Two-Color, Cytokeratin-Labeled DNA Flow Cytometric Analysis of 332 Breast Cancers. Lack of Prognostic Value With 12-Year Follow-up

Two-Color, Cytokeratin-Labeled DNA Flow Cytometric Analysis of 332 Breast Cancers Lack of Prognostic Value With 12-Year Follow-up Anil R. Prasad, MD; George Divine, PhD; Richard J. Zarbo, MD, DMD Context. DNA flow cytometry of breast cancer is a proposed tumor marker of prognostic significance that is of controversial clinical utility because of lack of standardization and confirmatory studies. Objective. To evaluate the prognostic significance of the more informative technique of multiparametric 2-color DNA flow cytometry as recommended by the 1992 DNA Cytometry Consensus Conference. Design. Three hundred thirty-two breast carcinomas with 7 to 12 years of follow-up were prospectively analyzed as fresh tumors that were mechanically dissociated into whole cell suspensions. These suspensions were dual fluorescence labeled with propidium iodide (DNA) and antibodies to cytokeratin (epithelium) and leukocyte common antigen (internal leukocyte control) for gated analysis of subpopulations. Multicycle software with histogram-dependent algorithms employing background, aggregate, and debris correction were used in DNA and cell-cycle quantitation. Data were analyzed according to the DNA Flow Cytometry Consensus Conference recommendations. Results. DNA ploidy and proliferation stratified into 3 categories were not predictive of overall or disease-free survival. Sixty-five percent of tumors were nondiploid, and 35.4% were diploid. Two hundred six tumors were able to be evaluated for synthesis-phase fraction (SPF) analysis, with 74 of 206 cases in the low range ( 13.4%), 36.4% in the intermediate range ( 13.5 to 25.4%), and 27.6% in the high SPF ( 25.5%) category. Aneuploid tumors tended to have a higher SPF. Univariate survival analysis showed prognostic significance of the following: tumor size, stage, TNM components, vascular invasion, nuclear grade, and histologic grade. Only T classification, presence of positive axillary lymph nodes, and distant metastases were significant independent predictors of survival in multivariate Cox regression models. Age and hormone receptor status showed no prognostic significance. Synthesisphase fraction was significantly correlated with tumor size, stage, T classification, nuclear and histologic grade, presence of estrogen or progesterone receptors, and axillary lymph node status. None of the histologic parameters showed any significant association with DNA aneuploidy, except for high nuclear and histologic grade and the absence of estrogen receptors. Conclusions. Despite the use of state-of-the-art processing and flow cytometry analytic techniques, DNA ploidy and proliferation measurements were not predictive of survival in any stage of breast cancer. However, select histopathologic parameters and TNM stage were significant predictors of survival in univariate and multivariate analyses. We conclude that DNA ploidy and proliferation measurements do not provide significant prognostic information for clinicians to integrate into therapeutic decision making for patients with breast cancer. (Arch Pathol Lab Med. 2001;125:364 374) Approximately 180 000 women in the United States are diagnosed with breast cancer each year, and roughly 1 in 9 women living to the age of 85 years will develop breast cancer during her lifetime. The recognition of significant prognostic factors for patients with breast cancer has been the focus of numerous research studies as clinicians strive to identify optimal treatment strategies for these patients. The emphasis is now on the utility of pre- Accepted for publication September 20, 2000. From the Departments of Pathology (Drs Prasad and Zarbo) and Biostatistics and Epidemiology (Dr Divine), Henry Ford Hospital, Detroit, Mich. Presented in part at the United States and Canadian Academy of Pathology Annual Meeting, New Orleans, La, March 29, 2000. Reprints: Richard J Zarbo, MD, DMD, 2799 W Grand Blvd, Detroit, MI 48202. emptive chemotherapy for individuals with clinically localized tumors at high risk for relapse, 1 and additional validated prognostic tumor markers are especially needed in this group of patients. 2 Conventional clinicopathologic features, such as tumor size 3,4 and axillary lymph node involvement, 5,6 are statistically significant prognostic parameters in invasive breast carcinomas. Other histomorphologic features, for example, histologic grade, 7,8 vascular invasion, 9,10, and tumor necrosis, 11 have been found to be significant predictors of recurrence and mortality. However, these factors are often highly interrelated, either with each other, with disease stage and nodal status, or with both. 9,11 In addition, some of these features are confounded by interobserver variation and subjectivity. 11,12 In the last decade, the study of flow cytometric data especially with regard to the prog- 364 Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al

nostic value of nuclear DNA content (ploidy) and proliferative activity (synthesis-phase fraction [SPF]) in breast cancer has spawned numerous publications with great variation in study design and analytic techniques (Tables 1 and 2). These conflicting conclusions and a lack of standardization and quality control leave the routine use of DNA flow cytometric analysis for treatment decisions in breast cancer controversial. The DNA Flow Cytometry Consensus Conference of 1992 13 evaluated studies up to that time and found that DNA ploidy was, on the whole, a poor prognostic marker, whereas SPF was considered to be an important prognostic factor in both node-negative and node-positive patients. The tumor marker expert panel of The American Society of Clinical Oncology in its consensus report concluded that ploidy derived from DNA flow cytometry should not be used for the management of breast cancer patients, and that data were insufficient to recommend the use of flow cytometry derived proliferation measurements for treatment purposes. 14 More recently, the College of American Pathologists (CAP) Conference on Solid Tumor Prognostic Factors, held in June 1999, 15 ranked flow cytometry derived SPF measurements under category II (factors extensively studied biologically and clinically but whose import remains to be validated in statistically robust studies) and observed that SPF correlates with clinical outcomes of patients with primary breast cancer. DNA ploidy analysis was placed in category III (all other factors not sufficiently studied to demonstrate their prognostic value), indicating a lack of sufficient data to demonstrate prognostic value. Contradictory flow cytometric findings may result from differences in study design and technical, analytic, and interpretive methods. Guidelines of the 1992 DNA Flow Cytometry Consensus Conference 13 were applied in this study, and our methodology conformed to the technical recommendations of the CAP Conference on Solid Tumor Prognostic Factors, June 1999, 15 which included making whole-cell preparations from fresh tissues and using multiparametric dual fluorescence markers to assist in identification of subpopulations and enrichment of neoplastic cells for analysis. 