Role of DNA Flow Cytometryand Image Cytometry onEffusion FluidIndranil Saha, M.B.B.S.,1 Pranab Dey, M.D., M.I.A.C.,2* Harpreet Vhora, Ph.D.,3and Raje Nijhawan, M.D.2

The objective of the study was to assess the value of DNA flowcytometry (FCM) and image cytometry (ICM) as an adjunct toroutine diagnostic cytology. In this prospective study, 100 consecu-tive effusion fluids were studied for routine cytology, DNA FCM,and in selected cases, ICM. One half of the centrifuged fluid samplewas used for routine cytology and the remaining portion was usedfor DNA FCM. Nuclear area, nuclear diameter, nuclear perimeter,nuclear convex perimeter, nuclear roundess, and nuclear convexarea were measured on at least 100 cells by ICM in cytologicallymalignant or DNA aneuploid cases along with control cases.Clinical follow-up was done in all cases. There were 22 cytologi-cally malignant cases and 78 cytologically benign cases. Amongthe 22 cytologically malignant cases, there were 11 aneuploid anddiploid cases each by DNA FCM. Out of 78 cytologically benigncases, six (7.7%) were aneuploid by DNA FCM. Smears of thesecases showed predominantly reactive mesothelial cells, but theDNA histograms showed hypodiploid (one), hyperdiploid (three),tetraploid (one), and hypertetraploid (one) aneuploidy. Follow-upof these cases showed clinical or histologic features of malignancyexcept in one case of tetraploid aneuploidy, which did not show anyfeatures of malignancy and responded well to antituberculartherapy. Therefore, out of 27 malignant effusions, DNA FCMpicked up 16 cases and routine cytology detected 22 cases.Sensitivity and specificity of DNA FCM were thus 59.25% and98.63%, respectively. There was a statistically significant differ-ence (Student’s unpaired t-test, P, 0.05) between cytologicallymalignant cases and control benign cases in all the nuclearmorphometric parameters except for nuclear roundness. Therewas, however, no statistically significant difference of nuclearmorphometric parameters between cytologically benign vs. DNAaneuploid cases and control benign cases. DNA FCM is a usefuladjunct for routine diagnostic cytology. Visual diagnostic cytologyand morphometric digital microscopy miss some cases of malig-nancy which can be detected by DNA flow cytometry.Diagn.Cytopathol. 2000;22:81–85.r 2000 Wiley-Liss, Inc.

Key Words: DNA flow cytometry; image cytometry; effusion;cytology

Effusion in one or more of the body cavities is an importantsign of disease, which requires rapid diagnosis and classifi-cation. Cytologic examination of the sediment of effusionfluid has been the method of choice in this regard for manyyears. However, problems arise frequently, particularly inthe differentiation between reactive mesothelial cells andadenocarcinoma cells.1 Thus, there is a need for improved orcomplementary methods to identify malignant cells withinthe body cavity fluids. A variety of experimental techniqueshave been applied to effusions to improve the diagnosticyield; among them, DNA flow cytometry (FCM),2 imagemorphometry (ICM),3,4 and immunocytochemistry5 are im-portant. Sensitivity and specificity of DNA FCM in effusionfluids varied widely in different studies.2,6,7,8There are veryfew studies on the role of automated ICM in detection ofmalignancy in effusion fluids. There is no study combiningthese techniques with light microscopic findings to show anyincrease of diagnostic accuracy in case of effusion fluids. Inthis study, DNA FCM and ICM were performed on effusionfluids to assess their diagnostic role as an adjunct to lightmicroscopy.

Materials and MethodsThis was a prospective study consisting of 100 effusion fluidcases. The samples included 33 pleural fluids and 67peritoneal fluids. The samples studied were obtained fromthe fresh effusion fluid specimens submitted to the Depart-ment of Cytopathology, Postgraduate Institute of MedicalEducation and Research, Chandigarh, India, for routineconventional diagnostic cytology during 1995–1996.

