synopsis
Pathology of Ovarian Cancers in BRCA1 and BRCA2 Carriers
Sunil R. Lakhani,1,2 Sanjiv Manek,3
Frederique Penault-Llorca,4 Adrienne Flanagan,5
Laurent Arnout,6 Samantha Merrett,1
Lesley McGuffog,7 Dawn Steele,1 Peter Devilee,16
Jan G. M. Klijn,17 Hanne Meijers-Heijboer,17
Paolo Radice,8 Silvana Pilotti,9 Heli Nevanlinna,25
Ralf Butzow,25,26 Hagay Sobol,18
Jocylyne Jacquemier,18
Dominique Stoppa Lyonet,23
Susan L. Neuhausen,24 Barbara Weber,22
Teresa Wagner,10 Robert Winqvist,11
Yves-Jean Bignon,4 Franco Monti,27
Fernando Schmitt,12 Gilbert Lenoir,13
Susanne Seitz,14 Ute Hamman,15 Paul Pharoah,21
Geoff Lane,20 Bruce Ponder,21
D. Timothy Bishop,19 and Douglas F. Easton7
1
The Breakthrough Toby Robins Breast Cancer Research Centre,
Institute of Cancer Research, London, United Kingdom; 2
The Royal
Marsden Hospital, London, United Kingdom; 3
Department of Cellular
Pathology, John Radcliffe Hospital, Oxford, United Kingdom; 4
Centre
Jean Perrin, Clermont-Ferrand, France; 5
Department of Pathology,
Royal Free & University College, London, United Kingdom; 6
Service
d’Anatomie Pathologique, Centre G-F Leclerc, Dijon, France;
7
Cancer Research U.K. Genetic Epidemiology Unit, Cambridge,
United Kingdom; 8
Department of Experimental Oncology Istituto
Nazionale Tumori, Milan, Italy; 9
Department of Pathology, Istituto
Nazionale Tumori, Milan, Italy; 10Department of Obstetrics and
Gynaecology, General Hospital, University of Vienna, Vienna,
Austria; 11Department of Clinical Genetics, Oulu University Hospital/
University of Oulu, Oulu, Finland; 12Institute of Pathology and
Molecular Immunology, University of Porto, R. Robert Frias, Porto,
Portugal; 13Laboratoire de Genetique, Lyon, France; 14Mac-DelbruckCentrum,
Berlin, Germany; 15Deutsches Krebsforschungszentrum,
Divisions of Epidemiology and Molecular Genome Analysis,
Heidelberg, Germany; 16Department of Genetics and Pathology,
Leiden University, Leiden, the Netherlands; 17Department of Clinical
Genetics and Medical Oncology, Daniel den Hoed Cancer Centre,
Erasmus University Medical Centre Rotterdam, Rotterdam, the
Netherlands; 18Department of Genetic Oncology and Cancer Control,
Paoli-Calmettes Institute, Marseille, France; 19Cancer Research UK
(CRUK) Genetic Epidemiology Laboratory and 20Department of
Gynaecological Oncology, St. James University Hospital, Leeds,
United Kingdom; 21CRUK Human Cancer Genetics Research Group,
Cambridge, United Kingdom; 22Abramson Family Cancer Research
Institute, University of Pennsylvania, Philadelphia, Pennsylvania;
23Unite de Genetique Oncologique, Institut Curie, Paris, France; 24Division Epidemiology, Department of Medicine, University of
California Irvine, Irvine California; 25Department of Obstetrics and
Gynaecology, Helsinki University Central Hospital, Helsinki, Finland;
26Department of Pathology, University of Helsinki, Helsinki, Finland; 27Laboratorio di Biologia Oncologica, Ospedale Infermi, Rimini, Italy
ABSTRACT
Purpose: Germline mutations in the BRCA1 and
BRCA2 genes confer increased susceptibility to ovarian cancer.
There is evidence that tumors in carriers may exhibit a
distinct distribution of pathological features, but previous
studies on the pathology of such tumors have been small.
Our aim was to evaluate the morphologies and immunophenotypes
in a large cohort of patients with familial ovarian
cancer.
Experimental Design: We performed a systematic review
of ovarian tumors from 178 BRCA1 mutation carriers,
29 BRCA2 mutation carriers, and 235 controls with a similar
age distribution. Tumors were evaluated by four pathologists
blinded to mutation status. Both morphological features
and immunochemical staining for p53 and HER2 were
evaluated.
Results: Tumors in BRCA1 mutation carriers were
more likely than tumors in age-matched controls to be invasive
serous adenocarcinomas (odds ratio, 1.84; 95% confidence
interval, 1.21–2.79) and unlikely to be borderline or
mucinous tumors. Tumors in BRCA1 carriers were of higher
grade (P < 0.0001), had a higher percentage solid component
(P 0.001), and were more likely to stain strongly for
p53 (P 0.018). The distribution of pathological features in
BRCA2 carriers was similar to that in BRCA1 carriers.
Conclusions: Use of pathological features can substantially
improve the targeting of predictive genetic testing.
Results also suggest that BRCA1 and BRCA2 tumors are
relatively aggressive and may be expected to have poor
prognosis, although this may be treatment dependent.
INTRODUCTION
The BRCA1 and BRCA2 genes are the most important
known predisposition genes for ovarian cancer. Mutations in
these genes cause a high lifetime risk of both breast and ovarian
cancer; the risk of ovarian cancer in BRCA1 mutation carriers is
40% by age 70, with the corresponding risk in BRCA2 carriers
being 10% (1). Mutations in these genes account for 5–13% of
ovarian cancer cases in Western countries (2, 3) and for the
majority of the familial aggregation of this disease (4).
Ovarian neoplasms can be subdivided into three main
groups: epithelial/stromal, germ cell, or sex cord/stromal. The
Received 7/11/03; revised 12/16/03; accepted 12/29/03.
Grant support: Cancer Research U.K. and NIH Grant CA74415.
