- Email: [email protected]
Flow cytometric measurements of DNA index and S-phase on paraffin-embedded early stage endometrial cancer: An important prognostic indicator
GYNECOLOGIC
ONCOLOGY
35, 50-54
(1989)
Flow Cytometric Measurements of DNA Index and S-Phase on ParaffinEmbedded Early Stage Endometrial Cancer: An Important Prognostic Indicator PER ROSENBERG,* Departments
STEN
WINciREN,t
of *Gynecological
ERNST SIMONSEN, * OLLE STAL,~
BJGRN RISBERG,$
Oncology,
University
tOncology,
and $Pathology,
Hospital,
AND Bo NORDENSKJ~LD~ Linktipping,
Sweden
Received February 2, 1989
Paraffin-embeddedcurettage material from 120 patients with stage I and II endometrial cancer were analyzed by flow cytometry. The follow-up time in all caseswas at least 5 years or until death. It waspossibleto determineDNA ploidy in 111cases and S-phasewasalsoevaluatedin 92/l 11 cases.DNA ploidy and S-phasewere then comparedwith the FIG0 grading and clinical outcome.DNA ploidy resultsshowedthat 11%of grade1 tumors were aneuploid, 14% of grade 2, 42% of grade 3, and 85% of the casesof uterine papillary serouscarcinoma (UPSC) were aneuploid. Patientswith tumors that showedan S-phasefraction below5% had a cumulative S-yearcancermortality of 7% whereas 49% of the patients with tumors having S-phasefraction above 10% died of their cancer. The prognostic significanceof DNA and S-phasecorrelated fairly well with the FIG0 grading as determined by a pathologist with specialinterest in gynecologic oncologyand the DNA parametersaddedprognosticinformation independentof the FIG0 grading. Q 1989 Academic PM, IIK. INTRODUCTION
Endometrial carcinoma is one of the most common female malignancies, and the age-adjusted incidence is increasing in the Western world. A majority of the patients have tumors that are diagnosed and treated in early stages and the treatment results have generally been very good. However, a minority of these patients still will die from their disease despite early detection and treatment. We therefore need parameters that indicate those patients with a poor prognosis. Conventional prognosticators are, according to FIG0 (1971), stage, age, clinical condition of the patient, the uterine size, and the histological grade [ 11. Despite the fact that the FIG0 grading system is a simple method emphasizing only the architectural pattern, there is a remarkable variability of 288% in the frequency of grade 1, stage I endometrial cancer from different institutions [2]. Nuclear grading has also been shown to have prognostic significance [3,4]
MATERIAL
AND METHODS
Patients There were 120 cases of endometrial carcinoma stages 1 and II selected from a total of 300 early stage endometrial cancer patients treated in Linkoping, Sweden from 1979 to 1981 (stage II defined as endometrial cancer growth within the cervical mucosa shown in the curettage material). Only curettage material obtained before 50
0090-8258189 $1 SO Copyright 0 1989 by Academic Press, Inc. All rights of reproduction in any form reserved.
but it is somewhat subjective. The prognostic value of morphometrical methods must be assessed in cooperation with the pathologist. A more objective method is clearly needed. The prognostic value of DNA index, when determined by flow cytometry on fresh or fresh frozen material, is well established for several different tumor types, including endometrial carcinoma [5-71. For some years we have used different methods of performing flow cytometry on paraffin-embedded archival cancer material from different tumor types. This method allows retrospective analysis of tumors from groups of patients whose outcome is known and also has the advantage of minimizing the difficulties connected with the handling of fresh or fresh frozen tumor samples. We have found that the method for single cell preparation proposed by Schtitte [8], followed by cell staining according to Vindelov [9] has given good results as judged by minimal amounts of debris, low coefficient of variation, and reliable S-phase measurements. In a recent study [lo], we have employed the Schtitte method and shown a strong correlation between DNA indices and S-phases from fresh and fixed breast cancers. Encouraged by these results we performed the present study on archival material from 120 patients with endometrial cancer and analyzed the prognostic value of the DNA index and S-phase fraction.
