Purpose
Cervical cancer is the second most common cancer among Indian women, and it ranks fourth worldwide according to recent data [1]. The definitive management of locally advanced cervical carcinoma (LACC) consists of pelvic external beam radiotherapy (EBRT) with concurrent chemotherapy, with or without para-aortic irradiation, followed by intracavitary and/or interstitial brachytherapy. Brachytherapy (BT) remains an integral component of the curative management of cervical cancer, despite advances in techniques, such as intensity-modulated radiotherapy or stereotactic body radiotherapy [2-4]. Traditionally, cervical brachytherapy planning has been performed based on two-dimensional orthogonal films normalizing the prescription dose to point A [5].
The utilization of three-dimensional imaging, such as computed tomography (CT) in brachytherapy application and planning, has many advantages. It helps avoid uterine perforation, ensures accurate applicator geometry, provides volumetric coverage of target volumes, and enables evaluation of doses to critical organs at risk (OARs), such as the rectum and bladder, resulting in favorable clinical outcomes and toxicity profiles [6, 7]. Most international centers, including Indian, predominantly use CT scan for image-guided brachytherapy (IGBT) planning in carcinoma of the cervix. However, CT scans are less sensitive than magnetic resonance imaging (MRI) in delineating high-risk clinical target volume (HR-CTV) [8]. Hence, MRI-based IGBT is considered the gold standard imaging modality for cervical cancer brachytherapy. The Groupe Européen de Curiethérapie and European Society for Radiotherapy and Oncology (GEC-ESTRO) have provided recommendations and contouring guidelines for MRI-based IGBT to aid physicians [9-11].
Although MRI-based IGBT is the standard approach, its use is limited in many centers. Most radiotherapy departments do not have an in-house MRI scanner, while other logistical challenges, such as increased scanning time, requirement for special MRI-compatible applicators, and increased treatment cost, make routine utilization of MRI-based IGBT unfavorable. These factors are particularly challenging in implementing MRI-based IGBT in high-volume or resource-constrained settings, such as in most parts of India.
On the other hand, with widely available CT-based planning, some studies have shown that HR-CTV D90 (dose received by 90% of the volume) on CT is significantly lower than on MRI [12, 13], while others reported no statistically significant difference between CT- and MRI-based plans [14]. Therefore, it is important to investigate whether an improvised CT-based IGBT approach can substitute for MRI-based IGBT. We designed this prospective study with the pragmatic approach of using MRI-based IGBT at the first fraction and CT-based IGBT for subsequent fractions. The primary objective of our study was to compare dose-volume parameters for the target volume and OARs (the rectum, bladder, and sigmoid) achieved with MRI- vs. CT-based IGBT plans. The secondary objectives were to evaluate late toxicities, loco-regional control (LRC), disease-free survival (DFS), and overall survival (OS).
Material and methods
This single arm, prospective, interventional trial (CTRI/2019/01/01691) enrolled a total of 22 patients with histologically proven squamous cell carcinoma (SCC) of the cervix with stages IIA-IIIB (FIGO 2010 staging) between January 2019 and January 2020. Patients younger than 25 years old or older than 60 years old, those with stage IV disease, a positive para-aortic node on imaging, adenocarcinoma histology, pregnant or lactating women, patients suffering from active HIV, hepatitis B/C infection and patients with a history of any other malignancy in the last five years, were excluded from the study. Also, patients unfit for general anesthesia and MRI procedure, or those who did not provide consent, were excluded.
