- Top of page
Although the overall mortality of patients with carcinoma of the uterine cervix has decreased over the years on account of the widespread availability of effective screening programs, cervical cancer ranks third worldwide among gynecological malignancies, with an age-standardized incidence of 8.9 per 100 000 women/year, and an estimated 9710 new cases per year in the United States. Overall, the 5-year survival rate has been reported to be 73%, but the prognosis for the subgroup with locally advanced cervical cancer still remains unsatisfactory1.
Assessment of the extent of the disease is crucial in planning the optimal treatment strategy. The International Federation of Gynecology and Obstetrics (FIGO) recommends a clinical staging system for cervical carcinoma based on findings from physical examination performed under anesthesia, colposcopy with biopsies of the lesion, chest radiography, cystoscopy, sigmoidoscopy, intravenous excretory urography and barium enema studies2. When comparing the FIGO staging with the surgical and pathological data, there is an underestimation of the extent of the disease in 17–32% of cases classed as stage IB, rising to 67% in stages II–IV of the disease3–5. The use of sophisticated radiological examinations, such as computed tomography (CT) and magnetic resonance imaging (MRI), has entered into routine practice, even if these methods are not regarded as obligatory in the assessment of clinical staging according to FIGO. Although acceptable figures of diagnostic accuracy have been reported for each of these procedures, there is still no general consensus on which represents the best option in the assessment of the extent of the disease6. The issues of costs, availability and dedicated radiological training call for the evaluation of imaging approaches other than CT and MRI. Given the great advances in ultrasound technology and equipment documented in recent decades, and considering its low cost and wide availability, it is reasonable to consider ultrasound as a potential diagnostic tool for cervical cancer staging.
The aim of this study was to prospectively compare the diagnostic performance of transvaginal ultrasound and MRI, using histology as the gold standard, with regard to the presence, size and extent of invasive cervical cancers and the detection of metastatic lymph nodes.
- Top of page
A consecutive series of women scheduled for surgery for early-stage invasive cervical cancer or triaged for surgery after neoadjuvant treatment of locally advanced cervical cancer were examined using transvaginal ultrasound and MRI within 1 week before surgery. All the patients had been admitted to the Department of Obstetrics and Gynecology, Division of Gynecological Oncology, at the Catholic University in Rome, between May 2002 and August 2005.
Patients who underwent a cervical cone biopsy were not eligible, and patients who underwent surgery > 7 days after ultrasound and MRI examination were also excluded. Written informed consent to take part in the study was obtained from all the women at enrollment.
All the patients were staged according to FIGO criteria and underwent a physical examination, tumor biopsy and gynecological examination under anesthesia and chest X-ray. Abdominal–pelvic MRI was carried out as routinely performed in our institution and the results were made available to the clinicians. For this study, the examiner of the MR images had access to the patient history but was blinded to the FIGO staging. Pelvic transvaginal sonography was performed only for research purposes and the sonographic results were not available to the clinicians. The ultrasound examiner had access to the patient's history but she neither performed gynecological palpation nor examined the patient by colposcopy, and was blinded to the FIGO staging and findings on MRI. The examiner of the MR images was unaware of the ultrasound findings. Cystoscopy and sigmoidoscopy were performed when indicated.
Patients with early-stage tumor (IB1 and IIA with tumor size < 4 cm) were treated surgically (Piver's types II or III radical hysterectomy and pelvic lymphadenectomy), whereas patients with stage IB2, IIA with tumor size greater than 4 cm, IIB, III and IV were given neoadjuvant treatment (neoadjuvant chemoradiation or neoadjuvant chemotherapy according to the clinicians' decision). Chemoradiation was given using the following protocol: whole pelvic external radiotherapy (22 fractions at a dose of 1.8 Gy/day, totaling 39.6 Gy) in combination with cisplatin (20 mg/m2) and 5-fluorouracil (1000 mg/m2) (on days 1–4 and days 27–30)7. Neoadjuvant chemotherapy consisted of two or three 21-day cycles of epirubicin (100 mg/m2), paclitaxel (175 mg/m2) and cisplatin (100 mg/m2)8. Four weeks after the end of the neoadjuvant treatment, the patient was reassessed using the same clinical and imaging procedures described above, and the response was recorded according to the World Health Organization criteria. The patients who showed complete or partial response to treatment underwent surgery (pelvic systemic lymphadenectomy and Piver II or III radical hysterectomy, depending on the preoperative assessment of response to neoadjuvant treatment) while patients experiencing no change or progression of the disease were treated with salvage chemotherapy. If pelvic nodes were intraoperatively suspected to be positive for tumor metastasis, para-aortic lymphadenectomy up to the inferior mesenteric artery was carried out.