DNA AND S-PHASE AS PROGNOSTIC
the start of therapy was used. We analyzed a11 FIG0 grade 3 tumors (n = 15) and all cases of UPSC (n = 22) where curettage material was available in our archives. Then 40 grade 1 and 43 grade 2 cases were additionally selected at random from the archives without prior knowledge of the clinical outcome. UPSC was diagnosed according to the histological criteria of Hendrickson and Kempson and co-workers [ 11,121. No attempt was made to grade UPSC tumors according to FIGO. All cases were examined by the same pathologist (B.R.) without any knowledge of the clinical course. The follow-up time for all patients was at least 5 years or until the death of the patient. The mean age was 65 years which did not differ significantly between different grades or between patients with UPSC and those with nonUPSC tumors. The principal treatment was consistent throughout the period studied (for stage I grade I tumors: abdominal hysterectomy and bilateral salpingoophorectomy (TAH + BSO), followed by one vaginal intracavitary radiation treatment; for stage I grade 2 or 3 tumors, and stage II (all grades) and UPSC: three preoperative intracavitary radium applications by the Heymans packing technique at 2-weeks intervals, followed by TAH + BSO). Preparation
and Flow Cytometry
51
INDICATORS
TABLE 1 DNA Ploidy Correlated to Histology Histology
No.
Aneuploid
Grade 1 Grade 2 Grade 3 UPSC
37 43 12 20
II 14 42 85
%
was estimated by multiplying the number of channels between the G,/G, and GZ/M peaks by the mean number of cells per channel in the mid-S-phase range. Statistical
Methods
The statistical calculations were made with the BMDP software program [13] using Cox regression models and life tables [ 141. RESULTS DNA Analysis Correlated
to Histology
In 9 cases there was insufficient amounts of material for the analysis or the histogram showed too low quality to be acceptable. An aneuploid DNA histogram was seen in 32 of 111 tumors. The mean CV was 5.7%. The Sphase fraction (i.e., the fraction of cells with a DNA content between Go/G, and Gz/M values), corresponding to the DNA content expected to be found in proliferating cells, was possible to be determined in 92 cases. Ploidy correlated with the different histological groups. Grade 1 tumors showed the lowest percentage of aneuploidy and UPSC the highest (Table 1). The mean Sphase percentage also correlated with the histology (Table 2). Diploid tumors (n = 79) had a mean S-phase of 5.6% (range 1.6-I I .4%) and aneuploid tumors (n = 32) showed a mean S-phase frequency of 13.9% (range 5.9-
The method we employed has been described elsewhere [S]. In brief, it involves the preparation of SO-pm sections of formalin-fixed, paraffin-embedded curettage specimens. We used the entire 50 sections without attempting to identify and cut out specific areas. After deparaffinization with xylene, the specimen was rehydrated stepwise in ethanol of decreasing concentrations. The tissue samples were then treated with trypsin overnight to create a single cell suspension. After filtration through a nylon mesh, the cell suspension was prepared as described by Vindelov using propidium iodide as the 22.8%). fluorescent dye [9]. We used a Leitz MPV flow cytometer DNA Analysis and Mortality and a Monroe OC 8888 microcomputer. Only histograms including at least 20,000 cells, with a CV of less than The cancer mortality rate was closely related to both 8%, and with at most a moderate level of debris, were DNA ploidy and S-phase fraction (Tables 3 and 4, Figs. recorded. The peak with the lowest DNA value was 1 and 2). Patients with aneuploid tumors had a relative considered to contain diploid cells and served as a diploid cancer mortality rate four times higher than that of their internal reference. The coefficient of variation (CV) was calculated from the approximation for a Gaussian disTABLE 2 tribution estimating the width of the peak at half the S-phases (Mean Value) Correlated to Histology maximum height. The S-phases was analyzed after a background correction was performed by the subtraction S-phase (mean %) Histology No. of a constant number of cells from all channels in the Grade I 33 4.3 DNA histogram. The constant was taken as the mean 7.1 Grade 2 35 number of cells per channel at an interval to the right Grade 3 8 10.5 of the Gz/M peak. Then assuming a rectangular distriUPSC 17 12.7 bution of cells in S-phase, the number of S-phase cells
52
ROSENBERG
TABLE 3 Cancer Mortality Related to Ploidy and S-Phase: Univariate Analysis No. of No. of Cum. 5-year pts. deaths cancer mortality
Rel. death rate
Test for trend
DNA ploidy Diploid Aneuploid
79 32
9 12
13% 42%
1.0 4.2
P = 0.0013
S-phase 4% 5-10% >lO%
33 40 19
2 6 9
7% 17% 49%
1.0 2.5 9.9
P = 0.0010
TABLE 4 Cancer Mortality Related to Ploidy and S-Phase: Multivariate Analysis No.
