Prognostic value of pathologic features and DNA analysis in endometrial carcinoma

GYNECOLOGIC

ONCOLOGY

39, 272-276

(1990)

Prognostic Value of Pathologic Features and DNA Analysis in Endometrial Carcinoma DANIEL Department

of Laboratory Department

Medicine, of Pathology,

The Union Memorial The Johns Hopkins

A.

SYMONDS,

Hospital, University

M.D.

201 East University Parkway, Baltimore, Maryland School of Medicine, Baltimore, Maryland 21205

21218;

and

Received May 30, 1990

In a seriesof 38 nonmetastasizingand 29 metastasizingendometrialcancers,histologicsubtype, aneuploidy,nuclear grade, and depth of myometrial invasion were the features most distinctly separatingthese two groups of patients. The most important factor was histologic type: papillary, papillary serous, adenosquamous, clear cell, and diffusely infiltrating carcinomas werestrongly correlatedwith metastasis usingchi squareanalysis. Nuclear grade 3 and deep myometrial invasion were similar to eachother in sensitivity and predictive value, and both werealso strongly correlated with metastatic risk. Aneuploidy, also correlated with metastasis,wassimilar to nuclear grade 3 and deep myometrial invasion in predictive value but was more sensitive. DNA analysisappearedto be mostusefulwhenhistologicstudies were inconclusivein assigninga high degreeof risk. Twice as many aneuploidcarcinomashaving nucleargrade2 (intermediate risk) morphology proved to be metastatic. A similar relationship was found with deep myometrial invasion. A quantitative correlation betweenthe DNA index and the risk of metastasiswas found; a DNA index above 1.5 was a strong attribute of the metastaticgroup. 0 1990 Academic Press, Inc. INTRODUCTION

Endometrial carcinoma in the majority of patients is an indolent form of cancer cured by uterine ablation, but in a significant number, it pursues an aggressive course. It is not surprising, then, that pathologic features have been scrutinized over the years to identify this subgroup. Motivation to do so is particularly spurred by the availability of adjuvant therapy. Segregation of endometrial cancer patients into high and low risk categories, as has been done in some studies [I], has great clinical appeal. The most widely reported influential parameter is the degree of differentiation as measured by grading [2,3]. While the statistical relationship is variable from study to study, a clear association between advanced grade and adverse outcome is documented, using several dif272 009%82X3/90 $1 SO Copyright 0 1990 by Academic Press. Inc. All rights of reproduction in any form reserved.

ferent grading schemes, including Broders [4], architectural, and nuclear grading [1,5]. Another major determinant of prognosis is the histologic subtype of the carcinoma, with papillary, papillary serous, adenosquamous, and clear cell variants having a diminished survival [5]. A third factor associated with virulence is depth of myometrial invasion. Analysis of DNA content is found in some studies to correlate with behavior [6,7], as well as morphometric features such as nuclear axis and size [8]. This study evaluates selected features simultaneously with DNA analysis by cytometry to assess their prognostic potential, comparing metastasizing and nonmetastasizing carcinomas. SUBJECTS AND METHODS

Pathology records of the Union Memorial Hospital were searched for patients with hysterectomies for endometrial carcinoma over a IO-year span. Slides were reviewed and cases were selected for inclusion which satisfied criteria for unequivocal carcinoma and in which there was at least 5-year follow-up, or in which metastasis was documented. In general, the criteria used for diagnosis of carcinoma parallel those of Kurman and Norris [9] usually in the form of confluent glandular pattern; borderline proliferations were excluded. All nonmetastasizing endometrial carcinomas (38 cases) were stage I on presentation. Metastasizing carcinoma (29 cases) included 17 stage I, one stage II, and 11 stage III carcinomas. Most stage I carcinomas with moderate differentiation and more than minimal myometrial invasion were treated with postoperative radiation. Patient followup was provided by the medical record and the patient’s physician. Without regard to the original pathologic interpretation, each tumor was regraded by the FIG0 system [lo],