Fresh fluid was collected in ammonium oxalate anticoagu-lant and was centrifuged at 1,500 revolutions per min, andthe supernatant was discarded. One portion of the depositwas used for routine Papanicolaou and May-Gru¨nwald-Giemsa stain. Another portion of the cell pellet was used forDNA FCM. Image morphometry was done on the Papanico-

1Department of Pathology, Post Graduate Institute of Medical Educationand Research (PGIMER), Chandigarh, India

2Department of Cytology, PGIMER, Chandigarh, India3Department of Experimental Medicine and Biotechnology, PGIMER,

Chandigarh, India*Correspondence to: Pranab Dey, M.D., M.I.A.C., Department of

Cytology, PGIMER, Chandigarh, India. E-mail: pranab@ch1.dot.net.inReceived 18 May 1999; Accepted 31 August 1999

r 2000 WILEY-LISS, INC. Diagnostic Cytopathology, Vol 22, No 2 81

laou-stained smears in selected cases. Cytological diagnosis,DNA flow cytometric results, and image morphometricresults were correlated and analyzed statistically.

DNA Flow CytometryPreparation and staining.The effusion fluid was centri-

fuged, and the cell pellet was washed twice with phosphate-buffered saline. A single-cell suspension was made byrepeatedly passing the sample through a nylon mesh of 50µm pore size. The cell concentration was adjusted to 13 106

cells per ml. One milliliter of cell suspension was stained byadding 1 ml of Krishan’s solution (containing 0.0125 gpropidium iodide (Sigma Chemical Co., St. Louis, MO),0.005 g RNAse (Sigma), 0.75 ml NP-40 (Sigma), and 0.25 gsodium citrate in 250 ml of water, pH 7.6), and kept in darkfor 30 min by covering with aluminum foil at 4°C in arefrigerator.9 Peripheral blood lymphocytes from healthypersons were processed in the same way and were used ineach case as a diploid external reference control. Thefluorochrome-stained cells were analyzed in the DNA flowcytometer (Becton Dickinson flow cytometer, cell questprogram, Becton Dickinson, San Jose, CA) at a rate of 100cells per sec in a gated population of at least 10,000 cells.

Only samples with a coefficient of variation (CV) lessthan 6% were included in this study. Average CV of thediploid peak of these 100 samples was 3.25. DNA aneu-ploidy was considered when there was an additional G0–G1

peak other than normal diploid.The DNA index (DI) was calculated from the following

equation:

DI 5Mean channel number of aneuploid G0/G1

Mean channel number of diploid G0/G1.

Image morphometry, selection of cases.Image morphom-etry was done in selected cases where there was DNA

aneuploidy and/or positivity for malignancy by cytology.The cases which were cytologically benign, diploid by DNAflow cytometry, and without clinical history of malignancywere used as controls.

Machine and software.A Leica image cytometer (Leica,Cambridge, UK) with the help of Quantimet 600 softwaresupplied by the company was used in this study to measurenuclear area, nuclear diameter, nuclear perimeter, nuclearconvex perimeter, nuclear roundness, and nuclear convexarea. At least 100 cells were studied in each case.

ResultsThere were 22 cytologically malignant cases and 78 cytologi-cally benign cases. Among the 22 cytologically malignantcases, there were 11 aneuploid and diploid cases each byDNA FCM. Out of 78 cytologically benign cases, six (7.7%)were aneuploid by DNA FCM. Various cytologic diagnosesoffered in malignant effusions were adenocarcinoma (11),lymphoma (5), breast carcinoma (1), squamous-cell carci-noma (1), sarcoma (1), leukemia (1), and positive formalignancy (2). Primary sites of the tumors included ovary(10), lung (1), breast (3), lymphoreticular system (6), cervix(1), and foot (1). Table I shows details of clinical and DNAFCM features of cytologically malignant and DNA aneu-ploid cases. Out of 22 cytologically malignant cases, 11cases (50%) showed DNA aneuploidy.

