Although the correlation between overall clinical and MRI stagings by percent agreement in this study was moderate (73.9%), the kappa coefficient (K = 0.000) showed a slight correlation. This might be due in part to the small sample size in this study. The previous study by Dhoot et al. showed a higher accuracy of 89.3% by MRI staging compared with 61.3% by clinical staging [4]. Another study by Ho et al. (1992) showed the overall accuracy rate of MRI in staging of cervical cancer was 75%, much higher than 55% by clinical staging [6, 7]. Ozsarlak et al. demonstrated that the overall accuracy of cervical cancer staging by clinical examination and by MRI was 47 and 86%, respectively [8]. Shirazi et al. showed 50% correlation between clinical and MRI stagings in stage IIIB patients (which is the main population in our study) [9]. According to the discrepancy between clinical and MRI stagings from previous studies, the slight correlation between clinical staging and MRI staging in cervical cancer in this study suggests the requirement of a large sample size study.
Other results such as vaginal invasion, pelvic sidewall invasion, adjacent pelvic organ invasion, and spreading to distant organs also showed moderate-to-strong correlation between clinical and MRI examinations by percent agreement (67.6 to 91.9%), although the correlation between them was only slight by kappa or weighted kappa coefficient (K = 0.000–0.128w). MRI sequences with other imaging modalities were used in the staging and follow the treatment of cervical cancer; i.e. relevant anatomy (including normal MRI appearance of the cervix, parametria, and pelvic ligaments), different stages of cervical cancer with prognostic and therapeutic implications [10].
For the parametrial invasion, both clinical and MRI examinations detected parametrial invasion in all 37 cases so that correlation analysis by both percent agreement and kappa coefficient was meaningless. One study reported that MRI has 74% and 93% sensitivity and specificity, respectively, to detect parametrial invasion [11]. Another study showed MRI accuracy in demonstrating parametrial involvement was 95%, with 73% sensitivity, 96% specificity, and 21–85% clinical staging accuracy [12]. According to the literature review by Thomeer et al. [13], MRI evaluation of parametrial invasion showed 84% pooled sensitivity (95% CI 76–90) and 92% pooled specificity (95% CI 90–95), whereas clinical examination showed 40% pooled sensitivity (95% CI 25–58) and 93% pooled specificity (95% CI 83–89).
In this study, clinical and MRI evaluations for hydronephrosis showed a statistically significant correlation by kappa coefficient (K = 0.749, P value < 0.001). This may be because clinical examination evaluated hydronephrosis by intravenous pyelography. Our results corresponded well with the study by Chung et al. in that all 18 patients with hydronephrosis who were identified by intravenous pyelography were also recognized by MRI or CT [14].
Advantages of pretreatment MRI
Clinical examination can define dimension in an axial plane. Multiplanar MRI with higher spatial soft tissue resolution can define pelvic tumor extent, including more accurate assessment of tumor size, stromal invasion depth, and parametrial invasion [15].
In this study, 8 patients (21.6%) had adjacent pelvic organ invasion by MRI, although the clinical examination cannot detect this finding. MRI has higher sensitivity to detect bladder invasion by early detection of bladder wall invasion, while clinical examination by cystoscopy needs to visualize intravesical tumor extension. Furthermore, the advantage of high spatial soft tissue resolution of MRI can also provide early detection of other adjacent pelvic organ invasions or distant organ metastases such as the rectum, colon, ovaries, etc.. This could potentially reduce staging costs and morbidity [16].
In our study, the pathological staging was not compared with the clinical and MRI findings, which is a gold standard to show sensitivity or specificity of the tests, although pathological staging can provide how accurate the staging was by clinical and MRI examinations.
There is a selection bias of some certain staging (all the cervical cancer patients in this study are locally advanced disease or advanced disease) of cervical cancer because the patients who received pretreatment MRI in our hospital are mostly locally advanced disease (FIGO stage IIB or greater). Moreover, early staging patients were assessed by clinical FIGO staging and underwent surgical treatment, thus those patients were not evaluated by MRI for pretreatment planning. Therefore, our study has a small and insufficient number of patients to show a solid conclusion about the correlation between clinical and MRI stagings.
Also, because this study is a retrospective review, we cannot standardize MRI protocol and time interval between clinical staging and pretreatment MRI examination, which might result in inaccurate comparison between clinical and MRI stagings.
Clinical examination limit to evaluate pelvic/para-aortic lymphadenopathy. Lymphadenopathy was evaluated by MRI in only 27 (73.0%) out of 37 patients. Although the accuracy of MR imaging was fairly high in the detection of pelvic node metastasis from uterine cervical carcinoma, morphology and short-axis size (such as only round-shaped lymph node with less than 0.8 cm in short-axis size) of lymph nodes sometimes are equivocal to diagnose metastatic lymph node [17]. It should be aware that MRI will fail to detect metastasis in normal-size lymph nodes [18].