All patients who underwent surgery (primary surgery or surgical treatment after neoadjuvant treatment) were examined by transvaginal sonography and MRI within 1 week before surgery. Both the ultrasound examiner and the examiner of the MR images assessed the following parameters and completed a dedicated form on paper: presence of cervical cancer, tumor diameters (craniocaudal length, and transverse and anteroposterior diameters), depth of stromal invasion, vaginal infiltration, parametrial infiltration, lymph node involvement, vesicovaginal septum infiltration, and rectovaginal septum infiltration. The results of the ultrasound and MRI examinations were compared with those of the histological examination of the surgical specimens.
All the surgical specimens were analyzed according to a predefined form by a pathologist (G. Z.) who had more than 10 years of experience in gynecological pathology.
The MRI examinations were performed using a 1.5T superconducting magnet (Echospeed, GE Medical Systems, Milwaukee, WI, USA). The pelvic phased-array coil was used in all patients. To reduce bowel peristalsis, 1 mg of butylscopolamine (Buscopan, Schering, Germany) was administered intramuscularly to all patients before beginning the examination.
Axial T1-weighted spin echo (SE) images were obtained with the following imaging parameters: repetition time/echo time (TR/TE) 500/14 msec, 4 mm slice thickness, 1 mm interslice gap, matrix 256 × 256 and acquisition time 4.24 min. Axial T2-weighted rapid acquisition with relaxation enhancement (RARE) (fast spin echo) images were obtained with the following parameters: TR/TE 4000/85 msec, echo train length (ETL) 12, 4 mm slice thickness, 1 mm interslice gap, matrix 256 × 256 and acquisition time 4.24 min. Sagittal T2-weighted RARE images were obtained with the following parameters: TR/TE 3500–4000/90 msec, ETL 12, 3 mm slice thickness, 1 mm interslice gap, matrix 256 × 256, acquisition time 5.52 min.
Oblique coronal (parallel to the main axis of the body of the uterus) and short axis (perpendicular to the main axis of the body of the uterus) T2-weighted RARE images of the uterus were also obtained with the same parameters used for the sagittal T2-weighted images, with an acquisition time of 4.24 min. Axial T2-weighted RARE images, with use of the body coil, were then acquired up to the renal hila, to assess the presence of lumboaortic lymphadenopathy. The parameters were as follows: TR/TE 4000/90 msec, 8 mm slice thickness, 1 mm interslice gap, matrix 256 × 192 and acquisition time 4.16 min.
The MR images were prospectively analyzed by two radiologists (R. M., B. G.) who had 10 and 5 years of experience, respectively, in MRI examination for gynecological pathologies. Interpretation discrepancies were resolved by consensus.
MRI analysis was performed according to diagnostic criteria previously reported in the literature9–12 (Figures 1 and 2). Lymph nodes with minimal axial diameter > 1 cm and with abnormal morphology were considered to contain metastatic tumors13.
Figure 1. (a) Sagittal T2-weighted magnetic resonance image of a cervical carcinoma (craniocaudal length × anteroposterior diameter, 21 × 23 mm) infiltrating less than two-thirds of the stroma (FIGO stage IB). (b) The tumor border, which is surrounded by free cervical stroma, is outlined.
Download figure to PowerPoint
Figure 2. Transverse (a) and coronal (b) T2-weighted magnetic resonance images of a cervical carcinoma (craniocaudal length × anteroposterior diameter, 37 × 31 mm) with infiltration of pericervical tissue (arrows) (FIGO stage IIB).
Download figure to PowerPoint
The first author, who had 10 years of experience in gynecological ultrasound, carried out all the ultrasound examinations. The women were examined in the lithotomy position with an empty bladder. Transabdominal and transvaginal sonographic examination was performed using an Esaote Technos (Esaote, Genova, Italy) ultrasound machine. A transvaginal scan of the pelvic organs was performed using a multifrequency endovaginal probe (ESAOTE: EC 123, 5.0–9.0-MHz) and was followed by a transabdominal scan using a 3- or 3.5–5.0-MHz convex transducer. A standardized examination technique was used. The transabdominal scan was performed in order to detect para-aortic lymph nodes and hydronephrosis.
