Ki67 Index and S-Phase Fraction in Human Breast Carcinomas Comparison and Correlations with Prognostic Factors PHILIPPE VIELH, M.D., SYLVIE CHEVILLARD, VERONIQUE MOSSERI, M.D., BRUNO DONATINI, M.D., AND HENRI MAGDELENAT, PH.D.
LONG-TERM FOLLOW-UP STUDIES of breast cancer patients showed that the proliferative index of the tumor is one of the most predictive biologic factors with respect to relapse-free and overall survival.21'22'35,36 In these studies, the standard method for measuring proliferation was the 3 H-thymidine labeling index (TLI), which allows one to count the proportion of cells that are synthesizing DNA at a given time. This procedure, however, requiring the use of tritiated thymidine and autoradiography, is time consuming and difficult to perform in a routine laboratory. Other techniques, including flow cytometry and immunohistology with the use of monoclonal antibodies, are currently available to study the cell cycle. The former
Departments of Pathology, Biology, and Biostatistics, Institut Curie, Paris, France
one allows one to rapidly analyze the DNA index of thousands of cells and to estimate, by computerized methods, the proportion of cells that are in the S-phase fraction (SPF). The latter one, with the use of the Ki67 monoclonal antibody that binds to a nuclear antigen only present in cycling cells,8 can be performed in frozen sections of tumors, thus allowing direct visual interpretation of labeling indices. The aims of this prospective study were, first, to compare proliferative indices obtained by the Ki67 immunohistologic staining (Ki67 index) and the flow cytometryderived SPF and, second, to correlate both methods with other pathologic and biochemical prognostic factors in a series of 148 consecutive breast adenocarcinomas treated by primary surgery. Patients and Methods Patients and Tumor Sampling
A prospective study of 148 consecutive patients with breast cancer undergoing primary surgery was conducted between September 1987 and February 1988 at the Institut Curie. Patients ranged in age from 28 to 88 years (mean, 56.5). All of the patients were women and 50 were premenopausal (34%). Clinical staging was performed according to the TNM classification:34 51 patients (34.4%) had stage I, 83 patients (56.1%) had stage II, and 11 patients (7.5%) had stage III disease. Three patients were not classified. Surgically removed breast tumor samples were obtained immediately after wide excision (76 cases) Received February 5, 1990; received revised manuscript and accepted or mastectomy (72 cases), with (141 cases) or without (7 for publication March 28. 1990. Supported in part by the Association pour la Recherche sur le Cancer cases) lateral axillary node dissection. Pathologists prepared samples of the same tumor for (1) estrogen and (grant 86241). Presented in part at the 80th Annual Meeting of the American Asprogesterone receptor dosage performed by radioligand sociation for Cancer Research, San Francisco, California, May 1989. assays using the classical dextran-coated charcoal method Address reprint requests to Dr. Vielh: Department of Pathology. Institut as previously described18; (2) histopathologic examination Curie, 75231 Paris Cedex 05, France.
681
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
In a prospective study of 148 consecutive breast adenocarcinomas, proliferative indices of the same surgical tumor sample were performed by immunohistologic staining (Ki67 index) with the use of the Ki67 monoclonal antibody, which binds to a nuclear antigen only expressed in cycling cells, and byflowcytometryderived S-phase fraction (SPF). Measurable Ki67 and SPF indices were obtained in 142 cases and 99 cases, respectively, and in 96 cases by both methods. In aneuploid tumors, a significant but low (P < 0.05, r = 0.3) relationship was observed between Ki67 index and SPF. When compared with clinical, pathologic, and biochemical parameters these two proliferative indices were shown to be associated with nuclear grading and mitotic index. Additionally, correlations were observed between Ki67 index and node involvement (P < 0.02) and between SPF and estrogen receptors (P = 0.002). These results show that (1) proliferative indices are obtained in 96% of surgical samples with Ki67 versus 67% with SPF and that (2) Ki67 index and SPF may provide complementary data with respect to prognosis. (Key words: Breast; Cancer; Cell; DNA; Kinetics; Ki67; Ploidy; S-phase fraction; Prognosis) Am J Clin Pathol 1990;94:681-686
682
A.J.C.P. • December 1990
VIELH ET AL.