D. Easton is a principal research fellow of Cancer Research U.K., and
S. Neuhausen is supported by NIH Grant CA74415.
The costs of publication of this article were defrayed in part by the
payment of page charges. This article must therefore be hereby marked
advertisement in accordance with 18 U.S.C. Section 1734 solely to
indicate this fact.
Note: This work was a United Kingdom Coordinating Committee on
Cancer Research (UKCCCR)/Breast Cancer Linkage Consortium
(BCLC) collaborative study.
Requests for reprints: Sunil R. Lakhani, The Breakthrough Toby
Robins Breast Cancer Research Centre, Institute of Cancer Research,
Mary-Jean Mitchell Green Building, Chester Beatty Laboratories, Fulham
Road, London SW3 6JB, United Kingdom. Phone: 44 020 7153
5525; Fax: 44 020 7153 5533; E-mail: Sunil.Lakhani@icr.ac.uk.
Vol. 10, 2473–2481, April 1, 2004 Clinical Cancer Research 2473
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
vast majority (90%) in the general population are epithelial in
origin, and a large proportion of these are benign. All studies
performed to date indicate that carcinoma (invasive epithelial
malignancy) is the usual histological diagnosis in BRCA1- and
BRCA2-associated ovarian cancer. The detailed pattern of histological
characteristics in mutation carriers compared with
ovarian cancer in noncarriers is less clear because most studies
have been based on relatively small numbers of cases. Most of
the available information relates to BRCA1-linked disease because
BRCA1 germline mutations are approximately four times
more common in ovarian cancer patients than BRCA2 mutations
(4). Most studies have reported that papillary serous adenocarcinoma
is the predominant type to occur in BRCA1 or BRCA2
carriers. Rubin et al. (5) reported that 43 of 53 women with
ovarian neoplasms who carried BRCA1 germline mutations had
papillary serous adenocarcinoma. They also found that the tumors
were of high grade. Stratton et al. (2) and Berchuck et al.
(6) obtained similar results in 12 of 13 and 15 of 15 individuals
studied, respectively. These data are further supported by results
reported by Risch et al. (3) and more recently by Shaw et al. (7).
However, three larger investigations have reported that papillary
serous carcinomas occurred with similar frequency in BRCA
mutation carriers compared with control groups (8).
Most studies have shown that malignant mucinous carcinoma
is underrepresented in BRCA1 mutation carriers (2), suggesting
that mutations in this gene do not generally play a role
in the development of this subtype of epithelial neoplasm.
However, occasional invasive (5) and borderline (2) mucinous
neoplasms have been described in BRCA1 mutation carriers.
In a large collaborative study carried out on behalf of the
Breast Cancer Linkage Consortium (BCLC), we characterized
the histopathological features of breast cancers arising in patients
harboring germline mutations in the BRCA1 and BRCA2
genes (9–11). The present study extends this approach to ovarian
cancer, using cases ascertained through the BCLC resource
together with cases identified from the United Kingdom Coordinating
Committee on Cancer Research Familial Ovarian Cancer
Study Group. It is a systematic blinded detailed review of
200 BRCA-associated ovarian cancers compared with population-based
controls carried out by specialist gynecological
pathologists. To our knowledge, this is the largest study on the
morphology and immunophenotype of these tumors.
MATERIALS AND METHODS
Ovarian Cancer Cases and Controls. We reviewed 223
“familial” tumors and 235 tumors unselected for family history.
Seventy-five of the familial cases were drawn from the UKCCR
study of familial ovarian cancer. All of these cases had at least
one first- or second-degree relative diagnosed with ovarian
cancer. The remaining familial cases were identified through
collaborating centers in the BCLC in the United Kingdom,
United States, the Netherlands, Ireland, Finland, Italy, France,
Germany, Austria, Portugal, Spain, Iceland, Switzerland, and
Hungary. These cases were identified on the basis of a family
history of breast and/or ovarian cancer. Of the 223 familial
cases, 178 were in women with germline BRCA1 mutations, 29
in women with BRCA2 mutations, and 16 had no definite
mutation in either gene. This latter group was not considered
further in the analysis because of its small size and the fact that
it was probably a heterogeneous mixture of mutation carriers
and noncarriers. For the purpose of these analyses we included
only those mutations that are classified as deleterious according
to the Breast Cancer Information Core (protein truncating
frameshift or nonsense mutations, large-scale rearrangements,
and splice-site and missense alterations classified as deleterious
by Breast Cancer Information Core). Nine of the 16 individuals
without definite mutations had possible disease-causing missense
variants or splice-site alterations in either BRCA1 or
BRCA2. Details on age at diagnosis and mutation type (but no
other identifying information) were collected.
Controls were drawn from a population-based study covering
West and North Yorkshire and Humberside over the
calendar year 1993 (190 tumors) and from a consecutive series
of patients from University College Hospital, London, over the
period 1980–1995 (45 tumors). Stratified random sampling by
age group (in decades) was used; a higher fraction of younger
cases was selected to minimize the difference in the overall age
distribution between the familial and unselected tumors.
We obtained specimens from case and control subjects in
the form of blocks or unstained 3-m-thick sections. All familial
case and control samples were allocated randomly generated
study numbers.
Morphological Analysis. Samples were analyzed for
morphological features by use of an agreed proforma similar to
that used in previous BCLC studies (9–11). The forms recorded
details of tumor subtype, histological grade (using the Silverberg
system), presence or absence of psammoma bodies, percentage
of solid component, presence of vascular invasion,
presence of necrosis, and total mitotic count. Each slide was
read independently by two pathologists (two of S. M., A. M. F.,
F. P-L., and L. A.). Because the slides were arranged and labeled
only by their study number, the pathologists were not
aware if the slide being read was from a case subject or a control
subject. The numbers in Tables 2 and 4 refer to the number of
observations of each histological category (counting the observations
by each pathologists separately) rather than the numbers
of tumors. No attempt was made to reconcile differences between
pathologists because it was difficult to design such a
process that would not introduce other biases.
