Predicting clinical outcome in patients diagnosed with synchronous ovarian and endometrial cancer.

Abstract

Patients with synchronous ovarian and endometrial cancers may represent cases of a single primary tumor with metastasis (SPM) or dual primary tumors (DP). The diagnosis given will influence the patient's treatment and prognosis. Currently, a diagnosis of SPM or DP is made using histologic criteria, which are frequently unable to make a definitive diagnosis.
In this study, we used genetic profiling to make a genetic diagnosis of SPM or DP in 90 patients with synchronous ovarian/endometrial cancers. We compared genetic diagnoses in these patients with the original histologic diagnoses and evaluated the clinical outcome in this series of patients based on their diagnoses.
Combining genetic and histologic approaches, we were able make a diagnosis in 88 of 90 cases, whereas histology alone was able to make a diagnosis in only 64 cases. Patients diagnosed with SPM had a significantly worse survival than patients with DP (P = 0.002). Patients in which both tumors were of endometrioid histology survived longer than patients of other histologic subtypes (P = 0.025), and patients diagnosed with SPM had a worse survival if the mode of spread was from ovary to endometrium rather than from endometrium to ovary (P = 0.019).
Genetic analysis may represent a powerful tool for use in clinical practice for distinguishing between SPM and DP in patients with synchronous ovarian/endometrial cancer and predicting disease outcome. The data also suggest a hitherto uncharacterized level of heterogeneity in these cases, which, if accurately defined, could lead to improved treatment and survival.

Full text

Page 1

Predicting Clinical Outcome in Patients Diagnosed with
Synchronous Ovarian and Endometrial Cancer
SusanJ.Ramus,1KarimElmasry,1ZhiyuanLuo,2AlexGammerman,2KarenLu,5AyseAyhan,6NaveenaSingh,3
W. GlennMcCluggage,7IanJ. Jacobs,1John C.Whittaker,4andSimon A. Gayther1
Abstract
Purpose: Patients with synchronous ovarian and endometrial cancers may represent cases of a
singleprimary tumorwithmetastasis(SPM)ordualprimary tumors(DP).Thediagnosisgivenwill
influencethepatient’streatmentandprognosis.Currently,adiagnosisofSPMorDPismadeusing
histologic criteria, which are frequently unable to make a definitive diagnosis.
Experimental Design: In this study, we used genetic profiling to make a genetic diagnosis of
SPMor DP in 90 patients with synchronous ovarian/endometrial cancers.We compared genetic
diagnoses in these patients with the original histologic diagnoses and evaluated the clinical
outcomein this series of patients basedon their diagnoses.
Results: Combining genetic andhistologic approaches, we were able make a diagnosis in 88 of
90 cases, whereas histology alone was able to make a diagnosis in only 64 cases. Patients
diagnosed with SPM had a significantly worse survival than patients with DP (P = 0.002).
Patients in which both tumors were of endometrioid histology survived longer than patients of
other histologic subtypes (P = 0.025), and patients diagnosed with SPM had a worse survival
if the mode of spread was from ovary to endometrium rather than from endometrium to ovary
(P = 0.019).
Conclusions: Genetic analysis may represent a powerful tool for use in clinical practice for
distinguishing between SPM and DP in patients with synchronous ovarian/endometrial cancer
andpredictingdisease outcome.The dataalsosuggest ahithertouncharacterizedlevelofhetero-
geneityinthese cases, which, ifaccuratelydefined, couldleadtoimprovedtreatment and survival.
Patients with synchronous cancers are a frequently encoun-
tered problem in clinical practice. Synchronous tumors of
the female genital tract occur in f10% of women diagnosed
with ovarian cancer and 5% of women with endometrial cancer
(1–3). Such cases may represent either two primary tumors
(DP) or a single primary tumor and associated metastasis
(SPM). There are significant clinical implications associated
with either diagnosis. Using the International Federation of
Gynaecology and Obstetrics guidelines, a patient diagnosed
with dual primary endometrioid tumors confined to the ovary
and uterus would represent two stage I cancers. The patient
would have a good prognosis and, depending on the substage
of the tumor, may not require adjuvant chemotherapy or
radiotherapy (4–9). By contrast, a patient with a primary
ovarian adenocarcinoma and an endometrial metastasis would
be classified, at the very least, as International Federation of
Gynaecology and Obstetrics stage IIA ovarian cancer, and a
patient with a primary endometrial adenocarcinoma and an
ovarian metastasis would be classified as a stage III endometrial
cancer. In both cases, the prognosis would be worse than for
a diagnosis of dual primary cancers, and so the patient would
receive more aggressive treatment.
Currently, distinguishing between a single primary cancer
with metastasis and dual primary cancers involves clinicopath-
ologic interpretation. Several pathologic criteria are assessed
including histologic type and tumor grade; presence and extent
of myometrial invasion; vascular invasion; presence of tumor
in the fallopian tube; ovarian tumor size; pattern of ovarian
involvement; and presence or absence of precursor lesions (e.g.,
ovarian endometriosis; refs. 10–12). However, establishing
the relationship between simultaneously occurring ovarian and
endometrial cancers with certainty remains a diagnostic
Imaging, Diagnosis, Prognosis
Authors’Affiliations:1Gynaecological Cancer Research Laboratory, University
College London Elizabeth Garrett Anderson Institute for Women’s Health,
University College London;2Department Computer Science, Royal Holloway,
University of London;3Department Pathology, St. Bartholomew’s andThe Royal
London;4London School of Hygiene and Tropical Medicine, London, United
Kingdom;5M. D. Anderson Cancer Center, Houston, Texas;6Department
Pathology, Seirei Mikatahara Hospital, Hamamatsu, Japan; and7Department
Pathology, RoyalVictoria Hospital, Belfast, United Kingdom
Received 2/12/08; revised 4/30/08; accepted 5/9/08.
Grant support: This work was undertaken at UCLH/UCL which received a
proportion of funding from the Department of Health’s National Institute for Health
Research Biomedical Research Centers funding scheme. S.J. Ramus is funded by
the Mermaid arm of the Eve Appeal; S.A. Gayther is a Higher Education Funding
Councilfor England ^ funded Readerin Cancer Genetics.
The costs of publicationof this article were defrayedinpartby thepaymentof page
charges.This article must therefore be hereby marked advertisement in accordance
with18 U.S.C. Section1734 solely toindicate this fact.
Note: Supplementary data for this article are available at Clinical Cancer Research
Online (http://clincancerres.aacrjournals.org/).
S.J. Ramus and K. Elmasrycontributed equally to this study.
Requests for reprints: Simon A. Gayther, Gynaecological Cancer Research
Laboratory, University College London Elizabeth Garrett Anderson Institute for
Women’s Health, University College London,The Paul O’Gorman Building, 72
Huntley Street, LondonWC1E 6BT, United Kingdom. Phone: 44-20-3108-2008;
Fax: 44-20-3108-2010; E-mail s.gayther@ucl.ac.uk.
F2008 American Associationfor Cancer Research.
doi:10.1158/1078-0432.CCR-08-0373
www.aacrjournals.org Clin Cancer Res 2008;14(18) September15, 20085840

