
Benign (n = 23) (%)
WDA: well differentiated adenocarcinoma; RBC: red blood cells.
Anisonucleosis (greater than four times)
Nuclear membrane irregularity
Nuclear enlargement (> 2 RBC)
Gap vs. confluent cell spacing
The sensitivity and specificity of FNAB for diagnosing pancreatic lesions were 98% and 100%, respectively. The calculation of the sensitivity and specificity is based on the cases available for review, which included 225 true positive cases, 4 false-negative cases, 27 true negative cases, and 0 false-positive cases. The cases reported as suspicious for malignancy were not included.
Ductal adenocarcinoma and its variants account for more than 90% of pancreatic malignancies.1 Fine-needle aspiration biopsy of pancreatic lesions has become a standard approach for establishing a diagnosis at M. D. Anderson and many other institutions.3, 9 Establishing a correct diagnosis from FNAB specimens generally presents few challenges to an experienced cytopathologist unless the tumor is a WDA, which may be extremely difficult to distinguish from reactive conditions such as pancreatitis.10
Over the past 15 years, the diagnostic sensitivity and specificity of FNAB have improved greatly as a result of several articles on the cytologic criteria for the diagnosis of pancreatic malignancy.3–8 In 1985, Mitchell and Carney4 published the first comprehensive study on diagnostic criteria. They focused on three-dimensional cellular fragments, nuclear enlargement, and nuclear membrane irregularity. Following this publication, several modified cytologic criteria based on those of Mitchell and Carney were reported. Cohen et al.3 identified anisonucleosis, nuclear molding, and large nuclei as the significant cytologic features for the diagnosis of pancreatic adenocarcinoma. Robins et al.8 were the first to distinguish major (overlapping nuclei/crowded groups, nuclear contour irregularity, and chromatin clearing and/or clumping) and minor criteria (single epithelial cells, necrosis, mitosis, and nuclear enlargement) for pancreatic adenocarcinoma. According to Robins et al., the sensitivity and specificity for diagnosing pancreatic adenocarcinoma are 100% when two or more major criteria or one major and three minor criteria are identified. Several other authors also considered single epithelial cells, necrosis, mitosis, and prominent nucleoli as significant cytologic features.5–8 Before the published criteria of Mitchell and Carney,4 Brits and Franz11 studied a small series of FNAB of the pancreas and suggested that pale nodular nuclei were a possible specific marker for pancreatic adenocarcinoma. Pale nodular nuclei were described as homogeneously hypochromatic nuclear chromatin, nuclear membrane irregularity with deep folds, and one or more large eosinophilic nucleoli. This description is similar to nuclear clearing observed in this study except that prominent nucleoli usually do not present in the nuclei with nuclear chromatin clearing. Pale nodular nuclei are not diagnostic of pancreatic adenocarcinoma because they are also present in benign pancreatic aspirate specimens as well as in the other types of malignancy, such as melanoma and lung carcinomas.1 Table 3 summarizes the cytologic criteria for pancreatic adenocarcinoma in the literature.
Table 3. Summary of Cytologic Criteria for Diagnosis of Pancreatic Carcinoma in Literature
Minor: Single epithelial cell, necrosis, mitosis, nuclear enlargement
The cytologic diagnosis of poorly differentiated (PDA) or even moderately differentiated adenocarcinoma (MDA) is usually straightforward. The problematic diagnostic cases generally involve WDA. Unfortunately, it is our impression that none of the previous studies specifically focused on WDA. The total number of cases included in each series ranged from 62 to 134 cases (Table 3). If the diagnostic categories of benign, suspicious, MDA, PDA, and metastasis are excluded, it is reasonable to assume that a limited number of WDA cases were actually included in each series.