No. of deaths
Rel. death rate
Test for trend
72 20
8 9
1.0 2.4
P = 0.16
33 40 19
2 6 9
1.0 2.1 5.2
P = 0.07
DNA ploidy Diploid Aneuploid S-phase <5% 5-10% >10%
ET AL.
diploid counterparts, and patients in the high S-phase groups (S-phase fraction >lO%) had almost a lo-fold increase in the cancer mortality rate compared to the low S-phase group (S-phase fraction 4%). We formed three risk groups by combining ploidy and S-phase: Group 1 included diploid tumors with S-phases 4%. Group 3 contained aneuploid tumors with S-phases >lO% and group 2 contained all other tumors. Patients in risk group 3 had a 15.5fold increase in the cancer mortality rate when compared to patients in group 1 (Table 5). Patients with group 1 tumors died of their cancer in only 7% of the cases, whereas patients having group 3 tumors succumbed to their malignancy in 63% of the cases. The group variables were entered into a stepwise Cox regression analysis together with stage, age, nuclear grade, and histology (FIG0 grade l-2 versus grade 3 or UPSC). Only FIG0 grade and ploidy/Sphase contributed as independent prognostic variables (Table 6). The other variables did not add any further significant prognostic information once corrected for histology and ploidy/S-phase. DISCUSSION
We did not try to cut out or isolate specific areas containing only cancer cells on the paraffin sections before the analysis because, even when there is a large homogenous area of pure cancer on the 5-pm stained glass sections, it is often virtually impossible to recognize the corresponding area on the 50-pm unstained paraffin sections. When paraffin-embedded material is analyzed, chicken or trout red blood cells, which we use with fresh tumors, are not reliable reference cells. As an alternative, nonmalignant cells contained in the tumor can serve as
Cumulative survival
S-phase
3020 lo-
12
I
1
1
24
36
48
I 6o
Time (months)
FIG. 1. Cumulative cancer-free survival correlated to ploidy and S-phase percentage.
10%
DNA AND S-PHASE AS PROGNOSTIC Cumulative survival % loo-
53
INDICATORS
I ,
I
go-
I_
ao-
I
Group
1
Group
2
Group
3
6050II---------
- ------
- -----
---_-------
20lo-
I 24
12
, 36
I 48
I 6o
lime (months)
FIG. 2. Cumulative cancer-free survival correlated to three risk groups. Group 1, diploid tumors with S-phase ~5%; group 3, aneuploid tumors with S-phase >lO%; and group 2, other tumors.
TABLE 5 Risk Groups According to Ploidy and S-PhaseValue
Group I: Diploid tumor with S-phase <5% Group 2: Other tumors Group 3: Aneuploid tumors with S-phase >lO%
Cancer deaths
Cum. S-year ca. death rate
Adj. rel. death rate
2/33
7%
I.0
IO/63
18%
2.7
9/15
63%
15.5
TABLE 6 StepwiseCox RegressionAnalysis” No. of pts.
No. of deaths
Adj. rel. death rate
Test for trend
Histology Grade 1-2 Grade 3 + UPSC
79 32
7 14
I.0 3.5
P = 0.03
Ploidy/S-phase Group 1 Group 2 Group 3
33 63 I5
2 IO 9
1.0 1.8 5.0
P = 0.05
” Variables included stage age, nuclear grade, FIG0 grade 1-2 versus 3, and UPSC, ploidy/S-phase. Only histology (grade 1-2 vs grade 3, UPSC) and ploidy/S-phase groups showed independent prognostic value. Once corrected for these two variables neither stage, age, nor nuclear grade had any additional prognostic significance.