Page 2

challenge. If two cancers have a similar morphologic appear-
ance, it may suggest that one is a metastasis of the other, but
they could still be independent primary cancers. Conversely,
two cancers can appear to be dissimilar when one is a
metastasis of the other due to tumor dedifferentiation or clonal
heterogeneity. When no clear diagnosis can be made, the
patient will usually be treated for the more aggressive disease.
To avoid the uncertainty associated with some histopathologic
diagnoses, a number of investigators have attempted
to characterize the relationship between tumors from the
same patient on the basis of their molecular profile. Several
different approaches have been tried, including immunohisto-
chemistry, flow cytometry (to compare chromosome
ploidy), X-chromosome inactivation, and sequence analysis of
genes frequently mutated during tumor progression (e.g., TP53).
However,thesehaveallbeenshowntohavelimiteduseinrefining
the relationship between dual site cancers (13–19).
Recent studies have suggested that the analysis of multiple
genetic changes in tumors, using either gene expression and/
or DNA microarray technologies or more traditional methods
such as microsatellite analysis, can be used to establish
genetic signatures of somatic changes in tumors (20–22).
Microsatellite marker genotyping has also been used in an
attempt to distinguish between dual primary tumors or
primary and metastatic tumors for several types of cancer
(19, 23–28).
Table 1. Clinical data of 90 patients with synchronous ovarian and endometrial cancer, grouped by
histopathologic diagnosis
Clinical characteristics of casesHistopathologic diagnosis.Total
DP (n = 34)*
SPM (n = 25)c
Uncertain (n = 26) DP (diff. HST; n = 5)b
Centre
United States
Belfast
Japan
London
Age at dx (y)
Average
Median
Histology
Endometrioid
Serous
Other
Grade (endometrial cancer)
1
2
3
ND
Grade (ovarian cancer)
1
2
3
ND
Stage (endometrial cancer)
I
II
III
IV
ND
Stage (ovarian cancer)
I
II
III
IV
ND
Chemotherapy
Yes
No
ND
Radiotherapy
Yes
No
ND
11
12
7
4
0
2
7
05
0
0
0
16
25
15
34
11
1
14 16
53.5
51.5
58.9
60
58.8
58
52
43
56.5
54.5
30
2
2
15
7
3
13
8
5
5 (mixed)
-
-
58 + 5
17
10
15
14
5
0
4
8
32
1
1
1
24
34
26
6
11
8
4
12
1
14
14
6
0
2
8
10
0
3
2
17
34
30
9
12
7
6
14
1
25
8
1
0
0
3
0
5
1
12
2
1
0
11
4
0
1
0
0
44
10
8
1
2716
25
3
6
0
0
8
0
1
0
12
1
3
0
10
3
0
2
0
0
48
4
12
0
2616
7
8
11
0
14
15
6
5
0
0
5
33
14
4319
6
8
2
2
60
0
5
14
21
55
11
920 21
Abbreviations: dx, diagnosis; ND, no data available.
*Diagnosis of dual primary cancers.
cDiagnosis of single primary cancer with metastasis.
bPatients in which the ovarian and endometrial cancers were of different histologic subtypes.
Outcomein Synchronous Ovarian/Endometrial Cancer
www.aacrjournals.orgClin Cancer Res 2008;14(18) September15, 20085841

