Development and validation of a novel scoring system to predict the risk of uterine perforation during intracavitary brachytherapy for cervical cancer

Cervical cancer is the second most common malignancy among women in India and the most common gynecological malignancy [1]. According to the Globocan 2020 data, the estimated age standardized incidence rate of cervical cancer among the Indian population is eighteen per one lakh population and the estimated age standardized mortality rate is sixty-one per one lakh population [2]. External beam radiation therapy (EBRT) followed by brachytherapy (BT) is the established standard of care in patients with cervical cancer International Federation of Gynaecology and Obstetrics (FIGO) stage IB3 to IVA [3]. In locally advanced disease, the treatment of central disease (cervix, vagina, and medial parametria) depends on the total dose received with EBRT and BT [4, 5, 6]. Intracavitary BT involves insertion of a uterine tandem into the cervix and uterus. Uterine perforation is a potential complication during tandem insertion, especially if the cervical architecture is distorted due to the tumor itself or due to response to EBRT. This can result in the tandem being inserted in a false passage through the vaginal fornices or through the myometrium of the uterus or it may also result in uterine perforation through the fundus if the uterus is atrophic. Sub-optimal insertion of the tandem can result in sub-optimal dose to the tumor while simultaneously leading to increased dose to nearby organs-at-risk. A perforation rate of 1.75%–15% (average=8%) has been reported in applicator insertions worldwide [7, 8, 9, 10]. Post-insertion computed tomography (CT) enables identification of applicator, its position, suboptimal insertions, and uterine perforation [7]. Avoiding uterine perforation is always better than early detection to avoid complications such as bleeding, infection, and treatment delay. Even with modern optimisation techniques, an application with a perforated uterus can seldom be salvaged and will invariably lead to abandonment of the procedure. This will in turn have an adverse impact on the overall treatment time which may lead to poor pelvic control in these patients [11].

Real-time ultrasound has been recommended by several authors to guide applicator placement [12, 13]. An internal audit done in our department in the year 2017–2018, revealed a 14% (26 out of 180) perforation rate when ultrasound guidance was not used and 5.6% (7 out of 125) with judicious use of ultrasound guidance during BT. Routine use of ultrasound for all applications leads to prolongation of procedure time which is undesirable when the theatre time and dedicated operating room space is limited. The aim of the study was to develop a novel scoring system to predict the risk of uterine perforation and stratify patients according to the risk score and to validate the scoring system on patients undergoing BT in the year 2021.

Data was collected from electronic medical records, treatment planning system and BT register from January 2018–December 2021 at a single institute following approval from the Institutional Review Board (Minute No: 12845 Dated: 1.05.2020). Patients with histopathologically confirmed carcinoma of the cervix diagnosed between 2018 and 2021 in accordance with the 2018 FIGO [14], who had completed EBRT with or without concurrent chemotherapy followed by high-dose-rate CT based intra cavitary BT were included. During January 2018 and December 2020, a total of 784 applications (Fig. 1) of intracavitary BT were done. For the test cohort we selected the first fifty consecutive applications with uterine perforation and the first fifty consecutive applications without uterine perforation. The most common reasons for perforation were identified based on analysis of CT images with the applicator in-situ and previously published literature with common causes of uterine perforation such as those published by Segedin et al. [7] who mentioned “age more than 60 years, necrotic cervical tumour, cervical polyp, submucosal fibroid, stenosis, or distortions of cervical canal (especially due to prior conization), retroflexed or extremely anteflexed uterus” as factors associated with increased incidence of uterine perforation. Using these previously published factors and our clinical experience from CT based BT we developed a scoring system (Table 1). The difference in the individual criteria and overall score were then compared with χ² test. Additionally on the test cohort, for statistical model evaluation, receiver operator characteristic (ROC) curves were generated and area under curve (AUC) of the individual scoring criteria and the overall score was calculated. Statistical analysis was done using SPSS Version 21.0 (IBM, Armonk, NY, USA).

Fig. 1
Flow diagram of test and validation cohort.
ROC, receiver operator characteristic.

• Click for larger image
• Download as PowerPoint slide

Table 1
Proposed scoring system for perforation risk during brachytherapy

• Click for larger image
• Click for full table
• Download as Excel file

For validation, the scoring system was applied on all newly diagnosed patients (n=120) who underwent BT in 2021 and χ² test was applied to differentiate those who had perforation (n=9) versus those who did not have perforation (n=111). ROC curves were generated, and AUC, sensitivity and specificity of the individual scoring criteria and the overall score was calculated.

