Context.—Differentiation of conventional renal cell carcinoma from other malignant neoplasms can be difficult, particularly with the limited sampling in fine-needle aspiration biopsies. Many studies have discussed the features of renal cell carcinoma in fine-needle aspiration biopsy specimens, but the significance of the cytologic features is not known.
Objective.—To define the significant cytologic findings that aid in differentiating renal cell carcinoma from other malignant neoplasms in fine-needle aspiration biopsy specimens.
Design.—Fine-needle aspiration biopsies from 35 patients with proven primary or metastatic conventional renal cell carcinoma and from 145 patients with proven primary or metastatic non–renal cell carcinoma malignant neoplasms were assessed for the presence or absence of the following cytologic features: the heterogeneous cell population, hemosiderin deposits, small cytoplasmic vacuoles, large cytoplasmic vacuoles, low nuclear-cytoplasmic ratio, prominent nucleoli, intranuclear inclusions, irregular nuclear membranes, and smooth nuclear membranes. Statistical analysis was performed to identify the features significant in distinguishing conventional renal cell carcinoma from non–renal cell carcinoma malignancies.
Results.—The presence of the heterogeneous cell population, hemosiderin deposits, small cytoplasmic vacuoles, and low nuclear-cytoplasmic ratio were each highly significant in conventional renal cell carcinoma when compared with non–renal cell carcinoma malignant neoplasms (P < .001) using univariate exact analysis. Features that were identified as being predictive of conventional renal cell carcinoma using multivariable logistic regression analysis included heterogeneous cell population, small cytoplasmic vacuoles, and hemosiderin deposits (P < .05).
Conclusions.—The presence of the heterogeneous cell population, small cytoplasmic vacuoles, hemosiderin deposits, and a low nuclear-cytoplasmic ratio aids in differentiating conventional renal cell carcinoma from other malignant neoplasms, including morphologically similar entities such as hepatocellular carcinoma, in fine-needle aspiration biopsy specimens.
Renal cell carcinoma (RCC) accounts for approximately 2% of all visceral cancers, with conventional renal cell carcinoma (cRCC) comprising approximately 75% of renal cancers in adults.1 Men are affected twice as frequently as women,2 and there is a striking predilection for metastases at various and unusual sites, including breast,3 thyroid,4,5 adrenal,6 parotid,7 and pancreas,8 with up to one third of patients having metastases at the time of initial diagnosis.2 As newer therapies emerge, and as smaller and asymptomatic masses are increasingly detected9 by improved diagnostic imaging techniques, fine-needle aspiration biopsy (FNAB) has assumed an increasingly important role in the diagnosis of primary and metastatic tumors.
Variation in the preservation of RCC features in FNAB specimens may be marked, leading to difficulty in its differentiation from other primary and metastatic tumors. In the literature, several studies have commented on the various cytologic features observed in RCC.10–15 The significance of these findings, however, has not been fully delineated. This study was undertaken to examine the cytologic features of cRCC, compare them with findings in other primary and metastatic malignant neoplasms, and define those features significant in establishing the differentiation between cRCC and non-RCC malignancies.
MATERIALS AND METHODS
Cellular FNAB smears from 35 patients with cRCC and from 145 patients with non-RCC malignant neoplasms were randomly selected for review. All cases were histologically confirmed by cell block, core biopsy, or surgical resection specimen. The FNABs were performed under radiologic guidance by radiologists or under superficial palpation by cytopathologists using 18- to 20-gauge and 23- to 25-gauge needles, respectively. The aspirated material was smeared onto glass slides; either air-dried or fixed in 95% ethyl alcohol; and stained by the Diff-Quik or standard Papanicolaou methods, respectively.
