Accepted Manuscript “Structured hysteroscopic evaluation of endometrium in women with postmenopausal bleeding” Margit Dueholm, Ina Marie D. Hjorth, Peter Secher, Annemette Jørgensen, Gitte Ørtoft PII:
S1553-4650(15)00517-8
DOI:
10.1016/j.jmig.2015.06.018
Reference:
JMIG 2607
To appear in:
The Journal of Minimally Invasive Gynecology
Received Date: 22 April 2015 Revised Date:
13 June 2015
Accepted Date: 17 June 2015
Please cite this article as: Dueholm M, Hjorth IMD, Secher P, Jørgensen A, Ørtoft G, “Structured hysteroscopic evaluation of endometrium in women with postmenopausal bleeding”, The Journal of Minimally Invasive Gynecology (2015), doi: 10.1016/j.jmig.2015.06.018. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Dueholm 1
Title
2
“Structured hysteroscopic evaluation of endometrium in women with postmenopausal bleeding”
3
Margit Dueholm1; Ina Marie D Hjorth1; Peter Secher2; Annemette Jørgensen2; Gitte Ørtoft1
4
From
5
Department of Obstetrics and Gynecology1
6
Aarhus University Hospital
7
Brendstrupgaardsvej 100
8
8200 Aarhus N, Denmark
9
Department of Obstetrics and Gynecology2
M AN U
SC
RI PT
1
Aalborg University Hospital
11
Reberbansgade 1
12
9100 Aalborg
13
Corresponding author:
14
Margit Dueholm, Department of Obstetrics and Gynecology1 ,Aarhus University Hospital
15
Brendstrupgaardsvej 100, 8200 Aarhus N, Denmark
16
Mail:[email protected]
17
Tlf: +45 24607091
18
Funding This study was supported with grants from the Danish Cancer Society.. No authors report
19
conflict of interest.
EP
AC C
20
TE D
10
21
Précis: Structured hysteroscopic evaluation of visual endometrial pattern terms increase diagnostic
22
accuracy of hysteroscopy for diagnosis of malignancy in women with postmenopausal bleeding.
ACCEPTED MANUSCRIPT Dueholm 2
Abstract
2
Study Objective a) To evaluate visual pattern parameters obtained with hysteroscopy for the
3
prediction of endometrial cancer b) To evaluate observer variation of these parameters c) To present
4
a scoring system of the parameters for the prediction of malignancy and compare to subjective
5
evaluation.
6
Design Prospective controlled study
7
Design Classification: Canadian task force II-1
8
Setting University Clinic
9
Patients 149 consecutive women with postmenopausal bleeding and an endometrium thickness ≥ 5
10
mm. 61 (41%) had endometrial cancer. 46 of 149 women were referred based on suspect malignancy.
11
Intervention_Endometrial pattern characteristics for endometrial cancer were evaluated in
12
hysteroscopic video clips. The reference standard was pathologic evaluation of resectoscopic samples
13
or hysterectomy. Using multivariate logistic regression, image parameters were correlated with the
14
presence of endometrial cancer. A scoring system of visual parameters for prediction of malignancy
15
was compared to subjective evaluation of malignancy.
16
Measurements and main results A score for lesion surface, necrosis and vessels had an (AUC) of 0.89,
17
0.89, and 0.87, respectively. A Hysteroscopic Cancer (HYCA) scoring system based on unsmooth lesion
18
surface, papillary projections, surface necrosis, “candy floss” necrosis, white hyperintense spots,
19
irregular branching vessels, and irregular distribution of irregular vessels was able to predict cancer
20
(AUC 0.964) with higher accuracy than subjective evaluation AUC 0.859 (p < 0.01). At a score value ≥
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 3
3, sensitivity was 89%, specificity 92% with moderate agreement between observers (kappa: 0.56
2
(0.42–0.71)).
3
Conclusions Systematic pattern evaluation of optimal parameters by a HYCA-scoring system based on
4
systematically defined terms may increase accuracy in the diagnosis of endometrial cancer and should
5
be further elaborated and external validity tested in unselected women with postmenopausal
6
bleeding.
SC
RI PT
1
7
EP
TE D
M AN U
Keywords: Hysteroscopy; Endometrial neoplasms; postmenopause; Uterine Hemorrhage
AC C
8
ACCEPTED MANUSCRIPT Dueholm 4
Introduction
2
Delay in the investigation of cancer patients, including those with endometrial cancer [1] has in
3
several countries led to introduction of fast-track schedules for women with postmenopausal
4
bleeding [2, 3]. These fast track regimens aim to minimize the time between first complaint of
5
bleeding to diagnosis and treatment of endometrial cancer. Patients at very low risk of endometrial
6
cancer may be identified by transvaginal ultrasonography (TVS) [4, 5] which is the most cost-effective
7
first test in the diagnostic work-up [6]. Recommended diagnostic strategy in women with an
8
endometrial thickness ≥ 4-5 mm are either endometrial samples[7], hysteroscopy in the presence of
9
localized pathology on contrast infusion sonography (by example saline infusion sonography or gel
M AN U
SC
RI PT
1
infusion sonography) [7-9] or office hysteroscopy for all women[8].
11
The visual assessment of the endometrial cavity, coupled with “oriented” or “targeted” tissue
12
collection, improves accuracy in the diagnosis of polyps, focal hyperplasia or carcinomas compared to
13
blind sampling procedures [10, 11].
14
To meet the demand for fast track regimens postmenopausal women with bleeding are increasingly
15
offered immediate see-and-treat office hysteroscopy performed with small dimensional mini-
16
hysteroscopes without anaesthesia[12-14]. Visually directed targeted biopsies of malignancy suspect
17
patterns are undertaken by small caliber (3 to 5 Fr) hysteroscopic forceps. Endometrial polyps,
18
hyperplasia and malignancy are often observed in different areas in the same endometrial cavity, and
19
the accuracy of visually directed biopsies is dependent on an observer related visual diagnosis of the
20
most suspect pattern. Endometrial polyps and small focal pathology can be removed by see-and-treat
21
hysteroscopy. However, to remove larger focal pathology as recommended[15], an additional
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 5
operative hysteroscopy may be needed in half of these postmenopausal patients [12]. Additional
2
operative hysteroscopy is costly [16] and stressful for these often elderly women. An accurate
3
subjective pattern evaluation of focal endometrial pathology during see-and-treat hysteroscopy may
4
reduce the need for additional operative hysteroscopy.
5
Several studies have evaluated the hysteroscopic view in postmenopausal women for diagnosis of
6
pathology [17-23]. Sugimoto [24] defined three hysteroscopic features of neoplastic morphology in a
7
study of 60 endometrial cancers. The features of malignancy were a nodular, polypoid, and
8
papillomatous type categorized irrespective of vascular atypia and tissue necrosis. However, mixtures
9
of these three distinctive features are often observed in endometrial cancer [25]. Based on prior
10
studies [26, 27], a classification system for hysteroscopic abnormalities defined malignancy by the
11
terms “whitish exophytic polypoid proliferation with atypical vascularization and hemorrhagic areas,
12
or papillary with a brain-like aspect and necrotic areas” [17]. Other unstructured terms with slight
13
modifications have been used [18-23] . A single case control study has evaluated the hysteroscopic
14
pattern in endometrial cancer[27]. However, there is no general recommended structured evaluation
15
and report of endometrial pattern obtained by hysteroscopy. We hypothesize, that a structured
16
evaluation of endometrial pattern may have higher accuracy, than subjective evaluation of pattern.
17
The aim of this study was to evaluate different visual parameters for the diagnosis of malignancy to
18
establish the parameters with the highest prediction of malignancy. Moreover, the aim was to clarify
19
the reproducibility of these terms in patients with postmenopausal bleeding and endometrial
20
thickness ≥5 mm, and to present an initial systematic score system for identification of endometrial
21
cancer.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 6
Methods
2
Hysteroscopy was performed in consecutive women with postmenopausal bleeding and an
3
endometrial thickness of ≥ 5 mm referred to Aarhus University Hospital, Denmark, from October 2010
4
to February 2012.
5
Hysteroscopy
6
A Storz resectoscope (26 Fr12 degree oblique optic) (Storz Endoskopi Danmark A/S, 2840 Holte ) was
7
used as an outpatient procedure under local or brief general anesthesia .
8
Two gynecologists with several years of experience in hysteroscopy performed hysteroscopy (live-
9
hysteroscopy) and obtained all video clips in the 164 women. At each examination, hysteroscopic
10
video-clips were obtained after a sufficient distension of the endometrial cavity was established by
11
sorbitol as distension media. The recording was started at the cervical canal at the entrance into the
12
endometrial cavity and systematically visualized the anterior wall, the fundus, the ostia, sidewalls, and
13
posterior wall. The panorama ended just before leaving the external os of the cervix and covered the
14
whole uterine cavity including the cervical canal. Often 2–3 clips were obtained in each patient. The
15
total recording time was 2–4 min.
16
Hysteroscopic video clips were evaluated independently by two gynecologists (PS) and (AMJ) from
17
another university hospital blinded to prior evaluations and the patients’ identities. Both had more
18
than 10 years of special experience in hysteroscopy.
19
Inclusion of patients
20
These women were included in a larger study where ultrasound was followed by hysteroscopy (live-
21
hysteroscopy) for the diagnosis of endometrial pathology. Diagnostic accuracy of two- and three-
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 7
dimentional ultrasound was evaluated [28, 29] and compared to accuracy of endometrial sample and
2
live hysteroscopy [30].
3
Patient characteristics and findings on transvaginal ultrasound (TVS) and gel infusion sonography (GIS)
4
were known to the operator of live-hysteroscopy. Therefore we used video evaluations in a blinded
5
set-up to eliminate observer bias induced by knowledge of prior imaging.
6
The exclusion criteria and selection of women used in the present study have in details been
7
described elsewhere [28, 29].
8
Consecutive women referred with postmenopausal bleeding (432 women) and an endometrial
9
thickness ≥ 5 mm (264 women) were included. 54 of 264 eligible women were not included: One had
10
an infection; 28 due to logistic problems or women did not want to participate; 10 patients had
11
hysteroscopy performed by other physician without systematic videos and in 15 patients, hysteroscopic
12
videos were not correctly stored and could not be evaluated.
13
Hysteroscopy was not performed in another 61 of these 264 women, leaving 149 patients for analysis. These
14
61 women not eligible for hysteroscopy had a prior diagnosis of malignancy: 9 women had hysteroscopic
15
removal of endometrial tumor before referral, 36 had other pelvic cancers, disseminated endometrial cancers,
16
serious comorbidity or advanced age and 16 women with endometrial cancer had a prior diagnosis of special
17
tumors ( 5 serous, 9 carcinosarcomas, 2 clear cell). In these 16 women hysteroscopy was not performed due to
18
a theoretic risk of spread of these rare tumors during hysteroscopy.
19
The study group of 149 women was either unreferred from the local area without prior imaging or
20
endometrial samples or referred from other hospitals or gynecologist often based on a diagnosis of
21
suspect malignancy (46 patients). Therefore this study included a higher numbers of patients with
22
endometrial cancer, than is normally seen in an unselected population of women with
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 8
postmenopausal bleeding. In 122 of 149 patients video-clips at TVS and gel infusion sonography(GIS)
2
were stored, and these patients were selected for evaluation of observer variation of endometrial
3
pattern parameters obtained by hysteroscopy. Thereby observer variation by hysteroscopy could be
4
compared to ultrasound, which will be presented elsewhere.
5
Visual Pattern analysis
6
The qualities of videos were evaluated (excellent, good, intermediate, poor, very poor) and
7
visualization of all the endometrium was noted. The endometrial pattern was systematically
8
evaluated according to a standard form. The different hysteroscopic patterns of cancer are outlined in
9
Figure 1.
M AN U
SC
RI PT
1
Observers were asked to note the presence or absence of the following pattern features:
11
Increased endometrial thickness, local lesions (≤ 25% of surface), diffuse lesions, (>25% of
12
endometrial surface involved).
13
The endometrial surface contour of the lesion (Figure 1 a):
14
Uneven surface texture pattern, polypoid surface, irregular surface and papillary projections.
15
Examples of an irregular surface are seen in Figure 1 (a1, a2, a3). Papillary projections (surface
16
papillarity with a fibrovascular stromal core) are displayed in Figure 1: a4, a5, a6.
17
Whitish (necrotic) areas (Figure 1 b):
18
Surface necrosis, candy floss endometrium, hyper intense white calcified spots in endometrium.
19
Surface necrosis are seen as whitish greyish areas without vessels on the surface of a lesion (b1), and
20
with characteristic features of white friable tissue (b2), shedding of clouds of whitish friable candy
21
floss-like tissue (b3, b4). Moreover hyper intense spots of whitish tissue may also be seen (b5, b6).