The samples were analyzed for two immunohistochemical
markers, p53 and HER2, using the antibodies DO7 (DAKO) and
CB11 (Novocastra), respectively, and protocols as described
previously (11). Proformas based on those used for the BCLC
breast cancer analysis (11) were used to score the slides. For
p53, the intensity of staining was recorded as negative, low,
moderate, or strong. The pathologists were provided with identical
color charts to aid consistency in scoring the intensity of
the staining [ranging from white (negative) to dark brown
(strong)]. The proportion of positive cells was divided into six
categories: 0 to 1%, 1–5%, 6–25%, 26–50%, 51–75%, and
75%. For HER2, tumors in which the majority (75%) of
cells showed a strong complete membrane staining (equivalent
to a score of 3 on the DAKO scoring system) were classed as
positive. All other cases were recorded as negative. The slides were
evaluated independently by two pathologists (S. M., F. P-L.).
Statistical Analysis. Statistical analyses were performed
in a manner similar to our previous analyses of breast tumors.
2474 Ovarian Cancers in BRCA1 and BRCA2 Carriers
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
We performed separate analyses comparing tumors in BRCA1
carriers and BRCA2 carriers with control tumors. The effects of
each morphological feature on cancer status were summarized
in terms of odds ratios (ORs). All analyses were adjusted for age
in groups of 30, 30–39, 40–49, 50–59, and 60–69 years and
by reviewing pathologist. These adjusted analyses were carried
out with multiple logistic regression analysis, using the program
Stata (version 7.0).
The main complication in the analysis is that the observations
by different pathologists on the same slide cannot be
considered independent. Use of standard logistic regression
therefore leads to unbiased OR estimates but underestimates the
SE and confidence intervals (CIs). To correct for this, we
computed confidence limits, using the robust sandwich estimator
for the variance-covariance matrix (12), with the “robust”
option in Stata. This approach allows for variation in scoring
individual samples between the pathologists without explicitly
modeling the error distribution. Significance levels for each
factor were derived from the parameter estimates and the covariance
matrix (adjusted using the sandwich estimator). For
those factors measured on an ordinal scale (e.g., grade) onedegree
of freedom tests based on testing for linear trends in log
(OR) with increasing category were derived. Heterogeneity 2
statistics (based on k 1 degrees of freedom for factors with k
levels) are also presented.
To determine which factors were independently predictive
of BRCA1 status, we also performed multiple regression analyses.
In these analyses, all factors that were significant at the 5%
level, together with pathologist and age of the patient, were
initially included. Factors (other than age and pathologist) were
then removed from the model on a stepwise basis until no
further factors could be removed at the 5% level. (The corresponding
analysis was not conducted for BRCA2 because the
number of tumors was too small and none of the risk factor
distributions were clearly different from controls.)
Concordance between pathologists was assessed using
statistics. For characteristics on an ordinal scale, weighted s
were used. Confidence limits were constructed by bootstrapping
using 1000 bootstrap replicates.
The predicted prevalence of BRCA1 mutations in ovarian
cancer cases with given pathological characteristics were calculated
as in previous BCLC analyses of breast cancer (11). If
there are n risk categories with frequencies p0, p1,… pn 1 and
the OR for category j versus category 0 according the best
model is j
, then the mutation prevalence for cases in category
j is given by:
qj q0j (A)
where q0 K/
pjj
; and K is the overall prevalence. For the
purpose of this analysis we present age-specific prevalences for
the age-groups 30–39, 40–49, and 50–69 years (the last of
these based on the average of the prevalences in the 50–59 and
60–69 years age groups). The overall prevalence of BRCA1
mutations in ovarian cancer cases in these age groups were
derived from the studies by Stratton et al. (2) and Antoniou et al.
(1). Stratton et al. (2) found an overall prevalence of mutations
in ovarian cancer cases without a previous breast cancer (and
assuming 70% mutation sensitivity in that study) of 4.2%. On
the basis of the penetrance estimates for breast and ovarian
cancer from the meta-analysis of Antoniou et al. (1), the probability
of a BRCA1 carrier being affected with ovarian cancer
before breast cancer in each age group was as follows: 30
years, 0.015%; 30–39 years, 2.2%; 40–49 years, 8.3%; 50–59
years, 4.9%; 60–69 years, 6.9%. On the basis of these figures
and the corresponding population risks for England and Wales,
an overall prevalence of 4.2% corresponds to a BRCA1 carrier
frequency of 0.3%, and the predicted age-specific prevalence of
BRCA1 mutations in ovarian cancers in the age groups 30–39,
40–49, and 50–69 years are 6.6, 9.3 and 3.7%, respectively.
Some studies, notably Risch et al. (3) have reported a higher
overall prevalence of ovarian cancer. However, because the
pathology-specific prevalence estimates are simply proportional
to the assumed overall prevalence, these estimates can be scaled
as required.
RESULTS
The age distributions of the BRCA1 and BRCA2 carriers
and controls are shown in Table 1. Thirteen of the controls were
under 30 years of age, whereas none of the tumors in carriers
were diagnosed in this age-group. These controls were therefore
excluded from all of the analyses. After this exclusion, women
with BRCA1 tumors were, on average, younger than the controls,
whereas women with BRCA2 tumors were, on average,
older than the controls. The University College Hospital, London
controls were (by deliberate selection), on average, younger
than the Yorkshire controls.
In the review, three BRCA1 tumors and nine controls were
scored benign by one of the pathologists but borderline/invasive
by the other. Six of the control tumors were scored as benign by
both pathologists. The three BRCA1 tumors scored as benign
were all scored “not-assessable” by the other pathologist. Of the
controls scored benign by one pathologist, the other scored
seven borderline, one invasive, and one not-assessable. For
consistency, we removed all of these 18 tumors from further
analyses.