At ultrasound examination cervical cancer appears as a solid lesion with mostly hypoechoic echostructure in comparison with the surrounding cervical stroma (Figure 3). Stromal invasion was categorized as being greater or smaller than two-thirds of the stromal width.
Figure 3. (a) Transvaginal ultrasound image showing a longitudinal section of a cervical carcinoma (craniocaudal length × anteroposterior diameter, 19 × 21 mm) infiltrating less than two-thirds of the stroma (FIGO stage IB). (b) The tumor border, which is surrounded by free cervical stroma, is outlined.
Download figure to PowerPoint
Tumor extension to the vagina was diagnosed in the presence of cervical neoplastic tissue obliterating and infiltrating the vaginal fornix/fornices when examined by moving the vaginal probe. Sonographic findings suggestive of parametrial invasion are extension of the cervical tumor beyond the cervical stroma and the presence of hypoechoic irregular tissue infiltrating the pericervical tissue (Figure 4). Infiltration of the vesicovaginal or rectovaginal septum was diagnosed on the findings of direct tumor extension and immobility of the fornix against the bladder or rectal wall; careful manipulation of the transvaginal probe permits the operator to analyze the mobility of the pelvic structure, thus obtaining a dynamic pelvic examination.
Figure 4. (a) Transvaginal ultrasound image showing a longitudinal section of a cervical carcinoma (craniocaudal length × anteroposterior diameter, 45 × 31 mm) with infiltration of the whole stromal thickness. (b) The tumor in (a) is outlined, with a transverse section (bold yellow line) of the cervical carcinoma demonstrated in (c). (d) The corresponding transverse section showing extension of the cervical tumor (outlined in (e)) beyond the cervical stroma with the presence of hypoechoic irregular tissue infiltrating the pericervical tissue (arrow).
Download figure to PowerPoint
A rounded shape and the absence of a fatty hilum14 of lymph nodes with a minimal axial diameter > 1 cm suggested metastatic involvement. Although color Doppler ultrasonography is routinely included in our sonographic examinations, with rich vascularization usually detected in the presence of cervical cancer, Doppler parameters were not considered in this study.
Ultrasound digital and photographic images were saved and stored on a hard disk for subsequent review and analysis.
The craniocaudal length of the tumor as measured by the pathologist was compared with those measured by ultrasound and MRI. The surgical specimen, fixed in formaldehyde, was examined as shown in Figure 5, with the maximum value of the craniocaudal length used for comparison with those obtained using the two imaging modalities.
Figure 5. (a) Illustration of the pathological examination of cervical cancer. The cervix has been amputated from the corpus with a sharp blade and opened with a pair of scissors through the endocervical canal at the 12 o'clock position. The entire cervix is then cut into parallel longitudinal sections, 2–4 mm apart, along the plane of the endocervical canal, starting at the 12 o'clock position and moving clockwise. For each section, the pathologist reported the craniocaudal extension (red line) and the stromal infiltration. The maximum value of the craniocaudal measurements was considered for the comparison between the craniocaudal tumor measurement obtained at ultrasound and MRI examination and the pathological findings. (b) Sonographic measurement of craniocaudal tumor length.
Download figure to PowerPoint
Sonographic and MRI data were independently compared with the histopathological data. The diagnostic performance of ultrasound and MRI was assessed by calculating the sensitivity, false-positive rate (1 − specificity), and positive and negative likelihood ratios (LR+ and LR−, respectively). The diagnostic performance (sensitivity and false-positive rate) of ultrasound and MRI was compared by using McNemar's test. Percent agreement and Cohen's kappa index were used as measures of concordance between the two methods. The number of cases where both ultrasound and MRI yielded a false-positive diagnosis (‘concordant’ false-positive diagnosis) and the number of cases where both ultrasound and MRI yielded a false-negative diagnosis (‘concordant’ false-negative diagnosis) were also calculated.