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 9
The vessel pattern (Figure 1 c):
2
An overall impression of irregular vessel pattern, irregular caliber of vessels, irregular branching of
3
vessels, and irregular distribution of vessels.
4
Characteristic features of irregular vessels are abnormal tortuous s-formed vessels, vessel loops and
5
vessels ending in small punctate “vessels sprouts” (c1, c2). Irregular caliber of vessels is characterized
6
by varying diameter of vessels with narrow parts of vessels (c3, c4). Irregular branching vessels have a
7
lack of hierarchy in branching from large vessels to often several very small vessels (c3, c4). Irregular
8
distribution of vessels (c5, c6) is characterized by areas with several irregular vessels and other areas
9
without vessels.
M AN U
SC
RI PT
1
Endometrial glands:
11
Dilated glands and glands with irregular openings.
12
Subjective diagnosis
13
At the end of systematic evaluation of the different parameters presence or absence of the following
14
subjective diagnosis was recorded on the standard form:
15
Hyperplasia, malignancy, benign polyps, myomas, proliferative endometrium, atrophy, and other
16
specified abnormalities.
17
The age of the patients was entered on the form. The subjective diagnosis of the endometrium was
18
performed according to specified definitions given on the form [17, 18, 30, 31].
19
Reference standard at pathology
20
After the video clips were recorded, all focal changes were removed by resectoscope. Three to five
21
resectoscopic biopsies were taken from diffuse areas with the most pronounced changes, or
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 10
resectoscopic biopsy was taken from the anterior and posterior wall of the normal uterine cavity.
2
Curettage was performed at the end of procedure to ascertain pathologic diagnosis of all endometrial
3
pathology.
4
The reference standard was pathology of the uterus if hysterectomy was performed or microscopic
5
evaluation of resectoscopic biopsy and curettage specimens [28, 30]. All specimens were evaluated by
6
one of two onco-pathologists.
7
Statistics and analysis
8
Data were analyzed using STATA (Statistic Data Analysis, STATA Corp, TX, USA). The diagnostic
9
performance with regard to discrimination between a benign and a malignant endometrium by
10
different hysteroscopic parameters was evaluated in terms of the areas under the curve (AUC),
11
sensitivity, specificity, positive likelihood ratios (LR+), and negative likelihood ratios (LR-). The
12
sensitivity, the specificity, and the LR+ and LR- of the most optimal cut-points of Receiver-operator
13
characteristic (ROC) curves were calculated. Optimal cut-points were calculated using a 45-degree
14
tangent line intersection or the smallest sum of residual sensitivity and specificity.
15
Then we evaluated which of the evaluated pattern features contributed to prediction of cancer on
16
multivariate regression analysis. If more than one variable for each of the three issues of parameters
17
(surface, necrosis, vessels) predicted cancer, a score variable was made with a score for each of the
18
relevant parameters present. Stepwise multivariate logistic regression was performed to build the
19
most optimal model that could be used to predict malignancy in hysteroscopy video clips. A maximum
20
of four fitting variables was allowed in the model.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 11
Based on the most optimal model calculated by logistic regression, a model based on scores (HYCA
2
score) was developed and compared to the most optimal model obtained on logistic regression
3
analysis and subjective evaluation; χ2 tests were used for discrete data. P < 0.05 was considered
4
statistically significant.
5
The inter-observer variation in image variable assessment was described by kappa analysis and the
6
standards for strength of agreement for the kappa coefficient were interpreted as follows: ≤ 0 = poor,
7
.01–.20 = slight, .21–.40 = fair, .41–.60 = moderate, .61–.80 = substantial, and .81–1 = almost perfect.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 12
Results
2
Pathologic findings in the 149 patients were 61 patients with endometrial cancer (56 endometroid
3
cancers, 1 mucinous, 2 serous, 1 clear cell, and 1 carcinosarcoma), 13 with complex or simple
4
endometrial hyperplasia, 6 with hyperplasia with atypia, and 44 with benign endometrial polyps. The
5
quality of the 149 hysteroscopic video clips was ranked as good or very good in 82%, intermediate in
6
11%, and of lower quality in 7%.
7
Table 1 lists all parameters evaluated for endometrial cancer on hysteroscopic video clips, and the
8
ability of each parameter to predict endometrial cancer. The overall subjective diagnosis had an AUC
9
of 0.859, and the AUC of subjective evaluation was comparable with the AUC of most optimal single
10
parameters (papillary projections, surface of uneven texture and irregular surface). All these
11
parameters had AUC values over 0.85.
12
Evaluation of surface parameters revealed that an irregular outline and papillary projections predicted
13
cancer, with a higher AUC when they were combined. Evaluation of necrosis parameters (surface
14
necrosis, candy floss, and hyper intense white spots) predicted cancer with a higher AUC for a
15
combination of all three parameters. For vessel parameters, a combination of irregular vessels with
16
irregular branching and irregular vessels with irregular distribution resulted in a higher accuracy for
17
the prediction of cancer.
18
Multivariate analysis including scores for surface, necrosis and vessels (Model M1, M2):
19
A model (M1) including surface score (irregular outline and papillary projections), necrosis score
20
(surface necrosis, candy floss, hyper intense white spots), and vessels score (irregular branching and
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 13
distribution) increased the diagnostic accuracy (AUC 0.966), indicating that most cancers could be
2
identified by combined evaluation of the pattern features: surface, necrosis, and vessel morphology.
3
In model M2, BMI was added to the graded score model. This increased AUC to 0.973. BMI had an
4
AUC (95% CI) of 0.651(0.56–0.75).
5
Multivariate analysis including single feature parameters (M3, M4, M5, M6):
6
When we included only a single ungraded parameter from each issue (surface, necrosis, vessels), the
7
most optimal model on multivariate logistic regression (M5) (Table 1) included the following features:
8
papillary projections, irregular distribution of vessels, surface necrosis, and BMI. The most optimal
9
model when only a single parameter was included (M5) had AUC (0.967) which had comparable
10
values with M1 (AUC:0.966). When AUC values were compared, all the regression models (M1, M2,
11
M3, M4, M5, M6) were more efficient than subjective evaluation.
12
HYCA-score:
13
Based on the scores obtained by hysteroscopy for each of the three graded parameters for surface,
14
necrosis, and vessels, we elaborated a hysteroscopic cancer score (HYCA score). In table 2 the
15
parameters used for assessment of endometrial cancer by the HYCA score system and most optimal
16
cut-points are displayed. The AUC of the score system was 0.964. At a cut-point of a score 3 or higher,
17
89% of the patients were correctly classified and 92% of cancers were identified. An example of HYCA-
18
score is given in the legend. The observer variation of the parameters is presented in Table 4. The
19
HYCA score system showed substantial agreement between observers. For individual parameters only
20
papillary projections had substantial observer agreement.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 14
Discussion
2
The systematic score system predicted cancers more efficiently than subjective evaluation. This
3
underlines the increased clinical value of hysteroscopy by systematic evaluation of endometrium and
4
elaboration of common terms to describe pattern and definitions of abnormal findings. The score
5
model had accuracy comparable with the most optimal model obtained by logistic regression. It
6
consisted of a count of numeric scores based on, the presence of lesions with irregular surface,
7
papillary projections, presence of surface necrosis of endometrium, candy floss shedding of necrosis,
8
hyper-intense white dots, irregular vessel pattern with irregular branching of vessels, irregular pattern
9
with irregular distribution of vessels in lesions. Using a score of 3 as a cut-point, nine of ten patients in
M AN U
SC
RI PT
1
the study were correctly classified with benign or malignant endometrial pathology.
11
In this population, some patients were referred based on a diagnosis of suspect malignancy, and we
12
had a high numbers of abnormal findings and endometrial cancer which is needed to evaluate the
13
accuracy of hysteroscopy to differentiate between abnormalities. However, the initial score system
14
developed in this population, has to be further tested and elaborated in a large population of
15
unselected women with postmenopausal bleeding and increased endometrial thickness before
16
recommending for use in general practice and should especially not be used in premenopausal
17
women before additional validation. Malignancy has been diagnosed by hysteroscopy with high
18
accuracy greater than in this study [18, 22, 32]. The lower accuracy in this study may be explained by
19
the high prevalence of pathology and other cut-points for malignancy but also by the use of video
20
clips. The accuracy of video clips was slightly lower than live hysteroscopy/ultrasound [30] although
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 15
the general quality of clips were good. We used recorded video clips instead of live hysteroscopy,
2
because a safe live hysteroscopy of cancer patients cannot be performed blinded to ultrasound
3
findings of myometrial invasion, which might affect evaluation of hysteroscopic parameters. We
4
wanted an evaluation of the most optimally clear visual markers of cancer, which could be easily
5
identified using panorama hysteroscopy in video clips. Major shortcomings were that the presence of
6
abnormal morphology was evaluated with a subjective cut-off and without any graded score system.
7
We selected parameters with a clear definition used in prior studies, but were unable to define any
8
reliable cut- offs for malignant morphology based on the present literature. Other terms could have
9
been relevant, and a further elaboration of terms may increase accuracy.
M AN U
SC
RI PT
1
To our knowledge, this is the first hysteroscopic scoring system for the diagnosis of endometrial
11
cancer. We evaluated the most commonly used hysteroscopic visual parameters for the diagnosis of
12
malignancy [17, 19-23, 33], and established that the systematic evaluation of three visual parameter
13
components, surface, necrosis, and vessels, were efficient for prediction of cancer. For each of the
14
three factors, addition of more specified parameters increased the accuracy of the score system. The
15
specified parameters lead to a two point score for surface (irregular, papillary projections), a three
16
point score for necrosis (surface necrosis, candy floss, hyper intense white dots), and a two points
17
score for irregular vessels (branching, distribution).
18
systematic evaluation of surface outline, presence of necrosis and vessel pattern, stated in a standard
19
form by a count of scores may standardize terms and reporting, which is essential for quality
20
improvement and learning.
AC C
EP
TE D
10
At panorama hysteroscopy, a standardized
ACCEPTED MANUSCRIPT Dueholm 16
“Irregular surface” is a subjective parameter that has to be used together with other parameters, and
2
is used during panorama hysteroscopy to focus on the areas for detailed inspection. Papillary
3
projection had both low observer variation and high accuracy and was the most reliable parameter,
4
but was not present in all cancers.
5
In some cancers, whitish necrosis seemed to be the most prominent characteristic. Stromal invasion
6
characterized by stromal desmoplasia or stromal necrosis [34-36] may be seen as whitish surface
7
areas, and may give rise to the shedding of clouds of white friable characteristic necroinflammatory
8
debris, that we have called “candy floss”.
9
Some cancers may have very hyperintense calcified white spots, which is a round collections of
10
calcium (psammoma body),that may be associated with the papillary morphology and are thought to
11
arise in areas of necrosis and inflammation[37]. Correlation with endometrial cancer has been
12
proposed in some studies [38, 39], but is questioned or rare in other [37, 40, 41]. These white spots
13
are easily identified and contributed to prediction of malignancy in the HYCA score. Calcified spots are
14
correlated with diseases other than cancer [40], and visual markers of neoplastic tissue may replace
15
this parameter in future studies.
16
Identification of abnormal vessel pattern is important in the visual diagnosis of cancer [42, 43].
17
Narrow band imaging [22] may increase visualization of abnormal microvessels [21, 22]. Moreover,
18
computer technology[44, 45] is an upcoming and promising development in endoscopic imaging for
19
the evaluation of microvessels and surface microstructure. Clear terms with definition of abnormal
20
vessel sprouts, tortuous vessels, vessel loops, and branching with angles over 90 degree, narrowing of
21
vessels, disorganized network, and an overall irregular distribution with areas with dense vessel
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 17
varying with areas without vessels may optimize diagnosis. However, in concordance with another
2
study, a simple increased vascular density[46] has to be combined with other parameters in the
3
diagnosis of cancer.
4
A further elaboration of the systematic pattern parameters and the HYCA-scoring system may be
5
essential to decrease the need for extensive hysteroscopic procedures and for progress of immediate
6
office hysteroscopy in postmenopausal women with bleeding. In see-and treat hysteroscopy with
7
small caliber hysteroscopes the need for removal of residual focal pathology by additional operative
8
hysteroscopy with resectoscope should be evaluated when benign pathology is suggested by
9
standardized visual evaluation of endometrial pattern and biopsies. However, we used resectoscopes
10
to have a valid reference standard, and the hysteroscopic view by small dimensional hysteroscopes
11
seems efficient [14] but has to be evaluated.