Consistency between pathologists was assessed on the remaining
403 tumors. As expected, the statistic was highest for
borderline/invasive (0.72) and lowest for vascular invasion
(0.14). values for the remaining morphological features varied
from 0.39 to 0.57. Agreement was good for p53 staining (
0.89) but weaker for HER2 ( 0.14).
The distribution of morphological characteristics is shown
Table 1 Numbers of tumors in the review, by mutation status and
age group
Age group
years
BRCA1,
n (%)
BRCA2,
n (%)
Controls (n)
Total Yorkshire UCLa
30 13 7 6
30–39 20 (11%) 16 10 6
40–49 65 (37%) 3 (10%) 49 30 19
50–59 64 (36%) 11 (38%) 67 55 12
60–69 22 (12%) 10 (35%) 56 54 2
70–79 7 (4%) 4 (14%) 27 27 0
80 0 1 (3%) 7 7 0
a UCL, University College Hospital, London.
Clinical Cancer Research 2475
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
in Table 2, and the corresponding ORs and significance levels,
adjusted for age and pathologist, are listed in Table 3. The
frequency of borderline tumors among BRCA1 carriers (1%)
was markedly lower than in the control group (10%; OR, 0.044;
P 0.0001). The frequency of borderline tumors was also
lower in the BRCA2 carriers than in the controls (OR, 0.57; 95%
CI, 0.15–2.21), but the difference was less marked and not
statistically significant.
The remaining analyses were restricted to tumors scored as
invasive. As anticipated from previous reports, the distribution
of histological type was markedly different among BRCA1
tumors than control tumors. Specifically, the frequency of serous
tumors was higher among BRCA1 tumors (OR, 1.84; 95%
CI, 1.21–2.79; P 0.004), whereas the frequency of mucinous
tumors was much lower (OR, 0.13; 95% CI, 0.05–0.34; P
0.0001). However, the frequency of other histological types in
BRCA1 tumors was higher than in previous reports. Even if
attention was restricted to tumors in which only one histological
type was recorded, only 45% of tumors were reported to be
serous. There was also some evidence of an increased frequency
of giant cell type in BRCA1 tumors (OR, 2.61; 95% CI, 1.17–
5.82). Endometrioid and clear cell tumors were less frequent in
BRCA1 carriers but not significantly so. The distribution of
histological types in BRCA2 tumors was very similar to that in
BRCA1 tumors, but (as a result of the small sample size) did not
differ significantly from the control distribution.
Both BRCA1 and BRCA2 tumors were of higher grade on
average than control tumors (P 0.0001 and P 0.028
respectively) and had a higher percentage solid component (P
0.0004 and P 0.056 respectively). The relationship with
mitotic count was less clear. There was some evidence of a
difference in the distribution of mitotic count between BRCA1
carriers and controls (heterogeneity P 0.049), but this was
mainly due to a higher frequency of the 20–29 mitoses/10hpf
Table 2 Distribution of morphological features in tumors from BRCA1 and BRCA2 carriers and controls
Factor Level
BRCA1 BRCA2 Controls
n % n % n %
Invasion Invasive 325 51 334
Borderline 4 1 5 10 69 20
NAa 15 2 1
Total 344 58 404
Histological type Serous 145 44 24 48 108 31
Mucinous 8 3 3 5 42 38
Endometroid 118 36 20 38 136 40
TCC 8 2 1 2 2 1
Sarcoma 2 1 1 2 2 1
Clear cells 51 16 6 16 71 23
Giant cells 21 6 3 6 9 3
Papillary 2 0.6 2 4 0 0
Squamous 3 1 0 0 9 3
Histological type
(one type only)
Serous 131 40 20 39 97 29
Mucinous 7 2 1 2 36 11
Endometroid 106 33 15 29 109 33
TCC 8 2 0 0 1 0.3
Sarcoma 2 0.6 1 2 2 0.6
Clear cell 32 10 2 4 44 13
Other/multiple 39 12 12 24 45 13
Psammoma bodies Present 60 18 9 19 60 22
Absent 265 42 274
Grade Well 5 1 2 3 36 10
Moderately 87 28 7 16 115 35
Poorly 181 57 26 52 153 46
Undifferentiated 52 15 16 29 30 9
Solid component Absent 32 9 4 8 62 19
25% 75 24 7 14 98 30
25–75% 127 40 19 41 115 34
75% 91 27 21 37 59 18
Vascular invasion None 229 42 242
Present 96 29 9 23 92 26
Necrosis None 105 32 13 26 102 31
Focal 116 36 17 31 114 35
Moderate 58 19 10 24 73 22
Marked 46 14 11 19 45 13
Mitotic count 10 58 18 4 8 73 22
10–19 67 22 1 5 70 21
20–29 66 21 9 19 51 16
30–39 35 11 10 17 46 14
a NA, not assessible, TCC, transitional cell carcinoma.
2476 Ovarian Cancers in BRCA1 and BRCA2 Carriers
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
category in the BRCA1 tumors, with no evidence of an elevated
frequency of tumors with 30 or more mitoses/10hpf. Mitotic
count was higher on average in BRCA2 carriers than controls;
again however the test for trend with increasing mitotic count
was not significant and the effect was only significant when
mitotic count was considered as an (unordered) categorical
variable (P 0.012). None of the other features considered
(presence of psammoma bodies, vascular invasion and necrosis)
differed significantly in frequency between BRCA1 or BRCA2
tumors and controls.
The results obtained by immunohistochemical examination
of the tumors and controls with the antibodies to HER2 and p53
are shown in Table 4; the corresponding ORs and significance
levels are shown in Table 5. No differences in HER2 expression
were identified. There was some evidence for an increased
frequency of p53 staining among BRCA1 tumors compared with
control tumors. There was no apparent effect for mild staining,
but the estimated OR for strong staining compared with no
staining was 2.96 (95% CI, 1.18–7.44). A similar pattern was
observed for the proportion of cells stained positive for p53. The
estimated ORs for BRCA2 were consistent with an effect similar
to BRCA1, but the numbers were too small to show a significant
difference from controls.