13,16 Here we present data from the evaluation of 332 breast cancer patients with 7 to 12 years of follow-up relating flow cytometry determined DNA content and cell proliferation to conventional pathologic parameters and patient outcome. MATERIALS AND METHODS Clinical Samples This study is based on 332 infiltrating ductal carcinomas (not otherwise specified) of the breast that were consecutively resected during the period August 1987 through June 1992. All tumors were received in a fresh, unfixed state in the Surgical Pathology Laboratory at Henry Ford Hospital, Detroit, Mich (see Laboratory and Analytic Exclusions ). These surgical specimens were handled in a consistent manner, with 1 or 2 fresh tissue samples taken from each tumor, placed in RPMI transport media, and either processed immediately or refrigerated at 4 C until the next day. All histopathologic assessments were made by applying standardized departmental protocols for gross and microscopic evaluation. The following histopathologic parameters were evaluated: histologic and nuclear grade, tumor size, presence or absence of lymph node and distant metastasis, peritumoral vascular invasion, and estrogen and progesterone hormone receptor status (dextran-coated charcoal method). All cases were staged according to the American Joint Committee on Cancer TNM system. 17 Clinical follow-up information was obtained from the Henry Ford Hospital Tumor Registry and from chart review. Because the senior author developed the DNA analysis technique employed in this laboratory since 1987, which was subsequently recommended by the 1992 DNA Cytometry Consensus Conference, the extended follow-up was possible. 13,16,18 Duration of follow-up for individual patients from the date of surgery to the date of death or to the end of this study was up to 12 years, with a median of 120 months (10 years). Six of the patients were lost to follow-up after 3 to 127 months. Sixteen percent of the patients in this cohort underwent lumpectomy only, without axillary dissection, whereas all the other patients had modified radical mastectomy with axillary dissection. Depending on the stage of disease and the status of lymph node and surgical margins, a proportion of patients (85.4%) in this series received adjuvant chemotherapy with or without tamoxifen and/or radiation therapy, according to the National Institutes of Health Consensus Development Panel on Adjuvant Chemotherapy and Endocrine Therapy for Breast Cancer. 19 Laboratory and Analytic Exclusions Of the 332 cases studied, 126 cases either had uninterpretable DNA histograms or did not meet the quality control criteria for cell cycle calculations in the ungated and cytokeratin (CK)-gated data set because of failure to satisfy DNA Flow Cytometry Consensus Conference criteria of histograms with a coefficient of variation (CV) less than 8%; background, aggregates, and debris less than 25%; or an abnormal peak composed of greater than 5% to 10% of total events after aggregate compensation. 13 Tumor Preparation and Staining Tumor dissociation, fixation, and staining methods, as well as spectral compensation for the 2-color flow cytometric DNA analysis method, have been described previously. 18,20 Briefly, fresh solid tumor slices from breast carcinomas were mechanically disaggregated into single-cell suspensions and resuspended in 1 ml RPMI and 1 ml fetal bovine serum. The cells were fixed by the slow addition of cold 70% ethanol (from absolute ethanol stock) to a final concentration of 50% ethanol while vortexing. Separate tumor aliquots were stained with titration-derived concentrations of (1) unconjugated nonimmune mouse immunoglobulin (Ig) G (Coulter Corp, Hialeah, Fla) as a green fluorescence-negative control, (2) unconjugated anti leukocyte common antigen antibody (Dako Corporation, Carpinteria, Calif), and (3) unconjugated anti-cytokeratin antibody, clone CAM 5.2 (Becton-Dickinson, Mountain View, Calif). Labeling with primary antibody was followed by staining with a fluorescein isothiocyanate conjugated goat anti-mouse (GAM-FITC) secondary antibody (Becton-Dickinson). Subsequent DNA staining was performed with 1 ml of propidium iodide (PI), 0.05 mg/ml (Sigma Chemical Co, St Louis, Mo), and 100 ml ribonuclease type 1-A (Sigma) solution, 1 mg/ml (approximately 80 Kunitz units/ml) per tube. Flow Cytometer Specifications All data were acquired on a FACScan flow cytometer (Becton- Dickinson) equipped with a single 488-nm argon laser. Fluorescein isothiocyanate fluorescence was detected through a 530 15-nm filter, and PI fluorescence was analyzed through a 585 22-nm filter. Forward-angle light scatter, 90 light scatter, log green fluorescence FITC (FL1), and linear red fluorescence PI (FL2) were acquired in list-mode format at low flow rate to a total of 20 000 events. Flow Cytometer Calibration Initial setup of the flow cytometer was performed with AutoCOMP software/calibrite fluorescent microspheres (Becton-Dickinson). Forward-angle light scatter and 90 light scatter photomultiplier settings were further adjusted based on the intrinsic light scatter properties of ethanol-fixed normal peripheral blood leukocytes. The G0/G1 peak of PI-stained peripheral blood Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al 365

Table 1. Studies of DNA Ploidy and Prognosis in Breast Cancer Source, y Tissue Type No. of Cases Aneuploid, % Significance Comments Schmidt et al, 45 1999 Fresh 106 56 OS, DFS-NS (u, m) FU NA Peiro et al, 83 1997 P 118 53 OS, DFS-NS (u, m) Node-negative patients; FU 108 mo (mean) Wyss-Desserich et al, 84 1997 P 57 60 (nondiploid) OS, DFS-NS (u, m) Node-negative patients; FU 73 mo (median) Romero et al, 31 1996 P 174 49 OS, DFS-NS (u, m) FU 55 mo (mean) Camplejohn et al, 40 1995 P 881 65 OS, DFS-S (u, m) FU NA Kute et al, 32 1995 P 397 NA OS, DFS-NS (u, m) FU 80 mo (median) Witzig et al, 53 1994 P 265 39 OS, DFS-NS (u, m) Node-negative patients; FU 60 mo Stal et al, 85 1994 Frozen 184 66 OS, DFS-S (u, m) FU 131 167 mo Wingren et al, 86 1994 Frozen 209 44 OS-NA; DFS-S (u); DFS-NS (m) Node-negative patients; FU 60 mo (median) Stal et al, 41 1993 Frozen 219 79 OS, DFS-NS (u, m) Node-negative patients; FU 60 mo (mean) Merkel et al, 51 1993 P 280 58 OS, DFS-NS (u, m) Node-negative patients; FU 76 mo (median) Clark et al, 58 1993 Frozen 851 56 OS-S (u), NS (u, m) Node-positive patients; FU 44 mo (median) Wenger et al, 28 1993 Frozen 127 220 47 DFS-S (u) FU 26 mo (median) Clark et al, 33 1992 P 274 49 OS, DFS-NS (u, m) Node-negative patients; FU 56 mo (median) Arnerlov et al, 34 1992 P 150 57 OS, DFS-NS (u, m) FU 78 mo (median) Dressler et al, 87 1992 P 565 64 OS, DFS-NS (u, m) Node-negative patients; FU 53 mo (median) Bosari et al, 35 1992 P 147 33 OS, DFS-NS (u, m) Node-negative patients; FU 108 mo (minimum) Stanton et al, 79 1992 P 281 64 OS, DFS-NS (u, m) FU 93 mo (minimum) Fisher et al, 72 1991 P 398 57 OS, DFS-NS (u, m) FU 80 mo (median) Ewers et al, 88 1991 Frozen 580 60 (nondiploid) OS, DFS-NS (u, m) FU 59 mo (median) Beerman et al, 42 1990 P, Fresh 690 73 OS, DFS-S (u, m) FU 84 mo (median) Joensuu et al, 39 1990 P 159 57 OS-S (m); DFS-NA FU 60 mo Keyhani-Rofagha et al, 22 1990 P 165 57 OS, DFS-NS (u) Node-negative patients; FU 36 180 mo Lewis, 89 1990 P 155 41 OS, DFS-S (u, m) Node-negative patients; FU 120 mo (median) O Reilly et al, 44 1990 P 140 69 OS, DFS-NS (u, m) FU 96 mo Uyterlinde et al, 56 1990 P 225 45 (nondiploid) OS-S (u); DFS-NA FU 60 120 mo Winchester et al, 82 1990 P 198 64 OS, DFS-NS (u, m) Node-negative patients; FU 80 mo (median) Arnerlov et al, 90 1990 P 99 53 OS, DFS-S (u), NS (m) FU 36 84 mo Kute et al, 91 1990 P 197 52 OS, DFS-NS (u, m) FU 41 mo (median) Eskelinen et al, 23 1989 P 117 65 OS, DFS-S (u) FU 137 mo Muss et al, 69 1989 Frozen 101 54 OS, DFS-NS (u, m) FU 51 mo (median) Toikkanen et al, 49 1989 P 351 68 (nondiploid) OS, DFS-NS (u, m) FU 324 mo (median) Stal et al, 68 1989 Fresh 472 63 OS, DFS-S (u); OS, DFS-NS FU 70 107 mo (m) Clark et al, 76 1989 Frozen 345 68 OS-S (u); DFS-S (u, m) Node-negative patients; FU 59 mo (median) Van der Linden et al, 92 1989 P 156 52 OS, DFS-S (u); OS-S (m) FU 43 mo (median) Roos et al, 24 1988 P 72 57 DFS-NS (u) FU 72 mo (mean) Hedley et al, 36 1987 P 473 72 OS, DFS-S (u), NS (m) FU 60 mo Cornelisse et al, 63 1987 P, Frozen 565 61 OS-S (u, m); DFS-NS (u, FU 33 mo (median) m); DFS-S (m); postmenopausal only Dowle et al, 93 1987 P 354 60 OS, DFS-S (u) (short-term FU 84 mo (median) FU only); OS, DFS-NS (m) Kallioniemi et al, 73 1987 P 93 60 OS-NS; DFS-S (u, m) FU 63 mo Kallioniemi et al, 38 1987 P 308 64 OS-S (u, m); DFS-NA FU 96 mo (mean) Owainati et al, 37 1987 P, Fresh 280 60 OS, DFS-NS (u, m) FU 96 156 mo Thorud et al, 26 1986 Frozen 59 54 DFS-NS (u) FU 48 mo (minimum) Ewers et al, 25 1984 Frozen 638 60 DFS-S (u) (early stage only) FU 16 mo (median) Coulson et al, 27 1984 Frozen 74 79 OS-S (u); DFS-NS (u) FU 36 mo * P indicates paraffin-embedded tissue; OS, overall survival; DFS, disease-free survival; NS, not significant; u, univariate analysis; m, multivariate analysis; S, significant; NA, not available; and FU, follow-up. 366 Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al

Table 2. Studies of Cell Proliferation and Prognosis in Breast Cancer Source, y Year No. of Cases Median SPF, % Significance Comments Schmidt et al, 45 1999 Fresh 106 14.2 OS, DFS-NS (u, m) FU NA Peiro et al, 83 1997 P 118 14.5 (diploid) 17.1 (aneuploid) OS-S (m); DFS-NS (m) Node-negative patients; FU 108 mo (mean) Wyss-Desserich et al, 84 1997 P 43 2.9 OS, DFS-NS (u, m) Node-negative patients; FU 73 mo (median) Romero et al, 31 1996 P 174 6.8 (mean) OS, DFS-S (u, m) FU 55 mo (mean) Pfisterer et al, 94 1995 P 59 3.2 DFS-NS (u) (diploid tumors only) Node-positive patients FU 83 mo (median) Camplejohn et al, 40 1995 P 802 7.2 OS, DFS-S (u, m) FU NA Kute et al, 32 1995 P 300 NA OS, DFS-S (u) FU 80 mo (median) Witzig et al, 53 1994 P 265 6.0 OS, DFS-S (u, m) Node-negative patients; FU 60 mo Wingren et al, 86 1994 Frozen 209 4.6 DFS-S (u, m) Node-negative patients; FU 60 mo (median) Stal et al, 85 1994 Frozen 116 6.6 (diploid) OS, DFS-S (u); OS, DFS-NS (m) FU 131 167 mo 11.4 (aneuploid) Stal et al, 41 1993 Frozen 219 5.5 (euploid) 8.0 (aneuploid) OS, DFS-S (u, m) Node-negative patients; FU 60 mo (mean) Clark et al, 58 1993 Frozen 851 6.9 OS-S (u, m) Node-positive patients; FU 44 mo (median) Silvestrini et al, 95 1993 Frozen 291 4.2 OS, DFS-NS (u, m) Node-negative patients; FU 48 mo (median) Wenger et al, 28 1993 Frozen 109 930 3.4 (diploid) DFS-S (u) FU 26 mo 10.7 (nondiploid) Merkel et al, 51 1993 P 280 4.8 (diploid) 11.7 (aneuploid) OS-S (u, m); DFS-S (u,), NS (m) Node-negative patients; FU 76 mo Bosari et al, 35 1992 P 136 7.3 (mean) OS-S (u), NS (m); DFS-S (u), NS (m) Node-negative patients; FU 108 mo (minimum) Dressler et al, 87 1992 P 565 6.97 OS, DFS-S (u, m) Node-negative patients; FU 53 mo (median) Clark et al, 33 1992 P 247 3.6 (diploid) 7.6 (aneuploid) OS-NS (u, m); DFS-S (u, m) Node-negative patients; FU 56 mo (median) Stanton et al, 79 1992 P 281 7.25 OS, DFS-NS (u, m) FU 93 mo (minimum) Arnerlov et al, 34 1992 P 122 8.2 OS, DFS-S (u, m) FU 78 mo (median) Fisher et al, 72 1991 P 398 2.4 (diploid) OS-S (u), NS (m); DFS-S (u, m) FU 80 mo (median) 7.9 (aneuploid) Ewers et al, 88 1991 Frozen 580 7.3 OS, DFS-S (u, m) FU 59 mo (median) O Reilly et al, 44 1990 P 140 7.1 OS, DFS-S (u); OS, DFS-NS (m) Node-negative patients; FU 60 mo Uyterlinde et al, 56 1990 P 225 NA OS-S (u); DFS-NA FU 60 120 mo Arnerlov et al, 90 1990 P 99 NA OS, DFS-NS (m) FU 36 84 mo Joensuu et al, 39 1990 P 159 12.0 (mean) OS-NS (u, m) FU 60 mo (median) Kute et al, 91 1990 P 197 8.0 OS, DFS-NS (u, m) FU 41 mo (median) Sigurdsson et al, 70 1990 Frozen 566 7.5 OS, DFS-S (u, m) Node-negative patients; FU 48 mo (median) Winchester et al, 82 1990 P 198 4.5 (diploid) 10.3 (aneuploid) DFS-S (u, m); (only diploid tumors) OS-NA Node-negative patients; FU 80 mo (median) Clark et al, 76 1989 Frozen 253 5.2 OS, DFS-S (u, m) Node-negative patients; FU 59 mo (median) Eskelinen et al, 23 1989 Frozen 117 4.8 OS, DFS-S (u) FU 137 mo Muss et al, 69 1989 Frozen 84 12.5 OS, DFS-S (u, m) Node-negative patients; FU 51 mo (median) Toikkanen et al, 49 1989 P 351 9.0 OS-S (u, m); DFS-NA FU 324 mo (median) Stal et al, 68 1989 Fresh 290 7.7 OS-S (m); DFS-NS (m) FU 70 107 mo Hedley et al, 36 1987 P 285 10.0 OS, DFS-S (u); NS (m) FU 60 mo (median) Kallioniemi et al, 73 1987 P 308 7.5 OS-S (u, m); DFS-NA FU 96 mo Coulson et al, 27 1984 Frozen 74 19.7 (mean) OS-S (u) FU 36 mo * P indicates paraffin-embedded tissue; SPF, S-phase fraction; NA, not available; OS, overall survival; DFS, disease-free survival; NS, not significant; u, univariate analysis; m, multivariate analysis; S, significant; and FU, follow-up. leukocytes was set to channel 70 of a 256-channel linear scale by adjustment of the FL2 detector voltage. A peripheral blood leukocyte G0/G1 coefficient of variation less than 2.0% was considered a measure of acceptable staining/cytometer resolution. System linearity was verified by calculating the doublet to singleton peak channel ratio for a PI-stained chicken erythrocyte nuclei suspension. Gating Method Cell subpopulations staining positively with leukocyte common antigen FITC or CK-FITC were identified by comparison of FL1 versus FL2 dot plots of the specifically stained samples with the corresponding nonspecific FITC fluorescence control. The antibody positive region for an individual tumor was based on the nonimmune mouse IgG-FITC stained sample, with the per- Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al 367