12
There was only moderate agreement on a subjective diagnosis of cancer and a high inter-observer
13
variation in the identification of several individual parameters. The agreement may expectedly be
14
even smaller between larger groups of not experienced investigators. Cut-offs for positive and
15
negative findings may be clarified by a clear definition or grading of visual parameters guided by a
16
pictogram and stored representative hysteroscopic images and videos. At present, recognition of
17
endometrial cancer obtained with hysteroscopy is learned either by evaluation of individually stored
18
(not necessarily representative) hysteroscopic videos, or most commonly, by performance of high
19
volumes of hysteroscopies and memorizing the pattern. In most studies of unselected hysteroscopies,
20
there are 1–6 cancers in 100 hysteroscopies [19, 47]. Thus, large volumes of hysteroscopies are
21
required to see the diversity of cancer pattern, and special attention may be needed to remember
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 18
different patterns in order to achieve competence for recognition of cancer patterns. The HYCA-
2
scoring is a systematic way to learn and classify the pattern. Consistent scoring using parameters
3
before subjective evaluation will most likely improve the raters’ accuracy. In the future video clips
4
obtained by hysteroscopy may be stored in databases and computers may perform the pattern
5
analysis [44]. The HYCA- scoring can be the first step in the evolution of pattern analysis.
6
Systematic classification systems have also been effective in endoscopic diagnosis and learning in
7
colorectal cancer [48-50]. Moreover, the accuracy of ultrasound with systematic evaluation of pre-
8
defined strictly IETA terms [28, 51] was only slightly lower than the accuracy of the HYCA-score.
9
Hysteroscopy would be expected to yield a higher diagnostic accuracy than ultrasound since
10
hysteroscopy has a higher resolution regarding evaluation of the endometrium. This experience
11
underlines the importance of strategies to improve the competence of visual diagnosis, and highlights
12
the importance of the elaboration of defined terms for visual parameters obtained with hysteroscopy.
13
Conclusions A HYCA-score system based on surface outline, necrosis, and vessel pattern predicted
14
cancer with higher accuracy than subjective evaluation. Observer agreement was only moderate for
15
most terms, and may be improved by clear definitions. The score system has to be evaluated and
16
further elaborated in a large population of unselected women with postmenopausal bleeding and
17
increased endometrial thickness before recommending for use. Further elaboration of this initial
18
HYCA score system by identification and definition of specific terms regarding the hysteroscopic
19
pattern to predict malignancy should be performed.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 19
Acknowledgements The authors gratefully acknowledge Charlotte Møller for collecting hysteroscopic
2
videos and the help of Elisabeth Melin and the staff at outpatient surgery unit.
3
Details of ethics approval The study was approved by The Ethics Committee of the Central Denmark
4
Region 9th of August 2010 (M-20100139). As hysteroscopy was a standard procedure in the
5
department, the committee found that no informed consent for this procedure was needed.
RI PT
1
SC
6 7 8
AC C
EP
TE D
M AN U
9
ACCEPTED MANUSCRIPT Dueholm 20
Legends
2
Figure 1: Figure 1a Surface: Surface irregularity is displayed in a1,a2,a3. Surface papillarity with a
3
fibrovascular stromal cores are called papillary projections (a4,a5,a6). These papillary projections may
4
be long, slender, villous papillae with thin fibrovascular cores (villo-glandular papillary projections) or
5
have a form of a bud-like to filiform appearance (non-villous papillae projections). A central vessel in
6
ciliated papillary projections may end in small vascular buds or sprouts (a6), which are common in
7
malignant tumors.
8
Figure 1b Necrosis: Surface necrosis (b1) are seen as whitish greyish areas without vessels on the
9
surface of a lesion. These areas may be composed of white friable tissue (b2), shedding of clouds of
10
whitish friable candy floss-like tissue (b3,b4). Hyper intense spots of whitish tissue may also be seen
11
(b5, b6). They may represent calcified material (psammoma bodies).
12
Figure 1 c. An irregular vascular pattern may represent slight irregularities to a totally chaotic vessel
13
pattern. Irregular vessels are (c1, c2) abnormal tortuous s-formed vessels, vessel loops and vessels
14
ending in small punctuates “vessels sprouts”. Irregular caliber of vessels is characterized by varying
15
diameter of vessels (c3, c4). Irregular branching vessels have a lack of hierarchy in branching (c3, c4)
16
with varying branching angles often with more than 90-degree branching angles. Irregular
17
distribution of vessels (c5, c6) are defined by presence of areas in which very large vessels are
18
present together with small vessels, and areas with increased density of often small micro vessels and
19
other areas with larger or no vessels. Small areas with hemorrhage may be seen.
20 21 22
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 21
References 1.
2.
6 7
8 3.
8 9
Olesen F, Hansen RP, Vedsted P. Delay in diagnosis: the experience in Denmark. Br.J.Cancer. 2009; 101:5-
Referral guidelines for suspected cancer NICE.Published May 2005. Accessed 1-3-2015; Available from: URL:https://www.nice.org.uk/guidance/cg27
4.
SC
5
Br.J.Cancer. 2011; 104:1836-9
RI PT
4
Johnson N, Miles T, Bailey D, et al. Delays in treating endometrial cancer in the South West of England.
Timmermans A, Opmeer BC, Khan KS, et al. Endometrial thickness measurement for detecting
M AN U
1 2 3
10
endometrial cancer in women with postmenopausal bleeding: a systematic review and meta-analysis.
11
Obstet.Gynecol. 2010; 116:160-7 5.
6.
15 16
endometrial cancer: cost-effectiveness of initial diagnostic strategies. BJOG. 2006; 113:502-10 7.
17 18
Munro MG. Investigation of women with postmenopausal uterine bleeding: clinical practice recommendations. Perm.J. 2014; 18:55-70
8.
19 20
Clark TJ, Barton PM, Coomarasamy A, Gupta JK, Khan KS. Investigating postmenopausal bleeding for
EP
14
cancer and other endometrial abnormalities. JAMA. 1998; 280:1510-7
TE D
13
Smith-Bindman R, Kerlikowske K, Feldstein VA, et al. Endovaginal ultrasound to exclude endometrial
Dreisler E, Poulsen LG, Antonsen SL, et al. EMAS clinical guide: Assessment of the endometrium in peri and postmenopausal women. Maturitas. 2013; 75:181-90
9.
AC C
12
Goldstein RB, Bree RL, Benson CB, et al. Evaluation of the woman with postmenopausal bleeding: Society
21
of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med. 2001;
22
20:1025-36
23 24
10.
Bakour SH, Jones SE, O'Donovan P. Ambulatory hysteroscopy: evidence-based guide to diagnosis and therapy. Best.Pract.Res.Clin.Obstet.Gynaecol. 2006; 20:953-75
ACCEPTED MANUSCRIPT Dueholm 22
11.
12.
4 5
successful? Eur.J Obstet.Gynecol Reprod.Biol. 2014; 178:70-3 13.
6 7
14.
Di Spiezio SA, Bettocchi S, Spinelli M, et al. Review of new office-based hysteroscopic procedures 20032009. J Minim.Invasive.Gynecol. 2010; 17:436-48
15.
10 11
Bettocchi S, Ceci O, Nappi L, et al. Operative office hysteroscopy without anesthesia: analysis of 4863 cases performed with mechanical instruments. J Am.Assoc.Gynecol Laparosc. 2004; 11:59-61
8 9
Gambadauro P, Martinez-Maestre MA, Torrejon R. When is see-and-treat hysteroscopic polypectomy
SC
3
assessing the diagnostic accuracy of office hysteroscopy. Fertil.Steril. 2002; 78:628-31
RI PT
2
Ceci O, Bettocchi S, Pellegrino A, et al. Comparison of hysteroscopic and hysterectomy findings for
M AN U
1
AnonymousAAGL practice report: practice guidelines for the diagnosis and management of endometrial polyps. J.Minim.Invasive.Gynecol. 2012; 19:3-10
16.
Saridogan E, Tilden D, Sykes D, Davis N, Subramanian D. Cost-analysis comparison of outpatient see-andtreat hysteroscopy service with other hysteroscopy service models. J Minim.Invasive.Gynecol. 2010;
13
17:518-25
14
17.
15
Dotto JE, Lema B, Dotto JE, Jr., Hamou J. Classification of microhysteroscopic images and their correlation with histologic diagnoses. J Am.Assoc.Gynecol Laparosc. 2003; 10:233-46
18.
Lasmar RB, Barrozo PR, de Oliveira MA, Coutinho ES, Dias R. Validation of hysteroscopic view in cases of
EP
16
TE D
12
endometrial hyperplasia and cancer in patients with abnormal uterine bleeding. J
18
Minim.Invasive.Gynecol. 2006; 13:409-12
19
19.
20 21 22
AC C
17
Garuti G, Sambruni I, Colonnelli M, Luerti M. Accuracy of hysteroscopy in predicting histopathology of endometrium in 1500 women. J Am.Assoc.Gynecol Laparosc. 2001; 8:207-13
20.
Surico D, Vigone A, Bonvini D, et al. Narrow-band imaging in diagnosis of endometrial cancer and hyperplasia: a new option? J Minim.Invasive.Gynecol. 2010; 17:620-5
ACCEPTED MANUSCRIPT Dueholm 23
21.
2 3
Cicinelli E, Tinelli R, Colafiglio G, et al. Reliability of narrow-band imaging (NBI) hysteroscopy: a comparative study. Fertil.Steril. 2010; 94:2303-7
22.
Tinelli R, Surico D, Leo L, et al. Accuracy and efficacy of narrow-band imaging versus white light
RI PT
1
4
hysteroscopy for the diagnosis of endometrial cancer and hyperplasia: a multicenter controlled study.
5
Menopause. 2011; 18:1026-9 23.
9
the Women's Clinic of Kyoto University Hospital. Am.J Obstet.Gynecol. 1975; 121:105-13 25.
11 12
26.
27.
Mencaglia L, Valle RF, Perino A, Gilardi G. Endometrial carcinoma and its precursors: early detection and treatment. Int.J Gynaecol.Obstet. 1990; 31:107-16
28.
17 18
Mencaglia L, Perino A, Hamou J. Hysteroscopy in perimenopausal and postmenopausal women with abnormal uterine bleeding. J Reprod.Med. 1987; 32:577-82
15 16
Garuti G, De GO, Sambruni I, Cellani F, Luerti M. Prognostic significance of hysteroscopic imaging in endometrioid endometrial adenocarcinoma. Gynecol Oncol. 2001; 81:408-13
13 14
Sugimoto O. Hysteroscopic diagnosis of endometrial carcinoma. A report of fifty-three cases examined at
TE D
10
SC
24.
Dueholm M, Møller C, Rydbjerg S, Hansen E. An Ultrasound Algorithm for the Identification of
EP
8
hyperplasia: a retrospective study of 323 patients. J.Minim.Invasive.Gynecol. 2005; 12:247-53
M AN U
7
Garuti G, Cellani F, Garzia D, Colonnelli M, Luerti M. Accuracy of hysteroscopic diagnosis of endometrial
Endometrial Cancer. Ultrasound Obstet.Gynecol. 2014; 43:557-68 29.
Dueholm M, Holm JW, Rydbjerg S, Hansen ES, Ortoft G. Two-dimensional and three-dimensional
AC C
6
19
transvaginal ultrasound with power Doppler angiography and gel infusion sonography for diagnosis of
20
endometrial malignancy. Ultrasound Obstet.Gynecol. 2015; 45:734-42
21
30.
Dueholm M, Marinovskij E, Hansen ES, Moller C, Ortoft G. Diagnostic methods for fast-track identification
22
of endometrial cancer in women with postmenopausal bleeding and endometrial thickness greater than
23
5 mm. Menopause. 2015; 22:616-26
ACCEPTED MANUSCRIPT Dueholm 24
31.
32.
4 5
hyperplasia: a systematic quantitative review. JAMA. 2002; 288:1610-21 33.
6 7
Clark TJ, Voit D, Gupta JK, et al. Accuracy of hysteroscopy in the diagnosis of endometrial cancer and
Zola FE, Nogueira AA, de Andrade JM, Candido Dos Reis FJ. Hysteroscopic appearance of malignant and benign endometrial lesions: a case-control study. Arch.Gynecol Obstet. 2007; 275:49-52
34.
8
Silverberg SG. Problems in the differential diagnosis of endometrial hyperplasia and carcinoma.
SC
3
pathologic findings on hysterectomy specimens. J.Minim.Invasive.Gynecol. 2006; 13:325-30
RI PT
2
Garuti G, Mirra M, Luerti M. Hysteroscopic view in atypical endometrial hyperplasias: A correlation with
Mod.Pathol. 2000; 13:309-27
M AN U
1
9
35.
Silverberg SG. Hyperplasia and carcinoma of the endometrium. Semin.Diagn.Pathol. 1988; 5:135-53
10
36.