To evaluate the independent predictive value of these morphological
and immunohistochemical features on BRCA1 positivity,
we next performed a multiple logistic regression analysis.
Tumor grade, histological type, and p53 staining remained independently
significant, whereas percentage of solid component
did not (Table 6). To test the adequacy of this model, we also
fitted models that included interactions between histological
type, grade, and p53 status and between any of these factors and
age at diagnosis. We found no significant evidence of interaction
(data not shown).
We used these results to compute the predicted BRCA1
Table 3 Estimated odds ratios for individual morphological features in BRCA1 and BRCA2 tumor versus control
Factor Level
BRCA1 BRCA2
ORa 95% CI OR 95% CI
Borderline (vs. invasive) 0.044 0.015–013 0.57 0.15–2.21
1
2 33.61 P 0.0001 1
2 0.67
Histological type (vs. all
other types)
Serous 1.84 1.21–2.79 1.72 0.85–3.51
Mucinous 0.13 0.05–0.34 0.56 0.06–1.90
Endometrioid 0.78 0.52–1.16 0.87 0.42–1.81
Clear cell 0.77 0.45–1.30 0.51 0.20–1.27
Giant cells 2.61 1.17–5.82 2.24 0.06–1.93
Histological type (one type
only)
Serous 1.0 1.0
Mucinous 0.10 0.03–0.31 0.18 0.02–1.43
Endometrioid 0.65 0.39–1.06 0.69 0.28–1.73
Clear cell 0.59 0.28–1.22 0.24 0.0.05–1.13
3
2 16.95 P 0.0007 3
2 4.28
Psammoma bodies Present 1.15 0.68–1.94 0.95 0.38–2.35
1
2 0.27 3
2 0.03
Grade Well 1.0 1.0
Moderately 5.98 2.12–16.88 1.04 0.21–5.06
Poorly 10.30 3.59–29.6 3.20 0.73–14.01
Undifferentiated 15.44 4.74–50.4 11.92 2.41–59.0
1
2 20.34 P 0.0001 1
2 4.80 P 0.028
3
2 23.83 P 0.0001 3
2 18.22 P 0.0001
Solid component Absent 1.0 1.0
25% 1.71 0.98–2.97 1.07 0.26–4.36
25–75% 2.62 1.43–4.80 2.52 0.61–10.43
75% 3.64 1.78–7.43 6.78 1.65–27.86
1
2 12.55 P 0.004 1
2 3.64 P 0.056
3
2 13.77 P 0.001 3
2 13.06 P 0.0014
Vascular invasion Present 1.02 0.67–1.56 0.47 0.20–1.10
1
2 0.02 1
2 3.00 P 0.08
Necrosis None 1.0 1.0
Focal 0.94 0.61–1.45 1.26 0.61–2.61
Moderate 0.80 0.46–1.39 1.27 0.48–3.33
Marked 1.16 0.59–2.26 2.11 0.68–6.51
1
2 0.00 3
2 1.53 1
2 1.32 3
2 1.70
Mitotic count 10 1.0 1.0
10–19 1.20 0.66–2.16 0.27 0.02–2.96
20–29 1.86 0.96–3.59 2.78 0.54–14.2
30–39 0.82 0.42–1.60 3.64 0.68–19.6
40 1.37 0.74–2.51 5.06 1.05–24.3
1
2 0.59 P 0.049 1
2 2.01 P 0.16
4
2 7.84 4
2 10.91 P 0.012
a OR, odds ratio; CI, confidence interval.
Clinical Cancer Research 2477
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
carrier probabilities among ovarian cancer patients with given
histological characteristics. Because the distributions of histological
types other than serous and mucinous were similar in
BRCA1 tumors and controls, we classified tumors for this purpose
as serous, mucinous, or other. Among tumors with more
than one histological type reported, serous was more likely to be
reported in BRCA1 tumors (OR, 2.45; 95% CI, 0.49–12.12), and
mucinous type was less likely to be reported (OR, 0.30; 95% CI,
0.032–2.81), similar to the pattern in tumors with a unique type.
We therefore classified tumors with serous and another type as
serous and tumors with mucinous and another type as mucinous.
Three tumors reported as both serous and mucinous were excluded
from this analysis.
Predicted BRCA1 carrier probabilities for different subgroups
based on age and pathology are given in Table 7. On the
basis of age, histological type, and grade, predicted carrier
probabilities exceeded 10% for serous tumors that were undifferentiated
or poorly differentiated in women 30–49 years of
age at diagnosis and moderately differentiated in women 40–49
years of age at diagnosis. They also exceeded 10% for tumors
diagnosed as “other histology” for undifferentiated tumors in
women diagnosed at age 30–49 or poorly differentiated tumors
in women 40–49. In contrast, carrier probabilities were 3%
for all categories of mucinous or well-differentiated tumors.
Table 7 also illustrates the additional predictive power of p53
staining. Thus, for poorly or undifferentiated serous tumors
diagnosed below age 50 with strong p53 staining, the predicted
carrier probability exceeded 20%.
DISCUSSION
The present study is the largest histopathological review of
its kind and clarifies and extends the morphological and immunological
profiles of familial ovarian cancers due to BRCA1 and
BRCA2.
Histopathological typing of tumors is commonly subject to
significant interobserver variation. Ovarian carcinoma is no
exception, and categorization is particularly difficult when a
lesion is high grade (13). Furthermore, consensus criteria for
grading ovarian carcinomas have not been agreed on and consequently
differ among individuals (14, 15), although a recently
proposed system by Shimizu et al. (16) and validated by Shaw
et al. (7) has helped provide uniformity among pathologists in
this area. The difficulty in subtyping ovarian carcinomas is
clearly shown in the publication by Pharoah et al. (8) in which
59% (61 of 133 cases) of the BRCA1-associated neoplasms and
36% (8 of 26 cases) of the BRCA2-associated cancers were
classified as unspecified carcinomas. The subjectivity of typing
and grading is likely to account, at least in part, for the different
results generated from studies undertaken to date. Systematic
reviewing of the slides included in familial cancer studies by a
group of histopathologists with a specialist interest in gynecological
pathology has the benefit of reducing the interobserver
diagnostic variation, but this has been performed only in studies
by Zweener et al. (17), Shaw et al. (7) and Werness et al. (18).