centage of cells in the positive region of this negative control sample generally comprising less than 2% of total events. Histogram Analysis Single-parameter DNA histograms generated from list-mode data were analyzed for both DNA ploidy and proliferative compartments of the cell cycle, including %SPF and G2/M phase with the Multicycle software package (Phoenix Flow Systems, San Diego, Calif). Both ungated and CK-gated data sets were evaluated. DNA index (DI) was calculated as the mean channel position of the tumor G0/G1 peak divided by the mean channel position of the diploid G0/G1 peak, as verified by leukocyte common antigen gating. In keeping with the 1992 DNA Flow Cytometry Consensus Conference 13 recommendations, a minimum of 5% to 10% of total events after aggregate compensation was required for an apparent minor aneuploid population to be recognized as such in the ungated histogram data set. If CK gating increased the frequency of a minor peak above the threshold value, aneuploidy was confirmed for the CK-gated data set. A conservative approach was taken in the estimation of SPF in applying zero-order and first-order polynomial shapes for modeling the S phase. Histogram debris was modeled in 2 manners for each histogram, and a separate SPF value was generated for background, aggregates, and debris. The sliced nuclei option was employed in the debris model to compensate for the creation of cut nuclei during mechanical disaggregation of fresh tissues with a scalpel. Coefficient of variation or G2/G1 constraints were used only if results without constraints proved to be biologically unreasonable. Flow Cytometry Data Sets Flow cytometry data were stratified into the following data set combinations for statistical analyses of correlation with clinicopathologic parameters and survival: diploidy, defined as DI 0.8 1.2; aneuploidy, defined as DI 0.8, DI 1.2 and 1.9; tetraploidy, defined as DI 1.9 and 2.1; ploidy derived from single DNA fluorescence (ungated) histograms; ploidy derived from 2-color fluorescence (DNA/CK) histograms (CK-gated); proliferation measured as SPF ( G2/M) derived from ungated histograms, CK-gated histograms, and based on a cutoff at the median and at tertile distributions of all tumors combined; the diploid tumor subset and the aneuploid tumor subset; Multicycle zero-order model of histogram-dependent curve fitting; and Multicycle first-order curve-fitting model. Statistical Methods Except for the exclusions cited, all patients were included in the analysis. For flow cytometry S-phase analysis, patients were excluded if the number of ungated or CK-gated events recorded in the histogram were less than 5000; patients were also excluded if the CV was greater than 8.0% or the percentage of background, aggregates, and debris was greater than 25%. Survival analysis was performed using proportional hazards modeling to test the association of clinical and flow cytometry variables versus all cause mortality. Analysis started with single predictor models to assess univariate associations with survival time. Actual survival with time to death was determined by univariate analysis (Kaplan-Meier method) and stratified by various clinical, histopathologic, and flow cytometric parameters to determine its clinical importance. The multivariate Cox regression model was used to evaluate the independent nature and relative predictive strength of these parameters in survival. Those alive at the end of the study had their survival time censored. A preliminary analysis of the individual variables was performed to estimate crude relative risks, and those with risk P values less than.10 were included in the multivariate model. Statistical significance was based on a P value less than.05. Spearman correlation coefficients were calculated for the flow cytometry variables that were continuous versus the clinical variables that were categorical. A 2-sample t test was used when Table 3. Pathologic Characteristics in 332 Breast Carcinomas Tumor size, cm 0 2.0 2.1 5.0 5.1 Axillary node status Positive Negative Distant metastasis Present Absent Nuclear grade 1 2 3 Histologic grade 1 (well differentiated) 2 (moderately differentiated) 3 (poorly differentiated) Hormone receptor status* ER positive ER negative PR positive PR negative Vascular invasion Present Absent Stage I II III IV T1 T2 T3 T4 N0 N1 N2 No. (%) 126 (38) 167 (50) 39 (12) 161 (56) 129 (44) 24 (7) 308 (93) 19 (6) 118 (36) 195 (58) 21 (7) 114 (34) 197 (59) 198 (64) 114 (36) 189 (61) 123 (39) 145 (48) 157 (52) 86 (26) 165 (50) 55 (16) 26 (8) 123 (37) 154 (46) 18 (6) 37 (11) 129 (44) 119 (41) 42 (15) * ER indicates estrogen receptor; PR, progesterone receptor. comparing the dichotomous clinical variables (lymph node and hormone receptor status, presence of distant metastasis, and peritumoral vascular invasion) and the continuous flow variables. A 2 test was also used to look at the distributions of the dichotomous clinical variables and the categorical flow data (SPF and DNA ploidy). RESULTS All patients included in this study by design were women, with ages ranging from 29 to 92 years (mean 63 years). The pathologic characteristics of the 332 invasive breast carcinomas are shown in Table 3. In 50% of the cases, the tumor measured between 2.1 and 5.0 cm. Histologic grade 3 was predominant in 59%, and grades 1 and 2 characterized 7% and 34% of tumors, respectively. Nuclear grade of the tumors showed a somewhat similar distribution pattern to histologic grade. Forty-four percent of the population was found to be pathologically node negative, with 41% having N1 disease and 15% with N2 disease. Distant metastases were present at the time of diagnosis in 24 patients. About half the tumors in this study were stage 368 Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al

Table 4. Multivariate and Univariate Analysis of the 12-Year Actuarial Rate for Overall Survival and Disease-Free Survival for Various Pathologic Parameters Overall Survival Risk Ratio P Value Disease-Free Survival Risk Ratio P Value Parameters (univariate) Tumor size 1.71.020 1.72.008 Axillary node status 3.39.001 3.02.001 T 1.75.001 1.63.001 N 2.48.001 2.16.001 M 29.11.001 60.10.001 Stage 3.88.001 3.48.001 Nuclear grade 1.65.044 1.68.021 Histologic grade 1.60.058 1.71.017 Angiolymphatic invasion 2.51.004 1.89.025 Age 1.00.728 1.00.707 ER positive 0.61.103 0.62.077 PR positive 0.59.069 0.64.091 Ploidy (ungated) 1.00.986 0.98.979 Ploidy (CK gated) 0.99.964 1.01.967 SPF (ungated) 1.17.375 1.11.523 SPF (CK gated) 1.19.334 1.10.547 Parameters (multivariate) Distant metastasis 24.05.001 44.68.001 T (TNM) 1.49.006 NS NS Axillary lymph node status NS NS 2.23.017 * ER indicates estrogen receptor; PR, progesterone receptor; CK, cytokeratin; and N, not significant. Figure 1. In univariate analysis, presence of axillary lymph node metastasis was an independent predictor of disease-free (DFS) and overall survival (OS) that persisted for DFS in multivariate analysis (P.001). Table 5. Comparison of Strength of Association (P Value) Between Flow Variables and Various Clinicopathologic Features* Parameters SPF (Ungated) SPF (CK Gated) Ploidy Ploidy (CK (Ungated) Gated) Tumor stage.0019.0153.4068.4129 Size.0001.0021.4391.4662 T.0003.0054.4494.4718 N.0474.1758.5473.5075 M.664.382.937.958 Nuclear