Test for trend
P = 0.0001
a reference. We found that the paraffin-embedded sections almost always contained sufficient amounts of normal cells to be used as an internal standard. This introduces, however, a risk of missing or misinterpreting hypodiploid tumors, which fortunately are rare. In nine cases there was either insufficient material or the histograms were of too low quality to allow an accurate analysis. However, the majority of the tumor material analyzed was obtained from a curettage done immediately before the first intracavitary treatment and not from the original diagnostic curettage performed 2-3 weeks earlier. This may in part explain why there is sometimes an insufficient quantity of material. We can not prove that the estimated S-phases concerned only proliferating tumor cells. Nevertheless the S-phases did correlate strongly with the clinical outcome. The only parameters in the multivariate analysis which yielded significant
54
ROSENBERG
prognostic information were histology (FIG0 grade l-2 vs FIG0 grade 3 and UPSC) and DNA ploidy/S-phase. The flow cytometric method is probably more objective and comparable among different institutions than the FIG0 grading system. In summary, we conclude that the prognostic value of this method is significant and seems to be comparable to the FIG0 grading, but without inherent subjectivity. Using paraffin-embedded material makes it possible to analyze curettage samples prepared from routine histopathologic examinations and is more convenient than working with fresh or frozen tumor tissue.
ET AL.
6.
7.
8.
9.
10.
REFERENCES 1. Cancer Committee of the General Assembly FIGO. Classification and staging of malignant tumors in female pelvis, ht. J. Gynecol. Obstet.9, 172-179 (1971). 2. Annual report on the results of treatment in gynecological cancer, Stockholm, FIGO, Vol. 18 (1982). 3. Christopherson, W. M., Conelly, P. J., and Alberhasky, R. C. Carcinoma of the endometrium. V. An analysis of prognosticators in patients with favourable subtypes and stage I disease, Cancer 51, 1705-1709 (1983). 4. Soderlin, E. Factors affecting prognosis of endometrial carcinoma, Acta Obstet. Gynecol. Scud. Suppl. 38 (1974). 5. Iversen, O., and Laerum, 0. Ploidy disturbances in endometrial
11.
12.
13. 14.
and ovarian carcinomas: A review, Anal. Quant. Cytol. Histol. 7, 327-336 (1985). Friedlander, M., Hedley, D., and Taylor, I. Clinical and biological significance of aneuploidy in human tumors, J. C/in. Pathol. 37, 961-974 (1984). Atkin, N. Prognostic significance of ploidy level in human tumors. I. Carcinoma of the uterus, J. Natl. Cancer Inst. 56, 909-910 (1976). Schutte, B., Reynders, M., Bosman, F., and Blijham, G. Flow cytometric determination of DNA ploidy level in nuclei isolated from paraffin-embedded tissue, Cytometry 6, 26-30 (1985). Vindelov, L., Christensen, I., and Nissen, N. A detergent-trypsin method for the preparation of nuclei for flow cytometric DNA analysis, Cytometry 3, 323-327 (1983). Wingren, S., Hatschek, T., Stal, O., Boeryd, B., and Nordenskjold, B. Comparison of static and flow cytomfluorometry for estimation of DNA index and S-phase in fresh and paraffin-embedded breast carcinoma. Acta Oncol. &and. 27, 793-797 (1988). Hendrickson, M., Ross, J., Eifel, P. J., Cox, R. S., Martinez, A., and Kempson, R. Adenocarcinoma of the endometrium: Analysis of 256 cases with carcinoma limited to the uterine corpus, Gynecol. Oncol. 13, 373-392 (1982). Hendrickson, M., Ross, J., Eifel, P. J., Martinez, A., and Kempson, R. Uterine papillary serous carcinoma, Amer. J. Surg. Pathol. 6, 93-108 (1982). BMDP. Statistical software, Univ. Cilifornia Press, Berkeley, pp. 187-220 (1981). Cox, D. Regression models and life tables, J. R. Stat. SOC. B. 34, 187-220 (1972).