We attempted to minimize the effects of interobserver variability
in the present study by blinding the pathologists with respect
to mutation status, by arranging for each slide to be scored by
two different pathologists, and by adjusting for pathologist as a
covariate in the analysis. The concordance as measured by
values was reasonably high for most features.
Accepting the limitations of morphological analysis, the
present results emphasize the greater frequency of serous carcinomas
in BRCA1-associated tumors, consistent with previous
studies by Rubin et al. (5), Berchuck et al. (6), and more
Table 4 Distribution of immunohistochemical features in tumors
from BRCA1 and BRCA2 carriers and controls
Antibody Level
BRCA1 BRCA2 Control
n % n % n %
HER2 Positive 24 17 4 22 45 14
Negative 120 14 273
p53 Negative 13 9 2 11 55 18
Low 17 12 1 6 68 22
Medium 37 26 7 39 79 25
Strong 73 52 8 44 111 35
1% 17 12 2 11 67 21
1–25% 17 12 2 11 62 20
26–100% 106 76 14 78 185 59
Table 5 Estimated odds ratios for immunohistochemical features in BRCA1 and BRCA2 tumors versus controls
Antibody Staining
BRCA1 tumors BRCA2 tumors
Odds ratio 95% CIa Odds ratio 95% CI
C-erb-b2 Positive 1.24 0.63–2.44 1.95 0.44–8.61
1
2 0.40 1
2 0.79
p53 Negative 1.0 1.0
Low 0.85 0.29–2.47 0.42 0.023–7.46
Moderate 1.71 0.64–4.57 2.59 0.27–24.5
Strong 2.96 1.18–7.44 1.97 0.21–18.5
1
2 2.54 P 0.11 1
2 0.22
3
2 10.01 P 0.018 3
2 3.59
1%b 1.0
1–25% 0.82 0.34–1.97
26–100% 2.35 1.08–5.10
1
2 1.61
2
2 9.06 P 0.01
b There were too few BRCA2 carriers to provide estimates by percentage of cells stained. a CI, confidence interval.
2478 Ovarian Cancers in BRCA1 and BRCA2 Carriers
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
recently by Shaw et al. (7). Conversely, the frequency of mucinous
tumors is much lower than among ovarian cancer patients
in general. The frequencies of endometrioid and clear cell
carcinomas were similar to, or slightly lower than, their frequencies
in controls, in accordance with other reports (5). These
types therefore represent a significant fraction of tumors in
BRCA1 carriers (36 and 18%, respectively). Although other
tumor types, including transitional cell carcinomas, papillary
and squamous carcinomas, and sarcomas, were observed, they
were rare, accounting for 10% of all tumors. A dysgerminoma
arising in a woman with a BRCA1 germline mutation has recently
been reported (19), but we found no examples of malignant
germ cell tumors.
We found that borderline tumors are much rarer (as a
proportion of all ovarian tumors) in BRCA1 carriers, in accordance
with previous observations (20). The age-adjusted frequency
of borderline tumors was 1/20th of the frequency in
unselected cases, whereas the incidence rates for ovarian cancer
in women older than age 30 are 50-fold greater than in
noncarriers (1). Given the wide confidence limits on these
estimates, it is thus possible that BRCA1 mutations confer little
or no increased risk of borderline ovarian cancer.
Our data demonstrate that BRCA1-associated tumors are of
higher grade, on average, than control tumors. This difference
has been found in several other studies (5, 7, 8, 21). In contrast,
Berchuck et al. (6) found that although the BRCA1 cases were
all of advanced stage (III/IV), they were less likely to be poorly
differentiated compared with cases without mutations, and Johannsson
et al. (22) did not identify a difference in grade
between the ovarian cancers in their BRCA mutation carriers and
the control population-based cancer registry group. We have
also found a greater proportion of solid tumor in BRCA1 tumors,
indicating poor differentiation, an effect also seen by Shaw et al.
(7). The other morphological features, such as vascular invasion,
necrosis, and mitotic count, were not significantly associated
with BRCA1 positivity in this study.
The requirement that the mutation-positive cases be tested
implies that cases with very poor survival may not have been
included in our study. Because such cases are likely to be high
grade, the effect of grade may, if anything, have been underestimated
in this study.
Consistent with the association with grade, we found a
higher frequency of strong p53 staining in BRCA1 and BRCA2
tumors. These results are consistent with those of Ramus et al.
(23), who analyzed both p53 immunohistochemistry and p53
mutations in 30 BRCA1, 18 BRCA2, and 33 sporadic ovarian
cancers. The frequencies of p53 overexpression in the three
groups were 70% for BRCA1, 67% for BRCA2, and 39% for
sporadic ovarian carcinomas, whereas the corresponding mutation
frequencies were 60, 50, and 30% respectively. In contrast,
our study did not reveal any difference in HER2 expression
between BRCA1 and BRCA2 ovarian cancers or controls. This
Table 6 Multiple logistic regression analysis of histological factors
in BRCA1 carriers vs. controlsa
Factor ORa 95% CI
Histological type
Serous 1.0
Mucinous 0.14 0.038–0.49
Endometrioid 0.82 0.45–1.49
Clear cell 0.25 0.082–0.79
3
2 12.38 P 0.006
Grade
Well differentiated 1.0
Moderately differentiated 1.83 0.61–5.50
Poorly differentiated 3.45 1.07–11.16
Undifferentiatedb 5.12 1.23–21.25
1
2 4.84 P 0.028
3
2 7.15 P 0.067
p53 staining
Negative/Low 1.0
Medium 2.03 0.86–4.82
Strong 3.67 1.56–8.61
1
2 7.80 P 0.005
2
2 8.97 P 0.011
a OR, odds ratio; CI, confidence interval.