grade.0001.0001.0008.0011 Histologic grade.0001.0001.0002.0003 Age.0782.0586.8868.9158 Lymph node status.204.047.886.810 Estrogen-receptor positive.001.001.034.036 Progesterone-receptor positive.001.001.197.226 Vascular invasion.447.444.550.503 * SPF indicates S-phase fraction; CK, cytokeratin. Spearman correlation coefficient. Chi-square test. II (165/332) at diagnosis, while 26% were stage I. No difference in outcome, either overall or disease-free survival, was detected between the 206 patients whose complete data were analyzed and the 126 patients excluded for flow cytometry quality control reasons. Estrogen receptor (ER) determinations were positive ( 9 fmol/mg) in 198 (64%) of 312 tumors for which hormone receptor status was available. There was a significant correlation between ER status and ploidy ( 2 ratio 6.656, P.036) and SPF measurements ( 2 ratio 31.218, P.001). Diploid tumors comprised 39% (56/142) of ER-positive and 23% (13/56) of ER-negative cases. High SPF values were noted in 18% (26/142) of ER-positive and 52% (29/56) of ER-negative tumors, whereas low SPF values were seen in 45% (64/142) of ER-positive and 11% (6/56) of ER-negative tumors. The SPF also showed significant association ( 2 ratio 27.922, P.001) with the presence of progesterone receptors. High SPF values were seen in 18% (24/132) of PR-positive and 47% (31/66) of PR-negative tumors, while 66% (61/132) of PR-positive and 14% (9/66) of PR-negative tumors had low SPFs. Unlike SPF, ploidy did not show a similar significant association ( 2 ratio 2.976, P.226) with progesterone receptors. In the nondiploid category, 62% (82/132) of the tumors were PR positive and 71% (47/66) were PR negative, whereas 38% (30/132) of PR-positive and 29% (19/66) of PR-negative tumors were diploid. Histopathologic Parameters and Survival Overall, 40% (135/332) of the patients in this series died; 25% died of disease, whereas 15% had no evidence of disease. At the end of the study, 3 patients were alive with disease, and 187 were alive with no evidence of disease. Table 4 shows the main clinicopathologic characteristics of the patients and the results of univariate analysis with overall survival and disease-free survival. A decrease in overall and disease-free survival paralleled the increase in tumor size, stage, histologic and nuclear grade, presence of nodal and distant metastasis, and presence of angiolymphatic invasion. Each additional stage had an estimated risk of death that was 3.88 times more likely than the previous stage. Hormonal status and age did not correlate with survival. Results of the multivariate analysis are also depicted in Table 4. Parameters without significance are excluded. Significant predictors of overall survival were tumor size and presence of distant metastasis. The presence of distant and nodal metastasis also constituted strong predictors of disease-free survival. Figure 1 shows patient survival in association with nodal status. Using Spearman correlation coefficients, some of the pathologic parameters, such as tumor size, stage, T classification, and nuclear and histologic grade, correlated significantly with SPF measurements, although only to a modest degree (Table 5). Nuclear grade and histologic grade showed significant association with ploidy. In ad- Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al 369

Table 6. Distribution of Cytokeratin-Gated Ploidy and S-Phase Fraction Values by Stage Ploidy (CK) and SPF (CK) Stage I Stage II Stage III Stage IV Total No. (%) Aneuploid, DI 0.8, 1.2 DI 1.9, DI 2.1 27 54 21 8 110 (53.4) Diploid, 0.8 DI 1.2 24 36 8 5 73 (35.4) Tetraploid, 1.9 DI 2.1 5 12 4 2 23 (11.7) Low SPF (CK) ( 13.4) 30 31 9 4 74 (35.9) Middle SPF (CK) ( 13.5 to 25.4) 15 39 13 8 75 (36.4) High SPF (CK) ( 25.5) 11 32 11 3 57 (27.6) Total No. (%) 56 (27.2) 102 (49.5) 33 (16.0) 15 (7.3) 206 (100) * CK indicates cytokeratin-gated; SPF, S-phase fraction; and DI, DNA index. Figure 2. Frequency distribution of synthesis-phase fraction (SPF) for the various ploidy categories. CK indicates cytokeratin-gated. dition, there was a significant, although weak, association between axillary lymph node status and SPF ( 2 ratio 6.123, P.047), but not with DNA ploidy ( 2 ratio 0.420, P.810). In the node-negative category, 44% (33/ 75) of the tumors had low SPFs compared to 29% (22/75) that had high SPFs. Among node-positive tumors, 28% (27/97) had low SPFs, whereas 30% (29/97) had high SPFs. Comparing ploidy and nodal status, 70% (52/75) of node-negative tumors were nondiploid, while 29% (22/75) were diploid; similarly, among node-positive tumors, 71% (69/97) were nondiploid, whereas again only 29% (28/97) were diploid. On applying the t test using node status as a dichotomous variable and comparing it to the flow variables as continuous data, this weak association between ungated and CK-gated SPF and lymph node status was lost. Table 6 shows the distribution of the tumors by stage according to the ploidy and proliferation values. Among the 206 tumors whose ploidy was assessed, 53.4% (110/ 206) were aneuploid and 11.2% (23/206) were tetraploid. Forty-nine percent (54/110) of the aneuploid tumors were categorized as stage II disease. Similarly, among the tumors with high SPF, 56% (32/57) were stage II. In this series, 36.4% of tumors (75/206) had an intermediate ( 13.5 to 25.4) SPF level, while 35.9% (74/206) and 27.6% (57/206) belonged to the low ( 13.4) and high ( 25.5) SPF categories, respectively. Figure 2 shows the frequency distribution of SPFs for the various ploidy categories. On analysis of the relationship between SPF and DNA ploidy using the Kruskal-Wallis procedure and employing the Wilcoxon rank sum tests, aneuploid tumors (73.7% [42/110]) tended to have a higher SPF compared to diploid tumors (14% [8/73]). This difference between aneuploid and diploid tumors in relation to SPF (P Figure 3. Disease-free survival (DFS) was not associated with flow cytometric DNA ploidy status, illustrated here as ploidy subsets derived from cytokeratin (CK)-gated histograms. Figure 4. Overall survival probability in breast cancer patients according to DNA ploidy status. CK indicates cytokeratin-gated..001), although modest, was not seen when tetraploid tumors (12.3% [7/23] had high SPFs) were compared with diploid tumors (P.101). DNA ploidy and proliferation indices were not predictive of overall or disease-free survival when analyzed by both univariate and multivariate analysis (Figures 3 6). This included ploidy determined from ungated and CKgated DNA histograms and classified by criteria outlined in the 1992 DNA Flow Cytometry Consensus Conference statements. 13 Assessment of proliferation, even when strat- 370 Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al