Page 3

In the current study, we used microsatellite analysis to (a)
establishtheclonalrelationshipbetweenovarianandendometrial
cancers from the same patient using genetic profiling, and (b)
compare genetic and histologic diagnoses for each patient
and establish whether either can predict patient outcome
(survival).
Materials and Methods
Patients.
and endometrium were identified from St. Bartholomew’s and The
Royal London Hospitals (London, United Kingdom; 34 cases), the
Royal Hospitals Trust (Belfast, United Kingdom; 25 cases), M.D.
Anderson Cancer Center (Houston, TX; 16 cases) and Seirei Mikatahara
Hospital (Hamamatsu, Japan; 15 cases). For 85 of 90 patients, both
tumors were of the same histologic subtype. Five patients in which the
two tumors were of different histologies had an unequivocal diagnosis
dualprimarycancerandwereincludedaspositivecontrols.Patientswere
diagnosed between 1989 and 2005. Patients from the United States were
all dual primary, endometrial, and ovarian cancer by histopathology
(29). For other centers, patients had previously been classified by
histopathology as dual primary tumors (DP), single primary tumor and
metastasis (SPM), or of uncertain diagnosis. Local research ethics
committee approval for the collection of samples was obtained at each
center and additional approval was obtained from University College
London Hospital to perform genetic analyses of tumor specimens.
The average age at diagnosis was 56 y (range, 20-88 y). The
histopathology reports and follow-up information were obtained from
medical records. Histopathologic typing and grading were done
according to WHO criteria. Clinical data for the patients grouped by
pathology diagnosis are given in Table 1. Multiple pathologists
contributed to the pathology reporting at each center. Therefore, a
‘‘within-center’’ pathology review was also undertaken by the lead
pathologist in each center to ensure that the pathology data were
consistent (Supplementary Table S1).
Patients were divided into two groups: a ‘‘training’’ set of 30 cases
including 14 cases of SPM and 16 cases of DP, and a ‘‘test’’ set of 60 cases
including 11 cases of SPM and 23 cases of DP. Cases with a known
diagnosis of SPM and DP in the training set were selected based on
unequivocal diagnoses between pathology reports and the independent
review.Sampleswere drawn equallyfrom allfour centers.The individual
making the genetic diagnosis did so without prior knowledge of the
histopathologic diagnosis of the patient.
Genetic profiling.Archival formalin-fixed, paraffin-embedded tissue
specimens were used for DNA extraction. H&E-stained sections were
analyzed to identify areas containing at least 70% tumor cells. DNA was
extracted as described previously (22). Two regions of tissue from each
section were extracted and tested. Microsatellite analysis was used to
score loss of heterozygosity and microsatellite instability for each tumor
at 22 different loci as previously described (22). Markers were evenly
distributedthroughoutthegenomeorwerelocatedinregionsoffrequent
somatic alteration (30–33). Automated nonsubjective calling of allelic
imbalance wasdone as previously published(22, 34). Theratios ofallele
peak heights for each tumor sample were compared with the allele peak
heights for both normal samples separately using the following formula:
L = (at2? an1) / (at1? an2), where L is the ratio, a is the height of the
peak,n1andn2arenormalallele1andnormalallele2,andt1andt2are
tumor allele 1 and tumor allele 2. The log of the values from the
comparison with the two normal samples was averaged. A value of L <
0.6 indicated loss of allele 2, and L > 1.67 loss of allele 1. When
additional alleles were seen, this was scored as microsatellite instability.
As a further measure of quality assurance, we analyzed two different
regions of neoplastic tissue for each tumor. Only changes seen in both
tumor regions were included in the analysis.
Each pair of tumors from each patient was scored as having
either the same or a different genetic pattern for each microsatellite
Ninety patients with synchronous tumors of the ovary
marker (Fig. 1). The genetic diagnosis is based on the ratio of
similarities/differences for each microsatellite marker between the
ovarian and endometrial tumor for each patient. Only informative
markers were included (i.e., where there was a change in at least one
of the tumors from the pair). Two levels of stringency were applied
in using genetic profiles to make a diagnosis. The ‘‘unrestricted’’
genetic diagnosis was independent of how many markers were
altered in the two tumors; the ‘‘restricted’’ genetic diagnosis was only
given if either >2 markers were informative for a genetic alteration or
when the ratio of genetic ‘‘similarities’’ to ‘‘differences’’ was >2 (for
SPM) or <0.5 (for DP). Where the ratio was 1 or when there were no
loss of heterozygosity or microsatellite instability events, a diagnosis
was not given.
This method of genetic diagnosis was preferred to statistical modeling
and machine learning, which we found gave poorer concordance
with known pathology (Supplementary Materials and Supplementary
Table S2).
Age of onset and survival analysis.
Cox proportional hazards model including a linear term for age at
Survival analysis was based on a
Fig.1. Examples of genetic profiling in two patients with synchronous ovarian and
endometrial cancer for two microsatellite markers innormal, ovarian tumor, and
endometrial tumor tissues in each patient. Arrows, genetic alterations.The ratio of
the two alleles in tumor samples is compared with the ratio of the two alleles in
normal samples from the same patient to establish whether one or other allele is
deleted (i.e., shows loss of heterozygosity).When there was >40% decrease in one
of the allele peaks in the tumor, this was called as loss of heterozygosity. Patient A:
For marker1, thenormal tissue shows two alleles; the ovarian tumor tissue shows
lossofalleleii;andtheendometrialtumors tissue showslossofallelei.Formarker2,
the normal and endometrial tumor tissues have two alleles, but the ovarian tissue
shows loss of allele ii.These results suggest that this patient represents a case of
dualprimary cancers. Patient B: For marker1, the normal tissue shows two alleles
and both endometrial and ovarian tumors show a similar allelic imbalance (loss of
heterozygosity) involving allele i. Both tumors also show loss of the same allele i,
for marker 2.Together, these results suggest a diagnosis of a single primary tumor
with metastasis.
Imaging, Diagnosis, Prognosis
www.aacrjournals.orgClin Cancer Res 2008;14(18) September15, 20085842