METASTATIC TABLE

RISK IN ENDOMETRIAL

1

Microscopic Features Metastasizing Nonmetastasizing Number of cases FIG0 grade I II III Nuclear grade 1 2 3 Myometrial invasion O-l/3 213-313 Lymphatic invasion Concurrent hyperplasia Nonconventional histology Papillary Papillary serous Adenosquamous Clear cell Diffusely infiltrating

Stage I

All stages

38

17

29

20 16 2

4 10 3

5 20 4

15 17 6

0 8 9

0 13 16”

30 8 3 21 2 1

8 9 4 5 11 3 2 3 2 1

12 17” 7h 8” 17” 7 3 3 2 2

1

I’ Significant at the 99.5% confidence level using xz test. ’ Significant at the 95% confidence level using x2 test. ’ Significant at the 97.5% confidence level using x2 test.

273

CANCER

optical density of Feulgen-stained nuclei was measured by computer-based video image analysis (Imagescan program: Karl Zeiss, Thornwood, NY). The DNA index was computed by dividing the modal value of tumor nuclei by modal value of stromal lymphocytes. Values were assigned to the diploid (~1.1) or aneuploid range (>1.2). The program simultaneously computes the area of the encircled field, which is a close approximation of the nuclear area. The results were tabulated subdividing the metastasizing cases into stage I and all stages combined. Sensitivity, specificity, and predictive value matrices were constructed. The following definitions were used: Sensitivity = MC/MC + MS Specificity = Ns/Ns + NC Predictive value of a positive finding = MC/MC + NC, where NC is nonmetastasizing cases with the feature studied; Ns is nonmetastasizing cases without the feature studied; MC is metastasizing cases with the feature studied; and MS is metastasizing cases without the feature studied. RESULTS

and by nuclear grading after the method of Christopherson et al. [5]. These systems are well-illustrated by Mittal et al. [1] and are not reillustrated here, except to note that nucleolar characteristics were included in nuclear grading. The histologic subtype was assessed and divided for simplification into two groups: conventional (not otherwise specified endometrioid) morphology; and nonconventional, unfavorable subtypes. The latter category included papillary, papillary serous, adenosquamous, clear cell, and diffusely infiltrating carcinomas, subtypes previously found to have unfavorable prognostic characteristics [5,11- 131. Adenoacanthoma and mutinous carcinoma were classified with the conventional group. Although all papillary carcinomas were placed in the nonconventional histology, this series did not appear to contain the very-well-differentiated villoglandular variant which has been described [3,14]. Other pathologic features recorded were concurrence of hyperplasia, presence of vascular space invasion, and presence and depth of myometrial invasion. The latter was further subdivided into two groups: minimal and upper one-third myometrial thickness (superficial), and twothirds to serosal (deep) invasion. DNA analysis was performed as previously described [ 151. Representative 8-pm sections were stained with the Feulgen reaction with 15 min of hydrolysis. Integrated

Table 1 contrasts the histologic and cytologic characteristics of metastasizing and nonmetastasizing cancers in this series. The majority of nonmetastasizing carcinomas were of FIG0 grades I and II, and nuclear grades 1 and 2. While a small number of FIG0 grade I tumors did metastasize, no nuclear grade 1 tumors did so. Nuclear grade proved to be a more sensitive and accurate predictor of metastatic status than the FIG0 grading system. Deep myometrial invasion had a similar degree of sensitivity and specificity to nuclear grade 3 in relating to the development of metastasis. The single strongest microscopic determinant of metastasis, however, was the histologic subtype of the tumor. Nonconventional morphology of the unfavorable subtypes was strongly correlated by x2 analysis with the metastatic group. As indicated in Table 2, nearly all cases of metastasis were associated with an aneuploid DNA index. However, aneuploidy was found in approximately one-third of nonmetastasizing cases. As a result, aneuploidy was more sensitive per se but had a similar predictive ability to nuclear grade 3 and deep myometrial invasion. On further analysis, there did appear to be a quantitative relationship between the DNA index and metastasis, as can be observed in the cluster diagram, Fig. 1. The metastatic group had a higher number of cases with DNA

