Medical records of 148 patients who underwent intracavitary brachytherapy for locally advanced cervical cancer (International Federation of Gynecology and Obstetrics [FIGO] stage IB3, IIA2, IIB and above) from October 2018 to December 2021 were retrospectively reviewed. Interstitial brachytherapy was not performed during this time period. Twenty-four patients without follow-up record were excluded. Seven patients with induction chemotherapy and 10 patients without concurrent chemotherapy were also excluded. Two patients with incomplete brachytherapy and one patient treated with both image-guided and conventional planning were ineligible for this study. The remaining 104 patients were included in the analysis. Among them, 50 patients (48.1%) underwent IGBT, and 54 patients (51.9%) underwent CBT.

All patients were treated with pelvic EBRT and concurrent cisplatin-based chemotherapy. Dose-fractionation schedule of pelvic EBRT was 45 to 50.4 Gy with 1.8 Gy of dose per fractions. Most EBRTs were planned by intensity-modulated radiation therapy (IMRT) technique. Twenty patients (19.2%) were treated by 3D conformal radiotherapy. Target volume delineation was based on previously published consensus guidelines [11]. Boost dose was delivered for gross metastatic lymph node. Parametrial or cervical tumor boost was performed in discretion of treating radiation oncologist or by protocol of other participating trials. Median total EBRT doses to gross lymph node, parametrium, and cervical tumor of the patients with boost to each site were 56.0 Gy (range, 48.6 to 64.0 Gy), 54.0 Gy (range, 50.0 to 55.0 Gy), and 50.5 Gy (range, 50.0 to 54.6 Gy), respectively. Paraaortic lymph node area was irradiated when suspicious metastatic common iliac or paraaortic lymph node was presented. Concurrent chemotherapy regimen was weekly cisplatin or tri-weekly cisplatin. Nine patients (18.0%) from the IGBT group enrolled in the randomized phase III CALLA trial, and durvalumab or placebo was administered during and after chemoradiation [12]. Fifty-eight patients (55.8%) underwent EBRT with concurrent chemotherapy at other institutions using similar principles of target delineation and dose prescription for radiotherapy. Thirteen patients (12.5%) underwent lymph node sampling or dissection on pelvic or paraaortic lymph node area before starting chemoradiation. The decision of lymph node surgery was made by the referring gynecologist when there was clinical suspicion of lymph node metastasis and exact staging was needed for decision of treatment modalities. Patients who were considered as unresponsive at 3 months after completion of chemoradiation by the referring gynecologist underwent additional treatment. Adjuvant chemotherapy was administered to two patients (1.9%) and total abdominal hysterectomy was performed for one patient (1.0%).

High-dose-rate (¹⁹²Ir) intracavitary brachytherapy with dose prescription of 25 to 30 Gy in 4 to 6 fractions was performed for all patients. Response to EBRT and disease extent were re-evaluated by diagnostic MRI scan before brachytherapy planning. Applicator (tandem and ovoids) insertion and planning were performed by an experienced radiation oncologist (H.-C.K.). Allocation of patients to IGBT or CBT was in discretion of the treating radiation oncologist, and mainly based on the availability of staff and 3D image-compatible devices. For CBT, simulation and the first treatment were performed on the same day with fluoroscopy. Point A, Point B, bladder point, and rectal point were set based on the definition from International Commission on Radiation Units and Measurements (ICRU) Report 38 [13], and the dose was prescribed to point A. For IGBT, the patient underwent CT simulation scan with applicator in place. MRI-compatible Fletcher-type applicator and hybrid applicator (Venezia, Elekta AB, Stockholm, Sweden) were used. MRI-based treatment planning with the applicator in place was performed for 30 patients (60.0%). MRI simulation was performed after CT simulation was completed. The patient transferred to MRI room with applicator in place and underwent MRI scan. A 0.35 T MRI installed in MRI-guided radiation therapy system (MRIdian, ViewRay Inc., Oakwood Village, OH) was used for the most patients, as this treatment system is capable of MRI simulation scan separated with actual treatment sessions. One patient underwent 1.5 T diagnostic MRI simulation scan. Delineation of high risk (HR) clinical target volume (CTV) and intermediate risk CTV was according to recommendations of the Gynaecological GEC-ESTRO working group [14]. The dose was prescribed to D_90% of HR-CTV. Rectum, sigmoid colon, and bladder were delineated as OARs.