Page 4

Fig. 2. The geneticprofilesofbothtumorsfromeachpatientandtheirdiagnosisofSPMorDP forbothgeneticandhistopathologyanalyses.Foreachcase,the geneticprofile
is given for each microsatellite marker for both tumors (O, ovarian; E, endometrial). For each marker, black indicates no genetic alterationin a tumor compared with the
normal tissue from the same patient. For eachinformative (heterozygous) marker, red represents loss of allele1and green represents loss ofallele 2.Yellow represents
microsatellite instability (MSI) and orange represents a different pattern of microsatellite instability for the same marker. Each genetic diagnosis was based on thenumber
of genetic similarities and differences between the two tumors in each patient. For example, in patient 5, we found10 genetic alterations in the ovarian cancer and the same
10 alterations in the endometrial cancer, indicating that these two tumors were related (i.e., a diagnosis of SPM). In patient10, we found19 different genetic alterations
between ovarian and endometrial cancers from the patient, and only1similarity.Therefore, this patient was diagnosed as DP. Finally, in patient 44, there were 4 genetic
similarities and6 geneticdifferences;wewereunabletomake ageneticdiagnosis ofDPorSPM withconfidenceinthis case.The ratio ofgenetic similarities to differenceswas
used to provide either an‘‘unrestricted’’genetic diagnosis (UG-Dx) or a restricted genetic diagnosis (RG-Dx). Additional keys: blue, diagnosis of dualprimary (D); yellow,
diagnosis of single primary cancer with metastasis (S); gray, diagnosis not known (NK); HISTDx, histologic diagnosis obtained from the pathology report.
Outcomein Synchronous Ovarian/Endometrial Cancer
www.aacrjournals.orgClin Cancer Res 2008;14(18) September15, 2008 5843

Page 5

diagnosis together with the specified exposures of interest (e.g.,
classification of subtype by histology). Wald tests were used to assess
significance. All models were fitted in the statistical language R.8
Results
Genetic and histopathologic diagnoses in synchronous ovarian
and endometrial cancer cases. We used genetic profiling to
establish a molecular diagnosis for 90 patients as either dual
primaryovarianandendometrialcancers(DP)orasingleprimary
cancer and associated metastasis (SPM). To test the accuracy of
genetic diagnoses,wefirstconductedthe analysisona trainingset
of 30 cases selected because pathologic data were able to indicate
with a high level of confidence a diagnosis of either SPM or DP.
The diagnosis made for each case is given in Fig. 2. Using the
unrestricted analysis, the call rate was 90% in the testing set and
the concordance with pathology diagnosis was 96%. When the
more stringent ‘‘restricted’’ analysis was applied, the call rate was
53% and the concordance was 94% (Tables 2 and 3).
We then went on to perform a genetic diagnosis for the
remaining cases; in 26 (43%) of these cases, the histopathology
reports had been unable to make a diagnosis. We were able to
make a genetic diagnosis in 51 (85%) cases using an
unrestricted analysis and in 30 (50%) cases using restricted
diagnosis. The diagnostic concordance in cases with an
accompanying histopathologic diagnosis was 85% and 92%,
respectively, for unrestricted and restricted genetic diagnoses.
Combining all samples, 57 of 90 (63%) patients were
diagnosed as DP, 21 patients (23%) as SPM, and the remaining
12 (13%) cases could not be classified as either using the
unrestricted genetic analysis. Using the restricted analysis, 10
(11%) cases were diagnosed as SPM, 36 (40%) cases as DP, but
44 (49%) cases could no longer be classified (Fig. 2; Table 2).
Histologic analysis had provided the following diagnoses: 25
cases were SPM; 39 cases were DP; and 26 cases were
unclassified. We compared overall diagnostic concordance
between histologic and genetic diagnoses. Using the unrestrict-
ed analysis, a diagnosis using both approaches was possible in
54 cases, of which 49 cases received the same diagnosis (91%
concordance). Using restricted genetic analyses, 27 of 29
patients received the same diagnosis (93% concordance; Table
3). Combining genetic and histologic diagnoses, it was possible
to make a diagnosis of SPM or DP in 88% compared with 71%
of cases using histology alone and 87% of cases using
unrestricted genetic analysis alone.
Finally, we further examined the five cases that were
discordant between genetic and histologic diagnoses. In all
five cases, a diagnosis of SPM was given in the histology reports
when the genetic analysis suggested a diagnosis of DP. For two
cases (10 and 59), the genetic analysis diagnosis seemed to be
unequivocal (Fig. 2). We conducted an additional histologic
analysis for all cases in this study, including a reanalysis of
tissue sections from all tumors by a gynecologic pathologist.
In four of five discordant cases, the original and the new
histologic diagnoses agreed, but in one patient (68), the revised
histologic diagnosis was DP when the original report suggested
a diagnosis of SPM. This was the only discordance in the 60
patients where two different histopathology reviews were able
to give a diagnosis.
Clinical features of synchronous ovarian and endometrial
cancer cases. Previous studies have suggested that patients
diagnosed with DP ovarian/endometrial tumors have a younger
age of onset than patients diagnosed with SPM (35). In this
study, patients with DP were generally younger at diagnosis
than patients with SPM. However, these differences were not
significant (P > 0.1).
We examined survival in patients with a diagnosis of SPM
or DP using histologic and genetic diagnoses (Fig. 3). After
adjusting for age, patients receiving a diagnosis of SPM based
on histology had significantly worse survival than patients
receiving a diagnosis of DP (P = 0.002). Patients diagnosed
with SPM using restricted and unrestricted genetic analyses
also have significantly worse survival than patients diagnosed
with DP (P = 0.004 and P = 0.044, respectively). A group of
44 patients that could not be classified by restricted genetic
diagnosis formed a group with intermediate survival; however,
this is not surprising when one considers that histology
diagnosed 15 of these cases with a SPM and 20 cases as DP.
Table 2. Comparison between histopathologic and genetic diagnoses (restricted and unrestricted) in 90
patients with synchronous ovarian and endometrial cancer
Genetic diagnosis (unrestricted) Genetic diagnosis (restricted)
Diagnoses
SPM (n = 21)DP (n = 57)Unclassified
(n = 12)
SPM (n = 10)DP (n = 36) Unclassified
(n = 44)
Training set* (n = 30)
SPM (n = 14)
DP (n = 16)c
Testing set (n = 60)
SPM (n = 11)
DP (n = 23)b
Unclassified (n = 26)
11
0
12
1
4
0
19
51511
6
0
4
41
6
2
4
0
2
1
8
6
17
20
15
9 15
*Samples selected for the training set were scored as either SPM or DP by both pathology report and pathology review.
cTwo cases with different histology and therefore known dual primary were included as an internal control.
bThree cases with different histology and therefore known dual primary were included as an internal control.
8RDevelopment CoreTeam. R: a language and environment for statistical comput-
ing.Vienna, Austria: R Foundation for Statistical Computing; 2005. ISBN 3-
900051-07-0. Available from: http://www.R-project.org.
Imaging, Diagnosis, Prognosis
www.aacrjournals.orgClin Cancer Res 2008;14(18) September15, 20085844

