Quantitative examination of mechanophysical tumor cell enrichment in fine-needle aspiration specimens


  • Linda M. Ernst M.D.,

    1. Department of Pathology, Yale University School of Medicine, New Haven, Connecticut
    Search for more papers by this author
  • David L. Rimm M.D., Ph.D.

    Corresponding author
    1. Department of Pathology, Yale University School of Medicine, New Haven, Connecticut
    • Department of Pathology, Yale University School of Medicine, 310 Cedar Street, BML-122, New Haven, CT 06520
    Search for more papers by this author
    • Fax: (203) 737-5089



The advent of cDNA microarrays and other molecular technologies necessitates the acquisition of tumor cell-enriched material because nonmalignant cells often decrease the sensitivity of the assays. Fine-needle aspiration (FNA) specimens from carcinoma have long been noted to be enriched in malignant cells. The current study quantitated the relative representation of tumor versus nontumor cells in FNA specimens compared with tissue sections using breast carcinoma as a model.


Five patients who had undergone both a diagnostic FNA and a surgical excision for breast carcinoma between January and July of 1996 were selected. Five random cellular fields from representative slides of the FNA (using the ThinPrep preparation of the wash) and surgical excision specimens were photographed digitally at ×20 power. The cells were judged as tumor or nontumor cells and then were counted manually in each field.


The calculated percentage of malignant cells in the FNA specimen (as represented on the ThinPrep slide) ranged from 66–93% compared with the calculated percentage of 37–78% noted in histologic sections. The average of all 5 fields from all 5 cases showed that 83.1% of the total cells were malignant in the ThinPrep preparation compared with 62.3% in the histologic sections. This difference was highly statistically significant when analyzed using the chi-square test (P = 0.0009).


The percentage of malignant cells on FNA specimens from breast carcinoma, as assessed by ThinPrep, was found to be significantly higher than that obtained by surgical excision. The results of the current study quantitatively confirm the impression of practicing cytopathologists, but also suggest that FNA will provide a good substrate for cDNA microarray and other molecular analyses. Cancer (Cancer Cytopathol) 2002;96:275–9. © 2002 American Cancer Society.

Fine-needle aspiration (FNA) is a long-standing and effective method with which to obtain samples of malignant tumors and is widely used around the world. Its low cost and high effectiveness have resulted in a progressive and continuing increase in its utilization.1 Another purported advantage of the technique is that FNA specimens from carcinomas are enriched in malignant cells compared with surgical biopsy material. This concept, although illustrated in a leading text in the discipline,2 is to our knowledge unsupported in the literature. In fact, a hypothetical graph was constructed for the illustration of the concept in the work by DeMay.2 The reasons for the lack of supportive data for this concept are unknown. We speculate that there has been no real need to prove that FNA specimens are enriched in tumor samples because FNA is a generally accepted method for obtaining diagnostic material. Furthermore, the quantitative assessment of tumor cells and benign cells on FNA and biopsy material is a tedious undertaking, one that is not amenable to image analysis or other computer-aided techniques.

Tumor cell enrichment is helpful for diagnostic cytology, but may be crucial for the success of some of the new molecular techniques. Specifically, cDNA microarray results are dependent on high-quality RNA that is as homogeneous as possible.3 Although current microarray technology generally requires more RNA that would be collected by a routine FNA, efforts concerning representative amplification are underway and it is likely that techniques soon will be available to allow microarray-based analysis of the quantities of material obtainable by FNA. A potentially more difficult problem is the heterogeneity of the material collected. The collection of a significant number of nonneoplastic cells with malignant cells in diagnostic procedures may limit the ability to detect molecular/genetic abnormalities specific for malignancy.3 Therefore, the current study was undertaken to test the hypothesis that there is a significant difference between the fraction of malignant cells obtained by FNA versus those obtained by surgical resection.

Because the breast is the most common site for FNA,1 breast carcinoma was used as the model for the current study. Furthermore, recent studies using cDNA microarray techniques have studied breast carcinoma.4–7 These studies have used breast carcinoma cell lines for the most part; however, a more recent study used clinical samples obtained from surgical specimens.8 The use of patient samples in these research protocols underscores the importance of these new technologies in the future treatment of patients with breast carcinoma. As laboratory and clinical medicine continue to develop molecular diagnostic testing, the use of smaller patient samples to obtain meaningful information will be desired. The data from the current study demonstrate that small breast FNA specimens are relatively enriched in malignant cells and, therefore, are a potentially useful source of material for molecular analyses.