To calculate the cumulative dose, dose from EBRT was assumed to be a prescribed dose to the pelvis. The equivalent dose in 2 Gy per fraction was calculated using α/β of 10 for tumor (EQD2₁₀) and 3 for OARs (EQD2₃). For CBT, Cumulative EQD2₃ of bladder and rectal point was intended to be limited to 75 Gy. For IGBT, cumulative EQD2₃ D_2cc was limited to 75 Gy for rectum and sigmoid colon and 90 Gy for bladder. A treating radiation oncologist reviewed the plan and decided whether to adjust dose-fractionation scheme or affirm the existing plan when these doses exceeded the limitation. Dose calculation and treatment planning were performed by Oncentra Brachy (Elekta AB, Stockholm, Sweden). For every treatment session, fluoroscopy was used to check the location of the applicator.

Clinical endpoints of this study were LC, locoregional control (LRC), progression-free survival (PFS), and overall survival (OS). An LC event was defined as a recurrence in the treated cervix or parametrium, while LRC event was defined as a recurrence in the pelvic or paraaortic regional lymph node area. For PFS, the event was defined as any recurrence or death of the patient. For OS, the event was defined as death of the patient. These clinical outcomes were measured from the date of the completion of brachytherapy for each defined event. Rates were calculated using the Kaplan-Meier method. Clinical endpoints of two groups were compared by log-rank test. Univariate and multivariate analyses were performed for LC, LRC, PFS, and OS to identify covariates potentially affecting clinical outcomes. Statistically significant (p < 0.05) or marginally significant (p < 0.1) variables in univariate analysis and brachytherapy technology (IGBT vs. CBT) were incorporated into multivariate analysis using the Cox proportional hazards model.

New occurrence or worsening of urinary or lower gastrointestinal (GI) toxicities were recorded and graded using the Common Terminology Criteria for Adverse Events (CTCAE) ver. 5 [15]. Toxicities occurred more than 3 months after completion of brachytherapy were considered as late toxicities and included for the analysis. Symptoms suspected of being caused by other reasons such as recurred tumor (e.g., rectovaginal fistula with cervical recurrence) were excluded. Crude rates of toxicity were compared between two groups by chi-square test. Grade ≥ 3 toxicities were considered as severe toxicities. Actuarial rate of occurrence of severe toxicities of two groups were calculated by the Kaplan-Meier method and compared by log-rank test. Dose-response analysis was done separately by treatment groups as the methodology of reporting dose to OARs was different between IGBT and CBT. For IGBT, relationship between cumulative EQD2₃ D_2cc of bladder and occurrence of any, grade ≥ 2, or ≥ 3 urinary toxicity was analyzed with a logistic regression model. The relationship between cumulative EQD2₃ D_2cc of rectum/sigmoid and occurrence of any, grade ≥ 2, or ≥ 3 lower GI toxicity was analyzed similarly. For CBT, dose-response relationship was analyzed with the same method. However, cumulative EQD2₃ D_2cc of bladder was substituted by cumulative EQD2₃ of bladder point and cumulative EQD2₃ D_2cc of rectum was substituted by cumulative EQD2₃ of rectal point. Receiver operative characteristic (ROC) analysis was performed for relationships that showed statistical significance (p < 0.05) or marginal significance (p < 0.1) in the logistic regression model. Optimal cumulative dose threshold was found using the Youden index method. Categorical variables were compared between two groups by chi-square test with or without Yates’s continuity correction and numerical variables were compared by t-test. All statistical analyses were performed using R ver. 4.2.0 (The R Foundation for Statistical Computing, Vienna, Austria).

Patient characteristics and treatment specifics are summarized in Table 1. The median follow-up period of all patients was 18.33 months (range, 3.25 to 38.43 months) from the completion of brachytherapy. There was no difference in follow-up period between the two groups. There were more patients referred from other institutions after chemoradiation in the CBT group (81.5% vs. 28.0%, p < 0.001). Baseline clinical factors were generally similar between the two groups. However, patients who underwent CBT were older (median, 59 years vs. 49.5 years, p < 0.001). Most patients in both groups had FIGO stage IIIC disease. A total of six patients (5.8%) had stage IVA disease. Five of them had disease spread to the bladder. One patient from the CBT group had a large tumor with adhesion to urethra, small bowel, and rectum. Three patients (2.9%) had distant metastasis at presentation. Two of them had supraclavicular lymph node metastasis and one patient from CBT group had suspected peritoneal seeding nodules in pelvis. These metastatic lesions were treated by EBRT.