b No controls with undifferentiated mucinous tumours were observed
in this study
Table 7 Predicted percentage of BRCA1 carriers among ovarian cancer patients with a given histological type
Type Age (years) p53 staining Well Differentiated Moderately differentiated Poorly differentiated Undifferentiated
Serous 30–39 1.9 7.8 13.6 20.5
40–49 2.6 10.7 18.8 28.2
50–59 0.9 3.9 6.8 10.1
Mucinous 30–39 0.2 0.9 1.6 a
40–49 0.03 1.3 2.3 a
50–59 0.1 0.5 0.8 a
Other 30–39 1.1 4.6 8.1 12.1
40–49 1.5 6.3 11.1 16.6
50–59 0.6 2.2 4.0 6.0
Serous 30–39 None/Low 2.2 3.7 5.9 9.1
Medium 4.4 7.3 11.7 18.3
Strong 7.7 12.9 20.5 32.1
Serous 40–49 None/Low 3.2 5.3 8.4 13.1
Medium 6.3 10.6 16.8 26.2
Strong 11.1 18.5 29.5 46.0
Serous 50–59 None/Low 1.0 1.6 2.5 4.0
Medium 1.9 3.2 5.1 7.9
Strong 3.4 5.6 8.9 13.9
a No controls with undifferentiated mucinous tumors were observed in this study.
Clinical Cancer Research 2479
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
contrasts with the pattern in breast cancer, in which HER2
overexpression is less frequent in BRCA-associated cancers than
in controls (11).
This study is the largest formal evaluation of ovarian
cancers in BRCA2 carriers, although the number of tumors is
still small. We found that the distribution of histology features
in BRCA2 carriers was very similar to those in BRCA1 carriers,
with a very low frequency of borderline and mucinous tumors,
a higher than average frequency of serous tumors, and smaller
but significant frequencies of endometrioid and giant-cell tumors.
This pattern has been reflected in other, smaller studies
(24). We also found that BRCA2 tumors were of higher than
average grade and solid component. This similarity in ovarian
cancer pathology between BRCA1 and BRCA2 carriers contrasts
with the breast cancer pathology, where there is a very marked
contrast between BRCA1- and BRCA2-associated disease. The
only notable differences between BRCA1- and BRCA2-related
ovarian cancer are the much lower risk in BRCA2 carriers and
the different age distributions, with BRCA2-associated disease
occurring later in life (1).
Although there is some disagreement regarding grade as a
prognostic factor, the increased frequency of high grade and
strong p53 staining, both of which have been shown in some
studies to be adverse prognostic factors in tumors in BRCA1
carriers, raises the possibility that the disease may have a poor
prognosis in these women. The direct evidence on the survival
in carriers is conflicting. Rubin et al. (5) found better survival in
carriers, but this effect may, at least in part, have been an artifact
because carriers needed to be alive to be tested. Aida et al. (25)
found it was twice as likely that women with BRCA-associated
cancers had a negative second-look operation compared with
their matched controls, whereas Boyd et al. (21) also found that
post-chemotherapy disease-free survival was extended compared
with that of the nonhereditary group. In contrast, Johannsson
et al. (22) found an initial survival advantage in their
BRCA1-associated ovarian cancer group that was lost over time.
Pharoah et al.(8) found essentially no difference in survival
between patients with BRCA1 or BRCA2 germline mutations
and noncarriers. The apparent absence of a survival disadvantage
in carriers might reflect an increased sensitivity to chemotherapy
in carriers. The hypothesis might be particularly pertinent
in BRCA2 carriers, given the role of BRCA2 in DNA
cross-link repair (26).
Morphological and immunohistochemical analysis provides
a powerful predictor of BRCA1 mutation status that could
aid genetic testing programs. Even in the absence of information
on family history, the mutation prevalence in women with
poorly differentiated and undifferentiated serous tumors exceeds
10% in most age groups, whereas the prevalence of mucinous
and well-differentiated tumors is low. The addition of p53
staining further improves prediction. Because carrier distributions
of tumor types are similar in BRCA1 and BRCA2 tumors,
similar predictions can be made for BRCA2 tumors.
ACKNOWLEDGMENTS
S. L. Neuhausen would like to acknowledge the help of Michael
Hoffman. Dr. Jorge Reis-Filho is acknowledged for critical comments
and discussions and S. Johnson for administrative help.
REFERENCES
1. Antoniou A, Pharoah PDP, Narod S, et al. Average risks of breast
and ovarian cancer associated with mutations in BRCA1 or BRCA2
detected in case series unselected for family history: a combined analysis
of 22 studies. Am J Hum Genet 2003;72:1117–30.
2. Stratton JF, Gayther SA, Russell P, et al. Contribution of BRCA1
mutations to ovarian cancer. N Engl J Med 1997;336:1125–30.
3. Risch HA, McLaughlin JR, Cole DE, et al. Prevalence and penetrance
of germline BRCA1 and BRCA2 mutations in a population
series of 649 women with ovarian cancer. Am J Hum Genet 2001;68:
700–10.
4. Gayther SA, Russell P, Harrington P, Antoniou AC, Easton DF,
Ponder BA. The contribution of germline BRCA1 and BRCA2 mutations
to familial ovarian cancer: no evidence for other ovarian cancersusceptibility
genes. Am J Hum Genet 1999;65:1021–9.
5. Rubin SC, Benjamin I, Behbakht K, et al. Clinical and pathological
features of ovarian cancer in women with germ-line mutations of
BRCA1. N Engl J Med 1996;335:1413–6.
