To determine the agreement between ultrasound and histological examination of the cervix in patients with early stage cervical cancer with regard to tumor size and local extent of the disease.
To determine the agreement between ultrasound and histological examination of the cervix in patients with early stage cervical cancer with regard to tumor size and local extent of the disease.
Eighteen patients with histologically proven cervical cancer Stage IB1–IIA according to traditional clinical staging (FIGO 1988) who were scheduled for radical surgery underwent a standardized transvaginal ultrasound examination. The maximum tumor length, anteroposterior tumor diameter, tumor width, tumor area, depth of cervical stroma invasion, and the minimal thickness of tumor-free cervical stroma on sagittal and transverse planes through the cervix were measured, and the local extent of the disease within the parametria and vagina were evaluated. The surgical specimens were examined using a specifically devised method of histopathological examination. The results of the ultrasound and histopathological examinations were compared.
Limits of agreement were wide and the intraclass correlation coefficient (ICC) was low (0.51–0.58) for three of the four measurements taken to represent the minimal depth of tumor-free cervical stroma, i.e. the results of the measurements taken posteriorly and laterally. However, the limits of agreement were narrower and the ICC values were higher (0.74–0.92) for the depth of cervical stroma invasion and for the tumor size measurements. Histological examination revealed parametrial cancer infiltration in four patients, which was detected during ultrasound examination, with no false-positive results.
Transvaginal sonography is acceptably accurate for evaluation of tumor size and depth of cervical stroma invasion in clinical practice. Copyright © 2011 ISUOG. Published by John Wiley & Sons, Ltd.
In cases of cervical cancer it is important to determine the tumor size, the location of the tumor in the cervix and the extent of local disease when choosing optimal treatment. Traditionally, pretreatment evaluation is based on FIGO (International Federation of Obstetrics and Gynecology) staging. This is a clinical system that relies on findings during pelvic examination, cystoscopy and/or proctoscopy, and simple imaging modalities such as chest radiography, transabdominal ultrasound examination, and intravenous pyelogram, the latter three being used to detect distant organ metastases or ureteral obstruction1, 2. Because of discrepancies between clinical staging and pathological findings3–6, more sophisticated imaging methods have been suggested: magnetic resonance imaging (MRI) for measuring tumor volume and local extent, and computed tomography (CT) and positron emission tomography (PET/CT) for evaluation of metastatic disease7, 8. However, the FIGO Gynecologic Cancer Committee has accepted these sophisticated methods only reluctantly, because MRI, CT and PET/CT are not universally available1, 8.
Transvaginal sonography is a simple, non-invasive and relatively inexpensive imaging technique. In articles published in the 1990s, it was reported to be not very helpful in the pretreatment evaluation of invasive cervical cancer because of its low contrast resolution7, 9–11. However, with the improved resolution of transvaginal sonography it might now be possible to measure reliably tumor size and evaluate the local extent of cervical cancer.
The aim of this prospective study was to determine the agreement between findings on ultrasound and histological examinations in cases of cervical cancer with regard to the extent of the disease (full stroma invasion, vaginal and parametrial infiltration, tumor size, tumor-infiltrated cervical stroma and tumor-free cervical stroma).
The local ethics committee approved the study, and all participating patients consented to the study protocol in writing.
Between October 2006 and November 2007, 94 patients with a histological diagnosis of invasive carcinoma of the cervix were treated at Kaunas University of Medicine Hospital, Kaunas, Lithuania. On the basis of traditional clinical staging, 22 (23%) of them were judged to have early stage cancers (IB1–IIA, FIGO 1988 classification) and were selected to undergo primary radical hysterectomy or radical vaginal trachelectomy with pelvic lymphadenectomy. These 22 patients were eligible for inclusion in our study and were examined by ultrasound as described below. Of the 72 patients not eligible for inclusion, 14 were diagnosed with invasive cervical cancer Stage IA and 58 patients with advanced cervical cancer underwent radiation or chemoradiation as primary treatment.