To assess the agreement between the sonographic and histopathological measurements, and between the MRI and histopathological measurements of craniocaudal cervical tumor diameter, the mean difference between the two measurements (ultrasound measurement subtracted from histopathological measurement and MRI measurement subtracted from histopathological measurement) and the limits of agreement were calculated as described by Bland and Altman15, 16. Patients in whom either histopathology or the imaging method showed no tumor at all were excluded from this analysis. To assess the relationship between the difference and the magnitude of the measured values, the differences between the measured values were plotted against the average of each pair of measurements16. To assess systematic bias, the 95% CI for the mean difference between the two methods was calculated; if zero lay within this interval, no systematic bias was assumed to exist between the two methods.
Statistical calculations were undertaken using the Statistical Package for Social Sciences (Version 12.02, SPSS Inc., Chicago, IL, USA). Two sided P < 0.05 was considered statistically significant.
- Top of page
We recruited 75 patients with the following FIGO stages: IA2 (two cases), IB1 (26 cases), IB2 (nine cases), IIA (two cases), IIB (25 cases), IIIA (two cases), IIIB (five cases), IV (four cases). Figure 6 is a flow-chart showing the treatment modalities of the included patients. Thirty-three patients were treated with primary surgery, including 28 patients with stage IA2 or IB1 tumors, four patients with stage IB2 tumors and one patient with a stage IIB tumor who refused the neoadjuvant treatment. After neoadjuvant treatment a clinical response was documented in 35/42 (83%) patients and these patients were scheduled for surgery. The clinical and pathological characteristics of the patients who underwent surgery (n = 68) are summarized in Table 1.
Figure 6. Flow-chart showing the staging and treatment of women with cervical cancer included in this study. NACT, neoadjuvant chemotherapy; NARCT, neoadjuvant radiochemotherapy.
Download figure to PowerPoint
Table 1. Clinical and pathological characteristics of cervical cancer patients who underwent surgery (n = 68)
|Characteristic||n (%) or mean (range)|
|Age (years)||51 (25–72)|
|FIGO stage|| |
| IA2||2 (2.9)|
| IB1||26 (38.2)|
| IB2||9 (13.2)|
| IIA||2 (2.9)|
| IIB||23 (33.8)|
| IIIA||1 (1.5)|
| IIIB||3 (4.4)|
| IVA||2 (2.9)|
| Squamous||53 (77.9)|
| Adenocarcinoma||9 (13.2)|
| Adenosquamous||4 (5.9)|
| Glassy cells||2 (2.9)|
|Grading of differentiation|| |
| G1, G2||23 (33.8)|
| G3||42 (61.8)|
| Unknown||3 (4.4)|
The diagnostic performance of ultrasound and MRI, with histopathological examination as the gold standard, is shown in Table 2. At histopathology an invasive cervical cancer tumor was confirmed in the 33 patients who underwent primary surgery, and a residual tumor mass was found in 27 of the 35 (77%) patients who underwent surgery after neoadjuvant treatment. No residual tumor was found in eight (23%) cases.
Table 2. Performance of preoperative ultrasound examination (US) and magnetic resonance imaging (MRI) in correctly detecting pathological characteristics of cervical cancer, using histology as a gold standard, in 68 patients (33 early cervical cancer patients triaged to primary surgery and 35 locally advanced cervical cancer patients treated with neoadjuvant therapy)
|Parameter||Sensitivity* (n) (95% CI)||False-positive rate* (n) (95% CI)||LR+ (95% CI)||LR− (95% CI)||Cohen's kappa†||Agreement for US and MRI (% (n))|