Several differences between treatment specifics were observed. More patients in the CBT group underwent EBRT with pelvic dose fractionation of 50.4 Gy in 28 fractions (18.5% vs 4.0%, p=0.013). Other dose fractionation schemes were 44 Gy in 22 fractions and 46 Gy in 23 fractions. Parametrial boost was more frequent in IGBT group (52.0% vs. 13.0%, p < 0.001) partly due to the EBRT protocol of the CALLA trial which requires parametrial boost of 54 Gy in 27 fractions and boost to gross lymph node of 58 Gy in 29 fractions. In the IGBT group, tri-weekly cisplatin was more often administered as a concurrent chemoradiation (20.0% vs. 1.9%, p=0.007) with more patients undergoing lymph node surgery (22.0% vs. 3.7%, p=0.013) due to institutional preference. Dosimetric parameters of brachytherapy are summarized in Table 2. Median average dose of bilateral point A of the IGBT group was 23.25 Gy (range, 14.38 to 44.49 Gy). The average dose of bilateral point A was lower than prescribed dose in the IGBT group (p < 0.001), indicating that prescribing the same dose to point A instead of HR-CTV may have been resulted overtreatment in general.

We also compared patient characteristics between MRI- and CT-based IGBT. There were trends for more FIGO stage III disease (86.7% vs. 60.0%, p=0.088) and larger tumors (median 5.8 cm vs. 5.1 cm, p=0.054) in patients underwent MRI-based IGBT. Most (n=28, 93.3%) of the patients underwent MRI-based IGBT had dose-fractionation regimen of 30 Gy in five fractions, while 12 patients (60.0%) underwent CT-based IGBT had such regimen. Patient characteristics of MRI- and CT-based IGBT were summarized in S1 Table.

Patterns of recurrence are summarized in Table 3. Crude rates of local recurrence, regional recurrence, and distant metastasis did not show significant difference. However, the CBT group showed higher crude rate of concurrent pelvic and paraaortic lymph node metastasis (0.0% vs. 11.1%, p=0.045). Kaplan-Meier curves of treatment outcomes are illustrated in Fig. 1. The 1-year and 2-year LC rates were 93.6% and 93.6% in the IGBT group and 94.2% and 86.5% in the CBT group, respectively, showing no statistically significant difference between the two groups (p=0.584). The 1-year and 2-year LRC rates were 83.4% and 78.8% in the IGBT group and 78.3% and 69.1% in the CBT group, respectively, showing no significant difference between the two groups (p=0.408). PFS rates at 1-year and 2-year were 77.3% and 72.7% in the IGBT group and 72.3% and 69.8% in the CBT group, respectively. OS rates at 1-year and 2-year were 93.8% and 88.4% in the IGBT group and 96.0% and 83.6% in the CBT group, respectively. No significant difference in PFS (p=0.777) or OS (p=0.475) was observed between the two groups. There was no difference of treatment outcomes between MRI- and CT-based IGBT (LC, p= 0.770; LRC, p=0.929; PFS, p=0.457; OS, p=0.347).

Results of univariate and multivariate analyses are summarized in S2 Table. In the multivariate analysis, lymph node dissection was associated with worse LC (hazard ratio, 13.24; 95% confidence interval [CI], 2.989 to 58.62; p=0.001). Distant metastasis at diagnosis was associated with worse PFS (hazard ratio, 4.738; 95% CI, 1.057 to 21.24; p=0.042). Initial tumor size was associated with worse LRC (hazard ratio, 1.288; 95% CI, 1.017 to 1.632; p=0.036) and OS (hazard ratio, 1.459; 95% CI, 1.090 to 1.952; p=0.011). Brachytherapy technology (IGBT vs. CBT) did not show statistically significant correlation with treatment outcomes in the multivariate analysis.