The ultrasound examinations and examination of the surgical specimens of the patients included followed a standardized research protocol, developed by the ultrasound examiner (A.G.) and the pathologist (A.C.) of our team, with the aim of defining the planes where measurements were to be taken. The results of the ultrasound and histopathological examinations were compared. In four patients the surgical specimen was not examined according to the research protocol and these patients were excluded. Thus, 18 patients are included in our statistical analysis.
Within 24 hours before surgery, the patients to be included in the study underwent a transvaginal ultrasound examination with the purpose of determining the size and local extent of the cervical tumor. All scans were performed by the same examiner (A.G.) using a Toshiba Xario ultrasound system (Toshiba Medical Systems, Tokyo, Japan) equipped with a multifrequency two-dimensional endovaginal microconvex probe (Toshiba PVT-674BT; range, 3.6–8.8 MHz). The ultrasound examiner neither performed gynecological palpation nor had access to the results of any other examination used in the clinical FIGO staging of the patient.
The patients were examined transvaginally in the lithotomy position with an empty bladder. The cervical tumor was identified as an area with echogenicity different from that of the surrounding cervical tissue according to real-time grayscale ultrasound examination (Figure 1). An isoechoic tumor was identified by pushing the probe against the cervix, in which case the tumor appeared as an uncompressible lesion. Color or power Doppler ultrasound examination was not used to identify the tumor. Measurements were taken with calipers on the frozen ultrasound image.
On a sagittal plane of the cervix joining the 12 o'clock (anterior) and 6 o'clock (posterior) positions (i.e. a mid-sagittal plane along the axis of the cervix) the maximum tumor craniocaudal length (Length) and the maximum anteroposterior diameter (AP diameter) of the tumor were measured. The tumor area in the sagittal plane (Area Sag) was automatically calculated by the ultrasound system after the contour of the tumor had been drawn by the ultrasound examiner on the frozen ultrasound image. To evaluate the maximum depth of cervical stroma invasion (Depth), measurements were taken in the sagittal plane of the cervix close to the 12 and 6 o'clock position, at the level of, or cranial to, the reflection of the vaginal fornices where the tumor extension seemed to be maximal. The thickness of the thinnest tumor-free cervical stroma in the anterior and posterior lip of the cervix (Free Ant and Free Post) was also measured in the sagittal plane of the cervix close to the 12 and 6 o'clock position at the level of, or cranial to, the reflection of the vaginal fornices where the tumor-free stroma, anteriorly and posteriorly, seemed to be at its minimum (Figure 2). Theoretically, the measurements of the maximum depth of cervical stroma invasion and the thickness of the thinnest tumor-free cervical stroma in the anterior and posterior lip of the cervix could be taken at three different craniocaudal levels.
The transverse plane of the cervix was obtained by rotating the probe 90 degrees clockwise or counterclockwise. Moving the probe, the largest tumor extension on a transverse plane was identified and the maximum tumor width (Width) was measured, drawing a line connecting the 3 o'clock and 9 o'clock positions (i.e. laterally from left to right). At the level of, or cranial to, the vaginal fornices, the minimal thickness of tumor-free cervical stroma was measured on the right, close to the 9 o'clock position and on the left, close to the 3 o'clock position (Free Rt and Free Lt) (Figure 3). At least three measurements were taken for each variable, the largest being used for determination of tumor size and the smallest for determination of the thickness of the tumor-free cervical stroma.
An ultrasound diagnosis of full-thickness cervical stroma infiltration was made when the cervical tumor was seen to extend to the external border of the cervical stroma (up to the hyperechoic line demarcating the cervix; also known as the paracervical fascia) at any of the four locations where the minimal thickness of the tumor-free cervical stroma was measured (minimal tumor-free cervical stroma would then be 0).
