Efficacy of transvaginal ultrasound versus magnetic resonance imaging for preoperative assessment of myometrial invasion in patients with endometrioid endometrial cancer: a prospective comparative study

Endometrial cancer (EC) is the most common malignancy in the female genital tract in developed countries and its incidence is increasing.1,2,3 Due to the early occurrence of abnormal uterine bleeding, most cases are diagnosed at first stage, when prognosis is very good, with a 5-year survival rate of 90%.4 The EC prognosis is determined by the FIGO (International Federation of Obstetrics and Gynecology) stage of disease, the histological type and grade of tumour, the depth of myometrial invasion (MI), cervical stromal invasion, and lymph nodes involvement.1,2,3 The depth of MI > 50%, which is considered to be one of the most important prognostic factors, highly correlates with lymph node metastases.1,2,3 The preoperative assessment of the depth of MI is crucial to determining the most effective therapeutic approach and to decide whether the patient should be referred for hysterectomy with bilateral adnexectomy or pelvic lymphadenectomy is need.1,2,3

The depth of MI can be evaluated by a number of imaging methods. These are magnetic resonance imaging (MRI), computer tomography (CT) and transvaginal ultrasonography (TVUS), among which the best results can be achieved by the MRI.1,3 TVUS has been used extensively to assess depth of MI by EC.3 TVUS when carried out by experienced hands has been shown to perform equally well as MRI, in the preoperative staging of EC.4

Several studies, systematic review and meta-analyses compared subjective assessment and objective TVUS measurements techniques2,5,6, TVUS (subjective assessment) and MRI3,7,8,9,10,11, three-dimensional TVUS and MRI12, and different MRI techniques.13,14,15 However, to the best of our knowledge this is the first study that compared two objective TVUS measurements techniques (Gordon’s and Karlsson’s method) with MRI on the same set of patients.

The aim of our study was to determine the accuracy of preoperative TVUS versus MRI for the assessment of MI depth in EC patients, while definitive histopathological diagnosis served as a reference method.

Of 72 cases diagnosed with endometrioid EC during the study period and who underwent preoperative TVUS and MRI, 60 cases were enrolled in the study, of mean (SD) age 60 (10) in range from 40–83 years. Forty-five (75%) out of the 60 were postmenopausal. Median body mass index was 32.05 kg/m2 (range 21.6-49.2). The most frequently encountered histological grade was grade 2 (58%, 35/60).

Twelve cases (16.6%) evaluated during the study period were excluded from the database, because of preoperatively made CT (morbid obesity and/or claustrophobia).

The patient demographics and tumour characteristics are summarized in Table 1.

Regarding final surgical procedure, 60 cases (100%) underwent open surgery. Hysterectomy with bilateral adnexectomy was performed in 60 cases (100%). Pelvic lymphadenectomy was performed in 32 cases out of 60 (53.3%). Pelvic lymphadenectomy was performed in eight cases out of 36 (22.2%) with MI ≤ 50%, in 24 cases out of 26 (92.3%) with MI > 50%, that is statistically significant frequently in group with MI > 50% (z = -5.29, p < 0.00001).

According to the gold standard, histopathological diagnostics, there were 34 (56.7%) cases in the study with MI ≤ 50%, and 26 (43.3%) with MI > 50%.

The depth of MI was correctly assessed by Gordon’s method in 50 (83.3 %) cases, overestimated in four (6.6%) and underestimated in six (10%) (Table 2).

The concordance coefficient between TVUS and histopathology was also statistically significant (p < 0.001) and kappa was 0.658. These data corresponded to a PPV of 83% and NPV of 83%, 77% sensitivity, 88% specificity, and 83 % overall accuracy (Table 3).

The depth of MI was correctly assessed by Karlsson’s method in 48 (80%) cases, overestimated in four (6.6%) and underestimated in eight (13.3%) (Table 2). The concordance coefficient between TVUS and histopathology was statistically significant (p < 0.001) and kappa was 0.585. Accordingly, Karlsson’s method calculating MI reached PPV of 82% and NPV of 79%, 69% sensitivity, 88% specificity, and 80% overall accuracy (Table 3).

MRI correctly assessed MI in 54 (90%) cases, overestimated it in five (8.3%) cases and underestimated it in one (1.6%) (Table 2). The concordance coefficient between MRI and histopathology was statistically significant (p < 0.001) and kappa was 0.80. Accordingly, MRI calculating MI reached PPV of 83% and NPV of 97%, 97% sensitivity, 85% specificity, and 90% overall accuracy (Table 3).

The diagnostic performance of two objective transvaginal ultrasonography methods assessment and MRI calculations in predicting deep MI as well as the statistical comparison of ultrasonography to MRI method are introduced in Table 3.

