J Clin Pathol 1984;37:1110-1113

Borderline or malignant ovarian tumour?A case report of decision making with morphometryJPA BAAK,* G VAN DER LEYt

From the *Department ofPathology, SSDZ, Reynier de Graefweg 7,2625 AD Delft, and the tDepartment ofGynaecology, Schieland Hospital, 3116 BA Schiedam, The Netherlands

SUMMARY A young woman presented with bilateral ovarian tumours. Multiple sections of eachtumour were shown to many pathologists for consultation; some considered the tumours to beborderline, whereas others thought that one or both of them was malignant. Morphometryshowed that the numerical classification probabilities for borderline tumour were 0-78 for the leftovarian tumour and 0 85 in the right. The lesions were therefore regarded as borderline tumours'and no additional chemotherapy was given.Three years after the second operation the patient is alive and well without clinical or biochem-

ical evidence of recurrence. Most patients with borderline tumours who die from the disease doso in the first two years after the operation. This young patient was prevented from severeovertreatment by the application of morphometry, illustrating its use in this area of diagnosticgynaecopathology.

Previous studies have shown considerable disagree-ment among pathologists in the subjective distinc-tion between borderline and malignant ovariantumours.23 Although this may be of little clinicalimportance, there are several conditions in whichdiscrepancies in diagnosis result in considerable dif-ferences in treatment. Application of morphometryto a predefined set of borderline and malignanttumours has, however, resulted in a classificationrule with a satisfactory diagnostic quality.' Routineapplication in our laboratory over a period of fouryears has regularly corrected the original grade oftumour assigned with qualitative subjective evalua-tion.4 We describe here a patient in whom mor-phometry was essential in the therapeutic decisionmaking process.

Material and methods

MORPHOMETRIC ANALYSISThe derivation of the morphometric classificationrule to distinguish between borderline and malig-nant ovarian tumours has been described in detailelsewhere.' Therefore, in this paper only a summaryof the morphometric analysis and classification willbe given.Four micrometre thick paraffin sections stained

with standard haematoxylin and eosin were used forthe measurements. Multivariate analysis showed

Accepted for publication 28 June 1984

that many of the 32 morphometric features investi-gated were interrelated. The most important featurefor the classification is the mitotic activity index (thenumber of mitoses at X400 magnification in 25fields, the diameter of each field used being 450,tm). The fields selected for counting were the mostcellular ones, with the severest atypicality and high-est mitotic rate. Having started in a certain area, 24additional contiguous fields were selected randomly.Only definite mitotic figures were counted, withdoubtful structures excluded.The second best discriminator, which adds to the

discriminating power of the mitotic rate, is the per-centage of epithelial tissue compared with thestroma. Measurements were performed with a 42point test grid (90 x 77 mm) according to Weibel.5The point grid was placed on a projection micro-scope at x 200 magnification.The three other features which are important are

the mean nuclear area (,um2), the mean nuclearperimeter (= circumference), and the mean nuclearaxis (both ,um). These were assessed in 25 epithelialcell nuclei photographed at x 1000 magnification.The nuclei were randomly selected in the most cellu-lar areas of the tumour, assessed subjectively. Themorphometric technique was used selectively-thatis, in the areas selected by a skilled pathologist asmorphologically the most atypical. Table I shows thedetailed measurements used.The multivariate classification rule calculates a

numerical probability, which ranges from 0-0 to 1-0,1110

copyright. on June 23, 2020 by guest. P

rotected byhttp://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.37.10.1110 on 1 October 1984. D

ownloaded from

Borderline or malignant ovarian tumour?

Table 1 Morphometric features used in the multivariateclassification rule to distinguish between borderline andmalignant cases

Features Remarks Units

Mitotic activity index Number of mitoses in 25 highpower field (x 400)

Volume percentage Volume % in relation toepithelium stroma (x 200 magnification) (%o)

Mean nuclear area Measured at x 1000magnification zm2

Mean nuclear perimeter x 1000 magnification ,umMean short nuclear axis x 1000 magnification ,um

that the tumour belongs to the borderline class. Ifthe numerical classification probability was above orequal to 0-60, the tumour was regarded as border-line; if below or equal to 0*40 as malignant; and inbetween these values as doubtful. Such a numericalprobability should not be confused with a nuclear tocytoplasmic area ratio. The numerical classificationprobability is calculated on the basis of the five his-tomorphometric and cytomorphometric featuresdescribed above. It is important to remember thatthese five features independently contribute to thediscriminating power in the multidimensional deci-sion space.The cases in the original study' consisted of 10

borderline and 22 malignant cases, with the clinicaloutcome known. The original diagnoses were madeby Professor Langley, Manchester. The criteria usedfor placing a tumour in the borderline categorywere:1 The absence of invasion