Sherman ME, Silverberg SG. Advances in endometrial pathology. Clin.Lab Med. 1995; 15:517-43
11
37.
Parkash V, Carcangiu ML. Endometrioid endometrial adenocarcinoma with psammoma bodies. Am.J Surg
38.
14 15
1991; 35:81-8 39.
16 17
Kern SB. Prevalence of psammoma bodies in Papanicolaou-stained cervicovaginal smears. Acta Cytol.
Clement PB, Young RH. Non-endometrioid carcinomas of the uterine corpus: a review of their pathology with emphasis on recent advances and problematic aspects. Adv.Anat.Pathol. 2004; 11:117-42
40.
EP
13
Pathol. 1997; 21:399-406
TE D
12
Truskinovsky AM, Gerscovich EO, Duffield CR, Vogt PJ. Endometrial microcalcifications detected by ultrasonography: clinical associations, histopathology, and potential etiology. Int.J Gynecol Pathol. 2008;
19
27:61-7
20
41.
21 22 23
AC C
18
Fausett MB, Zahn CM, Kendall BS, Barth WH, Jr. The significance of psammoma bodies that are found incidentally during endometrial biopsy. Am.J Obstet.Gynecol. 2002; 186:180-3
42.
Jichlinski P, Lovisa B. High magnification cystoscopy in the primary diagnosis of bladder tumors. Curr.Opin.Urol. 2011; 21:398-402
ACCEPTED MANUSCRIPT Dueholm 25
43.
3
clinical trials. Nat.Rev.Clin.Oncol. 2014; 11:566-84 44.
4 5
Biomed.Health Inform. 2014; 19:1129-36 45.
6 7
Neofytou M, Tanos V, Constantinou I, et al. Computer Aided Diagnosis in Hysteroscopic Imaging. IEEE J
RI PT
2
Emblem KE, Farrar CT, Gerstner ER, et al. Vessel calibre-a potential MRI biomarker of tumour response in
Vlachokosta AA, Asvestas PA, Gkrozou F, et al. Classification of hysteroscopical images using texture and vessel descriptors. Med.Biol.Eng Comput. 2013; 51:859-67
46.
Dias DS, Bueloni-Dias FN, Dias R, et al. Usefulness of clinical, ultrasonographic, hysteroscopic, and
SC
1
immunohistochemical parameters in differentiating endometrial polyps from endometrial cancer. J
9
Minim.Invasive.Gynecol. 2014; 21:296-302
10
47.
M AN U
8
de Wit AC, Vleugels MP, de Kruif JH. Diagnostic hysteroscopy: a valuable diagnostic tool in the diagnosis
11
of structural intra-cavital pathology and endometrial hyperplasia or carcinoma?. Six years of experience
12
with non-clinical diagnostic hysteroscopy. Eur.J Obstet.Gynecol Reprod.Biol. 2003; 110:79-82 48.
14 15
Uraoka T, Saito Y, Ikematsu H, Yamamoto K, Sano Y. Sano's capillary pattern classification for narrow-
TE D
13
band imaging of early colorectal lesions. Dig.Endosc. 2011; 23:112-5 49.
Sakamoto T, Saito Y, Nakajima T, Matsuda T. Comparison of magnifying chromoendoscopy and narrowband imaging in estimation of early colorectal cancer invasion depth: a pilot study. Dig.Endosc. 2011;
17
23:118-23 50.
19 20
Dai J, Shen YF, Sano Y, et al. Evaluation of narrow-band imaging in the diagnosis of colorectal lesions: is a
AC C
18
EP
16
learning curve involved? Dig.Endosc. 2013; 25:180-8 51.
Leone FP, Timmerman D, Bourne T, et al. Terms, definitions and measurements to describe the
21
sonographic features of the endometrium and intrauterine lesions: a consensus opinion from the
22
International Endometrial Tumor Analysis (IETA) group. Ultrasound Obstet.Gynecol. 2010; 35:103-12
23
ACCEPTED MANUSCRIPT Dueholm 26
1
AC C
EP
TE D
M AN U
SC
RI PT
Table 1. Diagnostic efficiency of parameters at hysteroscopy for prediction of endometrial cancer AUC (95% CI) Correctly Sens Spec N=149 classified % % % at (cut-point) Overall subjective impression cancer 61 .859 (.80- .92) 82.0 90.0 Parameter present: N Thickened endometrium 48 .630 (.55—.71) 65.8 47.5 78.4 Localized lesion 69 .330 (.25—.40) 34.2 26.2 39.8 Diffuse lesion 77 .715 (.64—.79) 70.5 77.1 65.9 Diffuse or localized lesion 133 .535 (.49—.58) 46.3 93.4 13.6 Surface Surface uneven texture 65 .852 (.79—.91) 85.2 85.2 85.2 Surface not polypoid 49 .624 (.55—.70) 65.1 47.5 77.3 Irregular surface (a) 69 .857 (.80—.91) 85.2 88.5 83.0 Papillary projections (b) 62 .856 (.80—.92) 85.9 83.6 87.5 Score surface (a)+ (b) .886 (.83—.94) 83.2 (≥1) 88.5 79.6 87.9 (≥2) 83.6 90.9 Necrosis Surface necrosis (c) 33 .646 (.58—.72) 69.1 39.3 89.8 Candy floss like endometrium(d) 44 .750 (.68—.82) 77.9 59.0 90.9 White dots or areas in endometrium(e) 40 .689 (.62—.76) 72.5 49.2 88.6 Score necrosis(c) + (d) + (e) .887 (.84—.94) 81.9 (≥1) 95.1 72.7 75.8 (≥2) 45.9 96.6 Vessels Irregular vascular pattern overall 70 .838 (.78—.90) 83.2 86.9 80.7 Irregular caliber of vessels 54 .846 (.79—.91) 85.9 77.1 92.1 Irregular branching of vessels(f) 51 .849 (.79—.91) 86.6 75.4 94.3 Irregular distribution of vessels(g) 52 .843 (.78—.90) 85.9 75.4 93.2 Dilated endometrial glands openings 31 .490 (.42—.56) 54.4 19.7 78.4 Score vessels: f + g .867 (.81—.93) 85.9 (≥1) 78.7 90.9 86.6 (≥2) 72.1 96.6 Multivariate logistic regression M1:Score vessel, score necrosis, score .966 (.94—.99) 89.3(.362) 91.8 87.5 surface M2: BMI, score vessel , score necrosis, .973 (.95—.99) 90.6(.381) 90.2 90.9 score surface M3: Age, BMI, Papillary projections .919 (.88—.96) 87.3(.348) 86.9 87.5 M4 : BMI, Papillary projections, Irregular .963 (.94—.99) 90.6(.512) 90.2 90.9 distribution of vessels M5: BMI, Papillary projections, Irregular .967 (.94—.99) 91.3(.468) 90.2 92.1 distribution of vessels, surface necrosis M6: Papillary projections, Irregular .948 (.91—.98) 86.6(.425) 93.4 81.8 distribution of vessels, surface necrosis
LR+
LR-
8.01
.201
2.20 .44 2.26 1.08
.669 1.86 .348 .480
5.77 2.09 5.19 6.69 4.33 9.20
.173 .679 .124 .187 .144 .180
3.84 6.49 4.32 3.49 13.47
.676 .451 .573 .068 .560
4.50 9.69 13.27 11.06 .911 8.66 21.16
.163 .249 .261 .264 1.02 .234 .289
7.34
.094
9.92
.108
6.95 9.92
.150 .108
11.34
.107
5.14
.080
ACCEPTED MANUSCRIPT Dueholm 27
EP
TE D
M AN U
SC
RI PT
Sens=Sensitivity. Spec=Specificity. LR+= Positive likelihood ratio, LR- = Negative likelihood ratio. AUC= Area under the ROC-curve. Model 1(M1) had odds ratio (95%CI): Score-surface 2.67 (1.3-5.4);Score necrosis: 5.51 (1.9-15.7); Score-vessels 5.09(2.4-10.9) and z= -3.8 + (0.98 x score-surface) +(1.71 x score necrosis) + (1.63 x score-vessels). Probability of malignancy (P) can be calculated as follows: P= (ez/(1 +ez)) where e=2.718 (base value of natural logarithms). Roc comparison M3 vs. subjective evaluation p < .05 chi2
AC C
1 2 3 4 5 6 7 8
ACCEPTED MANUSCRIPT Dueholm 28
1 2 3
Table 2: Hysteroscopy Cancer (HYCA) scoring system and diagnostic performance with regard to predicting endometrial cancer: score system (HYCA score) and most optimal cut-point.( N = 149)
Present
Scoring Absent
Value
1 1
0 0
n1 n2
Surface contour
Irregular surface Papillary projections
Vessel pattern
HYCA-score
6 7 8 9 10 11 12 13 14 15
0 0
n6 n7 N
% correctly classified at cut-point
LR+
LR-
76.5 85.2 90.6 89.9 86.6 73.8
2.51 4.57 11.13 23.56 60.59
0 0.082 0.125 0.204 0.315 0.639
TE D
1 1
EP
5
n3 n4 n5
Hysteroscopic cancer score (HYCA score) included: lesion with irregular surface (present = score 1), papillary projections (present = score 1), surface necrosis (present = score 1), candy floss necrosis (present = score 1), white hyper intense spots (present = score 1), irregular branching (present = score 1), irregular distribution of vessels in lesion (present = score 1). Example: A patient with irregular surface (score 1), no papillary projections (score 0), no surface necrosis (score 0), no candy floss (score 0), no white hyper intense spots (score 0), irregular distribution of vessels present (score 1), and irregular vascular branching (score 1) had a total HYCA score(N)=3. Sensitivity(Sens) (88.5), Specificity(Spec) (92.1), and Positive likelihood ratio(LR+) of cancer
AC C
4
0 0 0
Irregular branching Irregular distribution N= ∑ (n1+n2+n3+n4+n5+n6+n7)
Most optimal cut-point for HYCA score (N) Cut-point for Sens Spec score (N) % % ( >= 1 ) 100 60.2 ( >= 2 ) 93.4 79.6 ( >= 3 ) 88.5 92.1 ( >= 4 ) 80.3 96.6 ( >= 5 ) 68.9 98.9 ( >= 6 ) 36.1 100.0 AUC 0.964 (.94-.99)
1 1 1
M AN U
Surface necrosis Candy floss pattern Hyperintense white spots
SC
Evidence of necrosis
RI PT
Pattern Characteristic/Variable
11.13. Score system both with and without inclusion of BMI versus subjective evaluation Roc comparison (p < .01). Score system including BMI had AUC (95% CI): 0.973 (0.95-0.99)
ACCEPTED MANUSCRIPT Dueholm 29
1
Table 4: Agreement between two observers for the presence of different parameters during evaluation of hysteroscopic clips for a diagnosis of malignancy.
EP
5
54/33 57/23 40/21 51/30 35/17 28/19 31/22 59/36 46/25 42/25 42/15 24/41 49/31
77.9 70.5 66.4 82.8 82.0 77.9 74.6 76.2 76.2 76.2 72.3 64.8 80.3
RI PT
% agreement 74.6 73.0 63.9
M AN U
SC
0.53 (0.39 — 0.68) 0.39 (0.25 — 0.52) 0.13 (-0.04 — 0.31) 0.62 (0.49 — 0.76) 0.48 (0.31 — 0.65) 0.30 (0.09 — 0.50) 0.26 (0.07 — 0.45) 0.52 (0.38 — 0.66) 0.44 (0.29 — 0.60) 0.42 (0.25 — 0.59) 0.29 (0.13 — 0.45) 0.12 (-0.05 — 0.29) 0.56 (0.42— 0.71)
N1/n2 41/52 58/73 65/23
N1= numbers with presence of the parameter by observer 1, n2= numbers with presence of the parameter by observer 2.
AC C
2 3 4
Kappa (95% CI) 0.47 (0.31 — 0.62) 0.46 (0.31 — 0.62) 0.31 (0.18 — 0.43)
TE D
Parameter (N = 122) Thickened endometrium Localized lesion Diffuse lesion Areas in localized or diffuse lesions with: Surface uneven texture Irregular surface Surface not polypoid Papillary projections candy floss like endometrium Surface necrosis White dots in endometrium Irregular vascular pattern Irregular caliber of vessels Irregular branching of vessels Irregular distribution of vessels Dilated endometrial glands openings HYCA score at cut-point of ≥3
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE
D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
http://www.AAGL.org/jmig-22-6-JMIG-D-15-00241
ACCEPTED MANUSCRIPT
Précis: Structured hysteroscopic evaluation of visual endometrial pattern terms increase diagnostic
AC C
EP
TE D
M AN U
SC
RI PT
accuracy of hysteroscopy for diagnosis of malignancy in women with postmenopausal bleeding.