The factors considered for the perforation risk index included cervical os on assessment for BT, the response after EBRT which was further divided as good response (<2 cm residual), Moderate response (2–4 cm residual), Poor response (>4 cm residual), cervico-uterine angle on imaging, the uterine position, uterine length, any additional uterine pathology, and previous history of perforation.

The general characteristics of the patients in the 2 groups of the test cohort were well balanced and this has been depicted in Table 2. In the test cohort, the mean perforation risk score was 10.16 (range=7–14) among patients who had perforation as against 5.92 (range=5–8) among patients who did not have perforation. Also, 48 out of 50 (96%) patients in the perforation cohort had an overall score of 8 or more, thus classified as high risk as opposed to 5 patients of the 50(10%) patients in the no perforation cohort. Additionally, EBRT response, uterine version, cervico-uterine angle, identification of cervical os on pre-procedure assessment and uterine length were significantly different between the patients who had perforation when compared to those who did not (p-value <0.0001). Presence of an additional uterine anomaly was also significant (p=0.013). History of previous perforation was not significant in our cohort of patients (p=0.315). To assess the performance of the scoring system on the Test cohort ROC curves were generated (Fig. S1). and AUC, sensitivity, and specificity of the individual criteria in the scoring system and the overall scores were calculated (Table 3).

Table 2
Characteristics of the 2 groups in the “Test Cohort”

• Click for larger image
• Click for full table
• Download as Excel file

Table 3
Statistical model on the test cohort: AUC, sensitivity and specificity of the various criteria and overall scoring system

• Click for larger image
• Click for full table
• Download as Excel file

In the validation cohort we included 120 consecutive patients who underwent BT in 2021, the clinical characteristics of these patients have been depicted in Table 4. Of the 120 patients, 111 did not have perforation (92.5%) and nine had perforation (7.5%). The mean perforation index score was 6.9 (5–10) among those who did not have perforation as against 9.33 (7–11) among those who had perforation. Among the patients who had perforation, 7 out of 9 (77.78%) had score more than 8, thus classified as high perforation risk. In the remaining 111 patients,34 (30.6%) had a score of 8 or more.

Table 4
Clinical characteristic of the patients in the “Validation Cohort”

• Click for larger image
• Click for full table
• Download as Excel file

Among the individual criteria in the perforation risk index response to EBRT (Fig. S2) (p=0.036), uterine version (Fig. S3) (p=0.007), cervico uterine angle (uterine flexion) (Fig. S4) (p=0.020), identification of cervical os at BT assessment (p=0.0001), and the total score (p=0.0001) were statistically significant. However, uterine length (p=0.779) and additional uterine anomaly (Fig. S5) (p=0.966) were not statistically significant.

Based on the significant findings of the statistical model from the test cohort, we generated ROC curves (Fig. S6) to validate the scoring criteria and the total score. Cervical os at BT, response at the time of BT, uterine version at the time of BT and the overall score as a sum of all the criteria were validated as significant predictors of perforation. Uterine length, cervico uterine angle, and presence of additional uterine anomaly were not significant (AUC <0.5) in predicting the risk of perforation in this cohort of patients (Table 5).

Table 5
AUC, sensitivity and specificity of the various criteria and overall scoring system for the “Validation Cohort”

• Click for larger image
• Click for full table
• Download as Excel file

BT is an integral component of curative treatment in cervical cancer [15]. Historically BT planning was done with orthogonal radiographs. Adequacy of application was based only on the position of the tandem and ovoids with respect to each other and the bony pelvic anatomy. With the advent of cross-sectional imaging-based BT it is possible to visually appreciate the position of the applicators with respect to uterine anatomy. Hence, uterine perforation is being picked up more often. Mahantshetty et al. [16] reported that perforation during BT is commonly seen as a false passage through the fornices, missing the uterine canal or even through the wall of the uterus. Further they recommended that in cases of perforations through the fornices or through the anterior/posterior walls of the uterus, the BT treatment should be abandoned and in cases where perforation occurs through the uterine cavity and fundus, BT could be delivered only after adequate modification of source loading pattern, avoiding loading at the top of the tandem. It is also recommended to monitor these patients closely with abdominal girth charting and treatment with a course of antibiotics, most commonly a combination of Ciprofloxacin and Metronidazole.