ONCOLOGY
35, 50-54
(1989)
Flow Cytometric Measurements of DNA Index and S-Phase on ParaffinEmbedded Early Stage Endometrial Cancer: An Important Prognostic Indicator PER ROSENBERG,* Departments
STEN
WINciREN,t
of *Gynecological
ERNST SIMONSEN, * OLLE STAL,~
BJGRN RISBERG,$
Oncology,
University
tOncology,
and $Pathology,
Hospital,
AND Bo NORDENSKJ~LD~ Linktipping,
Sweden
Received February 2, 1989
Paraffin-embeddedcurettage material from 120 patients with stage I and II endometrial cancer were analyzed by flow cytometry. The follow-up time in all caseswas at least 5 years or until death. It waspossibleto determineDNA ploidy in 111cases and S-phasewasalsoevaluatedin 92/l 11 cases.DNA ploidy and S-phasewere then comparedwith the FIG0 grading and clinical outcome.DNA ploidy resultsshowedthat 11%of grade1 tumors were aneuploid, 14% of grade 2, 42% of grade 3, and 85% of the casesof uterine papillary serouscarcinoma (UPSC) were aneuploid. Patientswith tumors that showedan S-phasefraction below5% had a cumulative S-yearcancermortality of 7% whereas 49% of the patients with tumors having S-phasefraction above 10% died of their cancer. The prognostic significanceof DNA and S-phasecorrelated fairly well with the FIG0 grading as determined by a pathologist with specialinterest in gynecologic oncologyand the DNA parametersaddedprognosticinformation independentof the FIG0 grading. Q 1989 Academic PM, IIK. INTRODUCTION
Endometrial carcinoma is one of the most common female malignancies, and the age-adjusted incidence is increasing in the Western world. A majority of the patients have tumors that are diagnosed and treated in early stages and the treatment results have generally been very good. However, a minority of these patients still will die from their disease despite early detection and treatment. We therefore need parameters that indicate those patients with a poor prognosis. Conventional prognosticators are, according to FIG0 (1971), stage, age, clinical condition of the patient, the uterine size, and the histological grade [ 11. Despite the fact that the FIG0 grading system is a simple method emphasizing only the architectural pattern, there is a remarkable variability of 288% in the frequency of grade 1, stage I endometrial cancer from different institutions [2]. Nuclear grading has also been shown to have prognostic significance [3,4]
MATERIAL
AND METHODS
Patients There were 120 cases of endometrial carcinoma stages 1 and II selected from a total of 300 early stage endometrial cancer patients treated in Linkoping, Sweden from 1979 to 1981 (stage II defined as endometrial cancer growth within the cervical mucosa shown in the curettage material). Only curettage material obtained before 50
0090-8258189 $1 SO Copyright 0 1989 by Academic Press, Inc. All rights of reproduction in any form reserved.
but it is somewhat subjective. The prognostic value of morphometrical methods must be assessed in cooperation with the pathologist. A more objective method is clearly needed. The prognostic value of DNA index, when determined by flow cytometry on fresh or fresh frozen material, is well established for several different tumor types, including endometrial carcinoma [5-71. For some years we have used different methods of performing flow cytometry on paraffin-embedded archival cancer material from different tumor types. This method allows retrospective analysis of tumors from groups of patients whose outcome is known and also has the advantage of minimizing the difficulties connected with the handling of fresh or fresh frozen tumor samples. We have found that the method for single cell preparation proposed by Schtitte [8], followed by cell staining according to Vindelov [9] has given good results as judged by minimal amounts of debris, low coefficient of variation, and reliable S-phase measurements. In a recent study [lo], we have employed the Schtitte method and shown a strong correlation between DNA indices and S-phases from fresh and fixed breast cancers. Encouraged by these results we performed the present study on archival material from 120 patients with endometrial cancer and analyzed the prognostic value of the DNA index and S-phase fraction.