S1553-4650(15)00517-8
DOI:
10.1016/j.jmig.2015.06.018
Reference:
JMIG 2607
To appear in:
The Journal of Minimally Invasive Gynecology
Received Date: 22 April 2015 Revised Date:
13 June 2015
Accepted Date: 17 June 2015
Please cite this article as: Dueholm M, Hjorth IMD, Secher P, Jørgensen A, Ørtoft G, “Structured hysteroscopic evaluation of endometrium in women with postmenopausal bleeding”, The Journal of Minimally Invasive Gynecology (2015), doi: 10.1016/j.jmig.2015.06.018. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Dueholm 1
Title
2
“Structured hysteroscopic evaluation of endometrium in women with postmenopausal bleeding”
3
Margit Dueholm1; Ina Marie D Hjorth1; Peter Secher2; Annemette Jørgensen2; Gitte Ørtoft1
4
From
5
Department of Obstetrics and Gynecology1
6
Aarhus University Hospital
7
Brendstrupgaardsvej 100
8
8200 Aarhus N, Denmark
9
Department of Obstetrics and Gynecology2
M AN U
SC
RI PT
1
Aalborg University Hospital
11
Reberbansgade 1
12
9100 Aalborg
13
Corresponding author:
14
Margit Dueholm, Department of Obstetrics and Gynecology1 ,Aarhus University Hospital
15
Brendstrupgaardsvej 100, 8200 Aarhus N, Denmark
16
Mail:[email protected]
17
Tlf: +45 24607091
18
Funding This study was supported with grants from the Danish Cancer Society.. No authors report
19
conflict of interest.
EP
AC C
20
TE D
10
21
Précis: Structured hysteroscopic evaluation of visual endometrial pattern terms increase diagnostic
22
accuracy of hysteroscopy for diagnosis of malignancy in women with postmenopausal bleeding.
ACCEPTED MANUSCRIPT Dueholm 2
Abstract
2
Study Objective a) To evaluate visual pattern parameters obtained with hysteroscopy for the
3
prediction of endometrial cancer b) To evaluate observer variation of these parameters c) To present
4
a scoring system of the parameters for the prediction of malignancy and compare to subjective
5
evaluation.
6
Design Prospective controlled study
7
Design Classification: Canadian task force II-1
8
Setting University Clinic
9
Patients 149 consecutive women with postmenopausal bleeding and an endometrium thickness ≥ 5
10
mm. 61 (41%) had endometrial cancer. 46 of 149 women were referred based on suspect malignancy.
11
Intervention_Endometrial pattern characteristics for endometrial cancer were evaluated in
12
hysteroscopic video clips. The reference standard was pathologic evaluation of resectoscopic samples
13
or hysterectomy. Using multivariate logistic regression, image parameters were correlated with the
14
presence of endometrial cancer. A scoring system of visual parameters for prediction of malignancy
15
was compared to subjective evaluation of malignancy.
16
Measurements and main results A score for lesion surface, necrosis and vessels had an (AUC) of 0.89,
17
0.89, and 0.87, respectively. A Hysteroscopic Cancer (HYCA) scoring system based on unsmooth lesion
18
surface, papillary projections, surface necrosis, “candy floss” necrosis, white hyperintense spots,
19
irregular branching vessels, and irregular distribution of irregular vessels was able to predict cancer
20
(AUC 0.964) with higher accuracy than subjective evaluation AUC 0.859 (p < 0.01). At a score value ≥
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 3
3, sensitivity was 89%, specificity 92% with moderate agreement between observers (kappa: 0.56
2
(0.42–0.71)).
3
Conclusions Systematic pattern evaluation of optimal parameters by a HYCA-scoring system based on
4
systematically defined terms may increase accuracy in the diagnosis of endometrial cancer and should
5
be further elaborated and external validity tested in unselected women with postmenopausal
6
bleeding.
SC
RI PT
1
7
EP
TE D
M AN U
Keywords: Hysteroscopy; Endometrial neoplasms; postmenopause; Uterine Hemorrhage
AC C
8
ACCEPTED MANUSCRIPT Dueholm 4
Introduction
2
Delay in the investigation of cancer patients, including those with endometrial cancer [1] has in
3
several countries led to introduction of fast-track schedules for women with postmenopausal
4
bleeding [2, 3]. These fast track regimens aim to minimize the time between first complaint of
5
bleeding to diagnosis and treatment of endometrial cancer. Patients at very low risk of endometrial
6
cancer may be identified by transvaginal ultrasonography (TVS) [4, 5] which is the most cost-effective
7
first test in the diagnostic work-up [6]. Recommended diagnostic strategy in women with an
8
endometrial thickness ≥ 4-5 mm are either endometrial samples[7], hysteroscopy in the presence of
9
localized pathology on contrast infusion sonography (by example saline infusion sonography or gel
M AN U
SC
RI PT
1
infusion sonography) [7-9] or office hysteroscopy for all women[8].
11
The visual assessment of the endometrial cavity, coupled with “oriented” or “targeted” tissue
12
collection, improves accuracy in the diagnosis of polyps, focal hyperplasia or carcinomas compared to
13
blind sampling procedures [10, 11].
14
To meet the demand for fast track regimens postmenopausal women with bleeding are increasingly
15
offered immediate see-and-treat office hysteroscopy performed with small dimensional mini-
16
hysteroscopes without anaesthesia[12-14]. Visually directed targeted biopsies of malignancy suspect
17
patterns are undertaken by small caliber (3 to 5 Fr) hysteroscopic forceps. Endometrial polyps,
18
hyperplasia and malignancy are often observed in different areas in the same endometrial cavity, and
19
the accuracy of visually directed biopsies is dependent on an observer related visual diagnosis of the
20
most suspect pattern. Endometrial polyps and small focal pathology can be removed by see-and-treat
21
hysteroscopy. However, to remove larger focal pathology as recommended[15], an additional
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 5
operative hysteroscopy may be needed in half of these postmenopausal patients [12]. Additional
2
operative hysteroscopy is costly [16] and stressful for these often elderly women. An accurate
3
subjective pattern evaluation of focal endometrial pathology during see-and-treat hysteroscopy may
4
reduce the need for additional operative hysteroscopy.
5
Several studies have evaluated the hysteroscopic view in postmenopausal women for diagnosis of
6
pathology [17-23]. Sugimoto [24] defined three hysteroscopic features of neoplastic morphology in a
7
study of 60 endometrial cancers. The features of malignancy were a nodular, polypoid, and
8
papillomatous type categorized irrespective of vascular atypia and tissue necrosis. However, mixtures
9
of these three distinctive features are often observed in endometrial cancer [25]. Based on prior
10
studies [26, 27], a classification system for hysteroscopic abnormalities defined malignancy by the
11
terms “whitish exophytic polypoid proliferation with atypical vascularization and hemorrhagic areas,
12
or papillary with a brain-like aspect and necrotic areas” [17]. Other unstructured terms with slight
13
modifications have been used [18-23] . A single case control study has evaluated the hysteroscopic
14
pattern in endometrial cancer[27]. However, there is no general recommended structured evaluation
15
and report of endometrial pattern obtained by hysteroscopy. We hypothesize, that a structured
16
evaluation of endometrial pattern may have higher accuracy, than subjective evaluation of pattern.
17
The aim of this study was to evaluate different visual parameters for the diagnosis of malignancy to
18
establish the parameters with the highest prediction of malignancy. Moreover, the aim was to clarify
19
the reproducibility of these terms in patients with postmenopausal bleeding and endometrial
20
thickness ≥5 mm, and to present an initial systematic score system for identification of endometrial
21
cancer.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 6
Methods
2
Hysteroscopy was performed in consecutive women with postmenopausal bleeding and an
3
endometrial thickness of ≥ 5 mm referred to Aarhus University Hospital, Denmark, from October 2010
4
to February 2012.
5
Hysteroscopy
6
A Storz resectoscope (26 Fr12 degree oblique optic) (Storz Endoskopi Danmark A/S, 2840 Holte ) was
7
used as an outpatient procedure under local or brief general anesthesia .
8
Two gynecologists with several years of experience in hysteroscopy performed hysteroscopy (live-
9
hysteroscopy) and obtained all video clips in the 164 women. At each examination, hysteroscopic
10
video-clips were obtained after a sufficient distension of the endometrial cavity was established by
11
sorbitol as distension media. The recording was started at the cervical canal at the entrance into the
12
endometrial cavity and systematically visualized the anterior wall, the fundus, the ostia, sidewalls, and
13
posterior wall. The panorama ended just before leaving the external os of the cervix and covered the
14
whole uterine cavity including the cervical canal. Often 2–3 clips were obtained in each patient. The
15
total recording time was 2–4 min.
16
Hysteroscopic video clips were evaluated independently by two gynecologists (PS) and (AMJ) from
17
another university hospital blinded to prior evaluations and the patients’ identities. Both had more
18
than 10 years of special experience in hysteroscopy.
19
Inclusion of patients
20
These women were included in a larger study where ultrasound was followed by hysteroscopy (live-
21
hysteroscopy) for the diagnosis of endometrial pathology. Diagnostic accuracy of two- and three-
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 7
dimentional ultrasound was evaluated [28, 29] and compared to accuracy of endometrial sample and
2
live hysteroscopy [30].
3
Patient characteristics and findings on transvaginal ultrasound (TVS) and gel infusion sonography (GIS)
4
were known to the operator of live-hysteroscopy. Therefore we used video evaluations in a blinded
5
set-up to eliminate observer bias induced by knowledge of prior imaging.
6
The exclusion criteria and selection of women used in the present study have in details been
7
described elsewhere [28, 29].
8
Consecutive women referred with postmenopausal bleeding (432 women) and an endometrial
9
thickness ≥ 5 mm (264 women) were included. 54 of 264 eligible women were not included: One had
10
an infection; 28 due to logistic problems or women did not want to participate; 10 patients had
11
hysteroscopy performed by other physician without systematic videos and in 15 patients, hysteroscopic
12
videos were not correctly stored and could not be evaluated.
13
Hysteroscopy was not performed in another 61 of these 264 women, leaving 149 patients for analysis. These
14
61 women not eligible for hysteroscopy had a prior diagnosis of malignancy: 9 women had hysteroscopic
15
removal of endometrial tumor before referral, 36 had other pelvic cancers, disseminated endometrial cancers,
16
serious comorbidity or advanced age and 16 women with endometrial cancer had a prior diagnosis of special
17
tumors ( 5 serous, 9 carcinosarcomas, 2 clear cell). In these 16 women hysteroscopy was not performed due to
18
a theoretic risk of spread of these rare tumors during hysteroscopy.
19
The study group of 149 women was either unreferred from the local area without prior imaging or
20
endometrial samples or referred from other hospitals or gynecologist often based on a diagnosis of
21
suspect malignancy (46 patients). Therefore this study included a higher numbers of patients with
22
endometrial cancer, than is normally seen in an unselected population of women with
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 8
postmenopausal bleeding. In 122 of 149 patients video-clips at TVS and gel infusion sonography(GIS)
2
were stored, and these patients were selected for evaluation of observer variation of endometrial
3
pattern parameters obtained by hysteroscopy. Thereby observer variation by hysteroscopy could be
4
compared to ultrasound, which will be presented elsewhere.
5
Visual Pattern analysis
6
The qualities of videos were evaluated (excellent, good, intermediate, poor, very poor) and
7
visualization of all the endometrium was noted. The endometrial pattern was systematically
8
evaluated according to a standard form. The different hysteroscopic patterns of cancer are outlined in
9
Figure 1.
M AN U
SC
RI PT
1
Observers were asked to note the presence or absence of the following pattern features:
11
Increased endometrial thickness, local lesions (≤ 25% of surface), diffuse lesions, (>25% of
12
endometrial surface involved).
13
The endometrial surface contour of the lesion (Figure 1 a):
14
Uneven surface texture pattern, polypoid surface, irregular surface and papillary projections.
15
Examples of an irregular surface are seen in Figure 1 (a1, a2, a3). Papillary projections (surface
16
papillarity with a fibrovascular stromal core) are displayed in Figure 1: a4, a5, a6.
17
Whitish (necrotic) areas (Figure 1 b):
18
Surface necrosis, candy floss endometrium, hyper intense white calcified spots in endometrium.
19
Surface necrosis are seen as whitish greyish areas without vessels on the surface of a lesion (b1), and
20
with characteristic features of white friable tissue (b2), shedding of clouds of whitish friable candy
21
floss-like tissue (b3, b4). Moreover hyper intense spots of whitish tissue may also be seen (b5, b6).
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 9
The vessel pattern (Figure 1 c):
2
An overall impression of irregular vessel pattern, irregular caliber of vessels, irregular branching of
3
vessels, and irregular distribution of vessels.
