Unclear tumor border in magnetic resonance imaging as a prognostic factor of squamous cell cervical cancer

Magnetic resonance imaging (MRI) is used for pretreatment staging in cervical cancer. In the present study, we used pretreatment images to categorize operative cases into two groups and evaluated their prognosis. A total of 53 cervical cancer patients with squamous cell carcinoma who underwent radical hysterectomy were included in this study. Based on MRI, the patients were classified into two groups, namely clear and unclear tumor border. For each patient, the following characteristics were evaluated: overall survival; recurrence-free survival; lymph node metastasis; lymphovascular space invasion; and pathological findings, including immunohistochemical analysis of vimentin. The clear and unclear tumor border groups included 40 and 13 patients, respectively. Compared with the clear tumor border group, the unclear tumor border group was associated with higher incidence rates of recurrence (3/40 vs. 3/13, respectively), lymphovascular space invasion (24/40 vs. 13/13, respectively), lymph node metastasis (6/40 vs. 10/13, respectively), and positivity for vimentin (18/40 vs. 10/13, respectively). Despite the absence of significant difference in recurrence-free survival (p = 0.0847), the unclear tumor border group had a significantly poorer overall survival versus the clear tumor border group (p = 0.0062). According to MRI findings, an unclear tumor border in patients with squamous cell cervical cancer is linked to poorer prognosis, lymph node metastasis, and distant recurrence of metastasis.

In the present study, we used MRI to categorize cases into two groups, and evaluated the patient characteristics, tumor characteristics, and disease prognosis.

Materials and methods
Patients.A total of 328 cervical cancer patients, who underwent primary treatment in Nagoya University Hospital (Nagoya, Japan) between January 2009 and December 2013, were analyzed.Of those, 53 patients diagnosed FIGO stage 1 preoperatively who underwent abdominal radical hysterectomy with identifiable lesions limited in the cervix (according to MRI) were included in the study.Patients with pathologies other than squamous cell carcinoma, those who had received neoadjuvant therapy, as well as those with lesions growing outside the cervix, were excluded (Fig. 1).Subsequently, patients were categorized into two groups based on T2-weighted images, namely clear and unclear tumor borders (Fig. 2).Adjuvant concurrent chemoradiotherapy (CCRT) (50.4 Gy whole pelvic irradiation plus three cycles of 70 mg/m 2 of cisplatin and 2800 mg/m 2 of 5-fluorouracil)  Image analysis and definition.The sagittal and horizontal planes of T2-weighted images were used.MRI was performed at multiple institutes; despite differenced in the devices used, 1.5-T scanners were mainly used with a slice thickness of 5.0-6.0 mm.Group were defined as follows: (i) clear tumor border: cancer with a clear outline; (ii) unclear tumor border: cancer with an unclear tumor border between the tumor and normal tissue (Fig. 2).Two gynecologists with > 10-year experience of gynecological cancer treatment interpreted the MRI findings; radiologists also analyzed the data.The gynecologists were blinded to the staging.In case of mismatched categorization, the gynecologists reached a consensus through discussion.Tumor diameter was measured using the T2 sagittal plane, and the maximum diameter was utilized in the analysis.
Immunohistochemical staining.Paraffin-embedded blocks were sliced into sections (thickness: 4 μm) and mounted onto slides.Following deparaffinization and antigen retrieval, the tissues were incubated with antibodies against vimentin [1:200; #5741; Cell Signaling Technology (CST)] overnight at 4 °C.After washing with phosphate-buffered saline with Tween 20, the tissues were incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG (CST) for 1 h at 37 °C.Finally, the sections were stained using a 3,3'-diaminobenzidine staining solution and counterstained with hematoxylin.All images were captured using a Zeiss Axio Imager.A1 microscope (Carl Zeiss, Tokyo, Japan).Immunoreactivity was evaluated by three gynecologists who were blinded to clinical data.Staining of ≥ 10% cells denoted positivity for vimentin.

Statistical analysis.
The SPSS version 29.0 software (IBM Corp., Armonk, NY, USA) was used to perform the statistical analysis.The chi-squared test and Student t-test were used to evaluate differences in patient characteristics between the groups.To evaluate the power, post-hoc power analysis was performed.Survival curves were plotted using the Kaplan-Meier method, and statistical significance was assessed using the log-rank test.The associations of patient characteristics and MRI subtypes with OS and recurrence-free survival (RFS) were evaluated based on the univariate and multivariate Cox proportional hazards regression models.Results with p-values < 0.05 denoted statistically significant differences.

Results
Patient characteristics and pathological differences are summarized in Tables 1 and 2, respectively.All patients in the unclear tumor border group (13/13) had LVSI compared with 60% (24/40) in the clear tumor border group.Furthermore, 92% (12/13) of patients exhibited postoperative upgraded staging in the unclear tumor border group compared with 37.5% (15/40) in the clearer border group, and 76.9% (10/13) of patients in the unclear tumor border group exhibited pathological lymph node metastasis versus 15% (6/40) in the clear tumor border group.Because of LVSI and lymph node metastasis, the rate of patients who underwent concurrent chemoradiotherapy as adjuvant therapy was higher in the unclear tumor border group compared with the clear tumor border group (11/13 vs. 24/40, respectively).The recurrence rate also differed between the groups (3/13 vs. 3/40, respectively).The unclear tumor border group had a higher recurrence rate compared with the clear tumor border group (23.1% vs. 7.5%, respectively), both in paraaortic lymph nodes and parenchymal organs.
Table 3 shows the univariate and multivariate analyses for overall survival.Correlation analysis excluded LVSI, lymph node metastasis, and postoperative treatment because they were correlated with tumor boundaries.The tumor border finding was an independent poor prognostic factor for overall survival.
Figure 3 shows the OS and RFS stratified by the type of tumor border.Although there was no significant difference in RFS (p = 0.0847), the unclear tumor border group was associated with a significantly poorer OS versus the clear tumor border group (p = 0.0062).
Figure 4 shows the immunohistochemical findings for vimentin.Based on the results, 37 of the 53 patients exhibited positivity for vimentin expression.As shown in Table 2, positivity for vimentin was significantly higher in the unclear tumor border group compared with the clear tumor border group (p = 0.045).