EBRT
All patients underwent contrast-enhanced CT simulation using Siemens Somatom 16-slice CT simulator in the supine position with arms crossed over the chest and a knee rest for immobilization. A bladder protocol was followed with 500 ml of normal saline applied 20-30 minutes prior to CT simulation. An intra-vaginal, perineal radio-opaque marker was placed to aid in delineation of volumes. CT images were then imported to treatment planning system (TPS) for volume delineation and treatment planning. Elective lymph node volume included bilateral common iliac, external iliac, internal iliac, obturator, and pre-sacral lymph nodes contoured as per Taylor et al. [15]. Clinical and imaging findings were used to contour clinical target volume (CTV), which included entire uterus, cervix, bilateral parametrium, and vagina (the extent of vaginal contours was inferiorly dependent on the extent of vaginal involvement), according to Bansal et al. guidelines [16]. Gross nodal disease with a margin of 1 cm was included in CTV. Planning target volume (PTV) was generated by adding a margin of 1-1.5 cm all around CTV, and 7 mm around the elective nodal CTV. PTV was prescribed a dose of 50 Gy in 25 fractions over 5 weeks using three-dimensional conformal radiation therapy. Patients with positive pelvic nodes were further boosted with AP/PA fields with midline shielding to an additional EBRT dose of 5.4 Gy in 3 fractions. All EBRT plans were accepted if > 95% of the PTV received at least 95% of the prescribed dose. Elective para-aortic irradiation was not performed in any patient. Treatment was delivered using Elekta (Infinity, Synergy) linear accelerator, with a multi-leaf collimator (MLC) leaf width of 1 cm at the isocenter. All patients received weekly cisplatin-based concurrent chemotherapy at doses of 40 mg/m2.
Brachytherapy
Each patient underwent MRI at the first intracavitary brachytherapy (ICBT) application, while CT-based ICBT was performed at the second and third applications. Brachytherapy procedure was carried out one week after completion of EBRT. The first brachytherapy application was done under spinal anesthesia and subsequent applications were completed under short general anesthesia. Trans-rectal ultrasound scans were performed before and after procedure, to guide and confirm central tandem placement. MRI/CT compatible Fletcher-Suit applicators were used according to the patient’s cervical anatomy.
All patients underwent non-contrast MRI at the first fraction using a 3.5 Tesla machine with applicators in situ. Just before imaging, the urinary bladder was emptied, and 20 ml of normal saline was injected into the bladder through Foley’s catheter and clamped. T2-weighted fast spin-echo sequences were acquired with pelvic coil, using slice thickness of 3 mm and no inter-slice gap. The imaging data were obtained in DICOM format and imported into the Oncentra system (Nucletron-Elekta). Non-contrast CT scans were performed in all patients at the second (CT-1) and third (CT-2) fractions using Siemen’s Somatom 16-slice CT simulator, after placing CT dummy source catheters in MRI/CT-compatible applicators in situ. Twenty milliliters of diluted contrast agent were injected through Foley’s catheter into the bladder and clamped just before imaging. HR-CTV and OARs (the rectum, bladder, and sigmoid colon) were delineated on MRI [10, 11]. For CT-based plans, HR-CTV was contoured as per Vishwanathan et al. recommendations [12]. The prescription dose was 7 Gy per fraction normalized to point A, and further optimized with manual dwell time optimization with or without graphical optimization, to ensure adequate sparing of OARs without affecting the dose to HR-CTV. Plan acceptance criteria were as follows: the D90 for HR-CTV should be at least > 90% of the prescribed dose, and the cumulative equivalent dose at 2 Gy per fraction (EQD2) D2cc for the bladder, rectum, and sigmoid should be < 90 Gy, < 75 Gy, and < 75 Gy, respectively, according to the American Brachytherapy Society (ABS) references [17].
Dose-volume histogram parameters for HR-CTV and OARs were evaluated. For HR-CTV, D90 was recorded, and for OARs, D0.1cc, D1cc, and D2cc (the minimum doses to the most exposed 0.1, 1, and 2 cc volumes, respectively) were recorded. High-dose-rate (HDR) brachytherapy using iridium-192 was delivered with a remote afterloading microSelectron unit to all patients in the study.