The 35 cases of cRCC included 9 primary tumors and 26 metastatic tumors. The 145 cases of non-RCC malignant neoplasms included 25 cases of hepatocellular carcinoma and 61 cases of adenocarcinoma, including 11 from the lung and 10 each from the breast (ductal type), gastrointestinal tract, pancreas, thyroid (papillary type), and prostate. Also included were 10 cases each of islet cell tumor, melanoma, transitional cell carcinoma, and squamous carcinoma; 11 cases of leiomyosarcoma; and 8 cases of adrenocortical carcinoma (Table 1). All cases were retrospectively reviewed in a blind fashion by 2 pathologists and evaluated for the presence or absence of 9 cytologic features. The features studied included the heterogeneous cell population, defined as the presence of 3 separate components: (1) epithelioid tumor cells, (2) endothelial cells of relatively thick and tortuous vascular structures, and (3) an admixture of neutrophils and lymphocytes intimately associated with either or both of the former components. This feature was considered positive when all 3 components were present concurrently in a single cellular cluster or dispersed separately in the smear (ie, endothelial-lined, thick, tortuous vessels present in one area of the smear and epithelioid tumor cells with admixed inflammatory cells in another area of the smear). Other features studied included prominent nucleoli, smooth nuclear membrane, irregular nuclear membrane, low nuclear-cytoplasmic (N/C) ratio (≤1:2), small (frothy) cytoplasmic vacuoles, large (distinct) cytoplasmic vacuoles, intranuclear inclusions, and hemosiderin deposits. The percentage of cases showing each feature was calculated for each tumor type and for the combined category of non-RCC tumors.
Statistical analysis, including a logistic regression model, was done on the resulting data (with likelihood ratio P values reported) to identify the features significant in distinguishing cRCC from non-RCC malignant neoplasms.
Table 2 lists the cytologic features considered in the statistical analysis. It begins with the most commonly found cytologic feature of cRCC and continues through the least common of the 9 variables studied. Table 2 also lists the occurrence in percentage of these cytologic features in non-RCC tumors considered both singly and as a group.
Low N/C Ratio
Tumor cells having a low N/C ratio were seen in 97% of cRCC cases and in 10% of non-RCC cases in this study. This criterion was not present in the following categories of non-RCC tumors: squamous carcinoma; transitional cell carcinoma; leiomyosarcoma; islet cell tumor; and adenocarcinomas of the lung, breast, prostate, and pancreas. Forty-four percent of hepatocellular carcinomas; 13% of adrenocortical carcinomas; and 10% each of papillary thyroid carcinomas, gastrointestinal adenocarcinomas, and melanomas showed this feature.
Smooth Nuclear Membrane
A smooth nuclear membrane was observed in 97% of cRCC cases and in 81% of non-RCC cases in this study. Only 10% of papillary thyroid carcinomas showed this feature. The positive percentages ranged from 50% to 100% for all other non-RCC tumors.
The Heterogeneous Cell Population
The heterogeneous cell population, as previously defined, was present in 83% of cRCCs and in 2% of non-RCC tumors in this study. This finding was seen in 10% of papillary thyroid carcinoma cases and in 8% of hepatocellular carcinoma cases. This feature was observed in smears having moderate cellularity, defined as at least 1 cellular cluster of 50 or more cells. Although this feature could be seen on the Diff-Quik smears, it was more readily observed on the alcohol-fixed Papanicolaou-stained preparations. Some of the cases of hepatocellular carcinoma, adrenocortical carcinoma, papillary thyroid carcinoma, and pancreatic islet cell tumors contained fine, thin capillary structures. However, in contrast to the thick, tortuous/multidirectional nature of the vessels defined in the heterogeneous cell population (Figure 1), these vascular structures were delicate and unidirectional (Figure 2). Additionally, although some non-RCC tumors showed the presence of admixed neutrophils and lymphocytes, none was accompanied by the presence of the aforementioned vascular structures. It should be noted that the presence of inflammatory cells was considered positive only when their presence was intimately admixed with tumor cells or vascular structures with an absence of significant background inflammatory cells (ie, peripheral blood contamination from the FNAB procedure itself) (Figures 3 and 4).