Page 6

Thus, the survival data from genetic diagnoses are consistent
with those from histologic diagnoses.
Interestingly, a diagnosis of SPM using a restricted genetic
analysis was associated with a worse survival than a diagnosis
of SPM made by either histologic or unrestricted genetic
analysis. There also seems to be a subset of cases diagnosed as
DP by genetics, which could not be diagnosed by histology,
which have worse survival than cases of DP confirmed by both
genetic and histology.
A previous study has suggested that patients with DP
diagnosed by histopathology, in which both tumors are of
endometrioid subtype, have a better survival than patients with
other tumor subtypes (29). Therefore, we investigated the
effects of histologic subtype on clinical outcome. The majority
of cases in this study were matched for histologic subtype.
Patients with endometrioid ovarian/endometrial cancer had
significantly better survival than patients with nonendometroid
tumors (P = 0.025; Fig. 4). This was also true when only
patients diagnosed as DP by genetic analysis were considered
(P = 0.02). We also examined whether the survival differences
we observed in patients diagnosed with DP by genetics or
histology were due to differences in the proportion of tumors
that were endometrioid compared with other histologic sub-
types. The ratio of endometrioid to nonendometrioid histologic
subtypes in DP cases diagnosed by histology was 3.9:1 (n = 39)
and 2.1:1 for patients diagnosed by genetics (n = 56). This is
consistent with a better survival in the group of patients
diagnosed by histology. These data also suggest that pathologists
havegreaterdifficultyinclassifyingcasesofDPcancersiftheyare
nonendometrioid as opposed to endometrioid subtypes.
It is well documented that confounding factors can influence
the analyses of survival. Therefore, where there were sufficient
data, we adjusted for other clinical factors (i.e., stage and grade).
Asexpected,bothstageandgradevarybetweendiagnosisgroups,
both being higher in tumors diagnosed as SPM (P < 0.05). Thus,
stage and grade variation may explain some of the diagnosis
related variation in survival. However, this data set is not large
enough to allow us to reliably estimate the relative contributions
of stage, grade, and residual unmeasured diagnosis effects on
survival. Data on surgical intervention (optimal or sub optimal
debulking), extent of dissemination, treatment, and recurrent
disease,whicharealsolikely to affect survival, were not available
on sufficient numbers of patients to enable analyses of sufficient
statistical power.
Finally,weinvestigatedtheassociationbetweensurvivalandthe
mode of metastatic spread in cases of SPM (i.e., whether
metastastic spread went from the ovary to endometrium or
vice versa). This could only be done for cases diagnosed by
histopathology.Metastaticspreadfromendometriumtotheovary
was associated with significantly better survival than ovarian
cancer spreading to the endometrium (P = 0.019; Fig. 4). In both
groups, survival was significantly worse than for patients
diagnosed with DP cancers (P = 0.017 and P = 0.00004,
respectively).
Discussion
Adjuvant therapeutic regimens for patients with synchronous
tumors of the ovary and endometrium vary and depend on
several factors, based largely on histopathologic criteria. Thus, a
correct diagnosis is crucial to ensure the appropriate treatment
is given. Management of endometrial cancer alone is primarily
surgical, but depending on the stage and histologic subtype,
some patients may require adjuvant radiotherapy to reduce
vaginal recurrence. Whereas the rate of serious complications
with radiotherapy is low, about 20% of patients have long-term
symptoms that influence their quality of life. For ovarian
cancer, current clinical management depends largely on tumor
stage and grade and usually combines surgery with adjuvant
chemotherapy. The standard first-line chemotherapy is carbo-
platin +/- paclitaxel, which is associated with significant
morbidity. Patients diagnosed with SPM would receive aggres-
sive treatment with adjuvant therapy. For patients with DP, the
treatment tends to be less aggressive and adjuvant therapy may
be unnecessary. However, when a diagnosis of SPM or DP
cannot be made with certainty, it is likely that the treatment
regimen will be based on a worse case scenario comprising
both adjuvant chemotherapy and radiotherapy.
In the current study, we analyzed 90 patients diagnosed with
synchronous ovarian and endometrial cancer. The original
histopathologic reports on which the patients’ treatment
regimens were based failed to provide a diagnosis in 26
(29%) cases. We have previously established the principle that
genetic analysis could be used to classify patients with either DP
or SPM ovarian/endometrial cancers (23). Here, we have
shown that genetic analysis can be used to both confirm the
histopathology diagnoses in patients with SPM or DP cancers
and, perhaps more importantly, provide a diagnosis in >80% of
cases in which pathology could not make a diagnosis. Most
cases of uncertain diagnosis from pathology were DP by genetic
analysis, which suggests that some of these patients may have
Table 3. Call rates for genetic diagnoses (restricted and unrestricted) and concordance with histopathologic
diagnosis in the training set, test set, and all patients
Genetic diagnosis (unrestricted) Genetic diagnosis (restricted)
CalledCall
rate (%)
Concordance
with pathology
Concordance
(%)
CalledCall
rate (%)
Concordance with
pathology
Concordance
(%)
Training set* (n = 30)c
Test set (n = 60)b
Total (n = 90)
27
51
78
90
85
87
26/27
23/27
49/54
96
85
91
16
30
46
53
50
51
15/16
12/13
27/29
94
92
93
*Samples selected for the training set were scored as either SPM or DP by both pathology report and pathology review.
cTwo cases with different histology and therefore known dual primary were included as an internal control.
bThree cases with different histology and therefore known dual primary were included as an internal control.
Outcomein Synchronous Ovarian/Endometrial Cancer
www.aacrjournals.orgClin Cancer Res 2008;14(18) September15, 20085845