Statistical analysis
Descriptive statistics were employed for patient characteristics. Quantitative data were expressed as mean (standard deviation) or median (range), while qualitative data were presented as proportions. Shapiro-Wilk test was used to check for data normality. Paired t-tests were applied to compare volumes of HR-CTV and OARs, and compare doses with HR-CTV and OARs. A two-tailed p-value of < 0.05 was considered statistically significant for all analyses. SPSS (version 20.0) was used for analysis, whereas Common Terminology Criteria for Adverse Events (CTCAE) version 4.0 was employed for late toxicity analysis. Loco-regional control (LRC) was defined as the time from treatment completion to a loco-regional recurrence. Disease-free survival (DFS) was described as the time from treatment completion to either a loco-regional or distant recurrence. Overall survival was defined as the time from registration in the hospital to death.
Results
The median patients’ age was 50 years (range, 32-60 years), while the median Karnofsky performance scale was 90 (range, 80-90). The stage-wise distribution (FIGO 2010) for IIA, IIB, and IIIB stages were 8 (36.3%), 6 (27.7%), and 8 (36.4%), respectively. Four patients (18.2%) had node positive disease, while the median EBRT dose was 50 Gy. All patients received weekly concurrent chemotherapy with intravenous cisplatin at 40 mg/m2, and the median overall treatment time was 57 days (range, 43-74 days).
The mean HR-CTV volumes on MRI (the first fraction), CT (the second fraction), and CT (the third fraction) were 24.23 cc ±9.76 cc, 27.82 cc ±15.37 cc, and 24.82 cc ±11.14 cc, respectively (Table 1). The volumes of OARs (the bladder and rectum) were comparable between MRI- and CT-based plans. However, CT-based volumes of the sigmoid at the third fraction were statistically larger as compared with MRI-based volumes at the first fraction (MRI: 25.28 cc vs. CT-2: 34.61 cc, p = 0.01), and also larger than CT-based volumes at the second fraction, likely due to inter-fraction variations in sigmoid filling (Table 1).
Table 1
Comparison of volumetric assessment of tumor volume and organ at risk volumes
The mean HR-CTV D90 was significantly higher in MRI-based plans at the first fraction compared with CT-based plans at the second fraction (MRI: 108.86 ±24.21% vs. CT-1: 98 ±23.18%, p = 0.03), but similar to CT-based plans at the third fraction (MRI: 108.86 ±24.21% vs. CT-2: 106.86 ±17.36%, p = 0.69). The mean doses to OARs were comparable between MRI- vs. CT-based plans, as shown in Table 2.
Table 2
Comparative doses to HR-CTV and OARs
The median equivalent dose at 2 Gy per fraction (EQD2) for HR-CTV was 84.7 Gy (range, 75-111 Gy). The median cumulative EQD2 for the bladder, rectum, and sigmoid were 80.2 Gy (range, 71.5-95 Gy), 66.7 Gy (range, 59.9-73.3 Gy), and 64.5 Gy (range, 52.5-79.2 Gy), respectively.
The median follow-up was 65.7 months (range, 5-74 months). During this period, four patients developed recurrences: three patient experienced loco-regional failures, and one had liver and lung metastases without local recurrence. The 5-year LRC, DFS, and OS rates were 85.2%, 80.7%, and 79.0%, respectively. Grade ≥ 2 late gastrointestinal and genitourinary toxicities were observed in 5 (22.7%) and 2 (9%) patients, respectively. One patient (4.5%) developed grade 3 late gastrointestinal toxicity. No grade 4 late toxicity was observed among the patients.