Figure 5. There was no correlation between overall survival and synthesis-phase fraction (SPF), shown here by tertile distributions of proliferation from cytokeratin (CK)-gated histograms. Figure 6. Disease-free survival (DFS) probability according to synthesis-phase fraction (SPF) status. ified by tertile distribution, DNA ploidy status, and disease stages, failed to correlate with survival. COMMENT Our extensive review of the literature (summarized in Tables 1 and 2) identified few studies showing flow cytometric DNA ploidy, DNA index, and proliferation (SPF) to have prognostic value in breast carcinoma, but did identify numerous studies that failed to confirm this clinical significance (Tables 1 and 2). These discrepancies have been explained by the low number of cases in most studies or by use of paraffin material that provides less reliable cell cycle data. 21 In addition, many studies have been based on relatively short periods of follow-up for a disease characterized by significant late recurrence and mortality. A number of investigations have failed to apply multivariate analysis to control for the effect of known powerful prognostic factors within the study groups. 22 27 Other important factors contributing to contradictory results include simplistic staining methods that fail to enrich for tumor populations and lack of standardized cell cycle/ histogram analysis methodology and hardware or software methods to correct for histogram background, aggregates, and debris. 18,21,28 In this study, we attempted to correct for each of these variables using fresh tissues from a large number of breast carcinomas and employing cell cycle analysis methodology strictly adhering to the recommendations of the 1992 DNA Flow Cytometry Consensus Conference. 13 Multiparametric DNA and cell-cycle analysis with 2-color phenotypic labeling (previously developed and quality controlled by this laboratory 18,20 ) was employed to provide precise identification of the patient-specific intrinsic diploid standard for determination of ploidy and to limit kinetic analysis to epithelial populations. This method has been shown to enhance histogram interpretation of overlapping peaks in the G0/1, near-diploid, or G2/M and near tetraploid regions with near-diploid or small aneuploid populations readily detected. 18,20 In addition, our study population had the benefit of randomized treatment assignment, wherein most of the patients in this unselected series belonged to various National Cancer Institute recommended controlled clinical trials, 29,30 and others received adjuvant chemotherapy and or tamoxifen according to the guidelines and recommendations of the National Institutes of Health Consensus Development Panel on Adjuvant Chemotherapy and Endocrine Therapy for Breast Cancer. 19 In this extensive analysis, after 12 years of ongoing prospective study, we are unable to validate any prognostic significance for flow cytometric measurements of DNA ploidy status or proliferation in these uniformly analyzed breast carcinomas. None of the flow cytometry variables showed any significant association with overall or diseasefree survival (Table 4). Our study indicates that flow cytometric DNA and cell-cycle analysis is not predictive of outcome independent of conventional staging and morphologic parameters. This finding is similar to the conclusions of other studies with long-term follow-up. 26,31 39 We conclude from our literature review that the studies reporting ploidy, SPF, or both to be of independent predictive value in at least some patient subsets can be considered exploratory in nature with multiple arbitrary cutoff points for defining high- and low-risk populations. 28,40 42 Based on our extensive analyses, determination of DNA ploidy and SPF is not a significant indicator of prognosis for any stage of breast carcinoma. The lack of association between ploidy and nodal status and the weak association between SPF and nodal metastases in our series underscores this conclusion. The absence of correlation between DNA ploidy and/or SPF and nodal metastasis has been noted by others 43 47 and may indicate that DNA content and proliferation fractions are measures of tumor growth rate and are not predictive of metastatic potential. 43,47,48 The exceptions in the literature include the study of Toikkanen et al, 49 who demonstrated a significant correlation between DNA ploidy and presence of both axillary node and distant metastasis (relationship between SPF and node status was not addressed), Hedley et al, 36 and Dressler et al, 50 in whose large series only weak correlation between SPF and lymph node metastasis was observed. However, in these latter studies no significant association was seen between aneuploid populations and nodal metastasis. The results of many other studies indicate no association as well. 33,34,51 53 In this study, the prognostic value of conventional pathologic parameters, such as tumor size, TNM classification Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al 371

components, nuclear and histologic grade (differentiation), and the presence of nodal and distant metastases, is confirmed through univariate and multivariate analysis (Table 4). Age and hormone receptor status appear to have no prognostic significance. Disease stage components are unequivocally useful in predicting disease progression in breast cancer. The importance of lymph node status and tumor size in predicting the clinical course of most patients with breast cancer is universally accepted as the gold standard, with the likelihood of axillary node metastasis strongly related to tumor size. 54 We found peritumoral vascular invasion to be significantly associated with overall and disease-free survival by univariate analysis. However, this significance was lost in multivariate analysis. Like grade, vascular invasion is significantly associated with more advanced stage. 55,56 Presence of ER and PR receptor proteins has previously been shown to be significantly correlated with diseasefree interval in a few studies of lymph node negative disease. 57,58 However, the prognostic value of ER and PR is limited to short ( 5-year) follow-up intervals. 59,60 Hormone receptor status failed to correlate with disease outcome in this study with a long (7 12-year) follow-up. None of the histologic parameters correlated with DNA ploidy, except for nuclear and histologic grade and the presence of ER proteins. DNA aneuploidy has been shown to be weakly associated with tumor size and lymph node status. 22 In larger studies, a weak correlation between DNA content and staging parameters has been observed. 25,36,50 A weak trend toward aneuploidy in cases with regional metastasis, as well as larger tumor size, is apparent in the literature review. Close correlation between DNA content and stage of tumor was reported by some authors, 25,61 whereas others, like us, have shown no such relationship. 32,62,63 Our finding of a significant association between nuclear and histologic grade with ploidy endorses the results of many other studies that have compared tumor grade and DNA ploidy. In the literature, roughly 34.0% of the welldifferentiated cancers contained aneuploid range populations, but more than 80% of poorly differentiated tumors were aneuploid. 