Patient Selection

Patients who underwent both a diagnostic FNA and surgical excision for breast carcinoma between January 1996 and July 1996 were selected from the Yale-New Haven Hospital Pathology files. The first five patients whose FNA was accompanied by a good quality, ThinPrep®-prepared (Cytyc Corporation, Boxborough, MA), needle wash specimen were selected. The ages of these patients ranged from 47–61 years with a mean age of 53 years. Four of the patients had infiltrating ductal carcinoma of the breast and one patient had medullary carcinoma of the breast. The tumors all were at least three cm in greatest dimension. Four patients underwent axillary lymph node dissections, only one of whom was found to have lymph node metastases.

Preparation and Selection of Cytologic and Histologic Material

Breast FNA was performed in the traditional manner for a palpable mass in each case. One drop of the aspirated material was prepared as a smear and the remainder of the specimen was collected into Cytolyte™ (Cytyc Corporation, Boxborough, MA). ThinPrep slides were made using the Cytyc 2000 ThinPrep Processor and were stained with a conventional Papanicolaou stain. Although smears also were diagnostic on each of these cases, the wash specimen from the ThinPrep sample was used for analysis in the current study because it would be the material used for the preparation of nucleic acids or other ancillary techniques.

All histologic specimens were drawn from slides prepared during routine processing for diagnosis. Representative samples of the carcinoma within the breast tissue of the surgical resection specimen were fixed in 10% formalin, dehydrated, cleared, and embedded in paraffin. Five-micron thick sections were cut from the paraffin blocks and stained with hematoxylin and eosin. A representative slide of the carcinoma was selected for each patient.

Digital Photography, Cell Identification, and Categorization

Five random, cellular ×20 microscopic fields per slide (ThinPrep and histologic section) were captured with a color digital camera (Kontron Elektronik Prog/Res/3012 camera, Olympus Vanox AHBS3 microscope, and PC-based Autocyte Image Manager [Version 2.2] [Roche Imaging, Elon College, NC]) and saved in an Adobe Photoshop file. Each cell in every image was examined and judged by a pathologist (L.M.E.) to be either a malignant or benign cell based on features such as architectural arrangement of the cells, cell size, nuclear:cytoplasmic ratio, nuclear irregularity, and nuclear chromatin pattern. For the purpose of tracking and quantitation, a colored dot reflecting the categorization of each cell as benign (yellow dot) or malignant (green dot) was superimposed on the nucleus of each cell. The colored dots then were counted and recorded. A second pathologist (D.L.R.) reviewed representative “dotted” images for the accuracy and completeness of the quantitative analysis.

Statistical Analysis

Statistical analyses were performed using the Statview 4.5 software package (SAS Institute, Cary, NC) for Macintosh.


For FNA specimens, the total number of cells present in the 5 designated fields ranged from 305–1192 cells (mean, 572.2 cells). For surgical biopsy specimens, this number ranged from 3071–6049 cells (mean, 4525 cells). An example of the results of the cytologic interpretation and quantitative method using original and dotted images for one patient are shown in Figure 1.

Figure 1.

Sample photographs of the (A and C) surgical resections and (B and D) ThinPrep slides of the fine-needle aspiration specimens from one of the five patients in the current study. Malignant cells (green dots) and benign cells (yellow dots) are designated. Insets: higher power views of each field.

With the exception of the medullary carcinoma case (Patient 5), both the FNA and surgical biopsy specimens contained more malignant cells than nonmalignant cells. In surgical specimens, a total of 2.0 malignant cells (range, 0.6–3.5 malignant cells) were present for every benign cell compared with FNA specimens, in which the ratio was 6.7 malignant cells (range, 1.9–12.7 malignant cells) for every benign cell. When individual patient results were analyzed, there was a statistically significantly higher percentage of malignant cells in the FNA specimen in three of the five cases (Fig. 2). The material from the remaining two patients (Patients 1 and 2) demonstrated a higher percentage of malignant cells in the FNA sample that did not quite reach statistical significance. Table 1 shows the contingency tables used for the analysis of each patient.

Figure 2.