Discussion
Brachytherapy has been an integral component of the definitive management of cervical cancer management for several decades. In the last 20 years, the emergence and incorporation of IGBT have played an important role in improving the clinical outcomes in LACC. The French STIC study was the first prospective, multi-institutional, non-randomized study to compare two-dimensional point-based brachytherapy (2D-BT) with three-dimensional volume (3D-BT)-based ICBT [18]. Despite heterogeneity in treatment modalities, the STIC study demonstrated that 3D-BT, even without complex optimization procedures, improved the outcomes in cervical carcinoma, leading to a two-fold decrease in grade 3-4 toxicities compared with conventional 2D-BT, while also improving local control. In a single institutional retrospective study, Thomas et al. [19] showed that both sigmoid and bowel cumulative D2cc doses (EBRT plus BT) were significantly higher in the 2D-BT cohort; cumulative sigmoid D2cc was 83.9 Gy vs. 67.0 Gy (p < 0.001), and cumulative bowel D2cc was 75.5 Gy vs. 54.8 Gy (p < 0.001), correlating with significantly more grade 3 or higher gastrointestinal toxicities in the 2D-BT cohort.
MRI-based IGBT opened a new scope in the field of cervical cancer brachytherapy, owing to the fact that MRI is superior in soft tissue delineation than any other modality used for cervical cancer. As per ABS 2014 survey [20], MRI utilization has increased from 2% to 34% (p < 0.0001) for dose specification to the target in IGBT, while volume-based dose prescription (using MRI and/or CT) to the target has increased from 14% to 52% (p < 0.0001). However, the usage of MRI-based IGBT in many developing countries, and even in some developed countries, such as Japan [21], remains lower compared with the widely used CT-based IGBT. The reason being the increased time for scanning at each session, requirement of special MRI-compatible applicators, location of MRI scanners outside of radiotherapy department, and increased cost of therapy. This is particularly evident in resource-constrained countries, such as India. A survey by Gandhi et al. [22] found that less than one-sixth of practitioners in India use MRI-based IGBT.
With CT-based planning, some studies showed that the HR-CTV D90 on CT is significantly lower than that on MRI [12, 13], but evidence supporting clinical benefit of that difference is lacking. The GEC-ESTRO working group studies by Potter et al. have shown that CT is inferior to MRI in soft tissue delineation [9], thus our study employed a hybrid technique: MRI at the first fraction and CT for subsequent fractions. The HR-CTV volumes and D90 values from various studies using this hybrid approach are summarized in Table 3. From these studies, it is obvious that all CT-based fractions of a patient were combined together for evaluation and compared with a MRI-based fraction. The dosimetry for diverse fractions may vary due to slight differences in the applicator geometry, differential vaginal packing, inter-fraction variability in delineation and plan optimization, etc. Hence, we compared MRI with each CT fraction separately (MRI vs. CT-1; MRI vs. CT-2) as well as two CT fractions with each other (CT-1 vs. CT-2).
Table 3
Summary of various studies evaluating HR-CTV in hybrid approach used in cervical cancer brachytherapy
| Study [Ref.] | Type of study (year of publication) | No. of patients (n) | Mean or median HR-CTV MRI (volume) | Mean or median HR-CTV CT (volume) | p-value | Mean HR-CTV MRI D90 | Mean HR-CTV CT D90 | p-value (≤ 0.05) |
|---|---|---|---|---|---|---|---|---|
| Beriwal et al. [23] | Prospective, feasibility and efficacy study (2011) | 44 | 30.35 cc | 36 cc* | < 0.001 | N.A.** | N.A.** | N.A. |
| Choong et al. [25] | Retrospective study (2016) | 49 | 21 cc$ | – | N.A. | 97% ±11%$ | – | N.A. |
| Yip et al. [14] | Retrospective study (2017) | 11 | 33.2 cc | 50.7 cc* | 0.001 | 7.8 Gy ±0.4 Gy# | 8.1 Gy ±1.9 Gy# | 0.630 |
| Wang et al. [24] | Prospective, dosimetric study (2020) | 20 | 39.24 cc | 44.5 cc* | 0.006 | 5.63 Gy ±0.62 Gy | 5.23 Gy ±0.68 Gy | 0.035 |