on hematein and eosin-stained paraffin-embedded sections, pTNM 34 and World Health Organization classifications,14 and Scarff, Bloom, and Richardson (SBR) grading4; (3) immunohistologic Ki67 staining; and (4) DNA flow cytometry. Of the 148 breast adenocarcinomas, 135 were infiltrative (125 ductal, 10 lobular), 5 in situ (4 ductal and 1 lobular), 3 apocrine, 2 mucinous, and 1 medullary; 2 carcinomas were not classified. Immunohistologic Ki67 Staining
DNA Flow Cytometry The procedure has been described elsewhere.27 Briefly, tumor specimens (30 mg) were finely minced with curved scissors in 1 mL Young's buffer. The suspension was then transferred and homogenized in 2 mL of Young's buffer with a Dounce potter (pestle A), centrifuged (600 X g, 15 minutes), and digested by preboiled RNase A (Sigma; 1 mg/mL in PBS, 37 °C, 20 minutes). Samples were then repelleted, stained with propidium iodide (Sigma) (50 j*g/ mL in Isoton II; Coulter Electronics, Hialeah, FL), and filtered 18-24 hours later on nylon mesh (Graphosilk,
Statistical Analysis Nonparametric analyses based on the ranks of the observations and not on the data values were computed because of the subsamples' size and some extreme values. Correlations between continuous criteria were estimated by the Spearman rank correlation. Kruskal Wallis analysis of variance was performed to compare mean values of these criteria according to clinical, pathologic, and biochemical parameters. Results Ki67 Index and DNA Flow Cytometry Data Ki67 index was obtained in 142 cases, DNA index in 126 cases, SPF in 99 cases, and both KJ67 index and SPF in 96 cases of the total of 148 breast adenocarcinomas (Table 1). Ki67 index calculated by ocular micrometry ranged from 0 to 43% of labeled cells (mean, 9.4 ± 0.74) and was highly correlated with the counting by computerized image analysis (P < 0.0001, r = 0.84) in the un-
Table 1. Ki67 Index and DNA Flow Cytometry Analysis of 148 Consecutive Breast Carcinomas
Ki67 index DNA index S-phase fraction (SPF) Ki67 index and SPF
No
%
142 126 99 96
96 85 67 65
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Cryostat sections (5 ^m) were immediately prepared in embedding medium specific for frozen sections (TissueTek®, Miles Scientific) and fixed in cold acetone (-20 °C, 10 minutes), air-dried, and stored until use at —80 °C. For immunostaining, sections were washed briefly (1-2 minutes) and incubated for 30 minutes with the Ki67 (DAKOPATTS, Copenhagen, Denmark) murine monoclonal antibody (dilution 1/10). Slides were then incubated successively with the biotinylated horse antimouse antibody and the avidin-biotinylated peroxidase complex (Vector Laboratories, Burlingame, CA). Incubations were done at room temperature. Washings and dilutions were performed with phosphate-buffered saline (PBS), pH 7.6. Staining with the diaminobenzidine substrate was developed until labeling was clearly detectable. Control slides were incubated with an unrelated monoclonal antibody. Quantification of Ki67 staining (Ki67 index) was achieved either by ocular micrometry (142 cases) on a Leitz microscope (Orthoplan®) with the use of an ocular magnification of X40 with an eyepiece grid (Leitz) and also by computerized image analysis (53 cases) with the use of the Samba 200® cell image analyzer (TITN, Grenoble, France). Counterstaining was hematoxylin for ocular micrometry and/or toluidine blue for computerized image analysis. Ki67 index was based on a count of 300 consecutive tumor cells as identified by histopathologic criteria in the hematein and eosin-stained frozen sections and was expressed as a percentage of labeled cells per 100 cells.
Montreuil, France). Flow cytometric (FCM) study was performed on an FACS Analyzer I® (Becton Dickinson), and DNA histograms were derived from an analysis of 30,000 cells on a Hewlett-Packard computer with the program Consort 30® (Becton Dickinson). The diploid peak was set approximately on channel 55 with the use of lymphoblastoid or fibroadenomas cells. The coefficient of variation for G0/G1 peaks was close to 3%. Variations of the modal value of the peaks defined as diploid in each series never exceeded ± 1 channel from that defined by the external standard. In case of ambiguity, a different sample of the same tumor was used for analysis. Tumors displaying cells with DNA indices between 0.94 and 1.08 were considered to be DNA diploid and the presence of two aneuploid peaks was required to classify a tumor as multiploid. Flow cytometry-derived SPF was computed according to the rectangular model developed by Baisch and associates2,3 with exponential baseline substraction. SPFs were deemed uninterpretable in paucicellular samples, in histograms displaying overlapping stemlines, and in cases of large amounts of debris.