Mean percentage ± the standard error of benign cells (gray checkerboard bars) and malignant cells (black bars) in (top panel) fine-needle aspiration/ThinPrep (THIN PRE) and (bottom panel) surgical resection specimens in five individual patients. The last column shows the mean percentage of all five patients.

Table 1. Contigency Tables Used for the Analysis of Each Patient
 Patient 1Patient 2Patient 3Patient 4Patient 5
  1. Bx: biopsy; FNA: fine-needle aspiration.

% malignant69.480.977.988.570.592.756.687.636.965.7
% benign30.619.122.111.529.57.343.412.463.134.3
P valueP = 0.050P = 0.0598P < 0.0001P < 0.0001P < 0.0001

The total percentage of malignant cells in all patients ranged from 65.7–92.7% on FNA (mean, 83% ± 7%) and from 36.9–77.9% on surgical excision material (mean, 62.3% ± 5%). Overall, the FNA specimen contained a significantly higher mean percentage of malignant cells than the biopsy material (P = 0.0009). Because medullary carcinoma is so rich in nonepithelial cells, this analysis also was performed for the infiltrating ductal carcinomas only and again the FNA contained a significantly higher percentage of malignant cells than the biopsy material (P = 0.0021).


Similar to many tumors, tissue from breast carcinomas contains a heterogeneous combination of benign and malignant cell types. The benign cells, including adipocytes, fibroblasts, endothelial cells, and inflammatory cells, easily are distinguished from the malignant epithelial cells by histologic examination. In molecular studies of gene expression, it is necessary to distinguish the gene expression patterns of benign cells from the patterns of malignant cells. This often is accomplished by comparing the expression of a mix of cell lines with a test set of mRNA made from tumor tissue.8 The mRNA from the benign cells in the mix may dilute or dramatically skew the result because the differences detected also may reflect differences in amount or cell type of the benign component. Thus is it desirable to select tissue rich in malignant cells. Some investigators currently are testing microdissection techniques to achieve this goal.

The FNA technique is a combination of suctioning of the tissue into the needle tip and a cutting or scraping motion with the needle.9 During this procedure, it is the general belief that epithelial tumor cells are dislodged more easily during FNA, leaving connective tissue behind.2 In fact, one basic tenet of FNA cytology is that no diagnosis of malignancy should be made in the absence of single tumor cells.10 The common finding of numerous single malignant cells and small clusters of carcinoma cells may be based biologically on the down-regulation of adhesion that is associated frequently with invasive malignancy.11 In contrast, the dense sclerotic stroma surrounding a carcinoma has a more cohesive and organized structure.

The common abrogation of adhesion in malignancy, along with the mechanophysical properties of the FNA procedure, suggest that FNA specimens should be disproportionately rich in malignant cells. The data in the current study support this hypothesis, demonstrating that the FNA does indeed contain a statistically significantly higher percentage of malignant cells. Therefore, these data are consistent with the long-accepted but, to our knowledge, unsupported hypothesis that FNA obtains a higher percentage of malignant cells compared with routine surgical biopsy.

The current study was restricted to the study of breast carcinoma and therefore cannot necessarily be generalized to all carcinomas. It is interesting to note that a patient with medullary carcinoma of breast was included in the current study (Patient 5), and the surgical material from this particular case was the only instance in which a higher percentage of benign cells (mostly lymphocytes) were present than malignant cells. However, the FNA sample from this same patient contained a higher percentage of malignant cells than benign cells. Because this case was unusual in this regard, statistical analyses were performed with and without this case included and the results did not appear to differ significantly.

Although not surprising to many cytopathologists, the findings of the current study may be of importance for future molecular studies of breast carcinoma. To our knowledge to date, cDNA microarrays of breast carcinoma have relied on tumor cell lines and surgical resection specimens. This has been a result of the requirement for large quantities of RNA for each assay. However, as linear amplification methods improve, it may become advantageous to use material obtained by FNA. FNA is performed more easily with less patient morbidity and mortality because no general surgery is involved. Thus, larger cohorts of patients may be examined, including cases in which tumors are treated with adjuvant therapy. In these cases, expression pattern changes may be examined both before and after therapy.

The data from the current study demonstrate quantitatively that FNA specimens of breast carcinomas are enriched significantly with regard to the percentage of tumor cells when compared with surgical resection specimens. The recovery of a higher fraction of malignant cells may provide a superior specimen for advanced molecular diagnostic studies.