In 11 cases, cytologic diagnoses were malignant but theDNA histogram showed diploidy. Table II shows DNA flowcytometric features and clinical data of cytologically benignand DNA aneuploid cases. There were in total 6 aneuploidcases out of the 78 cytologically benign cases. Smears ofthese cases showed predominantly reactive mesothelialcells, but the DNA histogram showed hypodiploid (one),hyperdiploid (three) (Fig. 1), tetraploid (one), and hypertet-raploid (one) aneuploidy. Follow-up of all these casesshowed clinical or histological features of malignancyexcept in one case of tetraploid aneuploidy, which did notshow any features of malignancy and responded well to

Table I. Detailed Clinical and DNA Flow Cytometric Features of Cytologically Malignant and DNA Aneuploid Cases

Age/sexSite of

effusion Primary site of malignancy Cytological diagnosisDNAindex Aneuploidy

45/F Ascitic fluid Ovary Adenocarcinoma 0.72 Hypodiploidy60/M Pleural fluid Lung Adenocarcinoma 1.28 Hyperdiploidy50/F Ascitic fluid Ovary (adenocarcinoma) Positive for malignant cells 0.59 Hypodiploidy35/M Pleural fluid Lymphoid cells Immature lymphoid cells, NHL 1.78 Hyperdiploidy55/F Ascitic fluid Ovary with peritoneal metastasis Positive for malignancy, carcinoma 3.04 Hypertetraploidy6/M Pleural fluid High-grade (immunoblastic) NHL Consistent with NHL 1.31 Hyperdiploidy

45/F Pleural fluid Breast carcinoma, right pleuraleffusion

Positive for malignancy 1.19 Hyperdiploidy

45/F Ascitic fluid Malignant ovarian tumor Positive for malignancy 2.05 Tetraploid aneuploidy65/F Ascitic fluid Ovarian tumor Adenocarcinoma 0.64 Hypodiploidy45/M Ascitic fluid Soft-tissue sarcoma of the thigh Positive for malignancy 0.72 Hypodiploidy60/F Pleural fluid Carcinoma cervix Squamous-cell carcinoma 2.07 Tetraploid aneuploidy

NHL, non-Hodgkin’s lymphoma.

SAHA ET AL.

82 Diagnostic Cytopathology, Vol 22, No 2

antitubercular therapy. It was difficult to measure S-phasecells in DNA aneuploid cases. The 11 cytologically malig-nant and DNA diploid cases showed S1 G2M percentage ofcells between 0.72–11%, and the mean S phase was 4.79%.Two such cases showed more than a 10% population of S1G2M cells. The cytologically benign and DNA diploid cases(72 cases) showed S-phase cells between 0.21–8.6% with amean S phase of 2.37%. None of the cases showed a highS 1 G2M percentage of cells (i.e.,.10%). Out of the 78cytologically benign cases, DNA FCM picked up fivemalignant cases (6.4%). The rest of the 72 cases werecytologically benign, and DNA histograms showed diploidywith a DNA index of one.

Considering 27 malignant effusions (22 cytologicallypositive and 5 on follow-up had malignancy in some parts ofthe body or died due to malignancy or had omentalmetastasis and showed DNA aneuploidy), DNA FCM pickedup 16 cases. Sensitivity and specificity of DNA flowcytometry were thus 59.25% and 98.63%, respectively.Similarly, sensitivity and specificity of cytology were 78%and 100%, respectively.

Image MorphometryCases selected for image morphometry were divided intothree groups: group A, cytologically malignant cases(n 5 20); group B, cytologically benign but DNA aneu-ploid cases (n5 2); group C, control group, i.e., clinicallyand cytologically benign, and diploid by DNA flow cytome-try (n 5 10).

Table 3 shows the means6 SD of nuclear area, diameter,perimeter, convex perimeter, roundness, and convex area ofthe above groups.

Student’s unpaired t-test was done and showed a statisti-cally significant difference (P , 0.05) between group A andgroup C in all the parameters except for nuclear roundness.There was, however, no statistically significant differencebetween group B (cytologically benign but DNA aneuploidcases) and group C (control benign cases).