|Presence of cervical tumor||88 (60/68)|
| US||0.93 (56/60) (0.84–0.97)||0.5 (4/8) (0.38–0.62)||1.86 (0.93–3.75)||0.13 (0.04–0.43)||0.43|| |
| MRI||0.88 (53/60) (0.78–0.94)||0.38 (3/8) (0.26–0.49)||2.35 (0.96–5.79)||0.18 (0.08–0.45)||0.42|| |
|Stromal invasion greater than two-thirds||88 (60/68)|
| US||1 (16/16) (0.81–1)||0.25 (13/52) (0.15–0.25)||4.0 (2.50–6.40)||—||0.58|| |
| MRI||0.94 (15/16) (0.72–0.99)||0.15 (8/52) (0.07–0.24)||6.0 (3.18–11.67)||0.07 (0.01–0.49)||0.68|| |
|Parametrial infiltration||87 (59/68)|
| US||0.60 (3/5) (0.23–0.88)||0.11 (7/63) (0.04–0.19)||5.40 (1.99–14.68)||0.45 (0.15–1.32)||0.33|| |
| MRI||0.40 (2/5) (0.12–0.77)||0.11 (7/63) (0.04–0.19)||3.60 (1–12.96)||0.67 (0.33–1.39)||0.21|| |
|Vaginal infiltration||94 (64/68)|
| US||0.33 (1/3) (0.06–0.79)||0.03 (2/65) (0–0.09)||10.83 (1.32–88.73)||0.68 (0.31–1.53)||—|| |
| MRI||0 (0/3) (0–0.56)||0.05 (3/65) (0.01–0.11)||—||1.04 (0.99–1.11)||—|| |
|Vesicovaginal septum infiltration||97 (66/68)|
| US||1 (1/1) (0.21–1)||0 (0/67) (0–0.04)||—||—||—|| |
| MRI||1 (1/1) (0.21–1)||0.03 (2/67) (0–0.09)||33.5 (8.56–131.19)||—||—|| |
|Rectovaginal septum infiltration||99 (67/68)|
| US||0||0 (0/68)||—||—||—|| |
| MRI||0||0.01 (1/68) (0–0.07)||—||—||—|| |
|Lymph node metastases‡||91 (60/66)|
| US||0.09 (1/11§) (0.02–0.38)||0 (0/55) (0–0.05)||—||0.9 (0.75–1.10)||—|| |
| MRI||0.27 (3/11) (0.10–0.57)||0.04 (2/55) (0.01–0.11)||7.5 (1.41–39.76)||0.75 (0.52–1.09)||—|| |
Ultrasound examination detected the presence of a cervical tumor mass with a sensitivity of 93%, a false-positive rate of 50%, an LR+ of 1.86 and an LR− of 0.13. MRI detected the presence of a tumor with a sensitivity of 88%, a false-positive rate of 38%, an LR+ of 2.35 and an LR− of 0.18. Ultrasonography yielded four false-negative cases and four false-positive cases, and MRI yielded seven false-negative cases and three false-positive cases. There were three concordant false-negative cases and two concordant false-positive cases. The four patients with a false-negative finding on ultrasound had received neoadjuvant treatment. Microscopic disease was observed in three of the four cases. In the fourth case a residual tumor infiltrating 0.8 cm of 1 cm cervical stroma thickness was not detected at ultrasound examination. Six out of seven patients with a false-negative diagnosis on MRI examination had received neoadjuvant treatment and five of these had microscopic disease. In the sixth case a residual tumor infiltrating 0.7 cm of 1 cm cervical stroma thickness was missed. In one patient with early cervical cancer, who underwent primary surgery, MRI failed to detect a cervical tumor measuring 5 × 4 mm (stage IA2 at histology). All patients with a false diagnosis on either MR or ultrasound imaging had received neoadjuvant treatment. Median sonographic and MRI craniocaudal diameter of suspected tumors was 15 (range, 14–19) mm and 13 (range, 13–35) mm, respectively. In the false-positive cases, histopathology revealed only fibrosis and reactive tissue.
At histopathology, a tumor infiltrating more than two-thirds of the cervical stroma was found in 16 cases (23%), a tumor infiltrating no more than two-thirds of the cervical stroma was found in 44 (65%) cases, and no cervical tumor was found in eight cases (12%). Ultrasound examination correctly identified all 16 cases where the tumor infiltrated more than two-thirds of the cervical stroma, with a false-positive rate of 25% (13/52) and an LR+ of 4 (the LR− could not be calculated). The corresponding figures for MRI were 94% (15/16), 15% (8/52), 6 and 0.07. There were seven concordant false-positive cases.