4
Characteristic features of irregular vessels are abnormal tortuous s-formed vessels, vessel loops and
5
vessels ending in small punctate “vessels sprouts” (c1, c2). Irregular caliber of vessels is characterized
6
by varying diameter of vessels with narrow parts of vessels (c3, c4). Irregular branching vessels have a
7
lack of hierarchy in branching from large vessels to often several very small vessels (c3, c4). Irregular
8
distribution of vessels (c5, c6) is characterized by areas with several irregular vessels and other areas
9
without vessels.
M AN U
SC
RI PT
1
Endometrial glands:
11
Dilated glands and glands with irregular openings.
12
Subjective diagnosis
13
At the end of systematic evaluation of the different parameters presence or absence of the following
14
subjective diagnosis was recorded on the standard form:
15
Hyperplasia, malignancy, benign polyps, myomas, proliferative endometrium, atrophy, and other
16
specified abnormalities.
17
The age of the patients was entered on the form. The subjective diagnosis of the endometrium was
18
performed according to specified definitions given on the form [17, 18, 30, 31].
19
Reference standard at pathology
20
After the video clips were recorded, all focal changes were removed by resectoscope. Three to five
21
resectoscopic biopsies were taken from diffuse areas with the most pronounced changes, or
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 10
resectoscopic biopsy was taken from the anterior and posterior wall of the normal uterine cavity.
2
Curettage was performed at the end of procedure to ascertain pathologic diagnosis of all endometrial
3
pathology.
4
The reference standard was pathology of the uterus if hysterectomy was performed or microscopic
5
evaluation of resectoscopic biopsy and curettage specimens [28, 30]. All specimens were evaluated by
6
one of two onco-pathologists.
7
Statistics and analysis
8
Data were analyzed using STATA (Statistic Data Analysis, STATA Corp, TX, USA). The diagnostic
9
performance with regard to discrimination between a benign and a malignant endometrium by
10
different hysteroscopic parameters was evaluated in terms of the areas under the curve (AUC),
11
sensitivity, specificity, positive likelihood ratios (LR+), and negative likelihood ratios (LR-). The
12
sensitivity, the specificity, and the LR+ and LR- of the most optimal cut-points of Receiver-operator
13
characteristic (ROC) curves were calculated. Optimal cut-points were calculated using a 45-degree
14
tangent line intersection or the smallest sum of residual sensitivity and specificity.
15
Then we evaluated which of the evaluated pattern features contributed to prediction of cancer on
16
multivariate regression analysis. If more than one variable for each of the three issues of parameters
17
(surface, necrosis, vessels) predicted cancer, a score variable was made with a score for each of the
18
relevant parameters present. Stepwise multivariate logistic regression was performed to build the
19
most optimal model that could be used to predict malignancy in hysteroscopy video clips. A maximum
20
of four fitting variables was allowed in the model.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 11
Based on the most optimal model calculated by logistic regression, a model based on scores (HYCA
2
score) was developed and compared to the most optimal model obtained on logistic regression
3
analysis and subjective evaluation; χ2 tests were used for discrete data. P < 0.05 was considered
4
statistically significant.
5
The inter-observer variation in image variable assessment was described by kappa analysis and the
6
standards for strength of agreement for the kappa coefficient were interpreted as follows: ≤ 0 = poor,
7
.01–.20 = slight, .21–.40 = fair, .41–.60 = moderate, .61–.80 = substantial, and .81–1 = almost perfect.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 12
Results
2
Pathologic findings in the 149 patients were 61 patients with endometrial cancer (56 endometroid
3
cancers, 1 mucinous, 2 serous, 1 clear cell, and 1 carcinosarcoma), 13 with complex or simple
4
endometrial hyperplasia, 6 with hyperplasia with atypia, and 44 with benign endometrial polyps. The
5
quality of the 149 hysteroscopic video clips was ranked as good or very good in 82%, intermediate in
6
11%, and of lower quality in 7%.
7
Table 1 lists all parameters evaluated for endometrial cancer on hysteroscopic video clips, and the
8
ability of each parameter to predict endometrial cancer. The overall subjective diagnosis had an AUC
9
of 0.859, and the AUC of subjective evaluation was comparable with the AUC of most optimal single
10
parameters (papillary projections, surface of uneven texture and irregular surface). All these
11
parameters had AUC values over 0.85.
12
Evaluation of surface parameters revealed that an irregular outline and papillary projections predicted
13
cancer, with a higher AUC when they were combined. Evaluation of necrosis parameters (surface
14
necrosis, candy floss, and hyper intense white spots) predicted cancer with a higher AUC for a
15
combination of all three parameters. For vessel parameters, a combination of irregular vessels with
16
irregular branching and irregular vessels with irregular distribution resulted in a higher accuracy for
17
the prediction of cancer.
18
Multivariate analysis including scores for surface, necrosis and vessels (Model M1, M2):
19
A model (M1) including surface score (irregular outline and papillary projections), necrosis score
20
(surface necrosis, candy floss, hyper intense white spots), and vessels score (irregular branching and
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 13
distribution) increased the diagnostic accuracy (AUC 0.966), indicating that most cancers could be
2
identified by combined evaluation of the pattern features: surface, necrosis, and vessel morphology.
3
In model M2, BMI was added to the graded score model. This increased AUC to 0.973. BMI had an
4
AUC (95% CI) of 0.651(0.56–0.75).
5
Multivariate analysis including single feature parameters (M3, M4, M5, M6):
6
When we included only a single ungraded parameter from each issue (surface, necrosis, vessels), the
7
most optimal model on multivariate logistic regression (M5) (Table 1) included the following features:
8
papillary projections, irregular distribution of vessels, surface necrosis, and BMI. The most optimal
9
model when only a single parameter was included (M5) had AUC (0.967) which had comparable
10
values with M1 (AUC:0.966). When AUC values were compared, all the regression models (M1, M2,
11
M3, M4, M5, M6) were more efficient than subjective evaluation.
12
HYCA-score:
13
Based on the scores obtained by hysteroscopy for each of the three graded parameters for surface,
14
necrosis, and vessels, we elaborated a hysteroscopic cancer score (HYCA score). In table 2 the
15
parameters used for assessment of endometrial cancer by the HYCA score system and most optimal
16
cut-points are displayed. The AUC of the score system was 0.964. At a cut-point of a score 3 or higher,
17
89% of the patients were correctly classified and 92% of cancers were identified. An example of HYCA-
18
score is given in the legend. The observer variation of the parameters is presented in Table 4. The
19
HYCA score system showed substantial agreement between observers. For individual parameters only
20
papillary projections had substantial observer agreement.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 14
Discussion
2
The systematic score system predicted cancers more efficiently than subjective evaluation. This
3
underlines the increased clinical value of hysteroscopy by systematic evaluation of endometrium and
4
elaboration of common terms to describe pattern and definitions of abnormal findings. The score
5
model had accuracy comparable with the most optimal model obtained by logistic regression. It
6
consisted of a count of numeric scores based on, the presence of lesions with irregular surface,
7
papillary projections, presence of surface necrosis of endometrium, candy floss shedding of necrosis,
8
hyper-intense white dots, irregular vessel pattern with irregular branching of vessels, irregular pattern
9
with irregular distribution of vessels in lesions. Using a score of 3 as a cut-point, nine of ten patients in
M AN U
SC
RI PT
1
the study were correctly classified with benign or malignant endometrial pathology.
11
In this population, some patients were referred based on a diagnosis of suspect malignancy, and we
12
had a high numbers of abnormal findings and endometrial cancer which is needed to evaluate the
13
accuracy of hysteroscopy to differentiate between abnormalities. However, the initial score system
14
developed in this population, has to be further tested and elaborated in a large population of
15
unselected women with postmenopausal bleeding and increased endometrial thickness before
16
recommending for use in general practice and should especially not be used in premenopausal
17
women before additional validation. Malignancy has been diagnosed by hysteroscopy with high
18
accuracy greater than in this study [18, 22, 32]. The lower accuracy in this study may be explained by
19
the high prevalence of pathology and other cut-points for malignancy but also by the use of video
20
clips. The accuracy of video clips was slightly lower than live hysteroscopy/ultrasound [30] although
AC C
EP
TE D
10
ACCEPTED MANUSCRIPT Dueholm 15
the general quality of clips were good. We used recorded video clips instead of live hysteroscopy,
2
because a safe live hysteroscopy of cancer patients cannot be performed blinded to ultrasound
3
findings of myometrial invasion, which might affect evaluation of hysteroscopic parameters. We
4
wanted an evaluation of the most optimally clear visual markers of cancer, which could be easily
5
identified using panorama hysteroscopy in video clips. Major shortcomings were that the presence of
6
abnormal morphology was evaluated with a subjective cut-off and without any graded score system.
7
We selected parameters with a clear definition used in prior studies, but were unable to define any
8
reliable cut- offs for malignant morphology based on the present literature. Other terms could have
9
been relevant, and a further elaboration of terms may increase accuracy.
M AN U
SC
RI PT
1
To our knowledge, this is the first hysteroscopic scoring system for the diagnosis of endometrial
11
cancer. We evaluated the most commonly used hysteroscopic visual parameters for the diagnosis of
12
malignancy [17, 19-23, 33], and established that the systematic evaluation of three visual parameter
13
components, surface, necrosis, and vessels, were efficient for prediction of cancer. For each of the
14
three factors, addition of more specified parameters increased the accuracy of the score system. The
15
specified parameters lead to a two point score for surface (irregular, papillary projections), a three
16
point score for necrosis (surface necrosis, candy floss, hyper intense white dots), and a two points
17
score for irregular vessels (branching, distribution).
18
systematic evaluation of surface outline, presence of necrosis and vessel pattern, stated in a standard
19
form by a count of scores may standardize terms and reporting, which is essential for quality
20
improvement and learning.
AC C
EP
TE D
10
At panorama hysteroscopy, a standardized
ACCEPTED MANUSCRIPT Dueholm 16
“Irregular surface” is a subjective parameter that has to be used together with other parameters, and
2
is used during panorama hysteroscopy to focus on the areas for detailed inspection. Papillary
3
projection had both low observer variation and high accuracy and was the most reliable parameter,
4
but was not present in all cancers.
5
In some cancers, whitish necrosis seemed to be the most prominent characteristic. Stromal invasion
6
characterized by stromal desmoplasia or stromal necrosis [34-36] may be seen as whitish surface
7
areas, and may give rise to the shedding of clouds of white friable characteristic necroinflammatory
8
debris, that we have called “candy floss”.
9
Some cancers may have very hyperintense calcified white spots, which is a round collections of
10
calcium (psammoma body),that may be associated with the papillary morphology and are thought to
11
arise in areas of necrosis and inflammation[37]. Correlation with endometrial cancer has been
12
proposed in some studies [38, 39], but is questioned or rare in other [37, 40, 41]. These white spots
13
are easily identified and contributed to prediction of malignancy in the HYCA score. Calcified spots are
14
correlated with diseases other than cancer [40], and visual markers of neoplastic tissue may replace
15
this parameter in future studies.
16
Identification of abnormal vessel pattern is important in the visual diagnosis of cancer [42, 43].
17
Narrow band imaging [22] may increase visualization of abnormal microvessels [21, 22]. Moreover,
18
computer technology[44, 45] is an upcoming and promising development in endoscopic imaging for
19
the evaluation of microvessels and surface microstructure. Clear terms with definition of abnormal
20
vessel sprouts, tortuous vessels, vessel loops, and branching with angles over 90 degree, narrowing of
21
vessels, disorganized network, and an overall irregular distribution with areas with dense vessel
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 17
varying with areas without vessels may optimize diagnosis. However, in concordance with another
2
study, a simple increased vascular density[46] has to be combined with other parameters in the
3
diagnosis of cancer.
4
A further elaboration of the systematic pattern parameters and the HYCA-scoring system may be
5
essential to decrease the need for extensive hysteroscopic procedures and for progress of immediate
6
office hysteroscopy in postmenopausal women with bleeding. In see-and treat hysteroscopy with
7
small caliber hysteroscopes the need for removal of residual focal pathology by additional operative
8
hysteroscopy with resectoscope should be evaluated when benign pathology is suggested by
9
standardized visual evaluation of endometrial pattern and biopsies. However, we used resectoscopes
10
to have a valid reference standard, and the hysteroscopic view by small dimensional hysteroscopes
11
seems efficient [14] but has to be evaluated.