DiscussionIn the present study, sensitivity and specificity of DNA FCMwere 59% and 98%, respectively. These findings are compa-rable with to those in other studies.2,6,7,8Evans et al.6 in theirstudy of 41 cases of effusion fluids had 100% sensitivity and86% specificity. In their study, sensitivity was very high andspecificity was low compared to our present results. Evans etal.6 had two false-positive FCM diagnoses which werenegative by cytology. In comparison to their study, we hadonly one false-positive case (tetraploid aneploidy with DNAindex 2.18). Unger et al.8 in a study of 33 cases of effusionfluids had 80% sensitivity and 96% specificity. They had twofalse-negative diagnoses by FCM, both representing diploidtumors. In comparison to their study, we had 11 (50%)false-negative cases. Hedley et al.7 in a study of 119 effusionfluid samples had 55% sensitivity and 100% specificity.They had no false-positive FCM diagnoses, but 20 diag-noses were false-negative. These 20 false-negative sampleswere accounted for by the relatively low sensitivity of FCMin the study by Hedley et al.7 This study is highly compa-rable with the present study, in which we noted 11 false-negative cases by FCM. Stonesifer et al.2 in a study of 75effusion fluids had 88% sensitivity and 94% specificity ofFCM. They had three false-positive and four false-negativecases. This wide variation of sensitivity (55–100%) andspecificity (86–100%) may be explained by: 1) number ofcases examined, 2) types of cases included, 3) criteria ofaneuploid, and 4) various ways of processing of specimensfor FCM.

In the present study, there were 11 false-negative casesand one false-positive case. The false negativity can beexplained by: 1) the primary tumor may be diploid, 2) asmall aneuploid peak formed by a small number of malig-nant cells may be overshadowed by a large number ofdiploid reactive mesothelial cells, and 3) minor chromo-somal aberrations (1–3%) may not be picked up by DNAFCM.

There was only one false-positive case with a DNA indexof 2.18, and there was no doubt of tetraploidy. Tetraploidaneuploidy on DNA histograms should be carefully inter-preted. Clinically this patient had no evidence of malig-nancy, and he was put on antitubercular therapy empirically.Aneuploidy in benign cases has been described by variousauthors. Unger et al.8 reported a hypodiploid aneuploid peak

Fig. 1. DNA histogram of a cytologically negative case, showinghyperdiploid aneuploidy (M1 is a diploid peak, M2 is a hyperdiploidaneuploid peak, and M3 is a small hypertetraploid peak).

DNA FCM AND ICM IN EFFUSION FLUID

Diagnostic Cytopathology, Vol 22, No 2 83

in pleural effusion due to Nocardia pneumonia. Similarly,Katz et al.10 misinterpreted a small tetraploid peak asaneuploid. Rijken et al.11 found one false-positive aneuploidpeak in an effusion due to hemorrhagic pancreatitis and aninfected pseudocyst.

There were five cases (6.4%) out of 78 cytologicallybenign cases where DNA FCM picked up malignancy. Thisfinding is comparable with the studies mentioned by Zarbo,12

where it was shown that in 3% of cases, DNA FCM can pickup malignancy in cytologically benign cases. These casesshowed monomorphic round cells similar to mesothelialcells. Review of those smears showed no feature of malig-nancy. DNA flow cytometry was very helpful in those casesto further investigate for the primary site of malignancy.

To further increase the detection rate of malignancy byDNA FCM, a double-labeling method, i.e., propidium iodidefor DNA staining and keratin for labeling of epithelial cellswas performed by Croonen et al.13 A large number ofreactive mesothelial cells may dilute the malignant cells, andan aneuploid peak may be missed. Therefore, Schneller etal.14 performed image ploidy on the Feulgen-stained smearsof effusion fluids and selectively picked up the cancer cellsonly. They detected cases of aneuploidy by image ploidywhich were missed by DNA FCM.