| Present study | Prospective study | 22 | 24.23 cc | 1st CT: 27.82 2nd CT: 24.82 | 0.137 0.646 | 108.86% ±24.21% | 1st CT: 98% ±23.18% 2nd CT: 106.86% ±17.36% | 0.029 0.686 |
* Values calculated from all CT fractions, ** For contouring HR-CTV on CT, MRI images from the first fraction were used as guidelines, and mean HR-CTV D90 was not available from this study, $ HR-CTV defined from MRI-based fraction 1 was transferred onto subsequent fraction CT image sets using image registration techniques, assuming that HR-CTV was fixed in relation to applicator, # Planning aim per fraction ≥ 7.8 Gy and planning constraint per fraction ≥ 7.2 Gy
In our study, there was no statistically significant difference between HR-CTV volumes contoured on MRI at the first fraction, CT at the second fraction (CT-1), and CT at the third fraction (CT-2). Although HR-CTV on CT-1 tended to be larger than on HR-CTV MRI, there was no difference between HR-CTV on MRI and CT-2. This is in contrast to results of other studies [14, 23, 24], in which CT overestimated HR-CTV volumes significantly in all fractions, as shown in Table 3. We used a dedicated CT imaging protocol, with normal saline in the Foley’s catheter balloon and diluted contrast in the bladder, along with MRI/CT-compatible applicators for CT-based fractions. This approach led to a significant reduction in artifacts and improved image quality. Additionally, we used MRI at first fraction as a guide for delineating HR-CTV in subsequent fraction in terms of width and height. This may explain why the CT-based volumes were only slightly larger than the MRI-based ones, rather than showing large differences. Therefore, using MRI at the first fraction could significantly reduce inter-fraction variability in the HR-CTV delineation, further validating this pragmatic approach.
Table 4
Cumulative EQD2 doses of OARs (bladder, rectum, and sigmoid) in various studies and comparison with the present study
| Study [Ref.] | Type of study (year of publication) | No. of patients (n) | Total EQD2 bladder (Gy) | Total EQD2 rectum (Gy) | Total EQD2 sigmoid colon (Gy) |
|---|---|---|---|---|---|
| Beriwal et al. [23] (median with range) | Prospective, feasibility and efficacy study (2011) | 44 | 81.6 (61.76-86.60) | 56.65 (50.90-66.90) | 67.64 (57.80-80) |
| Choong et al. [25] (mean ±SD) | Retrospective study (2016) | 49 | 83.00 ±9.00 | 64.00 ±6.00 | 66.00 ±8.00 |
| Wang et al. [24] (mean ±SD) | Prospective, dosimetric study (2020) | 20 | 81.15 ±3.67 | 65.95 ±7.67 | 60.29 ±7.44 |
| The present study (median with range) | Prospective, clinical study (2019) | 22 | 80.2 (71.5-95.0) | 66.7 (59.9-73.3) | 64.5 (52.5-79.2) |
In our study, the mean HR-CTV D90 on MRI was higher than that on CT-1 (p = 0.03), but not statistically different from CT-2 (p = 0.68). The HR-CTV volume was slightly larger at the first CT fraction (although statistically not significant) as compared with the HR-CTV volume at the second CT fraction. This might be the reason for slight differences in the HR-CTV D90 doses, which came out to be statistically significant for MRI vs. CT-1 (p = 0.03). The HR-CTV D90 doses between CT-1 vs. CT-2 were not statistically significant (p = 0.07). Our study was not designed to detect this difference, and the small sample size may have created this effect. Additionally, as evident from Table 2, doses to OARs were also slightly lower at the CT-1 fraction. Using the MRI data from the first fraction and CT-1 planning data, our main goal in the last fraction was not to compromise the HR-CTV D90 doses. Thus, we ought to optimize this with differential packing and improvement in applicator geometry if required on a case to case basis. MRI images of the first ICBT fraction and CT images of the second fraction could serve as pre-planning surrogates to finally optimize the third or more subsequent fractions of IGBT. However, this is in contrast with other studies (Table 3), where the mean HR-CTV CT D90 was lesser than HR-CTV MRI D90 [14, 23-25]. Beriwal et al. reported the mean HR-CTV D90 by combining values from both MRI- and all CT-guided fractions, and their intent was to compare 3D-BT with 2D-BT rather than MRI vs. CT; this methodological difference could explain this discrepancy.