274

DANIEL A. SYMONDS TABLE 2 DNA Analysis and Nuclear Size Metastasizing Nonmetastasizing

Number of cases Ploidy Diploid range Aneuploid range DNA index > I.5 Average DNA index Average nuclear area (pm) Coefficient of variation for nuclear area

Stage I

All stages

38

17

26 12 5 I .25 45.2

2 I5 14 1.72 55.8

3” 26” 21” 1.7Sh 59.5”

Il.3

12.9”

9.8

29

DISCUSSION

” Significant at the 99.5% confidence level with ,$ test. ’ Significant at the P > 0.01 level using t test.

indices greater than 1.5. When stratified using a DNA index of 1.5 as a cutoff, abnormal ploidy retained sensitivity but had a much greater predictive value than nuclear grade 3 or deep myometrial invasion in stage I cases. Histologic subtyping, however, retained the highest predictive value. Sensitivity, specificity, and predictive value for the factors most strongly correlated with metastasis are summarized in Table 3. Nuclear area and the coefficient of variation for the nuclear area were both significantly greater (using Student’s t test) in the metastasizing patients. Positive but lesser degree of correlation was found for lymphatic space invasion, and an inverse relationship with concurrent hyperplasia and metastasis was found. In terms of correlation of microscopic features with the DNA analysis, the following observations were made. All nonconventional unfavorable subtypes were 2.50

2.25

X X

X 0 00 X

CONVENTIONAL HISTOLOGY

aneuploid. Likewise, all nuclear grade 3 carcinomas fell into the aneuploid range. All nuclear grade 1 carcinomas were diploid, but 4 of 5 metastasizing and 2 of 20 nonmetastasizing FIG0 grade I tumors were aneuploid. Interestingly, the stage I metastasizing nuclear grade 2 carcinomas had 75% aneuploidy, while nonmetastasizing nuclear grade 2 carcinomas had only 35% aneuploidy. Similarly, 92% of metastasizing tumors with deep myometrial invasion were aneuploid while only 38% of nonmetastasizing cases with deep myometrial invasion were aneuploid.

x

x

x

x

NONCONVENTIONAL HISTOLOGY

FIG. 1. The position of the DNA index is plotted for metastasizing (X) and nonmetastasizing (0) endometrial carcinomas, further subdivided by histologic type (see text for characterization of these groups).

The major objective of this study was to prioritize the prognostic importance of pathologic microscopic measurements and DNA analysis in the same study set of patients. A profile emerges from this study of the usual metastasizing endometrial carcinoma as one which is of unconventional morphology, of high nuclear grade, of aneuploid DNA complement, and deeply invasive into the myometrium. However, this summation is only applicable in the aggregate; significant exceptions occur in the individual case. For example, two polypoid nuclear grade 3 carcinomas without myometrial invasion demonstrated did develop metastases in this series. Thus, the significance of each individual factor studied needs to be ranked according to its prognostic utility. Based on the data generated by this study, a metastatic risk assessment hierachy can be constructed based on histologic type, nuclear grade, DNA index, and presence of deep myometrial invasion (Table 4). In the author’s view, the most unfavorable single feature would warrant assignment to the corresponding risk category. This study reaffirms the central importance of histologic evaluation. Most of the histologic variants construing unfavorable prognosis are recognized without too much difficulty. A possible exception is a papillary growth configuration which can be overlooked unless searched for at low power; this pattern should be diligently sought since it is an important prognostic finding. It is clear that there are significant differences in discriminating ability between different grading systems, with nuclear grading having more sensitivity than FIG0 grading in this and other studies [1,5]. For this reason, the grading system used in diagnostic reports should be specified. In addition, there appears to be considerable subjectivity even with the better defined nuclear grading system, since a good proportion of tumors are assigned a middling grade by default. Perhaps, further quantitative refinement could improve the predictive value of grading. DNA analysis by both flow cytometry and static cytometry is attractive because of its objectivity and reproducibility. At present, it divides patients into two