12
There was only moderate agreement on a subjective diagnosis of cancer and a high inter-observer
13
variation in the identification of several individual parameters. The agreement may expectedly be
14
even smaller between larger groups of not experienced investigators. Cut-offs for positive and
15
negative findings may be clarified by a clear definition or grading of visual parameters guided by a
16
pictogram and stored representative hysteroscopic images and videos. At present, recognition of
17
endometrial cancer obtained with hysteroscopy is learned either by evaluation of individually stored
18
(not necessarily representative) hysteroscopic videos, or most commonly, by performance of high
19
volumes of hysteroscopies and memorizing the pattern. In most studies of unselected hysteroscopies,
20
there are 1–6 cancers in 100 hysteroscopies [19, 47]. Thus, large volumes of hysteroscopies are
21
required to see the diversity of cancer pattern, and special attention may be needed to remember
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 18
different patterns in order to achieve competence for recognition of cancer patterns. The HYCA-
2
scoring is a systematic way to learn and classify the pattern. Consistent scoring using parameters
3
before subjective evaluation will most likely improve the raters’ accuracy. In the future video clips
4
obtained by hysteroscopy may be stored in databases and computers may perform the pattern
5
analysis [44]. The HYCA- scoring can be the first step in the evolution of pattern analysis.
6
Systematic classification systems have also been effective in endoscopic diagnosis and learning in
7
colorectal cancer [48-50]. Moreover, the accuracy of ultrasound with systematic evaluation of pre-
8
defined strictly IETA terms [28, 51] was only slightly lower than the accuracy of the HYCA-score.
9
Hysteroscopy would be expected to yield a higher diagnostic accuracy than ultrasound since
10
hysteroscopy has a higher resolution regarding evaluation of the endometrium. This experience
11
underlines the importance of strategies to improve the competence of visual diagnosis, and highlights
12
the importance of the elaboration of defined terms for visual parameters obtained with hysteroscopy.
13
Conclusions A HYCA-score system based on surface outline, necrosis, and vessel pattern predicted
14
cancer with higher accuracy than subjective evaluation. Observer agreement was only moderate for
15
most terms, and may be improved by clear definitions. The score system has to be evaluated and
16
further elaborated in a large population of unselected women with postmenopausal bleeding and
17
increased endometrial thickness before recommending for use. Further elaboration of this initial
18
HYCA score system by identification and definition of specific terms regarding the hysteroscopic
19
pattern to predict malignancy should be performed.
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 19
Acknowledgements The authors gratefully acknowledge Charlotte Møller for collecting hysteroscopic
2
videos and the help of Elisabeth Melin and the staff at outpatient surgery unit.
3
Details of ethics approval The study was approved by The Ethics Committee of the Central Denmark
4
Region 9th of August 2010 (M-20100139). As hysteroscopy was a standard procedure in the
5
department, the committee found that no informed consent for this procedure was needed.
RI PT
1
SC
6 7 8
AC C
EP
TE D
M AN U
9
ACCEPTED MANUSCRIPT Dueholm 20
Legends
2
Figure 1: Figure 1a Surface: Surface irregularity is displayed in a1,a2,a3. Surface papillarity with a
3
fibrovascular stromal cores are called papillary projections (a4,a5,a6). These papillary projections may
4
be long, slender, villous papillae with thin fibrovascular cores (villo-glandular papillary projections) or
5
have a form of a bud-like to filiform appearance (non-villous papillae projections). A central vessel in
6
ciliated papillary projections may end in small vascular buds or sprouts (a6), which are common in
7
malignant tumors.
8
Figure 1b Necrosis: Surface necrosis (b1) are seen as whitish greyish areas without vessels on the
9
surface of a lesion. These areas may be composed of white friable tissue (b2), shedding of clouds of
10
whitish friable candy floss-like tissue (b3,b4). Hyper intense spots of whitish tissue may also be seen
11
(b5, b6). They may represent calcified material (psammoma bodies).
12
Figure 1 c. An irregular vascular pattern may represent slight irregularities to a totally chaotic vessel
13
pattern. Irregular vessels are (c1, c2) abnormal tortuous s-formed vessels, vessel loops and vessels
14
ending in small punctuates “vessels sprouts”. Irregular caliber of vessels is characterized by varying
15
diameter of vessels (c3, c4). Irregular branching vessels have a lack of hierarchy in branching (c3, c4)
16
with varying branching angles often with more than 90-degree branching angles. Irregular
17
distribution of vessels (c5, c6) are defined by presence of areas in which very large vessels are
18
present together with small vessels, and areas with increased density of often small micro vessels and
19
other areas with larger or no vessels. Small areas with hemorrhage may be seen.
20 21 22
AC C
EP
TE D
M AN U
SC
RI PT
1
ACCEPTED MANUSCRIPT Dueholm 21
References 1.
2.
6 7
8 3.
8 9
Olesen F, Hansen RP, Vedsted P. Delay in diagnosis: the experience in Denmark. Br.J.Cancer. 2009; 101:5-
Referral guidelines for suspected cancer NICE.Published May 2005. Accessed 1-3-2015; Available from: URL:https://www.nice.org.uk/guidance/cg27
4.
SC
5
Br.J.Cancer. 2011; 104:1836-9
RI PT
4
Johnson N, Miles T, Bailey D, et al. Delays in treating endometrial cancer in the South West of England.
Timmermans A, Opmeer BC, Khan KS, et al. Endometrial thickness measurement for detecting
M AN U
1 2 3
10
endometrial cancer in women with postmenopausal bleeding: a systematic review and meta-analysis.
11
Obstet.Gynecol. 2010; 116:160-7 5.
6.
15 16
endometrial cancer: cost-effectiveness of initial diagnostic strategies. BJOG. 2006; 113:502-10 7.
17 18
Munro MG. Investigation of women with postmenopausal uterine bleeding: clinical practice recommendations. Perm.J. 2014; 18:55-70
8.
19 20
Clark TJ, Barton PM, Coomarasamy A, Gupta JK, Khan KS. Investigating postmenopausal bleeding for
EP
14
cancer and other endometrial abnormalities. JAMA. 1998; 280:1510-7
TE D
13
Smith-Bindman R, Kerlikowske K, Feldstein VA, et al. Endovaginal ultrasound to exclude endometrial
Dreisler E, Poulsen LG, Antonsen SL, et al. EMAS clinical guide: Assessment of the endometrium in peri and postmenopausal women. Maturitas. 2013; 75:181-90
9.
AC C
12
Goldstein RB, Bree RL, Benson CB, et al. Evaluation of the woman with postmenopausal bleeding: Society
21
of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med. 2001;
22
20:1025-36
23 24
10.
Bakour SH, Jones SE, O'Donovan P. Ambulatory hysteroscopy: evidence-based guide to diagnosis and therapy. Best.Pract.Res.Clin.Obstet.Gynaecol. 2006; 20:953-75
ACCEPTED MANUSCRIPT Dueholm 22
11.
12.
4 5
successful? Eur.J Obstet.Gynecol Reprod.Biol. 2014; 178:70-3 13.
6 7
14.
Di Spiezio SA, Bettocchi S, Spinelli M, et al. Review of new office-based hysteroscopic procedures 20032009. J Minim.Invasive.Gynecol. 2010; 17:436-48
15.
10 11
Bettocchi S, Ceci O, Nappi L, et al. Operative office hysteroscopy without anesthesia: analysis of 4863 cases performed with mechanical instruments. J Am.Assoc.Gynecol Laparosc. 2004; 11:59-61
8 9
Gambadauro P, Martinez-Maestre MA, Torrejon R. When is see-and-treat hysteroscopic polypectomy
SC
3
assessing the diagnostic accuracy of office hysteroscopy. Fertil.Steril. 2002; 78:628-31
RI PT
2
Ceci O, Bettocchi S, Pellegrino A, et al. Comparison of hysteroscopic and hysterectomy findings for
M AN U
1
AnonymousAAGL practice report: practice guidelines for the diagnosis and management of endometrial polyps. J.Minim.Invasive.Gynecol. 2012; 19:3-10
16.
Saridogan E, Tilden D, Sykes D, Davis N, Subramanian D. Cost-analysis comparison of outpatient see-andtreat hysteroscopy service with other hysteroscopy service models. J Minim.Invasive.Gynecol. 2010;
13
17:518-25
14
17.
15
Dotto JE, Lema B, Dotto JE, Jr., Hamou J. Classification of microhysteroscopic images and their correlation with histologic diagnoses. J Am.Assoc.Gynecol Laparosc. 2003; 10:233-46
18.
Lasmar RB, Barrozo PR, de Oliveira MA, Coutinho ES, Dias R. Validation of hysteroscopic view in cases of
EP
16
TE D
12
endometrial hyperplasia and cancer in patients with abnormal uterine bleeding. J
18
Minim.Invasive.Gynecol. 2006; 13:409-12
19
19.
20 21 22
AC C
17
Garuti G, Sambruni I, Colonnelli M, Luerti M. Accuracy of hysteroscopy in predicting histopathology of endometrium in 1500 women. J Am.Assoc.Gynecol Laparosc. 2001; 8:207-13
20.
Surico D, Vigone A, Bonvini D, et al. Narrow-band imaging in diagnosis of endometrial cancer and hyperplasia: a new option? J Minim.Invasive.Gynecol. 2010; 17:620-5
ACCEPTED MANUSCRIPT Dueholm 23
21.
2 3
Cicinelli E, Tinelli R, Colafiglio G, et al. Reliability of narrow-band imaging (NBI) hysteroscopy: a comparative study. Fertil.Steril. 2010; 94:2303-7
22.
Tinelli R, Surico D, Leo L, et al. Accuracy and efficacy of narrow-band imaging versus white light
RI PT
1
4
hysteroscopy for the diagnosis of endometrial cancer and hyperplasia: a multicenter controlled study.
5
Menopause. 2011; 18:1026-9 23.
9
the Women's Clinic of Kyoto University Hospital. Am.J Obstet.Gynecol. 1975; 121:105-13 25.
11 12
26.
27.
Mencaglia L, Valle RF, Perino A, Gilardi G. Endometrial carcinoma and its precursors: early detection and treatment. Int.J Gynaecol.Obstet. 1990; 31:107-16
28.
17 18
Mencaglia L, Perino A, Hamou J. Hysteroscopy in perimenopausal and postmenopausal women with abnormal uterine bleeding. J Reprod.Med. 1987; 32:577-82
15 16
Garuti G, De GO, Sambruni I, Cellani F, Luerti M. Prognostic significance of hysteroscopic imaging in endometrioid endometrial adenocarcinoma. Gynecol Oncol. 2001; 81:408-13
13 14
Sugimoto O. Hysteroscopic diagnosis of endometrial carcinoma. A report of fifty-three cases examined at
TE D
10
SC
24.
Dueholm M, Møller C, Rydbjerg S, Hansen E. An Ultrasound Algorithm for the Identification of
EP
8
hyperplasia: a retrospective study of 323 patients. J.Minim.Invasive.Gynecol. 2005; 12:247-53
M AN U
7
Garuti G, Cellani F, Garzia D, Colonnelli M, Luerti M. Accuracy of hysteroscopic diagnosis of endometrial
Endometrial Cancer. Ultrasound Obstet.Gynecol. 2014; 43:557-68 29.
Dueholm M, Holm JW, Rydbjerg S, Hansen ES, Ortoft G. Two-dimensional and three-dimensional
AC C
6
19
transvaginal ultrasound with power Doppler angiography and gel infusion sonography for diagnosis of
20
endometrial malignancy. Ultrasound Obstet.Gynecol. 2015; 45:734-42
21
30.
Dueholm M, Marinovskij E, Hansen ES, Moller C, Ortoft G. Diagnostic methods for fast-track identification
22
of endometrial cancer in women with postmenopausal bleeding and endometrial thickness greater than
23
5 mm. Menopause. 2015; 22:616-26
ACCEPTED MANUSCRIPT Dueholm 24
31.
32.
4 5
hyperplasia: a systematic quantitative review. JAMA. 2002; 288:1610-21 33.
6 7
Clark TJ, Voit D, Gupta JK, et al. Accuracy of hysteroscopy in the diagnosis of endometrial cancer and
Zola FE, Nogueira AA, de Andrade JM, Candido Dos Reis FJ. Hysteroscopic appearance of malignant and benign endometrial lesions: a case-control study. Arch.Gynecol Obstet. 2007; 275:49-52
34.
8
Silverberg SG. Problems in the differential diagnosis of endometrial hyperplasia and carcinoma.
SC
3
pathologic findings on hysterectomy specimens. J.Minim.Invasive.Gynecol. 2006; 13:325-30
RI PT
2
Garuti G, Mirra M, Luerti M. Hysteroscopic view in atypical endometrial hyperplasias: A correlation with
Mod.Pathol. 2000; 13:309-27
M AN U
1
9
35.
Silverberg SG. Hyperplasia and carcinoma of the endometrium. Semin.Diagn.Pathol. 1988; 5:135-53
10
36.