Overall, six urinary toxicities (12.0%) and 12 lower GI toxicities (24.0%) were reported in the IGBT group, and 12 urinary toxicities (22.2%) and 13 lower GI toxicities (24.1%) were reported in the CBT group. For severe toxicities with grade ≥ 3, there were three lower GI toxicities (6.0%) and 0 urinary toxicity (0.0%) in the IGBT group vs. nine lower GI toxicities (16.7%) and four urinary toxicities (7.4%) in the CBT group. All three patients with severe toxicities in the IGBT group underwent CT-based brachytherapy planning. There was no statistically significant difference in crude rate of overall or severe urinary/lower GI toxicities. Detailed toxicity profile is summarized in Table 4. The median time to occurrence of any urinary or lower GI toxicity was 11.93 months (range, 3.42 to 25.73 months). Treatment, details of toxicity, and disease status of severe toxicity cases are summarized in S3 Table. It should be noted that seven among 12 patients (1/3 in IGBT group, 6/11 in CBT group) with severe late lower GI toxicities had evidence of bleeding, perforation, or fistula involving GI tract above the rectum.

Actuarial rates of severe toxicities are illustrated in Fig. 2. The 1-year and 2-year rates for the occurrence of lower GI toxicities with grade ≥ 3 were 4.5% and 4.5% in the IGBT group and 17.8% and 20.4% in the CBT group, respectively, showing statistically significant differences between the two groups (p=0.046). For severe urinary toxicities, the 1-year and 2-year rates were 0.0% and 0.0% in the IGBT group and 7.1% and 13.7% in the CBT group, respectively. Marginally significant difference was observed between the two groups (p=0.061). The 1-year and 2-year actuarial rates of any grade ≥ 3 toxicities were 4.5% and 4.5% in the IGBT group and 17.8% and 25.7% in the CBT group, respectively, showing significant difference between the two groups (p=0.030). There was a trend for less any grade ≥ 3 toxicities in MRI-based IGBT when compared with CT-based IGBT (2-year rate, 0.0% vs. 10.6%; p=0.093).

Logistic regression was performed between occurrence of toxicity and cumulative EQD2₃ D_2cc of corresponding OAR in the IGBT group. Occurrence of grade ≥ 2 lower GI toxicities and cumulative EQD2₃ D_2cc of sigmoid colon was significantly correlated (odds ratio [OR], 1.158; 95% CI, 1.012 to 1.326; p=0.033), while the relationship between grade ≥ 2 lower GI toxicities and cumulative EQD2₃ D_2cc of rectum (OR, 1.160; 95% CI, 0.997 to 1.350; p=0.055) and that between grade ≥ 2 urinary toxicities and cumulative EQD2₃ D_2cc of the bladder (OR, 1.049; 95% CI, 0.996 to 1.104; p=0.069) showed marginal significance. ROC analysis showed that cumulative EQD2₃ D_2cc of rectum (area under the curve [AUC], 0.7542; 95% CI, 0.5975 to 0.9108) and sigmoid colon (AUC, 0.8173; 95% CI, 0.6777 to 0.9569) were predictive for grade ≥ 2 lower GI toxicities. Cumulative EQD2₃ D_2cc of bladder did not show significance in ROC analysis (AUC, 0.6711; 95% CI, 0.3558 to 0.9864). The optimal threshold of cumulative EQD2₃ D_2cc of rectum by Youden method was 69.7 Gy with a sensitivity of 85.7% and a specificity of 74.4%. The optimal threshold of cumulative EQD2₃ D_2cc of sigmoid colon was 70.8 Gy with a sensitivity of 85.7% and a specificity of 76.7%.

For the CBT group, logistic regression between cumulative EQD2₃ of bladder point and occurrence of urinary toxicities did not show a significant or marginally significant correlation. The relationship between cumulative EQD2₃ of rectal point and lower GI toxicities with grade ≥ 2/3 showed a marginal significance (OR, 1.037; 95% CI, 0.999 to 1.077; p=0.054). Setting the endpoint as grade ≥ 2 or ≥ 3 had the same effect because late lower GI toxicity with grade 2 was not reported in the CBT group. In ROC analysis, cumulative EQD2₃ of rectal point was predictive for lower GI toxicities with grade ≥ 2/3 (AUC, 0.7111; 95% CI, 0.5206 to 0.9016). The optimal threshold of cumulative EQD2₃ of rectal point was 101.0 Gy with a sensitivity of 55.6% and a specificity of 84.4%. ROC curves are illustrated in S4 Fig.