The presence of prominent nucleoli was seen in 83% of cRCC and in 43% of non-RCC tumors. This feature was present in all non-RCC categories (range, 9%–90%) except papillary thyroid carcinoma.
Small Cytoplasmic Vacuoles
Small, ill-defined (frothy) cytoplasmic vacuoles were seen in 74% of cRCC cases and in 7% of non-RCC cases. This feature was absent in all categories of non-RCC cases except 40% of hepatocellular carcinomas and 13% of adrenocortical carcinomas.
Hemosiderin deposits were observed in 60% of the cRCC tumors and in 11% of non-RCC malignant neoplasms in this study. Non-RCC categories showing this feature included papillary thyroid carcinoma (10%), transitional cell carcinoma (10%), hepatocellular carcinoma (16%), squamous carcinoma (20%), leiomyosarcoma (36%), and lung adenocarcinoma (36%). Morphologically, this feature was identified both intracellularly and extracellularly in close association with tumor cells and was more readily observed on the Papanicolaou-stained smears.
Large Cytoplasmic Vacuoles
Large, distinct cytoplasmic vacuoles were present in 6% of cRCC cases and in 13% of non-RCC cases, including 13% of adrenocortical carcinomas; 16% of hepatocellular carcinomas; and 36%, 60%, and 70% of lung, gastrointestinal, and pancreatic adenocarcinomas, respectively. Morphologically, this feature was as readily observable on the Diff-Quik preparations as on the Papanicolaou-stained smears.
Irregular Nuclear Membrane
Irregular nuclear contours were present in 6% of cRCC cases and in 33% of non-RCC cases, the majority of which were represented by the squamous carcinoma (90%), transitional cell carcinoma (80%), papillary thyroid carcinoma (80%), and pancreatic adenocarcinoma (70%) categories. This feature was more readily appreciated on the Papanicolaou-stained smears.
Intranuclear Cytoplasmic Inclusions
Intranuclear cytoplasmic inclusions were seen in only 3% of cRCC cases, as compared with 32% of non-RCC cases. In the non-RCC group, 90% of papillary thyroid carcinomas, 84% of hepatocellular carcinomas, 70% of melanomas, 46% of leiomyosarcomas, 25% of adrenocortical carcinomas, 20% of breast adenocarcinomas, and 9% of lung adenocarcinomas showed this feature. The inclusions had sharply delineated borders and occupied at least 10% of the nuclear diameter. This feature was appreciated on both the Diff-Quik and Papanicolaou-stained preparations.
Statistical Analysis and Probability of cRCC
Univariate analysis using exact methods showed the heterogeneous cell population, hemosiderin deposits, small cytoplasmic vacuoles, and low N/C ratio to be highly significantly associated with cRCC (P < .001). Using a significance threshold of P < .005, 3 additional features including nucleoli, nuclear inclusions, and irregular nuclear membranes would also have been significant. Only large cytoplasmic vacuoles (P = .22) and smooth nuclear membranes (P = .02) were found not to have a statistically significant association with cRCC.
Multivariable logistic regression analysis resulted in the selection of the heterogeneous cell population (P < .001), hemosiderin deposits (P = .02), and small cytoplasmic vacuoles (P < .001) as a set of independent predictive factors in distinguishing cRCC from non-RCC malignant neoplasms. Table 3 lists the probabilities of the presence of cRCC for the various combinations of these 3 FNAB variables.
Of the 35 patients with proven cRCC, the logistic model correctly predicted 89% as having this malignancy. Of the 145 cases that were not cRCC, 97% were correctly predicted as such.
Pathologists, and in particular cytopathologists, are often faced with the difficult task of differentiating primary and metastatic RCC from other malignant neoplasms. This task may prove particularly challenging in cancer center settings, where patients sometimes present with multiple primaries. The problem is compounded by the predilection of RCC to metastasize to various and sometimes unusual sites. Additionally, unlike some tumors, such as hepatocellular carcinoma, there are no particular biomarkers (eg, serum α-fetoprotein levels) that are helpful in the diagnosis of RCC.