26,36,37,54,64 Reviewing previous studies, about 55% of ER-positive breast cancers are aneuploid, compared to 73% of ER-negative cases. 26,27,36,43 In this study there was a significant association, albeit a small one, between ER status and ploidy. Sixty-eight percent of ER-negative tumors were aneuploid, compared to 58% of ER-positive tumors (P.036). Interestingly, although aneuploidy correlated with such poor prognostic features as advanced histologic grade and paucity of ERs, the failure to find a diminished survival rate for the aneuploid cases in our cohort might be related to the fact that adjuvant chemotherapy or hormonal therapy received by our patients could have affected the natural behavior of the aneuploid and diploid cancers encountered. Our data confirm previous studies showing a strong relationship between nuclear grade and SPF. 36,43 Similarly, our results corroborate previous data indicating a strong correlation between SPF and histologic grade. 38,43,64 66 Synthesis-phase fractions were approximately twice as high in poorly differentiated tumors as they were in well-differentiated tumors in many of these studies. 43,65,66 In the present study, we show a significant association between high SPF levels and negative steroid receptor status, which is in agreement with most other studies correlating ploidy and proliferation measurements with ER and PR status. 38,46,64,66 70 A significant association between tumor size and SPF was also seen in our study. A similar correlation has been noted by some authors 35,68,70 but not by others. 38,44,46,69 Our literature review revealed that SPF, but not DNA content, was generally more strongly related to other features of known prognostic value, such as tumor size, tumor stage components, and lymph node status. This finding indicates that SPF may be a more sensitive measure of biologic aggressiveness than DNA ploidy for breast cancer. 43 Similar observations were noted by Clark et al. 71 DNA ploidy and SPF did not correlate with age in our study, which is in agreement with some authors 22,50,68,72,73 but not with others. 26,39,70 The influence of flow cytometric SPF on outcome after adjuvant chemotherapy was studied by O Reilly et al 44 on 214 patients with node-positive breast carcinoma. Unlike their results, which showed an association between high SPF and poorer disease-free survival rates in premenopausal women, our study showed no significant improvement in survival, regardless of how the S phase was determined. The incidence of aneuploid tumors in our study was 53.4%. Frierson, 74 in an extensive review of 10 323 breast cancers, found 63% of tumors reported as aneuploid. The median value for tetraploidy in that review was 11%, which compares well with the incidence of 11.2% of tetraploidy in our series. Wenger and Clark, 75 in their review of 23 studies, reported a range of median SPF values from 4.8% to 14%. The intermediate range of SPF in our study, calculated as a tertile distribution of all values, was greater than 13.5% to less than 25.4%. It must be pointed out that there has been no standardization in methods or computerized algorithms for SPF calculation and whether more sophisticated 2-color gated analysis and correction for background, aggregates, and debris were undertaken, as in this study. Therefore, these literature values and ranges of SPF are not directly comparable and cannot be considered valid quality control benchmarks. In the literature, DNA aneuploid tumors have exhibited significantly higher mean SPFs than diploid range tumors. 43,50,54,64,65,76 A similar association between high SPF and aneuploid tumors was also seen in our study. This may lead one to speculate about the relationship between proliferation and neoplastic progression. It is believed that DNA content strongly correlates with total chromosome counts and fractional allelic loss in some tumors, and that invasion and metastasis involves progressive evolution of genetic and functional changes in neoplastic cells. 77 Accordingly, it is not unreasonable to conclude that uncontrolled cycling may occur simultaneously with genomic alterations, which are required for completion of the disease progression and the metastasis sequence. 60 Nevertheless, it must be pointed out that although tumor ploidy is related to its karyotype, the 2 concepts are not interchangeable. Tumor DNA index approximates the net chromosome change seen in gross karyotype alterations; however, subtle genomic abnormalities cannot be detected by assessment of ploidy. 78 The classic criticism of reliance on histopathologic observations is the human element of subjectivity involved. However, it must be recognized that there is also some amount of subjectivity in the numerous steps involved in obtaining flow cytometric results of both DNA ploidy and 372 Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al

proliferation. The assignment of DNA index and SPF is subject to a number of technical pitfalls. 79 The DNA index at which a tumor is determined as DNA aneuploid is arbitrary, depending on the quality of the flow cytometry. In instances in which the number of DNA aneuploid cells is low, the DNA aneuploid peaks may be obscured by background or debris counts, or by the DNA diploid G2/M peak. 80 It is incorrect to presume that there is little intratumoral variation of DNA content distribution, for there is good evidence in the literature of geographic DNA heterogeneity within tumors. 80 Likewise, proliferative fraction estimates show substantial intratumoral variation. 81 Synthesis-phase fraction calculations using computer models present additional problems. The SPF is not interpretable in all tumors and cannot be determined in approximately 25% or more of specimens. 29 It is also necessary to subtract background counts from the histogram, which must introduce an element of subjectivity into the process of SPF estimation. The presence of baseline debris, peak overlap (near diploid or near tetraploid populations), multiple DNA stemlines, and small aneuploid populations precludes accurate SPF calculations. There is great variability between laboratories in calculation methods of mean and median SPF values, including a lack of definitive criteria for defining high and low SPFs, with many using arbitrary cutoff values unrelated to their own patient outcomes. 