In 2013, a multi-institutional study by Petrič et al. [26] demonstrated that HR-CTV may be considered the most robust volume for dose prescription and optimization in cervical cancer IGABT due to lower delineation uncertainties when compared with GTV and IR-CTV. Therefore, based on this and other studies, we did neither delineate nor record IR-CTV in our study. Since, GTV cannot be delineated on CT images, we did not attempt to delineate or report doses to GTV in our research.
The volumes of OARs (the bladder, rectum, and sigmoid) were comparable between MRI- and CT-based plans. The CT-based sigmoid volumes at the third fraction (CT-2) was significantly larger than the MRI-based volume at the first fraction (MRI: 25.28 cc vs. CT-2: 34.61 cc, p = 0.01), likely due to interfraction variations in sigmoid filling. For the OARs dose parameters (D0.1cc, D1cc, and D2cc) in our study, no significant differences between CT and MRI were identified, which is similar to most of the studies comparing MRI- and CT-based IGABT [12, 14, 24-26]. The total doses (EQD2) received by OARs in our study are listed in Table 4 and compared with other research.
The current study is among the very few prospective clinical trials employing a hybrid approach, which reported clinical outcomes with a median follow-up of 5 years, unlike earlier similar studies with a shorter follow-up. The 5-year local control, pelvic control, and OS in a large prospective MRI-based RetroEMBRACE study was reported to be 91%, 87%, and 74%, respectively [6]. The 5-year loco-regional control rate in the current present study was 85.2%, which is comparable to control rates in many of the studies using MRI-based IGBT for LACC [27, 28]; albeit, with a smaller sample size than these studies.
One of the limitations of our study is the small number of patients and its single-institution design. Previous research using hybrid approach performed both MRI and CT at the first fraction and used image fusion to define the differences in HR-CTV between CT and MRI. In our study, we did not use this approach because of the presumed longer time for patient transportation: from brachytherapy suite to MRI location and back to CT scanner, which could lead to significant differences in application geometry and changes in OARs volume (e.g., the rectum). This is a potential confounding factor for disparity between MRI- and CT-based contours, and can be minimized by using dedicated brachytherapy IGBT (MR/CT) suite. Another limitation of our study is the lack of delineation and reporting of intermediate risk CTV (IR-CTV). The concept of IR-CTV is important, as it reduces the risk of marginal failures and accounts for coverage of microscopic spread of disease. However, this is often not delineated, as coverage and patterns of failure data corresponding to IR-CTV is lacking, its delineation is met with large inter-observer variability, dose constraints to IR-CTV are not defined, etc.
The 2021 joint IBS-GEC and ESTRO-ABS recommendations provide standardized guidelines for CT-based contouring in IGBT for cervical cancer, aiming to harmonize practice where MRI is not routinely available. The consensus outlines target and OARs delineation, ensuring safe and effective implementation of CT-guided brachytherapy in diverse clinical settings [29]. Additionally, incorporation of ultrasonography during implantation as well as planning may further refine the practices of CT-based IGABT [30]. Multi-institutional studies with longer follow-ups are required to fully validate the efficacy and safety of hybrid IGBT approaches.
Conclusions
MRI-based IGBT at the first fraction followed by CT-based planning used for the subsequent fractions is feasible and yields decent long-term clinical outcomes. This hybrid approach could be particularly valuable in low- and middle-income countries, such as India, and may potentially improve clinical outcomes for patients with LACC.