Vol. 94 • No. 6
Kl67 INDEX VS. SPF IN BREAST CANCER
selected subset of 53 cases where both methods were used. The mean value of Ki67 index observed in infiltrative ductal carcinomas was as follows: 10 ± 0.77 and was not modified when the five in situ carcinomas were excluded (range, 2-6%). Evidence of nuclear staining was seen in 112 cases (79%), and patterns of Ki67 immunostaining were found to be either diffuse in 20 cases or dotted and diffuse (Fig. 1) in 92 cases. FCM-derived DNA index was found to be diploid in 61 cases and aneuploid in 65 cases, including 2 hypodiploid and 4 multiploid cases. SPF ranged in percentage from less than 1 to 30 (mean, 7 ± 0.54) and from less than 1 to 30 (mean, 6.5 ± 0.5) in infiltrative ductal carcinomas. Comparison of Ki67 Index with SPF
FIG. 1. Ki67 immunostaining on frozen section. Dotted and diffuse patterns. Hematoxylin counterstaining (X320).
first one in which the Ki67 index was higher than the SPF, another one showing the converse, and a third one in which K.i67 index and SPF were similar. Comparison ofKi67 Index and SPF with Prognostic Factors No correlation was found between the two proliferative indices, the age of the patients, and the stage or the size of the tumor. Correlations with other prognostic factors are summarized in Table 2. As shown in Table 2, a positive relationship was found between the two proliferative indices and the nuclear grading and mitotic index. In addition, the KJ67 index was correlated with the presence of a node involvement (N + vs. N - ) in the axillary dissection but not with the level of estrogen receptors. Conversely, SPF showed an inverse relationship with the level of estrogen receptors but not with node involvement. Discussion The multivariate analysis of histopathologic factors influencing clinical outcome of breast cancer5 indicated that SBR grading of the tumor and node involvement of the axillary dissection were the main prognostic features. However, some difficulties in assigning reproducible scores to histologic grading prompted investigators to evaluate more quantitative biologic tumoral factors, such as estrogen and progesterone receptors, DNA index, and cell-cycle kinetics, to better estimate prognosis. Tumor cell kinetics has been studied extensively by the TLI method, and longterm results showed that the TLI is a strong stage-independent prognostic factor with regard to relapse-free and overall survival.23,3137 Because TLI determination requires the incubation of tumor cells with tritiated thymidine and the counting of the percentage of cells that take up the isotope by autoradiography that are difficult to apply in a routine laboratory, we attempted to compare in this prospective series two other nonradioactive methods of cell-cycle analysis and to correlate the results of both methods with histopathologic and biochemical prognostic factors. Addressing 148 surgical samples, K.i67 index and SPF data were obtained in 96% and 67% of the cases, respectively. These discrepancies resulted from difficulties encountered in obtaining interpretable SPFs that were sometimes obscured by debris of necrosis and overlapping peaks of aneuploid tumors. Comparison of K.i67 index with SPF showed a positive relationship between these two variables as pointed out by Walker and Camplejohn38 in a retrospective study on formalin-fixed and paraffinembedded breast cancer for the determination of SPF. Surprisingly, we observed that this significant correlation
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Ki67 index and SPF were compared in 96 cases (65%), showing a significant but low correlation (data not shown). Taking into account the DNA index (Fig. 2), this positive relationship was only found in aneuploid tumors (Fig. 2B). In these tumors three situations were identified: a
683
A.J.C.P. • December 1990
VIELH ET AL.
684
40
30
20
X
fc4
0
10
20
30
40
S-PHASE FRACTION FIG. 2. Comparison of Ki67 index with SPF in diploid (A) and aneuploid (B) breast carcinomas.