34,41,82 Thus, in addition to the cost of operating and maintaining a flow cytometer, the lack of standardization, quality control, and complexity of data interpretation are major limiting factors in the routine use of flow cytometric analysis for breast as well as other human cancers. Moreover, the fact that 38% of cases in this study were not able to be evaluated for DNA ploidy or SPF because of strict quality control criteria for analysis stands as a major drawback, considering that virtually all cases were completely analyzable pathologically, most reported within 48 to 72 hours after surgical excision. On the basis of this study, we conclude that DNA ploidy and proliferation measurements do not provide significant prognostic information for clinicians to integrate into therapeutic decision making for patients with breast cancer. There appears to be no justification for the inclusion of DNA flow cytometric analysis for the purpose of routine prognostication in human breast cancer. Our study, however, reemphasizes the clinical importance of the surgical pathologic assessment of pathologic tumor stage and the precise documentation of select histomorphologic parameters of prognostic relevance. The authors gratefully acknowledge Ron Brown and Pat Mackowiac for flow cytometric analysis and technical expertise, Julie Masura, MS, for assistance in statistical analysis, and Debra Pilarski for her excellent secretarial assistance. References 1. Hillner RE, Smith TJ. Efficacy and cost effectiveness of adjuvant chemotherapy in women with node-negative breast cancer. N Engl J Med. 1991;324: 160 168. 2. Goldhirch A, Wood WC, Senn H-J, Glick JH, Gelber RD. Meeting highlights: international consensus panel on the treatment of primary breast cancer. J Natl Cancer Inst. 1995;87:1441 1445. 3. Carter CL, Allen C, Henson DE. Relation of tumor size, lymph node status, and survival in 24,740 breast cancer cases. Cancer. 1989;63:181 187. 4. Silverstein MJ, Gierson ED, Waisman JR, Senofsky GM, Colburn WJ, Gamagami P. Axillary lymph node dissection for T1a breast carcinoma. Cancer. 1994; 73:664 667. 5. Rosen PP, Groshen S. Factors influencing survival and prognosis in early breast carcinoma (T1N0M0-T1N1M0): assessment of 644 patients with median follow-up of 18 years. Surg Clin North Am. 1990;70:937 962. 6. Osborne CK. Prognostic factors in breast cancer. Principles Pract Oncol. 1990;4:1 11. 7. Elston CW, Ellis IO. Pathological prognostic factors in breast cancer, I: the value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology. 1991;19:403 410. 8. Bloom HJG, Richardson WW. Histologic grading and prognosis in breast cancer: a study of 1409 cases of which 359 have been followed for 15 years. Br J Cancer. 1957;11:359 377. 9. Lee AKC, DeLellis RA, Silverman ML, Heatley GJ, Wolfe HJ. Prognostic significance of peritumoral lymphatic and blood vessel invasion in node-negative carcinoma of the breast. J Clin Oncol. 1990;8:1457 1465. 10. Rosen PP, Saigo PE, Braun DW, Weathers E, DePalo A. Predictors of recurrences in stage I (T1N0M0) breast carcinoma. Ann Surg. 1981;193:15 25. 11. Fisher ER, Redmond C, Fisher B. Pathologic findings from the National Surgical Adjuvant Breast Project (protocol No. 4), VI: discriminants for five-year treatment failure. Cancer. 1980;46:908 918. 12. Stenkvist B, Weekman-Naeser S, Vegelius J, et al. Analysis of reproducibility of subjective grading systems of breast carcinoma. J Clin Pathol. 1979;32: 979 985. 13. Hedley DW, Clark GM, Cornelisse DJ, Killander D, Kute T, Merkel D. Consensus review of the clinical utility of DNA cytometry in carcinoma of the breast. Cytometry. 1993;14:482 485. 14. American Society of Clinical Oncology, Tumor Marker Expert Panel. Clinical practice guidelines for the use of tumor markers in breast and colorectal cancer. J Clin Oncol. 1996;14:2843 2877. 15. Fitzgibbons PL, Page DL, Weaver D, et al. Prognostic factors in breast cancer: College of American Pathologists consensus statement 1999. Arch Pathol Lab Med. 2000;124:966 978. 16. Shankey TV, Rabinovitch PS, Bagwell B, et al. Guidelines for implementation of clinical DNA cytometry. Cytometry. 1993;14:472 477. 17. Beahrs OH, Henson DE, Hutter RVP, Kennedy BJ. Manual for Staging of Cancer. 4th ed. Philadelphia, Pa: JB Lippincott; 1992. 18. Zarbo RJ, Visscher DW, Crissman JD. Two-color multiparametric method for flow cytometric DNA analysis of carcinomas using staining for cytokeratin and leukocyte-common antigen. Anal Quant Cytol Histol. 1989;11:391 402. 19. National Institutes of Health Consensus Development Conference on Adjuvant Chemotherapy and Endocrine Therapy for Breast Cancer. Bethesda, Maryland, September 9 11, 1985. Introduction and conclusions. Natl Cancer Inst Monogr. 1986;1:1 4. 20. Brown RD, Zarbo RJ, Linden MD, Torres FX, Nakhleh RE, Schultz DS. Twocolor multiparametric method for flow cytometric DNA analysis: standardization of spectral compensation. Am J Clin Pathol. 1994;101:630 637. 21. Zarbo RJ. Quality control issues and technical considerations in flow cytometric DNA and cell cycle analysis of solid tumors. In: Keren DF, Hanson H, Hurtubise P, eds. Flow Cytometry in Clinical Diagnosis. Chicago, Ill: American Society of Clinical Pathologists; 1993:470 540. 22. Keyhani-Rofagha S, O Toole RV, Farrar WB, Sickle-Santanello B, DeCenzo J, Young D. Is DNA ploidy an independent prognostic indicator in infiltrative node-negative breast adenocarcinoma? Cancer. 1990;65:1577 1582. 23. Eskelinen MJ, Pajarinen P, Collan Y, et al. Relationship between DNA ploidy and survival in patients with primary breast cancer. Br J Surg. 1989;76:830 834. 24. Roos G, Arnerlov C, Emdin S. Retrospective DNA analysis of T3/T4 breast carcinoma using cytophotometry and flow cytometry: a comparative study with prognostic evaluation. Anal Quant Cytol Histol. 1988;10:189 194. 25. Ewers S-B, Langstrom E, Baldetorp B, Killander D. Flow-cytometric DNA analysis in primary breast carcinomas and clinicopathological correlations. Cytometry. 1984;5:408 419. 26. Thorud E, Fossa SD, Vaage S, et al. Primary breast cancer: Flow cytometric DNA pattern in relation to clinical and histopathologic characteristics. Cancer. 1986;57:808 811. 27. Coulson PB, Thornthwaite JT, Woolley TW, Sugarbaker EV, Seckinger D. Prognostic indicators including DNA histogram type, receptor content, and staging related to human breast cancer patient survival. Cancer Res. 1984;44:4187 4196. 28. Wenger CR, Beardslee S, Owens MA, et al. DNA ploidy, S-phase, and steroid receptors in more than 127,000 breast cancer patients. Breast Cancer Res Treatment. 1993;28:9 20. 29. Osborne CK, Rivkin SE, McDivitt RW. Adjuvant therapy of breast cancer: Southwest Oncology Group studies. Natl Cancer Inst Monogr. 1986;1:71 74. 30. Fisher B, Redmond C, Legault-Poisson S. Post-operative chemotherapy and tamoxifen compared with tamoxifen alone in the treatment of positive-node breast cancer patients aged 50 years and older with tumors responsive to tamoxifen: results from the National Surgical Adjuvant Breast and Bowel Project B-16. J Clin Oncol. 1990;8:1005 1018. 31. Romero H, Schneider J, Burgos J, Bilbao J, Rodriguez-Escudero FJ. S-phase fraction identifies high-risk subgroups among DNA-diploid breast cancers. Am J Clin Pathol. 1996;89:586 595. 32. Kute TE, Quadri Y, Muss H, et al. Flow cytometry in node-negative breast cancer: cancer and leukemia group B protocol 8869. Cytometry. 1995;22:297 306. 33. Clark GM, Mathier M-C, Owens MA, et al. Prognostic significance of S- phase fraction in good-risk, node-negative breast cancer patients. J Clin Oncol. 1992;10:428 432. 34. Arnerlov C, Emdin SO, Lundren B, et al. Mammographic growth rate, DNA Arch Pathol Lab Med Vol 125, March 2001 Flow Cytometric Analysis of Breast Cancer Prasad et al 373