Differences in the performance of the two TVUS imaging modalities (Karlsson vs. Gordon) were not statistically significant (p = 0.867). Differences in the performance between the Gordon’s method and MRI were not statistically significant (p = 0.417). Differences in the performance between of the Karlsson’s method and MRI were not statistically significant (p = 0.464).

The prevalence of myometrial pathology was the same in the incorrectly classified patients in TVUS and MRI methods of assessment of MI. Of the 10 patients in whom infiltration was misclassified on TVUS and of the 6 patients misclassified on MRI, 5 (50%) and three (50%) had benign myometrial pathologies (adenomyosis and leiomyoma), respectively.

TVUS and MRI correctly estimated MI in 45 (75%) patients, underestimated it in one (1.6%) and overestimated it in two (3.3%). If we consider only the cases in which the two techniques were in agreement (48 cases), the concordance with histology was 80%.

Evaluation of diagnostic accuracy was performed through Receiver Operating Characteristics (ROC) analysis and for percentage assessment of MI with TVUS by Gordon’s and Karlsson’s method, and with MRI. A graphical representation of this analysis is given in Figure 4, and a tabelar representation of the areas below the ROC curve (AUC) in the Table 4.

Figure 4

As can be seen, the best diagnostic accuracy according to the ROC analysis had MRI with an AUC of 0.911 (total accuracy 91.1%), which is in correlation with the already performed diagnostic accuracy analyses.

In this prospective, comparative, ultrasonographer-blinded study on patients with EC two objective TVUS methods (Gordon’s and Karlsson’s) and MRI were compared for the MI assessment in the same cohort of patients. All three tested approaches were found to be statistically significant predictors of the MI, exceeding AUC value of 0.85 and reaching final p value < 0.001. We found that MRI assessment of MI is better than any objective TVUS measurement technique in all measures of the diagnostic tests, but without statistically significance.

Median age in our patients is similar to Pineda et al. study17, 60 vs. 60.9, respectively, however there are studies with higher1,4,5,6,8, and lower median age.12,18

Postmenopausal in recent study were 75% patients which correlate with other studies, where postmenopausal patients were in the range from 70.2% to 92%.5,6,7,8,9,10,15,17,19 EC is disease of older and postmenopausal women, that was also confirmed by our study.3

Patients in recent study were obese with an median body mass index (BMI) of 32.05 which is similar to median BMI of 31 in Rei et al. study8, however there are studies with higher7,15 and lower median BMI.4,5,6,17,19 Obesity is a proven risk factor for developing EC, what was also confirmed by recent study.8

The most frequently encountered histological grade in recent study was grade 2 (58%) which is the same percentage as in Karatasli et al. study.15 In most of the reviewed studies, histological grade 1 is more common than other grades which agrees with the fact that endometrial cancer is usually a well-differentiated.1,5,6,7,10,14,17

In current study according to the gold standard, histopathological diagnostics, it is more common superficial MI (≤ 50%), which correlate with most reviewed studies and with fact that EC is detected at an early stage in most cases.4,5,6,7,8,10,11,12,14,15,18,22,23

Several studies evaluated objective measurements such as those proposed by Gordon et al. and Karlsson et al.2,3

Alcazar et al. in systematic review and meta-analysis found that the overall diagnostic performance of TVUS for Karlsson’s and Gordon’s method in detecting deep MI in women with EC gave a pooled sensitivity of 84 % and 85 %, pooled specificity of 82% and 80% which is higher sensitivity and lower specificity than in our study for both methods.2 They observed that both methods were similar, without statistical differences, in terms of diagnostic performance, similar as in our study.2

In reviewed studies diagnostic performance for Gordon’s method reached sensitivity from 69.6% to 92.3%, specificity from 65.9% to 79.2%, PPV from 56.7% to 61%, NPV from 77.1% to 96.1% and overall accuracy from 67.3% to 82.6%.1,5

Recent studies reported for Karlsson’s method sensitivity from 56.3% to 86.8%, specificity from 64.4% to 76.4%, PPV from 62.8% to 83.6%, NPV from 70% to 71.2% and overall accuracy 68.1%.5,6,19,21

Besides, current study did not find statistical differences between Gordon’s and Karlsson’s method. Although both TVUS objective calculations, Gordon’s and Karlsson’s method, had similar accuracy, the approach published by Gordon et al. have better sensitivity and accuracy in preoperative assessment of MI in EC. However, in our opinion Gordon’s method might be more difficult for assessment of MI.5 On the other side, under or overestimation of MI by Karlsson’s method is often caused by large polypoid EC, submucosal leiomyomas and adenomyosis which make longer anterio-posterior uterine diameter.5

In 2017 Alcazar et al. published a systematic review and meta-analysis based on preoperative detection of deep MI comparing TVUS and MRI on the same set of women.3 However, they found out that sensitivity and specificity for diagnosing deep MI were 75% and 82% for TVUS, and 83% and 82% for MRI, respectively. MRI showed a better sensitivity than TVUS for detecting depth of MI in women with EC, but without statistical differences, as in our study.3 However, none of these studies compared the three imaging methods altogether in one cohort of patients as is the case in our comparative study.