1111

2 The presence of two or more of the followingfeatures: multilayering of the epithelium; nuclearabnormality; "budding'; mitotic activity.The morphometric classification rule also had

prognostic importance.' Routine application of thismorphometric classification rule over a period offour years in our laboratory has corrected the origi-nal subjective rate assigned.4 So far, more than 200ovarian tumours of the common epithelial type havebeen investigated, including about 40 borderlinecases. Results have shown that it is also possible topredict which borderline tumours will survive or arelikely to have favourable outcomes (unpublishedobservations).

CASE REPORTA 24 year old woman presented with both ovariesenlarged, the left more than the right. The left ovarywas therefore removed, without spill. There were nosigns of malignancy in the peritoneal cavity, andextensive staging procedures were all negative.The tumour measured 9 x 8 x 6 cm, and the

outer surface was smooth. When cut open severalcysts were seen, with many papillary structures.Multiple sections were made; the most atypicalareas are shown in Figs. 1 and 2.At clinical examination six months later, the

remaining right ovary was more enlarged.Laparotomy was repeated in June 1981. There weremultiple adhesions, but again no signs of malig-nancy. Extensive staging was negative. A tumoursimilar to the first one was removed, without spill.

/

Fig. 1 First tumour. Note papillary structures and the wall of the cyst.Pseudoinvasion due to tangential cutting, erroneously taken for malignancy.Haematoxylin and eosin. Original magnification x 100.

copyright. on June 23, 2020 by guest. P

rotected byhttp://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.37.10.1110 on 1 October 1984. D

ownloaded from

Baak, Van Der Ley

Fig. 2 First tumour. Note the moderate nuclear atypia and pseudomultilayering dueto tangential cutting. Haematoxylin and eosin. Original magnification x 400.

This tumour measured 9 x 7 x 5 cm with a similarpattern (a multicystic tumour with papillary protru-sions). Again, multiple sections were made, and themost atypical area is shown in Fig. 3.

Results

The slides were shown to many pathologists, whose

~-s.

,* fX X q

%yf.w tf

opinions showed considerable disagreement: somethought that the tumours were borderline, othersmalignant.Table 2 summarises the clinical conditions in

which the distinction between borderline and malig-nant ovarian tumour may be clinically important.Clearly, the present case was in the second category.Indeed, we considered whether we should treat the

s.4.!

eYA,.N

Fig. 3 Second tumour. The dense lymphocytic infiltrate gives the impression ofhighcellularity and unfavourabk nuclear to cytoplasmic ratios (due to optical illusion).Haematoxylin and eosin. Original magnification x 100.

1112

copyright. on June 23, 2020 by guest. P

rotected byhttp://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.37.10.1110 on 1 October 1984. D

ownloaded from

Borderline or malignant ovarian tumour?

Table 2 Clinical conditions in which the distincionbetween borderline and malignant ovarian tumour can beimportant

1 FIGO IA with ruptureBO: no therapyGrade I: chemotherapy

2 FIGO I with later occurring contralaterality1st BO, 2nd grade I: no therapy1st grade 1: 2nd grade I: chemotherapy

3 FIGO II-IVBO: no therapyGrade I: debulking and chemotherapy

BO = borderline.FIGO = clinical stage (according to the International Federation ofGynecology and Obstetrics".

patient with multiple chemotherapy if both tumourswere malignant, because in that case the secondtumour could have been regarded as a metastasis.

Application of morphometry and calculation bythe computer of the numerical classification proba-bility of both tumours6' ' showed, however, thatboth tumours should be regarded as borderline. Theclassification probability of the first tumour was0-78, and of the second 0-85. None of the borderlinetumours in the basic material had classificationprobabilities below 0-60, and none of the malignanttumours above 0*40. Thus both tumours of thisyoung patient were clearly regarded as borderline.No further treatment was given, and a conserva-

tive approach was adopted. At regular follow up nosigns of malignancy were apparent, and three yearsafter the last operation the patient is without clinicalor biochemical evidence of recurrence.