To supplement previous studies, we took advantage of the large number of pancreatic adenocarcinoma cases available at M. D. Anderson. We collected 74 WDA cases from 225 malignant cases in which pancreatic FNAB was performed. The selection of WDA was based on an observance of cytologic smears together with available histology material (16) and cell blocks (20). One hundred ten cases of MDA and PDA, representing approximately 60% of the total primary pancreatic adenocarcinoma in this study series, were excluded from detailed evaluation. Anisonucleosis (Criteria 1), nuclear membrane irregularity (Criteria 2), nuclear crowding/overlapping/three-dimensionality (Criteria 3), and nuclear enlargement (Criteria 4) were observed in 92–99% of WDA cases versus 0–17% of the benign lesions. Six cases in the suspicious category with a subsequent histologic or clinical confirmation of adenocarcinoma met most of these criteria (67–100%). For example, anisonucleosis, nuclear membrane irregularity, nuclear crowding/overlapping/three dimensionality, and nuclear enlargement were seen in 100%, 83%, 67%, and 83% of the six cases, respectively. Scanty cellularity was the main reason why these cases remained in the suspicious category. It is noteworthy that the four false-negative cases also exhibited many of these four criteria (50–100%; Fig. 3). Figure 3A,C represents two cases showing a large sheet of ductal epithelial cells with marked anisonucleosis, nuclear crowding/overlapping/three-dimensionality, focal nuclear enlargement, nuclear membrane irregularity, and some nuclei containing small nucleoli. These features were sufficient for a diagnosis of WDA retrospectively. The other two cases with a false-negative diagnosis (Fig. 3B,D) demonstrated some of the diagnostic features, such as nuclear enlargement and anisonucleosis. Nuclear crowding/overlapping and nuclear membrane irregularities were very focal. These two cases should have been classified as suspicious for malignancy. The most likely reasons for the underdiagnoses of these cases were subtle cytologic changes and scanty cellularity.
The minimal number of cells necessary for a diagnosis of pancreatic adenocarcinoma is still being debated. Robins et al.8 proposed that at least six groups of atypical ductal epithelial cells are needed to confirm a diagnosis, which may be true in many instances. However, other factors should be taken into consideration, such as the number of cells in each group and the degree of cellular atypia. We believe that a diagnosis of WDA can be rendered in a specimen containing fewer than six groups of atypical ductal epithelial cells if cytologic criteria 1–4 are unequivocally present. Conversely, more than six groups of atypical ductal cells should be required for a definitive diagnosis if cytologic criteria 1–4 are not all present. However, due to the very few cases with a positive diagnosis that fall into this category, we are unable to provide an accurate number of groups of atypical ductal cells required for a definitive diagnosis if cytologic criteria 1–4 are not all present. As a recommendation, those cases should probably be considered as suspicious for adenocarcinoma.
It is important to recognize that anisonucleosis, nuclear crowding/overlapping, and nuclear membrane irregularity may occur focally in WDA, whereas nuclear enlargement usually involves the entire group of neoplastic ductal epithelial cells. It is also imperative to emphasize that none of these criteria, when present singly, is pathognomonic for WDA. Nuclear enlargement and focal crowding/overlapping are commonly observed in certain reactive conditions, especially pancreatitis. However, marked anisonucleosis (variation in nuclear size greater than four times in the same epithelial group) and nuclear membrane irregularities (deep notch, deep groove, popcorn, or rasinoid) are nearly always absent in reactive conditions. The combination of these two cytologic features in a benign lesion was not seen in this study or in other previous reports.3, 8 Anisonucleosis was probably first introduced by Cohen et al.3 in the cytology literature for diagnosing pancreatic adenocarcinoma, but it also served as one of the three major criteria for identifying pancreatic adenocarcinoma on the frozen section study by Hyland et al.12 We strongly suggest that marked anisonucleosis, variation in nuclear size greater than four times in the same epithelial group instead of three times proposed by Cohen et al., should be observed because this is the single most useful criterion and is somehow subjective among interobservers from our limited experience. It is our experience that the small-sized nuclei tend to present in the peripheral portion of the epithelial group if anisonucleosis is seen focally.