6. Berchuck A, Heron KA, Carney ME, et al. Frequency of germline
and somatic BRCA1 mutations in ovarian cancer. Clin Cancer Res
1998;4:2433–7.
7. Shaw PA, McLaughlin JR, Zweemer RP, et al. Histopathologic
features of genetically determined ovarian cancer. Int J Gynecol Pathol
2002;21:407–11.
8. Pharoah PD, Easton DF, Stockton DL, Gayther S, Ponder BA.
Survival in familial, BRCA1-associated, and BRCA2-associated epithelial
ovarian cancer. United Kingdom Coordinating Committee for Cancer
Research (UKCCCR) Familial Ovarian Cancer Study Group. Cancer
Res 1999;59:868–71.
9. Lakhani SR, Jacquemier J, Sloane JP, et al. Multifactorial analysis of
differences between sporadic breast cancers and cancers involving
BRCA1 and BRCA2 mutations. J Natl Cancer Inst (Bethesda) 1998;90:
1138–45.
10. Lakhani S, Easton D, Stratton MR. Pathology of familial breast
cancer: differences between breast cancers in carriers of BRCA1 or
BRCA2 mutations and sporadic cases. Breast Cancer Linkage Consortium.
Lancet 1997;349:1505–10.
11. Lakhani SR, Van De Vijver MJ, Jacquemier J, et al. The pathology
of familial breast cancer: predictive value of immunohistochemical
markers estrogen receptor, progesterone receptor, HER-2, and p53 in
patients with mutations in BRCA1 and BRCA2. J Clin Oncol 2002;20:
2310–8.
12. Huber PJ. The behaviour of maximum likelihood estimates under
nonstandard conditions. Proc Berkeley Symp Math Stat Probab 1967;
1:221–33.
13. Sakamoto A, Sasaki H, Furusato M, et al. Observer disagreement in
histological classification of ovarian tumors in Japan. Gynecol Oncol
1994;54:54–8.
14. Cramer SF, Roth LM, Ulbright TM, et al. Evaluation of the reproducibility
of the World Health Organization classification of common
ovarian cancers. With emphasis on methodology. Arch Pathol Lab Med
1987;111:819–29.
15. Lund B, Thomsen HK, Olsen J. Reproducibility of histopathological
evaluation in epithelial ovarian carcinoma. Clinical implications.
Apmis 1991;99:353–8.
16. Shimizu Y, Kamoi S, Amada S, Akiyama F, Silverberg SG.
Toward the development of a universal grading system for ovarian
epithelial carcinoma: testing of a proposed system in a series of 461
patients with uniform treatment and follow-up. Cancer (Phila) 1998;
82:893–901.
17. Zweemer RP, Shaw PA, Verheijen RM, et al. Accumulation of p53
protein is frequent in ovarian cancers associated with BRCA1 and
BRCA2 germline mutations. J Clin Pathol 1999;52:372–5.
18. Werness BA, Ramus SJ, Whittemore AS, et al. Histopathology of
familial ovarian tumors in women from families with and without
germline BRCA1 mutations. Hum Pathol 2000;31:1420–4.
2480 Ovarian Cancers in BRCA1 and BRCA2 Carriers
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
19. Werness BA, Ramus SJ, Whittemore AS, et al. Primary ovarian
dysgerminoma in a patient with a germline BRCA1 mutation. Int J
Gynecol Pathol 2000;19:390–4.
20. Gotlieb WH, Friedman E, Bar-Sade RB, et al. Rates of Jewish
ancestral mutations in BRCA1 and BRCA2 in borderline ovarian tumors.
J Natl Cancer Inst (Bethesda) 1998;90:995–1000.
21. Boyd J, Sonoda Y, Federici MG, et al. Clinicopathologic features
of BRCA-linked and sporadic ovarian cancer. JAMA 2000;283:
2260 –5.
22. Johannsson OT, Ranstam J, Borg A, Olsson H. Survival of BRCA1
breast and ovarian cancer patients: a population-based study from southern
Sweden. J Clin Oncol 1998;16:397–404.
23. Ramus SJ, Bobrow LG, Pharoah PD, et al. Increased frequency of
TP53 mutations in BRCA1 and BRCA2 ovarian tumours. Genes Chromosomes
Cancer 1999;25:91–6.
24. Zweemer RP, Verheijen RH, Gille JJ, van Diest PJ, Pals G, Menko
FH. Clinical and genetic evaluation of thirty ovarian cancer families.
Am J Obstet Gynecol 1998;178:85–90.
25. Aida H, Takakuwa K, Nagata H, et al. Clinical features of ovarian
cancer in Japanese women with germ-line mutations of BRCA1. Clin
Cancer Res 1998;4:235–40.
26. Tutt A, Ashworth A. The relationship between the roles of BRCA
genes in DNA repair and cancer predisposition. Trends Mol Med
2002;8:571–6.
Clinical Cancer Research 2481
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
Clin Cancer Res 2004;10:2473-2481.
Sunil R. Lakhani, Sanjiv Manek, Frederique Penault-Llorca, et al.
Pathology of Ovarian Cancers in BRCA1 and BRCA2 Carriers
Updated version
http://clincancerres.aacrjournals.org/content/10/7/2473
Access the most recent version of this article at:
Cited articles
http://clincancerres.aacrjournals.org/content/10/7/2473.full.html#ref-list-1
This article cites 24 articles, 8 of which you can access for free at:
Citing articles
http://clincancerres.aacrjournals.org/content/10/7/2473.full.html#related-urls
This article has been cited by 26 HighWire-hosted articles. Access the articles at:
E-mail alerts Sign up to receive free email-alerts related to this article or journal.
Subscriptions
Reprints and
Department at pubs@aacr.org.
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Permissions
Department at permissions@aacr.org.
To request permission to re-use all or part of this article, contact the AACR Publications
Research.
Downloaded from clincancerres.aacrjournals.org on May 4, 2016. © 2004 American Association for Cancer