Discussion

This case report illustrates the value of mor-phometry in diagnostic histopathology. In doubtfulcases, quantitative microscopical techniques can beused for objective decision making.We emphasise that morphometry should be

applied selectively-that is, a skilled pathologist hasto select the areas, cells, and nuclei of interest. Inthis selective diagnostic morphometry the followingsteps can be discerned:1 Selection of relevant areas, cells, nuclei, etc by askilled pathologist2 Measurement of relevant features3 Comparison of these quantitative features withthe typical "image" in the memory in the computer4 Classification on the basis of these results.

Clearly, the first step is important and in fact isused in everyday diagnostic pathology. This step ishighly reproducible providing the pathologist isaccurate and well trained. The advantage of the sec-ond step over routine qualitative subjective evalua-tion is that the features thus measured are quantita-tive and, when carefully controlled, highly reproduc-ible. In the third step, the typical image in the mem-

1113

ory of the computer is fixed. In contrast to the typi-cal images in the human mind, no distracting factorscan influence this reference (standard). The fourthstep is equally stable for the same reasons. Afterthree years of follow up the patient is free of clinicalor biochemical evidence of recurrence. If borderlinetumours do recur, it is usually within two years8 andthis has also been our own experience. In thispatient active treatment with chemotherapy wasconsidered but was rejected in the face of the mor-phometric evidence.Other examples of the clinical usefulness in indi-

vidual patient care have been described in the dis-tinction of endometrial hyperplasia from carcinoma9and in follicular thyroid carcinoma.'0The simplicity and usefulness of morphometric

techniques make them attractive. Moreover, themethods are now relatively inexpensive and appli-cable to a wide range of diagnostic problems.

This work was supported in part by grant 28-834 ofthe Praeventiefords.The authors thank Mrs DIM de Jong for typing

the manuscript.References

'Baak JPA, Agrafojo Blanco A, Kurver PHJ, et al. Quantitationof borderline and malignant mucinous ovarian tumours. His-topathology 1981;5:353-60.

2 Fox H. A study of consistency and value of histological grading ofborderline tumours of the ovary. Presentation at the XIIIthcongress international academy of Pathology, September,Paris, 1980.

Baak JPA, Lindeman Overdiep SH, Langley FA. Disagreementof histopathological diagnoses of different pathologists inovarian tumours-with some theoretical considerations.European Journal ofObstetrics, Gynecology and ReproductiveBiology 1982; 13:51-5.

Baak JPA, Kurver PHJ, Boon ME. Experience with routineapplication of morphometry in diagnostic pathology. In: Col-lan Y, Romppanen T, eds. Morphometry in morphologicaldiagnosis. Kuopio: Kuopio University Press, 1982.

Weibel ER. Practical methods for biological morphometry. Vol I.London: Academic Press, 1979.

6 Pauker SG, Kassirer JP. The threshold aproach to clinical deci-sion making. N Engl J Med 1980;302:1109-17.

Baak JPA, Langley FA, Hermans J. Classification of new cases:some aspects of single and multivariate analysis. In: Baak JPA,Oort J, eds. A manual ofmorphometry in diagnostic pathology.Heidelberg: Springer, 1983.

8Hart WR, Norris HJ. Borderline and malignant mucinoustumours of the ovary. Cancer 1973;31:1031-45.

9Baak JPA, Diegenbach PC, Kurver PHJ, Stolk HG, Harten J vander. An example of quantitative microscopy in individualpatient care. Mikroskopie (Wien) 1980;37 (suppl):305-7.

' Baak JPA, Kurver PHJ, Boon ME. Computer-aided applicationof quantitative microscopy in diagnostic pathology. PatholAnn 1982; 17: 287-306.

"Novak ER, Woodruff JD. Novak's gynecologic and obstetricpathology, Philadelphia: Saunders, 1974:353-6.

Requests for reprints to: Dr JPA Baak, Instititute ofPathology, Free University, De Boelelaan 1117, 1007 MBAmsterdam, The Netherlands.

copyright. on June 23, 2020 by guest. P

rotected byhttp://jcp.bm

j.com/

J Clin P

athol: first published as 10.1136/jcp.37.10.1110 on 1 October 1984. D

ownloaded from