By moving the endocavitary ultrasound probe in the anterior and posterior vaginal fornices, the infiltration of the vagina, vesicovaginal and rectovaginal spaces was assessed. Vaginal extension of the tumor was diagnosed if cervical neoplastic tissue obliterated and infiltrated the vaginal fornices. Infiltration of the vesicovaginal or rectovaginal spaces was diagnosed when the vaginal walls and fornices were not freely mobile against the bladder or rectum. Extension of the cervical tumor beyond the paracervical fascia with hypoechoic irregular solid tissue infiltrating the pericervical space suggested parametrial infiltration.
The ultrasound images and cineloops were stored and documented on hard disk in DICOM format. The electronically stored images and clips were used for subsequent review and analysis using OsiriX, version 2.7.5 DICOM software.
The resected specimen of the uterus was immediately fixed in 10% buffered formalin overnight and examined by a single pathologist using a methodology specifically designed for this research project (A.C.). The pathologist, who was blinded to the ultrasound findings, examined the specimen in the following manner. The cervix was amputated from the corpus with a sharp blade and opened with scissors through the endocervical canal at the 12 o'clock position (i.e. anteriorly). After that, the entire cervix was cut by making parallel longitudinal sections, 2–4 mm apart, along the plane of the endocervical canal starting at the 12 o'clock position and moving clockwise. Usually, three sections from every quadrant of the cervix, i.e. 12 sections in total, were taken (Figure 4).
During macroscopic examination of the 12 sections (Figure 5), the pathologist evaluated the location and the extent of the tumor and depicted them in a schematic drawing to be used for orientation during microscopic examination later. After that, cervical sections were processed and embedded in paraffin. The paraffin sections were cut on the microtome into thin sections (4 µm) and were stained with hematoxylin and eosin. When cervical sections were too large to fit the slide, they were cut into smaller portions. The pathologist interpreted the findings and marked the margins of the tumor on the slides under the microscope. To obtain measurements that could be used for comparison with the ultrasound measurements, sagittal and transverse plane views of the cervix were reconstructed from the microscope slides by scanning the slides with a high-resolution scanner (ScanJet G3010, Hewlett Packard Development Co., L.P., USA). The scanned images were then digitally processed using CanvasX and Photoshop CS3 (Adobe Systems Inc., USA) software and used for subsequent review and analysis using OsiriX, version 2.7.5 DICOM software. To reconstruct sagittal plane views, we used the histological sections close to the 12 o'clock and 6 o'clock positions with the largest tumor dimensions (Figure 6). In a reconstructed sagittal plane view through the cervix, the maximum tumor craniocaudal length and the maximum anteroposterior tumor diameter were measured. The tumor area in the sagittal plane was automatically calculated by the software after tracing the area of the tumor on the reconstructed sagittal plane view. To evaluate the maximum depth of cervical stroma invasion and the thickness of the thinnest tumor-free cervical stroma in the anterior and posterior lip of the cervix, measurements were taken in the reconstructed sagittal plane at the level of, or cranial to, the insertion of the vaginal fornices where the tumor invasion seemed to be maximal and where there was minimal tumor-free stroma anteriorly and posteriorly. Theoretically, measurements of the maximum depth of cervical stroma invasion and the thickness of the thinnest tumor-free cervical stroma in the anterior and posterior lip of the cervix could be taken at three different craniocaudal levels. In this manner measurements corresponding to the ultrasound measurements were obtained.
For reconstruction of the transverse plane view, the scans of the slides with tissue taken close to the 9 o'clock and 3 o'clock positions were used. Maximum tumor width was measured at the widest part of the tumor along a line connecting the 3 o'clock and 9 o'clock positions. At the level of, or cranial to, the reflections of the vaginal fornices, measurements of the minimal thickness of the tumor-free cervical stroma on the right and on the left (close to the 9 o'clock and 3 o'clock positions) were taken. A histological diagnosis of full thickness cervical stroma infiltration was made when the cervical tumor was seen to extend to the external border of the cervical stroma at any of the four locations where the minimal thickness of the tumor-free cervical stroma was measured (minimal tumor-free cervical stroma would then be 0).