DNA AND S-PHASE AS PROGNOSTIC
the start of therapy was used. We analyzed a11 FIG0 grade 3 tumors (n = 15) and all cases of UPSC (n = 22) where curettage material was available in our archives. Then 40 grade 1 and 43 grade 2 cases were additionally selected at random from the archives without prior knowledge of the clinical outcome. UPSC was diagnosed according to the histological criteria of Hendrickson and Kempson and co-workers [ 11,121. No attempt was made to grade UPSC tumors according to FIGO. All cases were examined by the same pathologist (B.R.) without any knowledge of the clinical course. The follow-up time for all patients was at least 5 years or until the death of the patient. The mean age was 65 years which did not differ significantly between different grades or between patients with UPSC and those with nonUPSC tumors. The principal treatment was consistent throughout the period studied (for stage I grade I tumors: abdominal hysterectomy and bilateral salpingoophorectomy (TAH + BSO), followed by one vaginal intracavitary radiation treatment; for stage I grade 2 or 3 tumors, and stage II (all grades) and UPSC: three preoperative intracavitary radium applications by the Heymans packing technique at 2-weeks intervals, followed by TAH + BSO). Preparation
and Flow Cytometry
51
INDICATORS
TABLE 1 DNA Ploidy Correlated to Histology Histology
No.
Aneuploid
Grade 1 Grade 2 Grade 3 UPSC
37 43 12 20
II 14 42 85
%
was estimated by multiplying the number of channels between the G,/G, and GZ/M peaks by the mean number of cells per channel in the mid-S-phase range. Statistical
Methods
The statistical calculations were made with the BMDP software program [13] using Cox regression models and life tables [ 141. RESULTS DNA Analysis Correlated
to Histology
In 9 cases there was insufficient amounts of material for the analysis or the histogram showed too low quality to be acceptable. An aneuploid DNA histogram was seen in 32 of 111 tumors. The mean CV was 5.7%. The Sphase fraction (i.e., the fraction of cells with a DNA content between Go/G, and Gz/M values), corresponding to the DNA content expected to be found in proliferating cells, was possible to be determined in 92 cases. Ploidy correlated with the different histological groups. Grade 1 tumors showed the lowest percentage of aneuploidy and UPSC the highest (Table 1). The mean Sphase percentage also correlated with the histology (Table 2). Diploid tumors (n = 79) had a mean S-phase of 5.6% (range 1.6-I I .4%) and aneuploid tumors (n = 32) showed a mean S-phase frequency of 13.9% (range 5.9-
The method we employed has been described elsewhere [S]. In brief, it involves the preparation of SO-pm sections of formalin-fixed, paraffin-embedded curettage specimens. We used the entire 50 sections without attempting to identify and cut out specific areas. After deparaffinization with xylene, the specimen was rehydrated stepwise in ethanol of decreasing concentrations. The tissue samples were then treated with trypsin overnight to create a single cell suspension. After filtration through a nylon mesh, the cell suspension was prepared as described by Vindelov using propidium iodide as the 22.8%). fluorescent dye [9]. We used a Leitz MPV flow cytometer DNA Analysis and Mortality and a Monroe OC 8888 microcomputer. Only histograms including at least 20,000 cells, with a CV of less than The cancer mortality rate was closely related to both 8%, and with at most a moderate level of debris, were DNA ploidy and S-phase fraction (Tables 3 and 4, Figs. recorded. The peak with the lowest DNA value was 1 and 2). Patients with aneuploid tumors had a relative considered to contain diploid cells and served as a diploid cancer mortality rate four times higher than that of their internal reference. The coefficient of variation (CV) was calculated from the approximation for a Gaussian disTABLE 2 tribution estimating the width of the peak at half the S-phases (Mean Value) Correlated to Histology maximum height. The S-phases was analyzed after a background correction was performed by the subtraction S-phase (mean %) Histology No. of a constant number of cells from all channels in the Grade I 33 4.3 DNA histogram. The constant was taken as the mean 7.1 Grade 2 35 number of cells per channel at an interval to the right Grade 3 8 10.5 of the Gz/M peak. Then assuming a rectangular distriUPSC 17 12.7 bution of cells in S-phase, the number of S-phase cells
52
ROSENBERG
TABLE 3 Cancer Mortality Related to Ploidy and S-Phase: Univariate Analysis No. of No. of Cum. 5-year pts. deaths cancer mortality
Rel. death rate
Test for trend
DNA ploidy Diploid Aneuploid
79 32
9 12
13% 42%
1.0 4.2
P = 0.0013
S-phase 4% 5-10% >lO%
33 40 19
2 6 9
7% 17% 49%
1.0 2.5 9.9
P = 0.0010
TABLE 4 Cancer Mortality Related to Ploidy and S-Phase: Multivariate Analysis No.