Discussion
The results of this study showed that patients with early-stage cervical cancer, for whom operation is considered a treatment option, can be classified using MRI into two groups based on the clarity of the tumor border.Moreover, it was shown that this classification may have prognostic importance.Tumor shape has been reported as a prognostic factor in other malignancies, such as breast cancer, bladder cancer, and meningioma [8][9][10] .To the best of our knowledge, this study is the first to examine the relationship between tumor border and prognosis in cervical cancer.MRI showed a significantly better diagnostic performance than clinical assessment for both overall staging and evaluation of prognostic factors 11 .As demonstrated in earlier reports, prognostic factors for cervical cancer are tumor size, patient age, stage, lymph node involvement and location, LVSI, histological type, and tumor grade [12][13][14] .However, some of these factors cannot be assessed unless the patient undergoes surgery.The current study revealed that the tumor border correlates with incidence of LVSI, postoperative upgrade in staging, lymph node metastasis, and distant metastasis.Therefore, these findings could be used to identify cases that would be benefited from radiotherapy and chemotherapy without surgery.Additionally, this evidence may facilitate the process of treatment planning for the preservation of fertility or ovarian function.In patients in whom lymph node metastasis is detected after radical trachelectomy or ovarian-sparing hysterectomy, concurrent chemoradiotherapy is mandatory, and efforts for the preservation of fertility or ovarian function must be abandoned.Hence, radical surgery instead of conservative surgery would be recommended for patients with an unclear tumor border.
Vimentin is a major constituent of the intermediate filament family.It is mainly expressed in mesenchymal cells, and play critical roles in cell adhesion, migration, and signaling 15 .In cancer, vimentin is used as a marker of EMT.EMT is a critical process for cancer metastasis 16 .It has been reported that vimentin is involved in various types of cancer.In prostate cancer, vimentin expression was mainly detected in poorly differentiated tumors and metastatic lesions 17 .In hepatocellular carcinoma, expression of vimentin was mainly associated with metastasis 18 .In non-small-cell lung cancer, vimentin overexpression was identified as an independent prognostic indicator 19 .In cervical cancer, Gilles et al. reported a clear association between vimentin expression and metastatic progression.This conclusion was based on the detection of vimentin in all invasive carcinomas and lymph node metastases, but not in cervical intraepithelial neoplasia 3 (CIN3) lesions 20 .Moreover, Lin et al. reported that vimentin expression is considered as an independent prognostic factor in cervical cancer 7 .Collectively, the currently available data suggests that cervical cancers with unclear tumor borders are associated with higher vimentin expression, lymph node metastasis rate, incidence of LVSI, and recurrence rate versus tumors with There are several limitations in the present study.Firstly, the number of patients with identifiable lesions by MRI who underwent operation was small; therefore, further investigation is warranted to validate the present data.Secondly, due to differences in the prognoses of squamous cell carcinoma and adenocarcinoma, this study focused only on the former type of cancer 21 .Hence, the present findings may not be generalizable to all cervical cancers.Finally, further investigation is warranted to elucidate the mechanism underlying the relationship between vimentin expression and unclear tumor borders.

Figure 1 .
Figure 1.Flow chart of patient selection.MRI, magnetic resonance imaging.

Figure 2 .
Figure 2. Example of classification based on the tumor border using magnetic resonance imaging.(a,b) A 60-year-old patient with a clear tumor border.The solid line indicates the clear borderline of the tumor.(c,d) A 46-year-old patient with a clear tumor border.The solid line indicates the clear borderline of the tumor.(e,f) A 29-year-old patient with an unclear tumor border.The solid line indicates the clear borderline, while the dotted line indicates the unclear borderline of the tumor.(g,h) A 35-year-old patient with an unclear tumor border.The solid line indicates the clear borderline, whereas dotted line indicates the unclear borderline of the tumor.

Figure 3 .
Figure 3. Progression-free survival (a) and overall survival (b) stratified according to the tumor border.

tumor border (n = 40) Unclear tumor border (n = 13) p value
clear borders.The potential implication of this study is that cervical cancer with unclear tumor border have a high expression of vimentin and enhanced EMT characteristics such as metastatic potential and invasiveness, and therefore have a high risk of upstaging and lymph node metastasis, and of recurrence and poor prognosis.

Table 2 .
Clinical and tumor pathological characteristics in each group.LVSI lymphovascular space invasion.*Pearson's Chi-square test.

Table 3 .
Uni-and multivariate Cox proportional hazard analysis with overall survival.MRI magnetic resonance imaging, HR hazard ratio, CI confidence interval.