Only one third of patients with RCC will present with the classic triad of flank mass, flank pain, and hematuria, and metastatic disease may be the initial presentation in up to one third of patients.2 The preservation of RCC features in FNAB specimens may vary markedly, leading to difficulty in its differentiation from other primary and metastatic malignant neoplasms.
The majority of the cytologic features examined in this study have been observed and reported on FNAB specimens of RCC.10–15 To our knowledge, these previous reports have not analyzed the significance of these features or defined any key criteria that may be helpful in distinguishing cRCC from other, non-RCC malignancies. The present study describes a unique feature, called the heterogeneous cell population, and shows that this feature, when accompanied by the presence of small cytoplasmic vacuoles and hemosiderin deposits, may be highly predictive of cRCC.
The unique characteristic referred to herein as the heterogeneous cell population deserves mention, because it was found in this study to be one of the most specific features of cRCC. Some studies have commented on the presence of vascular structures in other tumors such as hepatocellular carcinoma,14,16 islet cell tumor,17 and adrenocortical carcinoma.13 We found the vessels in these aforementioned tumors to be more commonly thin, delicate, and unidirectional in nature, rather than the thick, tortuous, and multidirectional vascular structures defined as part of the heterogeneous cell population. Additionally, the inflammatory component of the heterogeneous cell population that is observed intermingling in close association with the epithelioid tumor cells, vascular structures, or both is often lacking in vascular non-RCC malignant neoplasms. It is possible that inflammatory cells may be observed as a part of peripheral blood pronounced in the background of aspirates from vascular lesions. It should be emphasized, however, that only when these cells were present intermingled in close association with the vessels (sometimes seen within vessels) or epithelioid tumor cells and in relatively greater concentration than cells in the background were they considered as fulfilling part of the criteria for the heterogeneous cell population.
Additionally, in this study, the heterogeneous cell population was more readily observed in smears containing at least 1 group of 50 cells. In less-cellular specimens, it is possible that this feature may not be as readily observed. The criteria for the heterogeneous cell population, however, do not necessitate that its various components be present concurrently in a given cell group. In this study, its presence was noted as positive when the 3 components were present in separate areas of the smear or concurrently in a given cell cluster. It should be noted, however, that the inflammatory component was always present in close association with either the epithelioid tumor cells, the thick vascular structures, or both.
In our data set, 5 of the cytologic features studied were not helpful in distinguishing RCC from non-RCC malignancies when using a significance threshold of P < .001 in univariate analysis: nucleoli, intranuclear inclusions, and irregular nuclear membranes were significant only at the .005 level, and smooth nuclear membranes and large cytoplasmic vacuoles had P values greater than .01. A note of caution should be added, however, because the number of cases in each specific group was small.
The stepwise logistic regression analysis in this study showed the presence of the heterogeneous cell population, hemosiderin deposits, and small cytoplasmic vacuoles to be predictive of cRCC when compared with non-RCC malignancies. These features were found to aid in distinguishing primary and metastatic cRCC from other primary and metastatic malignant neoplasms on FNAB specimens and may be particularly helpful in cases where immunohistochemical analysis, which may often aid in differentiating cRCC from other non-RCC malignant neoplasms, is not useful because of inadequate cell block material.
We thank Ken Hess, PhD, for his assistance in statistical analysis.
Presented in part at the annual meeting of the United States and Canadian Academy of Pathology, Atlanta, Ga, March 2001.
The authors have no relevant financial interest in the products or companies described in this article.
Reprints: Z. Laura Tabatabai, MD, Department of Anatomic Pathology, 113B, UCSF/SFVAMC, 4150 Clement St, San Francisco, CA 94121 (email@example.com)