6th ESACP Congress, Heidelberg, April 7-11, 1999

A003
------> Abstract too long for 6thESACP abstract reprint in ACP. Authors have been asked to submit shortened abstract version <---------
DIAGNOSTIC, PROGNOSTIC AND THERAPEUTIC IMPACT OF QUANTITATIVE AND MOLECULAR PATHOLOGY IN ENDOMETRIAL HYPERPLASIA AND CARCINOMA
Baak JPA 1, Mutter GL 2

1) Departments of Pathology, University Hospital VU, Amsterdam and Medical Center Alkmaar, Netherlands, 2) Brigham and Women's Hospital, Harvard Medical School, Boston, Massachussets, USA

Endometrial hyperplasia (EH) is a frequently occurring disease. In the European Community (275 million inhabitants), an estimated number of 75,000 new cases occurs per year. 10 to 20% progress to cancer. Three major problems occur. 1) Diagnostic problem: to distinguish hyperplasias from cancer (considerable intra- and interobserver variability occurs; 2) Prognostic problem: prediction of progression-or-not to cancer is inaccurate, also with the WHO classification and 3) Discrimination of risk-bearing hyperplasias from benign conditions. As a result, considerable overtreatment occurs. This can have serious psychological and sexual problems, and also result in unnecessary costs for the national health care budget. Over the past 20 years, we have developed a computerized morphometric analysis (CMA) technique, that consists of a measurement tool, and an expert system based classification tool. With the 4-class rule, endometrial curettings or hysterectomy specimens are classified as morphometric low or high grade hyperplasia, morphometric low or high grade carcinoma. For each class, a numerical classification probability is calculated. Routine application of CMA over the past two decades in different laboratories has regularly corrected subjective classification by pathologists. In carcinomas, the 4-class rule correlates better with depth of myometrial invasion than subjective grade.
In subsequent studies, the D-score was developed to predict the outcome of EH. DNA ploidy assessed by FCM was of no value, but morphometric analysis is useful. With a multivariate D-score, specificity is equal to, or better than with the usual WHO classification, but the sensitivity of the D-score is better (1). This was confirmed in "the Philadelphia study" an independent blind multicenter collaboration (2) on 45 cases (sensitivity 100%, specificity 88.5%). Because of these findings, CMA generated D-score is routinely applied in EH in a number of laboratories and therapy is guided by the CMA classification. It is estimated that nationwide application in the Netherlands will save approximately 15 million USD per year, due to prevention of unnecessary hysterectomies.
Another promising approach to improve diagnostic accuracy is to the study the monoclonal nature of endometrial precancers that can be used as an informative biologic endpoint to evaluate the accuracy for morphologic diagnosis. We have tested the ability of objective computerized morphometry to diagnose monoclonal putative endometrial precancers. 93 nonmalignant areas of endometrium from 64 uteri of women with coexisting endometrial adenocarcinoma were scored as monoclonal (n=39) or polyclonal (n=54) by PCR analysis (monoclonal by either non-random X chromosome inactivation or clonal outgrowth of altered microsatellites). Polyps and progesterone altered endometrium were excluded. Morphometry of corresponding delineated regions on H&E stained sections was performed with the QProdit 6.1 image analysis expert system. Lesions were evaluated by 4 independent experienced gynaecopathologists. We also measured D-scores, incorporating standard deviation of short nuclear axis, % stromal volume, and gland outer surface density. The lesions then were classed as probable monoclonal (D<0, n=48), unknown (01, n=42).The D-scores were highly reproducible (inter-observer r>0.98), predicting monoclonal lesions with 95% sensitivity and 75% specificity, and polyclonal areas with 95% specificity and 77% sensitivity. Computerized image analysis is thus capable of recognizing monoclonal endometrial precancers with specificity and sensitivity at least comparable to (and usually higher than) that of experienced pathologists, with the additional advantage of excellent reproducibility.
In FIGO stage I endometrial cancer patients, histologic type and grade are correlated with prognosis and used for therapeutic decision making. However, assessment of these histologic features is subjective, and the results are not always perfectly reproducible. Contrarily, previous studies have shown that DNA-ploidy and morphometric features are highly reproducible and have a strong prognostic value in these cancers. Multivariate analysis has demonstrated that a combination of mean shortest nuclear axis (MSNA), DNA-ploidy and depth of myometrial invasion (the so-called ECPI-1 score) overshadowed the value of all other features investigated (3). An independent retrospective long-term study (10-15 years follow-up) confirmed the prognostic value of the ECPI-1 score in FIGO I Endometrial cancers (only 2 (=3%) of 64 cases with low ECPI-1 values died, in contrast to 11 of 13 cases with high values (p <0.0001, Mantel-Cox value = 51.1)(4).
In early 1987, we therefore started a prospective multicenter evaluation in the Netherlands, in which 26 hospitals participated. In total, 632 patients have been enrolled. An interim analysis in June 1998 again showed the overriding prognostic value of morphometric and cytometric analysis, also in subgroup of FIGO I. FIGO I cases with high ECPI-values have a prognosis similar to FIGO III cancers and therefore should be considered for adjuvant chemotherapy.
It is concluded that quantitative pathology has essential diagnostic, prognostic and therapeutic impact in the management of endometrial hyperplasias and cancers. Supported by grants #28-1203 to J.P.A.B. from the Dutch Praeventiefonds and # EDT-86 to G.L.M. from the American Cancer Society
References:
1 Path. Res. Pract. 1992; 188:856-859
2 Am. J. Obstet. Gynecol. 1996; 174:1518-1521
3 Cancer 1989; 63:1378-1387
4 Int. J. Gynecol. Cancer 1995;5:112-116