was restricted to aneuploid tumors (Fig. 2B) in which three situations were observed: a first with a strict correlation between the two variables; a second one in which the higher KJ67 index compared with SPF can be explained by the fact that Ki67 binds to a nuclear antigen expressed during the S-phase but also during the G l , G2, and Mphases of the cell cycle9; and a third one composed of carcinomas in which Ki67 index was lower than SPF. This feature could be explained in some cases by the overestimation of SPF when suspensions of tumor cells are contaminated by large amounts of inflammatory and stromal cells23,24; in other cases it could be explained by
Table 2. Comparison of KJ67 Index and SPF with Prognostic Factors Ki67 Index Nuclear grading* Mitotic index* Node involvement (N + vs. N") Estrogen receptors (RO + vs. RCT) SBR criteria (*) NS = not significant
LONG-TERM FOLLOW-UP STUDIES of breast cancer patients showed that the proliferative index of the tumor is one of the most predictive biologic factors with respect to relapse-free and overall survival.21'22'35,36 In these studies, the standard method for measuring proliferation was the 3 H-thymidine labeling index (TLI), which allows one to count the proportion of cells that are synthesizing DNA at a given time. This procedure, however, requiring the use of tritiated thymidine and autoradiography, is time consuming and difficult to perform in a routine laboratory. Other techniques, including flow cytometry and immunohistology with the use of monoclonal antibodies, are currently available to study the cell cycle. The former
Departments of Pathology, Biology, and Biostatistics, Institut Curie, Paris, France
one allows one to rapidly analyze the DNA index of thousands of cells and to estimate, by computerized methods, the proportion of cells that are in the S-phase fraction (SPF). The latter one, with the use of the Ki67 monoclonal antibody that binds to a nuclear antigen only present in cycling cells,8 can be performed in frozen sections of tumors, thus allowing direct visual interpretation of labeling indices. The aims of this prospective study were, first, to compare proliferative indices obtained by the Ki67 immunohistologic staining (Ki67 index) and the flow cytometryderived SPF and, second, to correlate both methods with other pathologic and biochemical prognostic factors in a series of 148 consecutive breast adenocarcinomas treated by primary surgery. Patients and Methods Patients and Tumor Sampling
A prospective study of 148 consecutive patients with breast cancer undergoing primary surgery was conducted between September 1987 and February 1988 at the Institut Curie. Patients ranged in age from 28 to 88 years (mean, 56.5). All of the patients were women and 50 were premenopausal (34%). Clinical staging was performed according to the TNM classification:34 51 patients (34.4%) had stage I, 83 patients (56.1%) had stage II, and 11 patients (7.5%) had stage III disease. Three patients were not classified. Surgically removed breast tumor samples were obtained immediately after wide excision (76 cases) Received February 5, 1990; received revised manuscript and accepted or mastectomy (72 cases), with (141 cases) or without (7 for publication March 28. 1990. Supported in part by the Association pour la Recherche sur le Cancer cases) lateral axillary node dissection. Pathologists prepared samples of the same tumor for (1) estrogen and (grant 86241). Presented in part at the 80th Annual Meeting of the American Asprogesterone receptor dosage performed by radioligand sociation for Cancer Research, San Francisco, California, May 1989. assays using the classical dextran-coated charcoal method Address reprint requests to Dr. Vielh: Department of Pathology. Institut as previously described18; (2) histopathologic examination Curie, 75231 Paris Cedex 05, France.
681
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
In a prospective study of 148 consecutive breast adenocarcinomas, proliferative indices of the same surgical tumor sample were performed by immunohistologic staining (Ki67 index) with the use of the Ki67 monoclonal antibody, which binds to a nuclear antigen only expressed in cycling cells, and byflowcytometryderived S-phase fraction (SPF). Measurable Ki67 and SPF indices were obtained in 142 cases and 99 cases, respectively, and in 96 cases by both methods. In aneuploid tumors, a significant but low (P < 0.05, r = 0.3) relationship was observed between Ki67 index and SPF. When compared with clinical, pathologic, and biochemical parameters these two proliferative indices were shown to be associated with nuclear grading and mitotic index. Additionally, correlations were observed between Ki67 index and node involvement (P < 0.02) and between SPF and estrogen receptors (P = 0.002). These results show that (1) proliferative indices are obtained in 96% of surgical samples with Ki67 versus 67% with SPF and that (2) Ki67 index and SPF may provide complementary data with respect to prognosis. (Key words: Breast; Cancer; Cell; DNA; Kinetics; Ki67; Ploidy; S-phase fraction; Prognosis) Am J Clin Pathol 1990;94:681-686
682
A.J.C.P. • December 1990
VIELH ET AL.