No. of deaths
Rel. death rate
Test for trend
72 20
8 9
1.0 2.4
P = 0.16
33 40 19
2 6 9
1.0 2.1 5.2
P = 0.07
DNA ploidy Diploid Aneuploid S-phase <5% 5-10% >10%
ET AL.
diploid counterparts, and patients in the high S-phase groups (S-phase fraction >lO%) had almost a lo-fold increase in the cancer mortality rate compared to the low S-phase group (S-phase fraction 4%). We formed three risk groups by combining ploidy and S-phase: Group 1 included diploid tumors with S-phases 4%. Group 3 contained aneuploid tumors with S-phases >lO% and group 2 contained all other tumors. Patients in risk group 3 had a 15.5fold increase in the cancer mortality rate when compared to patients in group 1 (Table 5). Patients with group 1 tumors died of their cancer in only 7% of the cases, whereas patients having group 3 tumors succumbed to their malignancy in 63% of the cases. The group variables were entered into a stepwise Cox regression analysis together with stage, age, nuclear grade, and histology (FIG0 grade l-2 versus grade 3 or UPSC). Only FIG0 grade and ploidy/Sphase contributed as independent prognostic variables (Table 6). The other variables did not add any further significant prognostic information once corrected for histology and ploidy/S-phase. DISCUSSION
We did not try to cut out or isolate specific areas containing only cancer cells on the paraffin sections before the analysis because, even when there is a large homogenous area of pure cancer on the 5-pm stained glass sections, it is often virtually impossible to recognize the corresponding area on the 50-pm unstained paraffin sections. When paraffin-embedded material is analyzed, chicken or trout red blood cells, which we use with fresh tumors, are not reliable reference cells. As an alternative, nonmalignant cells contained in the tumor can serve as
Cumulative survival
S-phase
3020 lo-
12
I
1
1
24
36
48
I 6o
Time (months)
FIG. 1. Cumulative cancer-free survival correlated to ploidy and S-phase percentage.
10%
DNA AND S-PHASE AS PROGNOSTIC Cumulative survival % loo-
53
INDICATORS
I ,
I
go-
I_
ao-
I
Group
1
Group
2
Group
3
6050II---------
- ------
- -----
---_-------
20lo-
I 24
12
, 36
I 48
I 6o
lime (months)
FIG. 2. Cumulative cancer-free survival correlated to three risk groups. Group 1, diploid tumors with S-phase ~5%; group 3, aneuploid tumors with S-phase >lO%; and group 2, other tumors.
TABLE 5 Risk Groups According to Ploidy and S-PhaseValue
Group I: Diploid tumor with S-phase <5% Group 2: Other tumors Group 3: Aneuploid tumors with S-phase >lO%
Cancer deaths
Cum. S-year ca. death rate
Adj. rel. death rate
2/33
7%
I.0
IO/63
18%
2.7
9/15
63%
15.5
TABLE 6 StepwiseCox RegressionAnalysis” No. of pts.
No. of deaths
Adj. rel. death rate
Test for trend
Histology Grade 1-2 Grade 3 + UPSC
79 32
7 14
I.0 3.5
P = 0.03
Ploidy/S-phase Group 1 Group 2 Group 3
33 63 I5
2 IO 9
1.0 1.8 5.0
P = 0.05
” Variables included stage age, nuclear grade, FIG0 grade 1-2 versus 3, and UPSC, ploidy/S-phase. Only histology (grade 1-2 vs grade 3, UPSC) and ploidy/S-phase groups showed independent prognostic value. Once corrected for these two variables neither stage, age, nor nuclear grade had any additional prognostic significance.