METASTATIC

RISK IN ENDOMETRIAL

275

CANCER

TABLE 3 Sensitivity, Specificity, Predictive Value” Feature studied

Sensitivity

Specificity

Predictive value of positive result

All stages Aneuploid DNA range DNA index > 1.5 Nuclear grade 3 Deep myometrial invasion Nonconventional subtype

26129 21129 16129 17129 17129

= = = = =

0.90 0.72 0.55 0.59 0.59

Aneuploid DNA range DNA index > 1.5 Nuclear grade 3 Deep myometrial invasion Nonconventional subtype

15/17 14117 14117 9117 11/17

= = = = =

0.88 0.82 0.82 0.53 0.65

26/38 33138 32/38 30138 36138

= = = = =

0.68 0.87 0.84 0.80 0.95

26138 21/26 16122 17125 17119

= = = = =

0.68 0.81 0.73 0.68 0.90

26138 33/38 32/38 30138 36/38

= = = = =

0.68 0.87 0.84 0.80 0.95

15127 14119 9115 9117 11/13

= = = = =

0.56 0.74 0.60 0.53 0.85

Stage I

” See text for definitions.

groups: diploid and aneuploid. This study suggests using

the DNA index in a semiquantitative fashion may provide more information. Iverson [6] observed a positive correlation between increasing grade and the DNA index. The current study indicates a positive correlation between increasing DNA index and the likelihood of metastasis. Using a DNA index above 1.5 increased the specificity and predictive value of DNA analysis without diminishing its sensitivity in this series of patients. Mittal et al. [l] pose a timely question: Does DNA analysis offer advantage over nuclear grading? Since all cases with nuclear grade 3 and unfavorable histologic subtype proved to be aneuploid, the DNA index may not be especially contributory unless it is used in a quantitative fashion in these patients. On the other hand, for the large number of patients with conventional histology who fall into the intermediate nuclear grade 2, ploidy studies offer definite prognostic advantage. In this study, stage I nuclear grade 2 carcinomas which subsequently metastasized were twice as likely to have DNA indices TABLE 4 Classificationof Risk Factors High risk Nonconventional histologic type Aneuploid DNA range, DNA index > 1.5 Nuclear grade 3 Deep myometrial invasion Indeterminate risk Conventional histologic type Aneuploid DNA range, DNA index < 1.5 Nuclear grade 2 Low risk Conventional histologic type Diploid DNA range Nuclear grade 1

in the aneuploid range. This may be a reflection of the imprecision and limitations of the grading system. However, the same relationship was found for deep myometrial invasion which is a relatively objective and reproducible measurement. Aneuploid tumors with deep myometrial invasion were twice as likely to metastasize as diploid tumors with deep myometrial invasion. This does imply that DNA content has some degree of independence as a prognostic variable. Nevertheless, some of the factors studied may be interrelated. For example, all of the unfavorable histologic subtypes and all nuclear grade 3 carcinomas were aneuploid. It is possible that high nuclear grade and aneuploidy are both manifestations of tumor heterogeneity, and hence both gauge deviations from stable cell populations. In conclusion, within limits quantitation of this deviation has considerable promise for risk assessment in endometrial carcinoma. ACKNOWLEDGMENTS The author thanks the following physicians for providing follow-up on their patients: Dr. Charles C. S. Chan; Dr. Shailaja M. Didolkar; Dr. Albert H. Dudley, Jr.; Dr. Andrew London; Dr. Robert E. Martin; Dr. William H. Spencer-Strong 11; and Dr. Clifford R. Wheeless. The author also thanks Mrs. Vivian McCoy, Ms. Shirin Golozar, and Ms. Elizabeth Schene for their expert assistance with the Feulgen studies.