on hematein and eosin-stained paraffin-embedded sections, pTNM 34 and World Health Organization classifications,14 and Scarff, Bloom, and Richardson (SBR) grading4; (3) immunohistologic Ki67 staining; and (4) DNA flow cytometry. Of the 148 breast adenocarcinomas, 135 were infiltrative (125 ductal, 10 lobular), 5 in situ (4 ductal and 1 lobular), 3 apocrine, 2 mucinous, and 1 medullary; 2 carcinomas were not classified. Immunohistologic Ki67 Staining
DNA Flow Cytometry The procedure has been described elsewhere.27 Briefly, tumor specimens (30 mg) were finely minced with curved scissors in 1 mL Young's buffer. The suspension was then transferred and homogenized in 2 mL of Young's buffer with a Dounce potter (pestle A), centrifuged (600 X g, 15 minutes), and digested by preboiled RNase A (Sigma; 1 mg/mL in PBS, 37 °C, 20 minutes). Samples were then repelleted, stained with propidium iodide (Sigma) (50 j*g/ mL in Isoton II; Coulter Electronics, Hialeah, FL), and filtered 18-24 hours later on nylon mesh (Graphosilk,
Statistical Analysis Nonparametric analyses based on the ranks of the observations and not on the data values were computed because of the subsamples' size and some extreme values. Correlations between continuous criteria were estimated by the Spearman rank correlation. Kruskal Wallis analysis of variance was performed to compare mean values of these criteria according to clinical, pathologic, and biochemical parameters. Results Ki67 Index and DNA Flow Cytometry Data Ki67 index was obtained in 142 cases, DNA index in 126 cases, SPF in 99 cases, and both KJ67 index and SPF in 96 cases of the total of 148 breast adenocarcinomas (Table 1). Ki67 index calculated by ocular micrometry ranged from 0 to 43% of labeled cells (mean, 9.4 ± 0.74) and was highly correlated with the counting by computerized image analysis (P < 0.0001, r = 0.84) in the un-
Table 1. Ki67 Index and DNA Flow Cytometry Analysis of 148 Consecutive Breast Carcinomas
Ki67 index DNA index S-phase fraction (SPF) Ki67 index and SPF
No
%
142 126 99 96
96 85 67 65
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Cryostat sections (5 ^m) were immediately prepared in embedding medium specific for frozen sections (TissueTek®, Miles Scientific) and fixed in cold acetone (-20 °C, 10 minutes), air-dried, and stored until use at —80 °C. For immunostaining, sections were washed briefly (1-2 minutes) and incubated for 30 minutes with the Ki67 (DAKOPATTS, Copenhagen, Denmark) murine monoclonal antibody (dilution 1/10). Slides were then incubated successively with the biotinylated horse antimouse antibody and the avidin-biotinylated peroxidase complex (Vector Laboratories, Burlingame, CA). Incubations were done at room temperature. Washings and dilutions were performed with phosphate-buffered saline (PBS), pH 7.6. Staining with the diaminobenzidine substrate was developed until labeling was clearly detectable. Control slides were incubated with an unrelated monoclonal antibody. Quantification of Ki67 staining (Ki67 index) was achieved either by ocular micrometry (142 cases) on a Leitz microscope (Orthoplan®) with the use of an ocular magnification of X40 with an eyepiece grid (Leitz) and also by computerized image analysis (53 cases) with the use of the Samba 200® cell image analyzer (TITN, Grenoble, France). Counterstaining was hematoxylin for ocular micrometry and/or toluidine blue for computerized image analysis. Ki67 index was based on a count of 300 consecutive tumor cells as identified by histopathologic criteria in the hematein and eosin-stained frozen sections and was expressed as a percentage of labeled cells per 100 cells.
Montreuil, France). Flow cytometric (FCM) study was performed on an FACS Analyzer I® (Becton Dickinson), and DNA histograms were derived from an analysis of 30,000 cells on a Hewlett-Packard computer with the program Consort 30® (Becton Dickinson). The diploid peak was set approximately on channel 55 with the use of lymphoblastoid or fibroadenomas cells. The coefficient of variation for G0/G1 peaks was close to 3%. Variations of the modal value of the peaks defined as diploid in each series never exceeded ± 1 channel from that defined by the external standard. In case of ambiguity, a different sample of the same tumor was used for analysis. Tumors displaying cells with DNA indices between 0.94 and 1.08 were considered to be DNA diploid and the presence of two aneuploid peaks was required to classify a tumor as multiploid. Flow cytometry-derived SPF was computed according to the rectangular model developed by Baisch and associates2,3 with exponential baseline substraction. SPFs were deemed uninterpretable in paucicellular samples, in histograms displaying overlapping stemlines, and in cases of large amounts of debris.