Sherman ME, Silverberg SG. Advances in endometrial pathology. Clin.Lab Med. 1995; 15:517-43
11
37.
Parkash V, Carcangiu ML. Endometrioid endometrial adenocarcinoma with psammoma bodies. Am.J Surg
38.
14 15
1991; 35:81-8 39.
16 17
Kern SB. Prevalence of psammoma bodies in Papanicolaou-stained cervicovaginal smears. Acta Cytol.
Clement PB, Young RH. Non-endometrioid carcinomas of the uterine corpus: a review of their pathology with emphasis on recent advances and problematic aspects. Adv.Anat.Pathol. 2004; 11:117-42
40.
EP
13
Pathol. 1997; 21:399-406
TE D
12
Truskinovsky AM, Gerscovich EO, Duffield CR, Vogt PJ. Endometrial microcalcifications detected by ultrasonography: clinical associations, histopathology, and potential etiology. Int.J Gynecol Pathol. 2008;
19
27:61-7
20
41.
21 22 23
AC C
18
Fausett MB, Zahn CM, Kendall BS, Barth WH, Jr. The significance of psammoma bodies that are found incidentally during endometrial biopsy. Am.J Obstet.Gynecol. 2002; 186:180-3
42.
Jichlinski P, Lovisa B. High magnification cystoscopy in the primary diagnosis of bladder tumors. Curr.Opin.Urol. 2011; 21:398-402
ACCEPTED MANUSCRIPT Dueholm 25
43.
3
clinical trials. Nat.Rev.Clin.Oncol. 2014; 11:566-84 44.
4 5
Biomed.Health Inform. 2014; 19:1129-36 45.
6 7
Neofytou M, Tanos V, Constantinou I, et al. Computer Aided Diagnosis in Hysteroscopic Imaging. IEEE J
RI PT
2
Emblem KE, Farrar CT, Gerstner ER, et al. Vessel calibre-a potential MRI biomarker of tumour response in
Vlachokosta AA, Asvestas PA, Gkrozou F, et al. Classification of hysteroscopical images using texture and vessel descriptors. Med.Biol.Eng Comput. 2013; 51:859-67
46.
Dias DS, Bueloni-Dias FN, Dias R, et al. Usefulness of clinical, ultrasonographic, hysteroscopic, and
SC
1
immunohistochemical parameters in differentiating endometrial polyps from endometrial cancer. J
9
Minim.Invasive.Gynecol. 2014; 21:296-302
10
47.
M AN U
8
de Wit AC, Vleugels MP, de Kruif JH. Diagnostic hysteroscopy: a valuable diagnostic tool in the diagnosis
11
of structural intra-cavital pathology and endometrial hyperplasia or carcinoma?. Six years of experience
12
with non-clinical diagnostic hysteroscopy. Eur.J Obstet.Gynecol Reprod.Biol. 2003; 110:79-82 48.
14 15
Uraoka T, Saito Y, Ikematsu H, Yamamoto K, Sano Y. Sano's capillary pattern classification for narrow-
TE D
13
band imaging of early colorectal lesions. Dig.Endosc. 2011; 23:112-5 49.
Sakamoto T, Saito Y, Nakajima T, Matsuda T. Comparison of magnifying chromoendoscopy and narrowband imaging in estimation of early colorectal cancer invasion depth: a pilot study. Dig.Endosc. 2011;
17
23:118-23 50.
19 20
Dai J, Shen YF, Sano Y, et al. Evaluation of narrow-band imaging in the diagnosis of colorectal lesions: is a
AC C
18
EP
16
learning curve involved? Dig.Endosc. 2013; 25:180-8 51.
Leone FP, Timmerman D, Bourne T, et al. Terms, definitions and measurements to describe the
21
sonographic features of the endometrium and intrauterine lesions: a consensus opinion from the
22
International Endometrial Tumor Analysis (IETA) group. Ultrasound Obstet.Gynecol. 2010; 35:103-12
23
ACCEPTED MANUSCRIPT Dueholm 26
1
AC C
EP
TE D
M AN U
SC
RI PT
Table 1. Diagnostic efficiency of parameters at hysteroscopy for prediction of endometrial cancer AUC (95% CI) Correctly Sens Spec N=149 classified % % % at (cut-point) Overall subjective impression cancer 61 .859 (.80- .92) 82.0 90.0 Parameter present: N Thickened endometrium 48 .630 (.55—.71) 65.8 47.5 78.4 Localized lesion 69 .330 (.25—.40) 34.2 26.2 39.8 Diffuse lesion 77 .715 (.64—.79) 70.5 77.1 65.9 Diffuse or localized lesion 133 .535 (.49—.58) 46.3 93.4 13.6 Surface Surface uneven texture 65 .852 (.79—.91) 85.2 85.2 85.2 Surface not polypoid 49 .624 (.55—.70) 65.1 47.5 77.3 Irregular surface (a) 69 .857 (.80—.91) 85.2 88.5 83.0 Papillary projections (b) 62 .856 (.80—.92) 85.9 83.6 87.5 Score surface (a)+ (b) .886 (.83—.94) 83.2 (≥1) 88.5 79.6 87.9 (≥2) 83.6 90.9 Necrosis Surface necrosis (c) 33 .646 (.58—.72) 69.1 39.3 89.8 Candy floss like endometrium(d) 44 .750 (.68—.82) 77.9 59.0 90.9 White dots or areas in endometrium(e) 40 .689 (.62—.76) 72.5 49.2 88.6 Score necrosis(c) + (d) + (e) .887 (.84—.94) 81.9 (≥1) 95.1 72.7 75.8 (≥2) 45.9 96.6 Vessels Irregular vascular pattern overall 70 .838 (.78—.90) 83.2 86.9 80.7 Irregular caliber of vessels 54 .846 (.79—.91) 85.9 77.1 92.1 Irregular branching of vessels(f) 51 .849 (.79—.91) 86.6 75.4 94.3 Irregular distribution of vessels(g) 52 .843 (.78—.90) 85.9 75.4 93.2 Dilated endometrial glands openings 31 .490 (.42—.56) 54.4 19.7 78.4 Score vessels: f + g .867 (.81—.93) 85.9 (≥1) 78.7 90.9 86.6 (≥2) 72.1 96.6 Multivariate logistic regression M1:Score vessel, score necrosis, score .966 (.94—.99) 89.3(.362) 91.8 87.5 surface M2: BMI, score vessel , score necrosis, .973 (.95—.99) 90.6(.381) 90.2 90.9 score surface M3: Age, BMI, Papillary projections .919 (.88—.96) 87.3(.348) 86.9 87.5 M4 : BMI, Papillary projections, Irregular .963 (.94—.99) 90.6(.512) 90.2 90.9 distribution of vessels M5: BMI, Papillary projections, Irregular .967 (.94—.99) 91.3(.468) 90.2 92.1 distribution of vessels, surface necrosis M6: Papillary projections, Irregular .948 (.91—.98) 86.6(.425) 93.4 81.8 distribution of vessels, surface necrosis
LR+
LR-
8.01
.201
2.20 .44 2.26 1.08
.669 1.86 .348 .480
5.77 2.09 5.19 6.69 4.33 9.20
.173 .679 .124 .187 .144 .180
3.84 6.49 4.32 3.49 13.47
.676 .451 .573 .068 .560
4.50 9.69 13.27 11.06 .911 8.66 21.16
.163 .249 .261 .264 1.02 .234 .289
7.34
.094
9.92
.108
6.95 9.92
.150 .108
11.34
.107
5.14
.080
ACCEPTED MANUSCRIPT Dueholm 27
EP
TE D
M AN U
SC
RI PT
Sens=Sensitivity. Spec=Specificity. LR+= Positive likelihood ratio, LR- = Negative likelihood ratio. AUC= Area under the ROC-curve. Model 1(M1) had odds ratio (95%CI): Score-surface 2.67 (1.3-5.4);Score necrosis: 5.51 (1.9-15.7); Score-vessels 5.09(2.4-10.9) and z= -3.8 + (0.98 x score-surface) +(1.71 x score necrosis) + (1.63 x score-vessels). Probability of malignancy (P) can be calculated as follows: P= (ez/(1 +ez)) where e=2.718 (base value of natural logarithms). Roc comparison M3 vs. subjective evaluation p < .05 chi2
AC C
1 2 3 4 5 6 7 8
ACCEPTED MANUSCRIPT Dueholm 28
1 2 3
Table 2: Hysteroscopy Cancer (HYCA) scoring system and diagnostic performance with regard to predicting endometrial cancer: score system (HYCA score) and most optimal cut-point.( N = 149)
Present
Scoring Absent
Value
1 1
0 0
n1 n2
Surface contour
Irregular surface Papillary projections
Vessel pattern
HYCA-score
6 7 8 9 10 11 12 13 14 15
0 0
n6 n7 N
% correctly classified at cut-point
LR+
LR-
76.5 85.2 90.6 89.9 86.6 73.8
2.51 4.57 11.13 23.56 60.59
0 0.082 0.125 0.204 0.315 0.639
TE D
1 1
EP
5
n3 n4 n5
Hysteroscopic cancer score (HYCA score) included: lesion with irregular surface (present = score 1), papillary projections (present = score 1), surface necrosis (present = score 1), candy floss necrosis (present = score 1), white hyper intense spots (present = score 1), irregular branching (present = score 1), irregular distribution of vessels in lesion (present = score 1). Example: A patient with irregular surface (score 1), no papillary projections (score 0), no surface necrosis (score 0), no candy floss (score 0), no white hyper intense spots (score 0), irregular distribution of vessels present (score 1), and irregular vascular branching (score 1) had a total HYCA score(N)=3. Sensitivity(Sens) (88.5), Specificity(Spec) (92.1), and Positive likelihood ratio(LR+) of cancer
AC C
4
0 0 0
Irregular branching Irregular distribution N= ∑ (n1+n2+n3+n4+n5+n6+n7)
Most optimal cut-point for HYCA score (N) Cut-point for Sens Spec score (N) % % ( >= 1 ) 100 60.2 ( >= 2 ) 93.4 79.6 ( >= 3 ) 88.5 92.1 ( >= 4 ) 80.3 96.6 ( >= 5 ) 68.9 98.9 ( >= 6 ) 36.1 100.0 AUC 0.964 (.94-.99)
1 1 1
M AN U
Surface necrosis Candy floss pattern Hyperintense white spots
SC
Evidence of necrosis
RI PT
Pattern Characteristic/Variable
11.13. Score system both with and without inclusion of BMI versus subjective evaluation Roc comparison (p < .01). Score system including BMI had AUC (95% CI): 0.973 (0.95-0.99)
ACCEPTED MANUSCRIPT Dueholm 29
1
Table 4: Agreement between two observers for the presence of different parameters during evaluation of hysteroscopic clips for a diagnosis of malignancy.
EP
5
54/33 57/23 40/21 51/30 35/17 28/19 31/22 59/36 46/25 42/25 42/15 24/41 49/31
77.9 70.5 66.4 82.8 82.0 77.9 74.6 76.2 76.2 76.2 72.3 64.8 80.3
RI PT
% agreement 74.6 73.0 63.9
M AN U
SC
0.53 (0.39 — 0.68) 0.39 (0.25 — 0.52) 0.13 (-0.04 — 0.31) 0.62 (0.49 — 0.76) 0.48 (0.31 — 0.65) 0.30 (0.09 — 0.50) 0.26 (0.07 — 0.45) 0.52 (0.38 — 0.66) 0.44 (0.29 — 0.60) 0.42 (0.25 — 0.59) 0.29 (0.13 — 0.45) 0.12 (-0.05 — 0.29) 0.56 (0.42— 0.71)
N1/n2 41/52 58/73 65/23
N1= numbers with presence of the parameter by observer 1, n2= numbers with presence of the parameter by observer 2.
AC C
2 3 4
Kappa (95% CI) 0.47 (0.31 — 0.62) 0.46 (0.31 — 0.62) 0.31 (0.18 — 0.43)
TE D
Parameter (N = 122) Thickened endometrium Localized lesion Diffuse lesion Areas in localized or diffuse lesions with: Surface uneven texture Irregular surface Surface not polypoid Papillary projections candy floss like endometrium Surface necrosis White dots in endometrium Irregular vascular pattern Irregular caliber of vessels Irregular branching of vessels Irregular distribution of vessels Dilated endometrial glands openings HYCA score at cut-point of ≥3
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE
D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
http://www.AAGL.org/jmig-22-6-JMIG-D-15-00241
ACCEPTED MANUSCRIPT
Précis: Structured hysteroscopic evaluation of visual endometrial pattern terms increase diagnostic
AC C
EP
TE D
M AN U
SC
RI PT
accuracy of hysteroscopy for diagnosis of malignancy in women with postmenopausal bleeding.