Page 7

received unnecessarily aggressive treatment for their disease if
these genetic diagnoses are correct. When we looked at
treatment data for these patients, we found that three patients
with stage I/grade 2 endometrioid tumors were given adjuvant
chemotherapy and/or radiotherapy, when their disease may
only have required surgery.
Genetic and histologic diagnoses agreed in 91% of cases in
which both approaches were able to make a diagnosis in the
same patient. The level of agreement between genetic and
histologic diagnoses was not dependent on using ‘‘unrestricted’’
rather than the more stringent ‘‘restricted’’ data in making a
genetic diagnosis. Histology and genetic data disagreed in their
diagnoses in five cases, all of which were diagnosed as DP by
geneticsbutSPMbyhistology.Intwocases,thegeneticdiagnosis
seemed to be unequivocal. These data probably reflect a degree
of difficulty in using only histologic criteria to distinguish the
relationship between tumors that are clonally unrelated if they
appear to be morphologically similar. We also identified one
case that was diagnosed as SPM using the original pathology
reportbutasDPusingarevisedhistologicreview.Together,these
data suggest that whereas histopathology is generally correct
whenitisabletogiveadiagnosis,thisdiagnosismaybeincorrect
in f10% of cases.
Our ability to make a diagnosis based on the genetic data was
limited by the number of genetic alterations that we were able to
identify in the two tumors from the same patient. We analyzed
22 different markers and, on average, identified 5.5 genetic
alterationspertumorpair(range,0-20).Itislikelythatthepower
of genetic analysis to define the relationship between dual
tumors would be increased by genotyping more markers than
were studied here. Recent developments in genotyping technol-
ogies should improve this further; microarray platforms now
enable the analysis of several thousands of single-nucleotide
polymorphisms in a single assay, providing information about
copynumberalterationsintumorsthroughoutthegenome(36).
Survival analysis suggested that both histologic and genetic
diagnoses were effective at distinguishing between patients with
SPM or DP. As expected, patients diagnosed with SPM had
much poorer survival than patients with DP, and these were
statistically significant differences using both approaches.
However, there were also differences in survival between SPM
and DP diagnoses and patients left unclassified. For example,
Fig. 3. Survival analysis in patients diagnosed with DP
or SPM, or cases unclassifiedusinghistopathology
report (A; probability of survival to 5 y for a patient
age 50 y at diagnosis was 0.92, 0.63, and 0.72,
respectively); unrestricted genetic diagnosis
(B; probability of survival to 5 y for apatient age 50 y at
diagnosis was 0.82, 0.66, and 0.87, respectively);
restricted genetic diagnosis (C; probability of survival
to 5 y for a patient age 50 y at diagnosis was 0.87,
0.50, and 0.79, respectively). Unbrokenlines, cases
diagnosed as DP; dashedlines, cases diagnosed as
SPM; dottedlines, unclassified cases.
Imaging, Diagnosis, Prognosis
www.aacrjournals.org Clin Cancer Res 2008;14(18) September15, 20085846