All statistical calculations were carried out using the Statistical Package for the Social Sciences (SPSS) Software, version 16.0 (SPSS, Chicago, IL, USA).
Agreement between ultrasound and histopathological measurements of the cervical tumor and tumor-free cervical stroma was expressed as the mean difference between the ultrasound and histopathological measurement results (histopathological measurements being subtracted from ultrasound measurements) and as limits of agreement12, 13. The 95% confidence intervals (CI) of the lower and upper limits of agreement13 were also calculated. To assess the relationship between the difference and the magnitude of the measured values, the differences were plotted against the mean of the two measurements (Bland–Altman scatter plots)12, 14. For evaluation of systematic bias, the 95% CI for the mean difference between the two methods was calculated; if zero fell inside this interval, no systematic bias was assumed to exist between the two methods. Agreement between the ultrasound and histopathological measurements was also expressed as the intraclass correlation coefficient (ICC)15.
The clinical and pathological characteristics of the 18 patients included in the study are summarized in Table 1. Most tumors were Stage IB1 (14/18; 78%) and of squamous type (16/18; 89%).
|Characteristic||Mean ± SD (range) or n (%)|
|Patient age (years)||39.5 ± 9.6 (20–56)|
|Endophytic or ulcerative||14 (77.8)|
|Squamous cell carcinoma||16 (89)|
|Adenosquamous carcinoma||1 (5.5)|
According to histopathological examination, one patient had vaginal infiltration of the cervical tumor and this was detected by ultrasound examination. There were three false-positive ultrasound diagnoses of vaginal infiltration.
Parametrial infiltration of the tumor was diagnosed by histopathological examination in four cases, the infiltration being nodular in one case and microscopic in the other three. All four were detected by ultrasound examination and none by clinical examination. There was no false-positive ultrasound diagnosis of parametrial infiltration.
According to histological examination, the tumor infiltrated the full thickness of the cervical stroma in eight patients. In five of these patients the ultrasound examination revealed coherent results but there were also four false-positive ultrasound diagnoses of full stroma infiltration. The minimum tumor-free stromal thicknesses as measured by histological examination in the four cases with a false-positive ultrasound diagnosis of full stroma invasion were 7.2, 3.1, 2.8 and 1.0 mm, respectively.
In one patient with clinically diagnosed cervical cancer Stage IIA, bladder and ureter infiltration was diagnosed during ultrasound examination. During transvaginal ultrasound examination this patient had anterior and lateral parametrial infiltration, infiltrated bladder and hydroureter on the right. Histology of the resected bladder and ureter confirmed the ultrasound findings.
Results of the ultrasound and histopathological measurements of tumor size and tumor-free cervical stroma are summarized in Table 2. Median tumor size according to ultrasound examination was 30 × 32 × 34 mm (Length × AP diameter × Width), and median tumor size according to histological examination was 28 × 30 × 32 mm, respectively. Differences between ultrasound and histological measurements of tumor size and tumor-free cervical stroma are presented in Table 3. Visual inspection of the Bland–Altman scatter plots revealed that the magnitude of the differences between ultrasound and histological measurements did not change with increasing values. The 95% CI of the mean differences showed no systematic bias between the two measurement methods except that ultrasound examination overestimated the depth of cervical stroma invasion and tumor area measured on the sagittal plane relative to the histological measurements. The ICCs varied between 0.75 and 0.94 for the measurements of tumor size and depth of infiltration of the cervical stroma. The ICC for the thickness of the tumor-free cervical stroma in the anterior lip of the cervix was 0.72; the ICC for the thickness of the tumor-free cervical stroma in the other locations varied between 0.51 and 0.58.