In contrast to cytologic criteria 1–4 (anisonucleosis, nuclear membrane irregularity, nuclear crowding, and nuclear enlargement), other cytologic features (e.g., macronucleoli, necrosis, and mitosis) were observed in only 7–22% of the WDA cases in this study. The presence of one or two mitoses does not favor a diagnosis of WDA. We encountered one case of chronic pancreatitis in which three mitotic figures were noted within a single small group of ductal epithelial cells.
Some investigators5, 8 considered that single intact malignant cells were one of the diagnostic criteria for adenocarcinoma. However, high-grade adenocarcinoma usually demonstrates many single malignant cells with obvious cytologic features for malignancy, such as marked pleomorphism, extensive necrosis, and macronucleoli. Therefore, we do not consider the presence of single malignant cells relevant in diagnosing WDA. Furthermore, the absence of macronucleoli, mitosis, and necrosis does not help to exclude the diagnosis of WDA.
Fraig et al.13 applied 10 cytologic criteria to evaluate the 33 cases of pancreatic adenocarcinoma with false-negative diagnoses. The cytologic criteria applied to their study were very similar to the criteria used in our study, including 1) loss of polarity, 2) nuclear enlargement, 3) nuclear membrane irrregularity, 4) pleomorphism, 5) pale or granular chromatin, 6) gaps between cells versus confluence, 7) increased cellularity, 8) hyperchromasia, 9) macronucleoli, and 10) necrosis. Their results indicated that loss of polarity, nuclear enlargement, and nuclear membrane irregularity were the most prevalent criteria. Hyperchromsia, macronucleoli, and necrosis were only observed in 22–27% cases.
The diagnostic sensitivity and specificity of pancreatic FNAB in the literature are reviewed and summarized in Table 4. The reported sensitivity and specificity are 50–98% and nearly 100%, respectively.4–8, 9, 14–16 In our study, the sensitivity was 98% and specificity 100%, which are comparable to the findings reported by David et al.9 and are more favorable than the findings in most previous studies. However, 35 cases were excluded from our study because the slides were not available for review. Among these 35 cases there were one false-positive case and four false-negative cases. If those cases had been included in this study, the diagnostic sensitivity and specificity probably would have been slightly lower.
Table 4. Reported Accuracy of Pancreatic FNA for Diagnosing Adenocarcinoma
Suspicious cases are usually excluded from calculations of sensitivity and specificity, as in our study. In the literature we reviewed, suspicious cases comprised 5.4–12% of evaluated cases (Table 4). The suspicious cases in our series made up 5% of the 291 cases we reviewed. Four cases in this category were not included in the calculations because the slides were not on file.
The objective of this study was to test and redefine the cytologic criteria for WDA. Our goal was to reduce the false-negative and false-positive rates, as well as the number of suspicious diagnoses. We believe that anisonucleosis (variation in nuclear size greater than four times in the same epithelial group), nuclear membrane irregularity, and nuclear crowding/overlapping/three-dimensionality, and nuclear enlargement should serve as the minimal cytologic criteria for making a diagnosis of WDA, even though these findings may be present only focally. With one exception, our study found that nuclear crowding/overlapping in WDA with mucin production was minimal. Instead, cytoplasmic mucin distends the cells and neoplastic cells have a low nucleus-to-cytoplasm ratio and form two-dimensional sheets. However, in this instance of WDA with mucin production, the neoplastic cells frequently showed marked nuclear pleomorphism and membrane irregularity. Therefore, one usually will not face a diagnostic challenge.
Our experience with this series of cases indicates that an accurate cytologic diagnosis of pancreatic WDA can be achieved when anisonucleosis, nuclear membrane irregularity, nuclear crowding/overlapping/three-dimensionality, and nuclear enlargement are present. Mitosis, macronucleoli, necrosis, hyperchromasia, and chromatin clearing are of limited significance because they are commonly absent in pancreatic WDA and are encountered more frequently in MDA and PDA. By using appropriate cytologic criteria in pancreatic FNAB, we believe that the false-negative and suspicious categorizations can be reduced and the false-positive cases can be avoided.