Page 8

patients diagnosed with DP by genetics had a worse survival
than patients with the same diagnosis from histology. Con-
versely, the group of patients that could not be classified as SPM
or DP fared worse if the diagnoses were made by histology
compared with genetic analyses. Finally, the group with the
worst survival were patients diagnosed with SPM using a
restricted genetic analysis. Combined, these data perhaps suggest
a degree of disease heterogeneity that may influence survival
following diagnosis.
In this study, both tumors from each patient were matched
for histologic subtype in most cases. Cases with the same
histology are the most difficult to diagnose as either SPM or
DP. Patients with an ovarian and endometrial tumor of
different histologic subtype are almost invariably cases of DP
(11, 12). We found that patients in which both tumors were of
the endometrioid subtype had a significantly better survival
than patients with other histologic subtypes. This is in keeping
with the well-known observation that endometrioid carcino-
mas, especially in the uterus, are generally less aggressive than
other subtypes such as serous. Intriguingly, patients also
survived significantly longer if the mode of tumor spread was
from the endometrium to the ovary rather than from the ovary
to endometrium. The biological explanation for these findings
is unclear, but the data indicate that greater subclassification of
dual primary ovarian/endometrial tumors by histologic sub-
type will be required in the future to specify treatment
regimens more effectively.
In conclusion, we have shown that genetic analysis represents
a powerful tool for distinguishing between primary and
metastatic tumors in patients with synchronous ovarian and
endometrial cancers. This suggests that such an analysis could
be used to complement current histopathologic approaches to
diagnosis in these patients. A future goal could be to
incorporate these tests, which can be done on routinely
processed paraffin-embedded tissue, into everyday clinical
practice. The methods used for genetic diagnosis are generic
Fig. 4. Survival analysis in patients based on
histologic subtype stratification. A, patients in
which both endometrial and ovarian tumors are
endometrioid histology (unbroken line) have a
significantly better survival than patients in which
tumors are of another histologic subtype (dotted
line; P = 0.025). Probability of survival to 5 y for a
patient age 50 y at diagnosis with endometrioid
was 0.85, and nonendometrioid 0.7. B, a similar
result is observed for only those patients
diagnosed as DP by unrestricted genetic analysis
(P = 0.02). Probability of survival to 5 y for a
patient age 50 y at diagnosis with endometrioid
was 0.89, and nonendometrioid 0.69. C, cases of
SPMin which thehistology reports indicate that
the mode of metastatic spread was from the
ovary to the endometrium (dashedline) had a
significantly worse survivalthanforcasesinwhich
the mode of spread was from endometrium to
ovary (dashed/dottedline; P = 0.019). Unbroken
line, cases diagnosed as DP cases; dottedline,
patients of unknown diagnosis by histology.
Probability of survival to 5 y for a patient age 50 y
at diagnosis with endometrium to ovary was 0.74,
and ovary to endometrium 0.37.
Outcomein Synchronous Ovarian/Endometrial Cancer
www.aacrjournals.orgClin Cancer Res 2008;14(18) September15, 2008 5847