Automated image morphometry can rapidly measurevarious parameters of a large number of cells. Only a fewstudies are available in the literature about ICM of effusionfluids. Walts et al.15 and Marchevsky et al.4 performedcomputerized interactive morphometry in effusion fluids toevaluate the diagnostic role of ICM. Walts et al.15 tried todifferentiate reactive lymphoid cells and lymphoma cells by

nuclear profile, with the use of computerized interactivemorphometry. Marchevsky et al.4 successfully diagnosed 21out of 24 test cases with the help of this technique. They onlymeasured nuclear and cytoplasmic profile diameters. In thepresent study, multiple nuclear parameters such as area,diameter, perimeter, convex perimeter, roundness, and con-vex area were measured by automated ICM.

Statistical analysis has shown significant differences be-tween cytologically malignant and control benign cases inregard to all parameters except nuclear roundness. Nuclearroundness indicates nuclear shape variation. Possibly this isbecause malignant cells suspended in effusion fluids tend toshow less variation with regard to shape than do the benigncells. Thus it is not surprising to get insignificant differencesbetween benign and malignant cells with respect to nuclearroundness.

Automated image morphometry data were not significantbetween cytologically benign but DNA aneuploid cases andcontrol benign cases. Thus, it was not surprising that theformer group of cases was missed by light microscopicexamination. Of the six cases of cytologically benign andDNA aneuploid cases, ICM was done in only two cases. Inthe rest of the cases, smears showed scanty cellularity andICM could not be performed.

Many difficulties have been faced during automatedimage morphometric analysis. It was very difficult to detectvesicular nuclei, which had a similar gray value to thecytoplasm or background. In many malignant cell clusters,there was nuclear overlapping and the clusters had to beexcluded from analysis. Debris, degenerate cells, and redblood corpuscles (RBCs) had to be excluded by image

Table II. Salient Cytological, DNA Flow Cytometric Features, and Clinical Data of Cytologically Benign and DNA Aneuploid Cases

Age/sexSite of

effusionCytologicaldiagnosis

DNAindex

S1 G2Mpercentage Aneuploidy Primary site

60/M Pleural fluid No malignant cells 0.63 3.18 Hypodiploidy Small-cell carcinoma, died within 1 mo60/F Ascitic fluid No malignant cells 1.52 1.22 Hyperdiploidy Adenocarcinoma of ovary with omental

metastasis60/F Pleural fluid No malignant cells 1.31 1.27 Hyperdiploidy Carcinoma lung with cerebral metastasis75/F Ascitic fluid No malignant cells 1.20 8.0 Hyperdiploidy Ovarian adenocarcinoma (operated)42/M Ascitic fluid No malignant cells 2.18 9.5 Tetraploidy Antitubercular therapy started65/M Ascitic fluid No malignant cells 2.30 2.0 Hypertetraploidy Postoperative sarcoma of the thigh, with

ascitis

Table III. Image Cytometry Findings of Effusion Samplesa

Nuclear area(square microns)

Nuclear diameter(microns)

Nuclear perimeter(microns)

Nuclear convexperimeter (microns)

Nuclearroundness

Nuclear convexarea (square microns)

Group A, cytologically malignantcases (n5 20) 46.736 27.95 8.636 2.67 27.806 9.21 24.146 7.14 1.426 0.20 53.566 33.55

Group B cytologically benign butDNA aneuploid cases (n5 2) 34.366 9.63 7.856 1.14 25.736 3.35 22.016 3.04 1.506 0.02 40.536 12.11

Group C, control benign cases(n 5 10) 30.256 8.48 7.146 0.92 22.566 3.49 20.306 2.90 1.316 0.12 33.476 9.47

aMean and standard deviation. Unpaired Student’s t-test. Group A vs. group C,P , 0.05 in all parameters (except nuclear roundness,P . 0.05), i.e., thedifference is significant. Group A vs. group B,P . 0.05 in all parameters; not significant. Group B vs. group C,P . 0.05 in all parameters; not significant.