Among published literature uterine perforation is a common complication with reported rates varying from as low as 1.75% to as high as 15%. Granai et al. [17] reported frank uterine perforation in 10% of their application (5/50). Matsuyama et al. [18] reported a 10% rate of uterine perforation (6 of 61 patients) despite metal clips placed on the uterine serosa to assess tandem position. Rotmensch et al. [19] reported uterine perforation in 6 of 20 implants with the use of ultrasound guidance during or after the application. Jhingran and Eifel [20] reported results from a centre with high volume of experience and the incidence of perforation was 2.79% (113/4,043).

Preventing uterine perforation and an optimal application not only reduces the risk of complications such as bleeding, infections and need for repeated applications and anaesthesia but also results in better disease outcomes by delivering appropriate dose to the tumor and reducing dose to normal structures [5]. The use of real-time ultrasound to guide applicator placement would help in significantly reducing the incidence of uterine perforation. In a case report and review of literature, Small et al. [13] concluded that, it is exceedingly difficult to detect uterine perforation clinically and hence recommended use of ultrasound guidance in all applications. Granai et al. [17] reported on the use of postoperative ultrasound in fifty consecutive applications (28 patients). In 34%, the tandem was found to be sub optimally positioned, it penetrated the myometrium in 12% and frank perforation in 10%. Wong and Bhimji [21], Tanaka et al. [22] also reported on the utility of intraoperative transabdominal ultrasound and concluded it as a useful adjunct, allowing to complete the planned treatment even in difficult cases. Davidson et al. [23] reported that the use of transabdominal US influenced the length and angle of intrauterine tandem chosen in close to half of all the patients undergoing BT (49%).

In practice however, the use of real time ultrasound during applicator placement requires a radiologist or a radiation oncologist with knowledge of interpreting and performing an ultrasound, requires additional workforce, and an ultrasound machine dedicated for intraoperative assistance. This might be challenging in centres without a dedicated ultrasound machine and in centres with high patient load, especially if it must be done for each application.

Risk adapted image guidance for BT applicator insertion has been described previously. Barnes et al. [24] in 2007 reported the incidence of perforation based on physician concern which was in turn based on age of the patient (>60 years) and tumor size at the time of BT. They reported an overall incidence of perforation of 13.7% on post application CT. It was interesting to note that the incidence of perforation was as high as 8.2% even when the radiation oncologist was clinically confident of correct tandem placement. Segedin et al. [7] in a report of 5-year experience of image guided BT from their centre identified age more than 60 years, necrotic cervical tumor, cervical polyp, submucosal fibroid, stenosis, or distortions of cervical canal (especially due to prior conization), retroflexed or extremely anteflexed uterus as factors associated with increased incidence of uterine perforation.

Al-Hammadi et al. [12] built on these findings and utilised offline MRI imaging to help in the intrauterine applicator placement in patients with 2 or more risk factors (necrotic cervical tumor, cervical polyp, submucosal fibroid, stenosis, or distortions of cervical canal, retroflexed or extremely anteflexed uterus). They reported zero incidence of perforation with the use of MRI imaging.

The findings of the present study are in concordance with the previously reported literature, and we found that identification of cervical os, response to EBRT, uterine position, and overall score were significant predictors of uterine perforation. To the best of our knowledge this is the first report which structures the risk factors into a scoring system, which can be readily applied to the pre BT imaging to aid in the judicious use of real time ultrasound guidance during BT applications.

The strength of this study is a novel structured risk stratification scoring system, which was also validated in a separate cohort of patients.

The limitations of this study are a small sample size from a single institution further constrained by the coronavirus disease 2019 pandemic. Further study among a larger sample size will aid in overcoming the selection bias.

In conclusion, the novel scoring system is a good predictor of perforation risk during BT and pre-emptive scoring at the time of assessment helps in optimising the need for ultrasound guidance during procedure.

Fig. S1

ROC curves on the test cohort.

Fig. S2

Response to EBRT. (A, B) Pre and post EBRT MRI with good response, (C, D) Pre and post EBRT MRI with poor response.

Fig. S3

Uterine version. (A) Retroverted uterus with tandem perforation, (B) Application done under ultrasound guidance.

Fig. S4

Uterine flexion/cervico-uterine angle. (A) Acutely flexed uterus with tandem perforation, (B) Acutely flexed uterus with tandem inside.

Fig. S5

Large sub serosal fibroid compressing the uterine canal. T2 weighted MRI (A) Axial, (B) saggital, and (C) coronal.

Fig. S6

ROC curves on the validation cohort.