In analysed studies diagnostic performance for MRI reached sensitivity from 70% to 92.6%, specificity from 71% to 95%, PPV from 65% to 92.2%, NPV from 70% to 98%, and overall accuracy from 74% to 89%.7,8,9,10,12,13,14,15,20,21,22

We have shown that two imaging modalities (contrast-enhanced MRI and TVUS) perform equally well in the assessment of MI, differences in the performance were not statistically significant.

Cubo-Abert et al. obtained similar results as in our study, when it comes to the diagnostic accuracy of TVUS versus MRI in estimating the depth of MI, with the difference that they used the TVUS method based on the measurement of the minimal distance to the uterine serosa.24

Costas et al. in systematic review and meta-analysis have not found a satisfying number of studies about the comparison of 2D-TVUS and 3D-TVUS methods for MI assessment in EC.25 They have identified lack of knowledge and studies regarding the objective methods used for 3D-TVUS assessment and their comparison with subjective 3D-TVUS methods.25

Although the definitive staging of endometrial cancer is based on histopathology, an accurate preoperative assessment of MI by TVUS and/or MRI provides the opportunity for surgical planning to provide an adequate type of surgery, the need of a multidisciplinary team, time management in the operating room, and avoid morbidity associated with unnecessary lymphadenectomy.8,9,23

Causes of over or understaging were similar for the TVUS and MRI: a polypoid EC, large exophytic tumors with distension and thinning of the myometrium with regular endometrial junction and without MI, fibromatosis, adenomyosis, leiomyomata, deep MI, small isolated glandular foci, uterine anomalies, uterine prolapse or retroversion, short time after previous endometrial biopsy, poorly defined endometrial borders on TVUS, and disappearance of the junctional zone on MRI.1,6,7,9,19

Considering that there are no statistically significant differences in diagnostic performance between TVUS and MRI and the cost and availability of MRI, TVUS may have a role as the first imaging technique for assessing MI in women with EC, especially as it is implemented in the everyday practice of gynecologists.1,2,4,8

As Miklos et al. concluded the diagnostic accuracy of the TVUS depends more on the individual experience and professional potential of the examiner than diagnostic accuracy of the MRI.1 Examiner experience, technological advances and different protocols for assessing MI among the studies, for both TVUS and MRI can contribute to the heterogeneity of published results of TVUS and MRI in the assessment of MI.1,3,8 Expert TVUS and MRI were comparable and superior to non-expert TVUS for assessing MI in EC.26

MRI, which is more expensive, time consuming and difficult to access, could be employed as a second-line imaging technique in patients in whom TVUS gives images of poor quality in case of obesity and factors above mentioned that may cause over or understaging.9

MRI is currently recommended for preoperative imaging in some guidelines, as the imaging modality of choice and most appropriate for the assessment of disease extent in patients with newly diagnosed EC.3,15,20

Iitsuka et al. in their study and review of literature found that their data and the pooled analysis with previous studies indicate that the frozen section diagnosis is sensitive as MRI assessment in predicting deep MI, and has a higher specificity compared with MRI.27

Contraindications for MRI should also be kept in mind, when MRI cannot be performed such as metal foreign body, pathologic obesity, contrast allergies and claustrophobia.

The main strength of our study is that, to the best of our knowledge, this is the first study that compares prospectively the two objective TVUS methods (Gordon’s and Karlsson’s) and objective MRI method, for assessing MI in patients with endometrioid EC in the same set of patients. A strength of our study are the prospective design and the fact that the gynaecologists performing TVUS and the radiologists performing MRI were blinded to each other’s results. Another strength is that all TVUS assessment of MI were made by the same physician, MRI assessment of MI by the same radiologist, and all surgeries and pathologic examinations were done at the same center. Strength of our study is also that we compared the three imaging methods altogether in one cohort of patients which is ideally, for comparing the diagnostic performance of different approaches to assess MI, and this is rarity in reviewed studies.

Limitations of our study are the relatively small number and short time of collection of participants, but this can be justified by the prospective design of the study, the strict inclusion and exclusion criteria, and in fact that the study was performed in the largest tertiary center in the country.

In conclusion, we found that objective TVUS assessment of myometrial invasion was performed with a diagnostic accuracy comparable to that of MRI in women with endometrial cancer. Further multi-centric studies with prospective designs and standardized protocols are needed to investigate which objective measurement techniques and MRI have the highest reproducibility, and how well they perform in the hands of examiners with more or less experience. In addition, studies evaluating inter-observer agreement as well as the impact of TVUS training would be of great interest.