Test for trend
P = 0.0001
a reference. We found that the paraffin-embedded sections almost always contained sufficient amounts of normal cells to be used as an internal standard. This introduces, however, a risk of missing or misinterpreting hypodiploid tumors, which fortunately are rare. In nine cases there was either insufficient material or the histograms were of too low quality to allow an accurate analysis. However, the majority of the tumor material analyzed was obtained from a curettage done immediately before the first intracavitary treatment and not from the original diagnostic curettage performed 2-3 weeks earlier. This may in part explain why there is sometimes an insufficient quantity of material. We can not prove that the estimated S-phases concerned only proliferating tumor cells. Nevertheless the S-phases did correlate strongly with the clinical outcome. The only parameters in the multivariate analysis which yielded significant
54
ROSENBERG
prognostic information were histology (FIG0 grade l-2 vs FIG0 grade 3 and UPSC) and DNA ploidy/S-phase. The flow cytometric method is probably more objective and comparable among different institutions than the FIG0 grading system. In summary, we conclude that the prognostic value of this method is significant and seems to be comparable to the FIG0 grading, but without inherent subjectivity. Using paraffin-embedded material makes it possible to analyze curettage samples prepared from routine histopathologic examinations and is more convenient than working with fresh or frozen tumor tissue.
ET AL.
6.
7.
8.
9.
10.
REFERENCES 1. Cancer Committee of the General Assembly FIGO. Classification and staging of malignant tumors in female pelvis, ht. J. Gynecol. Obstet.9, 172-179 (1971). 2. Annual report on the results of treatment in gynecological cancer, Stockholm, FIGO, Vol. 18 (1982). 3. Christopherson, W. M., Conelly, P. J., and Alberhasky, R. C. Carcinoma of the endometrium. V. An analysis of prognosticators in patients with favourable subtypes and stage I disease, Cancer 51, 1705-1709 (1983). 4. Soderlin, E. Factors affecting prognosis of endometrial carcinoma, Acta Obstet. Gynecol. Scud. Suppl. 38 (1974). 5. Iversen, O., and Laerum, 0. Ploidy disturbances in endometrial
11.
12.
13. 14.
and ovarian carcinomas: A review, Anal. Quant. Cytol. Histol. 7, 327-336 (1985). Friedlander, M., Hedley, D., and Taylor, I. Clinical and biological significance of aneuploidy in human tumors, J. C/in. Pathol. 37, 961-974 (1984). Atkin, N. Prognostic significance of ploidy level in human tumors. I. Carcinoma of the uterus, J. Natl. Cancer Inst. 56, 909-910 (1976). Schutte, B., Reynders, M., Bosman, F., and Blijham, G. Flow cytometric determination of DNA ploidy level in nuclei isolated from paraffin-embedded tissue, Cytometry 6, 26-30 (1985). Vindelov, L., Christensen, I., and Nissen, N. A detergent-trypsin method for the preparation of nuclei for flow cytometric DNA analysis, Cytometry 3, 323-327 (1983). Wingren, S., Hatschek, T., Stal, O., Boeryd, B., and Nordenskjold, B. Comparison of static and flow cytomfluorometry for estimation of DNA index and S-phase in fresh and paraffin-embedded breast carcinoma. Acta Oncol. &and. 27, 793-797 (1988). Hendrickson, M., Ross, J., Eifel, P. J., Cox, R. S., Martinez, A., and Kempson, R. Adenocarcinoma of the endometrium: Analysis of 256 cases with carcinoma limited to the uterine corpus, Gynecol. Oncol. 13, 373-392 (1982). Hendrickson, M., Ross, J., Eifel, P. J., Martinez, A., and Kempson, R. Uterine papillary serous carcinoma, Amer. J. Surg. Pathol. 6, 93-108 (1982). BMDP. Statistical software, Univ. Cilifornia Press, Berkeley, pp. 187-220 (1981). Cox, D. Regression models and life tables, J. R. Stat. SOC. B. 34, 187-220 (1972).