REFERENCES Mittal, K. R., Schwartz, P. E., and Barwick, K. W. Architectural (FIGO) grading, nuclear grading, and other prognostic indicators in stage I endometrial adenocarcinoma with identification of highrisk and low-risk groups, Cancer 61, 538-545 (1988). Kurman, R. J., and Norris, H. J. Endometrial carcinoma, in Blaustein’s Pathology of the female genital tract, (R. J. Kurman, Ed.), 3rd ed, Springer-Verlag, New York (1987).

276

DANIEL

A. SYMONDS

3. Hendrickson, M. R., and Kempson, R. L. Endometrial hyperplasia, metaplasia and carcinoma, in Haines and Taylor Obstetrical and gynaecological pathology (H. Fox, Ed.), 3rd ed, Churchill Livingstone, Edinburgh (1987). 4. Reagan, J. W., and Ng, A. B. P. The cells of uterine adenocarcinema, Williams & Wilkins, Baltimore (1965). 5. Christopherson, W. M., Connelly, P. J., and Alberhasky, R. C. Carcinoma of the endometrium. V. An analysis of prognosticators in patients with favorable subtypes and stage I disease, Cancer 51, 1705-1709 (1983). 6. Iversen, 0. E. Flow cytometric deoxyribonucleic acid index: A prognostic factor in endometrial carcinoma. Amer. J. Obstet. Gynecol. 155, 770-776 (1986). 7. Lindahl, B., Alm, P., Ferno, M., Killander, D., Langstrom, E., Norgren, A., and Trope, C. Prognostic value of flow cytometrical DNA measurements in Stage I-II endometrial carcinoma: Correlations with steroid receptor concentration, tumor myometrial invasion, and degree of differentiation, Anricancer Res. 7, 791-798 (1987). 8. vanderputten, H. W. H. M., Baak, J. P. A., Koenders, T. J. M., Kurver, P. H. J., Kurver, P. H. J., Stalk, H. G., and Stolte, L. A. M. Prognostic value of quantitative pathologic features and DNA content in individual patients with stage I endometrial adenocarcinoma, Cancer 63, l378- I387 (1989).

9. Kurman, R. J., and Norris, H. J. Evaluation of criteria for distinguishing atypical endometrial hyperplasia from well-differentiated carcinoma, Cancer 49, 2547-2559 (1982). IO. Poulsen, H. E., Taylor, C. W., and Sobin, L. H. Histological typing offemale genital s$cation of turnours,

tract

iumours.

International

histological

Vol. 13, World Health Organization,

clas-

Geneva

(1975). Il. Christopherson, W. M., Alberhasky, R. C., and Connelly, P. J. Carcinoma of the endometrium. II. Papillary adenocarcinoma: A clinical pathological study of 46 cases, Amer. J. Clin. Pathol. 77, 534-540 (1982). 12. O’Hanlan, K. A., Levine, P. A., Harbatkin, D., Feiner, C., Goldberg, Cl. L., Jones, J. G., and Rodriguez-Rodriguez, L. Virulence of papillary endometrial carcinoma, Gynecol. Oncol. 37, 112-l I9 (1990). 13. Mittal, K. R., and Bat-wick, K. W. Diffusely infiltrating adenocarcinoma of the endometrium, Amer. J. Surg. Pathol. 12, 754758 (1988). 14. Chen, J. L., Trost, D. C., and Wilkinson, E. J. Endometrial papillary adenocarcinomas: Two clinicopathological types, Int. J. Gynecol. Pathol. 4, 279-288 (1985). 15. Symonds, D. A., Johnson, D. P., and Wheeless, C. R. Feulgen cytometry in simultaneous endometrial and ovarian carcinoma, Cancer 61, 2511-2516 (1988).