Parametrial infiltration was documented at histopathological examination in five patients (one patient with an early cervical cancer who underwent primary surgery and four patients who had received neoadjuvant treatment). Both ultrasound and MRI had low sensitivity with regard to parametrial infiltration (three of five and two of five, respectively, P = 1) and both had a false-positive rate of 11% (7/63). The LR+ was 5.4 for ultrasound and 3.6 for MRI, while the LR− was 0.45 for ultrasound and 0.67 for MRI. There were two concordant false-negative cases and three concordant false-positive cases. Both ultrasound and MRI failed to detect a metastatic parametrial lymph node in one patient who underwent primary surgery. Both ultrasound and MRI failed to detect parametrial infiltration in a patient with a cervical cancer 2.5 cm in diameter, with whole stromal infiltration and 1 mm bilateral extension into the parametria after neoadjuvant chemoradiation. MRI failed to detect the parametrial infiltration in a patient who had undergone surgery after neoadjuvant chemotherapy and who had a 4 cm cervical cancer infiltrating the pericervical tissue. Both ultrasound and MRI correctly identified unilateral macroscopic parametrial tumor extension in two patients who had received neoadjuvant treatment. In three patients the cervical tumor involved the vagina. Ultrasound detected one of these cases, but MRI none. The false-positive rates for ultrasound and MRI were 3% (2/65) and 5% (3/65), respectively. In all false-positive cases an exophytic tumor protruding into the vagina was described. There were two concordant false-negative cases and one concordant false-positive case.
Ultrasound examination correctly identified the only case of infiltration of the vesicovaginal septum and accurately diagnosed the absence of rectovaginal septum infiltration in all cases in this series. MRI examination yielded two false-positive diagnoses of infiltration of the vesicovaginal septum and one false-positive diagnosis of cancer infiltration of the rectovaginal septum. In all three cases with a suspicion of infiltration of the vesicovaginal septum, cystoscopy was performed, with negative findings in all cases. The single case of infiltration of the vesicovaginal septum was characterized by disruption of the endopelvic fascia, as documented at histopathological examination.
At histopathological examination metastatic pelvic lymph nodes were diagnosed in 11 patients. The median number of removed lymph nodes in the whole series was 33 (range, 11–76) and the median number of metastatic nodes in patients with metastatic lymph nodes was two (range, 1–4) per woman. At histopathology the median diameter of the metastatic lymph nodes and the median diameter of the metastases within the lymph nodes were 0.8 (range, 0.6–1.4) cm and 0.5 (range, 0.1–1.3) cm, respectively. Ultrasound correctly detected metastatic lymph nodes in one of the 11 patients who had them, with no false-positive diagnoses, while MRI correctly detected metastatic lymph nodes in three of the 11 patients who had them (sensitivity 27%), with a false-positive rate of 4% (2/55). The differences in sensitivity and false-positive rate between ultrasound and MRI were not statistically significant for any of the parameters investigated (data not shown).
The Bland–Altman plot of ultrasound and histopathological measurements of craniocaudal tumor length, including only those patients in whom both histopathological examination and ultrasound examination showed a tumor (n = 56), is shown in Figure 7a, with the equivalent Bland–Altman plot of MRI and histopathological measurements (n = 56) shown in Figure 7b. Visual inspection of the scatterplots suggests that the magnitude of the differences did not change with the mean of the two measurements. The mean difference between histopathological measurements and ultrasound measurements of the craniocaudal diameter of the tumor was 0.62 mm (95% CI, − 1.96 to 3.21 mm; limits of agreement − 20.35 to 21.60 mm), and the mean difference between histopathological measurements and MRI measurements of the craniocaudal diameter of the tumor was 1.49 mm (95% CI, − 1.41 to 4.40 mm, limits of agreement − 21.85 to 24.83 mm).
Figure 7. Bland–Altman plots of measurements obtained at histopathological examination and at ultrasound (a) and at histopathological examination and magnetic resonance imaging (MRI) (b) for craniocaudal cervical tumor diameter. The diagrams include only cases in which both the histopathological examination and the imaging method showed a tumor.
Download figure to PowerPoint
- Top of page
In this study we evaluated the performance of two diagnostic techniques, i.e. ultrasound and MRI, for the preoperative evaluation of the presence, size and extent of invasive cervical cancer, with pathological findings used as the gold standard. Our results showed that ultrasound and MRI had a similar level of diagnostic performance, in contrast with studies in the past that have suggested a very limited role for ultrasound examination in the evaluation of cervical cancer17. Recently, a comparable accuracy of transrectal ultrasound and MRI in the staging of early cervical cancer has been reported in another study18, and there has also been a study published in which transrectal ultrasound and pathology-derived cervical tumor volume measurements correlated tightly and in which the accuracy of transrectal ultrasound was superior to MRI, especially in the detection of residual tumors following conization19.