SAHA ET AL.

84 Diagnostic Cytopathology, Vol 22, No 2

editing. In smears with scanty cellularity, it was laborious tostudy a large number of cells. However, ICM is a promisingtechnology and can be used in the future for detection ofmalignancy in test samples. Image ploidy in combinationwith image morphometry may have more potential todiagnose malignacy in effusion fluid. In conclusion, DNAFCM is a useful adjunct to routine diagnostic cytology,which may pick up malignancy in problematic cases. Thespecificity of DNA FCM is quite high, and it is a reliabletechnique. ICM may also provide additional informationwhich may be helpful.

References1. Koss LG. Effusions in presence of cancer. In: Koss LG, editor.

Diagnostic cytology and its histopathological basis. Philadelphia: J.B.Lippincott Co.; 1992. p 1116–1184.

2. Stonesifer KJ, Xiang J, Wilkinson EJ, Benson NA, Braylan RC. Flowcytometric analysis and cytopathology of body cavity fluids. ActaCytol 1987;31:125–130.

3. Marchevsky AM, Gil J, Caccamo D. Computerised interactive mor-phometry: a study of malignant mesothelioma and mesothelial hyper-plasia in pleural biopsies. Arch Pathol Lab Med 1985;109:1102–1105.

4. Marchevsky AM, Hauptman E, Gil J, Watson C. Computerizedinteractive morphometry as an aid in the diagnosis of pleural effusions.Acta Cytol 1987;31:131–136.

5. Menard S, Rilke F, Dellatorre G, Mariani-Costantini R, Regazzoni M,Tagliabue E, Alasio L, Conaghi MI. Sensitivity enhancement of the

cytologic detection of cancer cells in effusions by monoclonal antibod-ies. Am J Clin Pathol 1985;83:571–576.

6. Evans D, Thornthwaite J, Ng ABP, Surgarbaker E. DNA flowcytometryof pleural effusions: comparison with pathology for the diagnosis ofmalignancy. Anal Quant Cytol 1982;5:19–27.

7. Hedley DW, Philips J, Rugg C, Taylor IW. Measurement of cellularDNA content as an adjunct to diagnostic cytology in malignanteffusions. Eur J Cancer Clin Oncol 1984;20:749–752.

8. Unger K, Raber M, Bedrossian C, Stein DA, Barlogie B. Analysis ofpleural effusions using automated flow cytometry. Cancer 1983;52:873–877.

9. Krishan A. Rapid flow cytofluorometric analysis of mammalian cellcycle by propidium iodide staining. J Cell Biol 1975;66:188–193.

10. Katz RL, Johnson TS, Wilkinson KD. Comparison of cytologic andacridine-orange flow cytometric detection of malignant cells in humanbody cavity fluids. Anal Quant Cytol Histol 1985;7:227–235.

11. Rijken A, Dekker A, Taylor S, Hoffman P, Blank M, Krause JR.Diagnostic value of DNA analysis in effusions by flow cytometry andimage analysis: a prospective study on 102 patients as compared withcytologic examination. Am J Clin Pathol 1991;95:6–12.

12. Zarbo RJ. Flow cytometric DNA analysis of effusions: a new testseeking validation. Am J Clin Pathol 1991;95:2–4.

13. Croonen AM, Valk PVD, Herman CJ, Lindeman J. Cytology, immuno-pathology and flowcytometry in the diagnosis of pleural effusions. LabInvest 1988;58:725–732.

14. Schneller J, Eppich E, Greenebaum E, Elequin F, Sherman A, WerstoR, Koss LG. Flow cytometry and Feulgen cytophotometry in evalua-tion of effusions. Cancer 1987;59:1307–1313.

15. Walts AE, Morimoto R, Marchevsky AM. Computerized interactivemorphometry and the diagnosis of lymphoid-rich effusions. Am J ClinPathol 1993;99:570–575.

DNA FCM AND ICM IN EFFUSION FLUID

Diagnostic Cytopathology, Vol 22, No 2 85