|Ultrasound measurement||Histological measurement|
|Length of tumor (mm)||30.3||6.4||23.7||32.8||49.4||27.5||14.1||20.0||32.0||47.3|
|AP diameter of tumor (mm)||31.6||13.3||23.7||36.4||43.5||29.7||8.0||20.1||36.3||47.5|
|Depth of CSI (mm)||16.9||6.9||14.4||21.2||25.5||15.5||7.0||10.9||20.9||23.8|
|Free Ant (mm)||4.2||0.0||0.0||6.4||23.0||4.9||0.0||0.0||10.6||19.9|
|Free Post (mm)||4.2||0.0||1.3||7.6||21.3||4.4||0.0||0.0||10.9||16.9|
|Area Sag (mm2)||597.0||91.5||314.5||764.0||1663.1||483.3||118.0||265.6||724.4||1511.0|
|Width of tumor (mm)||33.8||10.8||25.3||37.6||45.0||31.8||15.0||24.7||41.1||44.5|
|Free Rt (mm)||3.9||0.0||0.0||6.9||11.0||4.4||0.0||0.0||9.9||20.5|
|Free Lt (mm)||3.6||0.0||0.0||7.0||12.9||4.4||0.0||0.0||10.5||21.8|
|Difference between ultrasound and histological measurements*|
|Variable||Mean (range)||95% CI of mean||Limits of agreement||95% CI of lower limit||95% CI of upper limit||ICC|
|Length of tumor (mm)||2.8 (−7.7 to 19.5)||− 0.65 to 6.15||− 10.64 to 16.14||− 13.43 to − 7.85||13.35 to 18.93||0.745|
|AP diameter of tumor (mm)||1.1 (−7.8 to 10.3)||− 1.09 to 3.21||− 7.45 to 9.57||− 9.22 to − 5.68||7.8 to 11.34||0.942|
|Depth of CSI (mm)||1.9 (−4.2 to 8.2)||0.1 to 3.78||− 5.33 to 9.21||− 6.84 to − 3.82||7.7 to 10.72||0.745|
|Free Ant (mm)||− 1.2 (−14.2 to 4.9)||− 3.39 to 0.99||− 9.86 to 7.46||− 11.66 to − 8.06||5.66 to 9.26||0.715|
|Free Post (mm)||0.3 (−15.2 to 9.8)||− 2.76 to 3.32||− 11.68 to 12.24||− 14.17 to − 9.19||9.75 to 14.73||0.510|
|Area Sag (mm2)||127.3 (−217.3 to 795.5)||12.1 to 242.6||− 327.0 to 581.6||− 421.6 to − 232.4||487.0 to 676.2||0.824|
|Width of tumor (mm)||0.2 (−9.2 to 10.3)||− 2.41 to 2.77||− 10.01 to 10.37||− 12.13 to − 7.89||8.25 to 12.49||0.827|
|Free Rt (mm)||− 2.1 (−16.7 to 9.6)||− 4.69 to 0.59||− 12.44 to 8.34||− 14.6 to − 10.28||6.18 to 10.5||0.578|
|Free Lt (mm)||− 2.2 (−21.8 to 4.2)||− 5.06 to 0.64||− 13.4 to 9.0||− 15.74 to − 11.06||6.66 to 11.34||0.513|
In this prospective study, we have determined the diagnostic performance of ultrasonography in pretreatment evaluation of the extent of early invasive cervical cancer using histopathological examination of the surgical specimen as a reference standard. The novelty and strength of this study is the direct comparison of ultrasound measurements with their histological counterparts. Comparison was possible only after we had developed a specific method for histopathological examination in close cooperation between the ultrasound examiner and the pathologist of our team. This method is not intended to be used clinically; it was developed only for this study to allow comparison of ultrasound and histological findings.