In our study most of the false-negative results of both methods were found in patients with only microscopic residual disease after neoadjuvant radiochemotherapy. These false-negative results are of no clinical importance, because both the clinical management and the prognosis of patients with microscopic partial response are identical to that in patients with complete pathological response7.
Both ultrasound and MRI detected most cases of deep stromal infiltration, but both methods had high false-positive rates. The true diagnostic performance of ultrasound and MRI with regard to parametrial, vaginal and septal infiltration is impossible to estimate accurately because there were few such cases. It is nevertheless possible to speculate that the dynamic nature of the ultrasound examination procedure, with the movement of the transvaginal probe, permits the operator to examine the sliding of the contiguous tissues against each other20–22, facilitating evaluation of the local extent of the tumor.
Ultrasound failed to identify most patients (91%) with metastatic lymph nodes, but this was also the case for MRI (73%). However, in most of the false-negative cases, the largest diameter of the nodal metastases was below the minimum diameter detectable by MRI examination.
The mean differences between ultrasound or MRI and histopathological measurements of the craniocaudal diameter of the tumor were small (0.62 mm for ultrasound and 1.49 mm for MRI), but the limits of agreement were wide. A limitation of comparing measurements taken in vivo with measurements taken on specimens that have been fixed in formaldehyde is that tissue shrinks after fixation, and so one cannot expect perfect agreement19. However, the mean differences obtained in our study between the imaging methods and the histopathological data did not significantly differ from zero, with the imaging methods in fact giving slightly lower readings on average.
In the present study we included a group of patients with locally advanced cervical carcinoma, who were examined after neoadjuvant treatment. These patients are difficult to examine with any imaging method because fibrosis and necrosis change the normal structure of the pelvic organs and disrupt the borders between organs and structures. Thus the ability of transvaginal sonography to analyze the uterine cervix in these patients, with an accuracy comparable to that of MR imaging, is of great relevance in supporting the validity of this technique.
The clinical impact of our results on the management of patients with cervical cancer is not easy to assess. In patients with early cervical cancer, an ultrasound examination is likely to provide sufficient diagnostic information (tumor size, local extent of the cancer) for planning appropriate surgical treatment. Ultrasound does not detect lymph-node metastases, but this should not represent a clinical obstacle for the following reasons: (1) in the presence of an early invasive cervical cancer, pelvic lymphadenectomy is mandatory, irrespective of the results of preoperative investigation of lymph nodes; (2) knowledge of the histology and size of the tumor is sufficient to estimate the risk of metastatic lymph nodes; and (3) a patient with a FIGO stage IB tumor has a theoretical risk of lymph-node metastases of 13–39%23–25. For a patient with a 13–39% risk of lymph-node metastases, MRI, with its false-positive rate of 33–44% and false-negative rate of 13–27%23–25 with regard to metastatic lymph nodes, does not seem more useful than ultrasonography.
Critical aspects of ultrasound examination are the operator's practical skill when carrying out the examination and expertise in interpreting the images obtained. On the other hand, even MRI images need to be interpreted, and this interpretation is liable to as much subjectivity as the interpretation of ultrasound images. Disadvantages of MRI are its cost and the fact that a dynamic examination is not possible. Moreover, it is also time consuming and not universally available. Some might suggest that three-dimensional (3D) ultrasound would overcome the problem with the need for ultrasound operator skill, because it allows off-line analysis of volumes25. However, the acquisition of an ultrasound volume and the interpretation of 3D images require at least as much skill as the acquisition and interpretation of two-dimensional images. It would be interesting to determine the feasibility of a method whereby the acquisition of an ultrasound volume of the cervix in a patient with cervical cancer is obtained by an operator, and the interpretation is made off-line by a second operator. However, the dynamic aspect of the ultrasound examination would be lost in such a method.
In conclusion, considering the great advances in ultrasound technology and equipment, its relatively low cost, its widespread availability and the rapidity of the procedure, the role of ultrasound in the preoperative work-up of cervical cancer should be re-evaluated. In the hands of an experienced examiner it may be considered as the first-line diagnostic method.