Page 9

and could also be applied to other instances of simultaneously
occurring cancers where histologic diagnosis may sometimes be
equivocal, including bilateral ovarian or breast cancers and
head and neck cancers. Finally, the data we present suggest a
hitherto uncharacterized level of heterogeneity in ovarian and
endometrial cancers from the same patient that could be used
to define different clinical phenotypes in this group of patients;
in the future, this might lead to improved clinical management
and outcome for the disease.
Disclosure of Potential Conflicts of Interest
Nopotential conflicts ofinterest were disclosed.
References
1. EisnerRF,NiebergRK,BerekJS.Synchronousprima-
ry neoplasms of the female reproductive tract. Gyne-
col Oncol1989;33:335^9.
2. Kline RC,Wharton JT, Atkinson EN, BurkeTW, Ger-
shenson DM, Edwards CL. Endometrioid carcinoma
of the ovary: retrospective reviewof145 cases.Gyne-
col Oncol1990;39:337^46.8.
3. Kurman RJ, Blaustein AU. editor Robert J. Kurman.
Blaustein’s pathology of the female genital tract. 4th
ed. NewYork: Springer-Verlag;1994.
4. Ayhan A,Yalcin OT,Tuncer ZS, GurganT, Kucukali T.
Synchronousprimarymalignanciesof thefemalegenital
tract.EurJObstet Gynecol ReprodBiol1992;45:63^6.
5. Zaino RJ, Unger ER,Whitney C. Synchronous carci-
nomas of the uterine corpus and ovary. Gynecol
Oncol1984;19:329^35.
6. Piura B, Glezerman M. Synchronous carcinomas of
endometrium and ovary. Gynecol Oncol 1989;33:
261^4.
7. HuangJ, Behrens C,Wistuba I, GazdarAF, JagirdarJ.
Molecular analysis of synchronous and metachronous
tumors of thelung: impact onmanagement andprog-
nosis. Ann Diagn Pathol 2001 ;5:321^9.
8. Pearl ML, Johnston CM, FrankTS, RobertsJA. Syn-
chronous dual primary ovarian and endometrial carci-
nomas. IntJGynaecol Obstet1993;43:305^12.
9. Eifel P, Hendrickson M, RossJ, Ballon S, Martinez A,
Kempson R. Simultaneous presentation of carcinoma
involving the ovary and the uterine corpus. Cancer
1982;50:163^70.
10. UlbrightTM, Roth LM. Metastatic andindependent
cancersoftheendometriumandovary:aclinicopatho-
logicstudyof34cases.HumPathol1985;16:28^34.
11. Ronnett BM, Kurman RJ. Endometrial Carcinoma.
In: Kurman RJ, editor. Blaustein’s pathology of the fe-
male genital tract. 5th ed. NewYork: Springer; 2002.
p. 501^60.
12. Young RH, Scully RE. Metaastatic tumors of the
ovary. In: Kurman RJ, editor. Blaustein’s pathology of
the female genital tract. 5th ed. NewYork: Springer;
2002. p.1063^102.
13. PratJ,Matias-GuiuX,BarretoJ.Simultaneouscarci-
nomainvolvingtheendometriumandtheovary.Aclini-
copathologic, immunohistochemical, and DNA flow
cytometricstudyof18cases.Cancer1991 ;68:2455^9.
14. SmitVT, Cornelisse CJ, DeJong D, Dijkshoorn NJ,
Peters AA, Fleuren GJ. Analysis of tumorheterogene-
ity in a patient with synchronously occurring female
genital tract malignancies by DNA flow cytometry,
DNA fingerprinting, and immunohistochemistry. Can-
cer1988;62:1146^52.
15. Symonds DA, Johnson DP,Wheeless CR, Jr. Feul-
gen cytometry in simultaneous endometrial and ovar-
ian carcinoma. Cancer1988;61:2511^6.
16. Shenson DL, Gallion HH, Powell DE, Pieretti M.
Loss of heterozygosity and genomic instability in
synchronous endometrioid tumors of the ovary and
endometrium. Cancer 1995;76:650^7.
17. Fujita M, EnomotoT,Wada H, Inoue M, OkudairaY,
Shroyer KR. Application of clonal analysis.Differential
diagnosis for synchronous primary ovarian and endo-
metrial cancers and metastatic cancer. Am J Clin
Pathol1996;105:350^9.
18. Shannon C, KirkJ, Barnetson R, EvansJ, et al.Inci-
dence of microsatellite instability in synchronous
tumors of the ovary and endometrium. Clin Cancer
Res 2003;9:1387^92.
19. Dinjens WN, van der Burg ME, Chadha S,
Sleddens HF, Burger CW, Ewing PC. Clinical im-
portance of molecular determinations in gynecolog-
ic patients with multiple tumors. Cancer 2003;97:
1766^74.
20. Gruvberger-SaalSK,CunliffeHE,CarrKM,Hedenfalk
IA. Microarrays in breast cancer research and clinical
practice-the future lies ahead. Endocr Relat Cancer
2006;13:1017^31.
21. van Beers EH, Nederlof PM. Array-CGH and breast
cancer. Breast Cancer Res 2006;8:210.
22. Khalique L, Ayhan A,Weale ME, Jacobs IJ, Ramus
SJ,GaytherSA.Geneticintra-tumourheterogeneityin
epithelial ovarian cancer and its implications for mo-
lecular diagnosis of tumours. J Pathol 2007;211:
286^95.
23. Brinkmann D, Ryan A, Ayhan A,et al. A molecular
genetic and statistical approach for the diagnosis of
dual-site cancers. J Natl Cancer Inst 2004;96:
1441^6.
24. Fujii H, MatsumotoT,Yoshida M, et al. Genetics of
synchronous uterine and ovarian endometrioid carci-
noma: combined analyses of loss of heterozygosity,
PTEN mutation, and microsatellite instability. Hum
Pathol 2002;33:421^8.
25. Kaneki E, OdaY, OhishiY, et al. Frequentmicrosatel-
lite instability in synchronous ovarian and endometrial
adenocarcinoma and its usefulness for differential di-
agnosis. Hum Pathol 2004;35:1484^93.
26. Nishimura N, HachisugaT, Nabeshima K, Kawara-
bayashiT.Synchronous endometrialandovariancarci-
nomas: analysis of genetic relationship of the tumors.
IntJOncol 2005;27:1519^26.
27. Leong PP, Rezai B, KochWM, et al. Distinguishing
second primary tumors from lung metastases in
patients with head and neck squamous cell carcino-
ma. JNatl Cancer Inst1998;90:972^7.
28. Kito H, Suzuki H, IgarashiT, et al. Distinct patterns
of chromosomal losses in clinically synchronous and
asynchronous bilateral renal cell carcinoma. J Urol
2002;168:2637^40.
29. Soliman PT, Slomovitz BM, Broaddus RR, et al.
Synchronous primary cancers of the endometrium
and ovary: a single institution review of 84 cases.
Gynecol Oncol 2004;94:456^62.
30. Ramus SJ, Pharoah PDP, Harrington P, et al.
BRCA1/2 mutation status influences somatic ge-
netic progression in inherited and sporadic epithe-
lial ovarian cancer cases. Cancer Res 2003;63:
417^23.
31.Yang-FengTL, Han H, Chen KC, et al. Allelic loss in
ovarian cancer. IntJCancer1993;54:546^51.
32. Cliby W, Ritland S, Hartmann L, et al. Human epi-
thelial ovarian cancer allelotype. Cancer Res1993;53:
2393^8.
33. Fabjani G, Kucera E, Schuster E, et al. Genetic alter-
ations in endometrial hyperplasia and cancer. Cancer
Lett 2002;175:205^11.
34. Canzian F, Salovaara R, Hemminki A, et al.Semiau-
tomated assessment of loss of heterozygosity and
replication error in tumors. Cancer Res 1996;56:
3331^7.
35. Nishimura N, Hachisuga T, Yokoyama M, Iwasaka
T, Kawarabayashi T. Clinicopathologic analysis of
the prognostic factors in women with coexistence
of endometrioid adenocarcinoma in the endometri-
um and ovary. J Obstet Gynaecol Res 2005;31:
120^6.
36. Dutt A, Beroukhim R. Single nucleotide polymor-
phism array analysis of cancer. Curr Opin Oncol
2007;19:43^9.
Imaging, Diagnosis, Prognosis
www.aacrjournals.orgClin Cancer Res 2008;14(18) September15, 20085848