Vol. 94 • No. 6
Kl67 INDEX VS. SPF IN BREAST CANCER
selected subset of 53 cases where both methods were used. The mean value of Ki67 index observed in infiltrative ductal carcinomas was as follows: 10 ± 0.77 and was not modified when the five in situ carcinomas were excluded (range, 2-6%). Evidence of nuclear staining was seen in 112 cases (79%), and patterns of Ki67 immunostaining were found to be either diffuse in 20 cases or dotted and diffuse (Fig. 1) in 92 cases. FCM-derived DNA index was found to be diploid in 61 cases and aneuploid in 65 cases, including 2 hypodiploid and 4 multiploid cases. SPF ranged in percentage from less than 1 to 30 (mean, 7 ± 0.54) and from less than 1 to 30 (mean, 6.5 ± 0.5) in infiltrative ductal carcinomas. Comparison of Ki67 Index with SPF
FIG. 1. Ki67 immunostaining on frozen section. Dotted and diffuse patterns. Hematoxylin counterstaining (X320).
first one in which the Ki67 index was higher than the SPF, another one showing the converse, and a third one in which K.i67 index and SPF were similar. Comparison ofKi67 Index and SPF with Prognostic Factors No correlation was found between the two proliferative indices, the age of the patients, and the stage or the size of the tumor. Correlations with other prognostic factors are summarized in Table 2. As shown in Table 2, a positive relationship was found between the two proliferative indices and the nuclear grading and mitotic index. In addition, the KJ67 index was correlated with the presence of a node involvement (N + vs. N - ) in the axillary dissection but not with the level of estrogen receptors. Conversely, SPF showed an inverse relationship with the level of estrogen receptors but not with node involvement. Discussion The multivariate analysis of histopathologic factors influencing clinical outcome of breast cancer5 indicated that SBR grading of the tumor and node involvement of the axillary dissection were the main prognostic features. However, some difficulties in assigning reproducible scores to histologic grading prompted investigators to evaluate more quantitative biologic tumoral factors, such as estrogen and progesterone receptors, DNA index, and cell-cycle kinetics, to better estimate prognosis. Tumor cell kinetics has been studied extensively by the TLI method, and longterm results showed that the TLI is a strong stage-independent prognostic factor with regard to relapse-free and overall survival.23,3137 Because TLI determination requires the incubation of tumor cells with tritiated thymidine and the counting of the percentage of cells that take up the isotope by autoradiography that are difficult to apply in a routine laboratory, we attempted to compare in this prospective series two other nonradioactive methods of cell-cycle analysis and to correlate the results of both methods with histopathologic and biochemical prognostic factors. Addressing 148 surgical samples, K.i67 index and SPF data were obtained in 96% and 67% of the cases, respectively. These discrepancies resulted from difficulties encountered in obtaining interpretable SPFs that were sometimes obscured by debris of necrosis and overlapping peaks of aneuploid tumors. Comparison of K.i67 index with SPF showed a positive relationship between these two variables as pointed out by Walker and Camplejohn38 in a retrospective study on formalin-fixed and paraffinembedded breast cancer for the determination of SPF. Surprisingly, we observed that this significant correlation
Downloaded from http://ajcp.oxfordjournals.org/ by guest on June 5, 2016
Ki67 index and SPF were compared in 96 cases (65%), showing a significant but low correlation (data not shown). Taking into account the DNA index (Fig. 2), this positive relationship was only found in aneuploid tumors (Fig. 2B). In these tumors three situations were identified: a
683
A.J.C.P. • December 1990
VIELH ET AL.
684
40
30
20
X
fc4
0
10
20
30
40
S-PHASE FRACTION FIG. 2. Comparison of Ki67 index with SPF in diploid (A) and aneuploid (B) breast carcinomas.
was restricted to aneuploid tumors (Fig. 2B) in which three situations were observed: a first with a strict correlation between the two variables; a second one in which the higher KJ67 index compared with SPF can be explained by the fact that Ki67 binds to a nuclear antigen expressed during the S-phase but also during the G l , G2, and Mphases of the cell cycle9; and a third one composed of carcinomas in which Ki67 index was lower than SPF. This feature could be explained in some cases by the overestimation of SPF when suspensions of tumor cells are contaminated by large amounts of inflammatory and stromal cells23,24; in other cases it could be explained by
Table 2. Comparison of KJ67 Index and SPF with Prognostic Factors Ki67 Index Nuclear grading* Mitotic index* Node involvement (N + vs. N") Estrogen receptors (RO + vs. RCT) SBR criteria (*) NS = not significant