The limitation of our study is that, despite our serious attempts to devise a method to make possible a direct comparison between ultrasound and histological measurements, our method does not ensure fully that the measurements by the ultrasound examiner and the pathologist were taken at exactly the same location. On the other hand, the measurements should be comparable because both the ultrasound examiner and the pathologist were asked to identify the maximum tumor dimensions and the thinnest tumor-free cervical stroma in the same sagittal plane (that joining the 6 o'clock and 12 o'clock positions), the maximum tumor width on a transverse section through the cervix using a line joining the 3 o'clock and 9 o'clock positions, and the thinnest tumor-free cervical stroma at the 3 o'clock and 9 o'clock positions. We consider it clinically relevant to estimate the agreement between ultrasound examination and histopathological examination with regard to maximum and minimal measurements, because the assessment of minimum and maximum is subjective, in the hands of both an ultrasound examiner and a pathologist.
Our results suggest that ultrasound examination is likely to be clinically useful for measuring tumor size and determining the local extent of early-stage cervical cancer. Ultrasound examination identified all four cases of histologically confirmed parametrial infiltration with no false-positive results. The differences between ultrasound and histological measurements of tumor size and depth of cervical stroma invasion were rather small, even though ultrasound examination led to a slight overestimation of the degree of cervical stroma invasion relative to histological examination. Although overestimation of tumor burden may seem preferable to underestimation, overestimation compromises the opportunity to offer fertility-preserving procedures where this might be appropriate.
Our findings are in agreement with those of other recently published studies which demonstrate the acceptable accuracy with regard to tumor size of both transvaginal and transrectal ultrasound examination when used for pretreatment evaluation of cervical cancer16, 17. One of these studies, in agreement with ours, revealed ultrasound examination to be associated with a high false-positive rate for full stromal infiltration17. The limited agreement between ultrasound and histopathological measurements of the minimal thickness of tumor-free cervical stroma is difficult to explain, but possibly neoplastic tissue and healthy tissue react differently to formalin fixation. Perfect agreement between measurements taken in vivo during an ultrasound examination and those taken on specimens fixed in formaldehyde where the tissues have shrunk cannot be expected. In three of our four cases, where ultrasound examination incorrectly suggested full stromal infiltration of the tumor, the histological examination revealed the thickness of the healthy stroma to be ⩽ 3.1 mm. The clinical relevance of incorrect classification according to ultrasound examination in these three cases may be questioned.
Lymph node metastasis, invasion of the lymph-vascular space, tumor size, and depth of invasion are significant independent prognostic factors for disease-free survival in patients with cervical cancer5, 18. Pelvic lymphadenectomy is always performed when patients undergo surgery for Stage IB–IIA cervical cancer. Pretreatment evaluation of tumor size and of location and depth of stromal invasion, however, is critically important for determining the extent of surgery if fertility-sparing surgery is desirable19–21. It is also critical when planning nerve-sparing surgery and tailored surgery, because the degree of radical surgery can be modulated on the basis of the tumor location22–26. Our results show that transvaginal ultrasound examination can provide reliable information on all three factors (tumor size, location of stromal invasion and depth of stromal invasion) which are important in selecting the extent of surgery while clinical examination has limitations in revealing these parameters accurately.
Our results, together with those of other recent studies that have shown ultrasound examination to have a diagnostic performance similar to that of MRI for determining the size and local extent of cervical cancer16, 17, suggest that ultrasound could be the first-line imaging technique for pretreatment evaluation of early cervical cancer. The relatively low cost of ultrasound examination and its widespread availability support this notion.
SUPPORTING INFORMATION ON THE INTERNET
The following supporting information may be found in the online version of this article:
Videoclip S1 Demonstration of three-dimensional reconstruction of the cervix, with schematic representation of tumor tissue in blue and healthy cervical stroma in red.
We dedicate this work to the memory of our colleague Dr S. Kajenas—brilliant surgeon and beloved friend, who left us on May 28, 2009, being in his early forties.