Comparison of Online-Onboard Adaptive Intensity-Modulated Radiation Therapy or Volumetric-Modulated Arc Radiotherapy With Image-Guided Radiotherapy for Patients With Gynecologic Tumors in Dependence on Fractionation and the Planning Target Volume Margin

Key Points Question What is the clinical benefit associated with online adaptive radiotherapy (ART) for gynecologic tumors compared with image-guided radiotherapy (IGRT)? Findings This comparative effectiveness study comprising 7 consecutive patients with gynecologic tumors found that ART was associated with improved deviation of the delivered generalized equivalent uniform dose (gEUD) for interfractional clinical target volume values from the original dose distribution on the planning computed tomography scan compared with IGRT, without increasing the gEUD for the bladder and rectum for planning target volume margins less than 5 mm. Meaning This study suggests that the advantage associated with ART depends on the number of applied dose fractions and planning target volume margins.


Treatment planning
Patients with histopathologically proven gynecologic tumors were presented to the department of radiation therapy. After individual case discussion in the interdisciplinary panel and assessment by a radiation oncologist, patients were scheduled for radiotherapy. Patients underwent treatment planning simulation and received a planning-CT with or without contrast-agent, and advanced imaging techniques such as iterative-reconstruction and automatic-dose-modulation.
For the reproducible positioning special techniques such as constant bladder-fill and additional vaginal applicators were used. Treatment planning was performed using the ETHOS-treatment-planning-system (Varian, Palo Alto, US). GTV and CTV were defined and delineated according to the MRDG-PreBT-MR/CTBT environment following IBS-GEC ESTRO-ABS recommendations for CT-based contouring in image-guided adaptive brachytherapy for cervical cancer [1]. With this approach a clinical MR was available at initial diagnosis (1.5-3TMRI at diagnosis (MRDG)) and within few days prior to boost planning with the applicator in place (1.5-3T MRI at boost and brachytherapy planning (PreBT-MR or MRBT)). The diagnostic 1.5-3T MRI was conducted with a dedicated coil in supine position. T1-, T1 fat-saturated, contrast-enhanced and T2weighted true-axial, sagittal, and coronal projections with 3-to 4mm thick slices, 0 to 1mm spacing, and depending on slice orientation 256*256 to 400*400 matrices. The MR-findings of residual tumor were taken side-by-side onto the high-resolution replanning-CT (CTBT), also acquired few days prior to boost planning, to delineate the CTV-boost volume with respect to anatomical margins, where possible [1]. If residual tumors on MRBT were too large for a brachytherapy boost or had residual infiltration of surrounding tissues incompatible for brachytherapy, patients underwent external beam therapy instead. The PTV was predefined as a 5-mm-expansion of the CTV to take potential set-up errors into account. Organs-at-risk were contoured on the planningand replanning-CT to avoid dose hot spots inside the vulnerable regions. The automatic adaptive planning module generated volumetric-modulated arc radiotherapy (VMAT) and static-field intensity-modulated radiation therapy (IMRT) plans, following the clinical goals defined in the ETHOS RT-Intent. The ETHOS-system applies inverse planning with manual or templated application of optimization costfunction based objectives [2]. Dose was calculated with the Ethos AcurosXB (version 1.1.2.44, primary fluence-mode FFF, 6MV).
Special considerations of adaptive radiation therapy (ART) planning and treatment delivery Treatment delivery on the ETHOS-system (Varian, Palo Alto, US) can be performed in image-guided radiotherapy (IGRT) or adaptive-mode [2]. While the former workflow allows for anatomic guidance based on the online patient position, the latter allows the onboard plan adaptation. Daily image guidance (IGRT) is an integral part of every radiation-therapy-sequence. Online target matching is achieved with the help of a 3degrees-of-freedom-table. Rigid registration leads to an update of the isocenter position.

Equivalent-uniform-dose
We examined the EUD according to the phenomenological power-law-model (gEUD) [3] with tissue-specific parameters for the tumor and organs-at-risk (a = -20 for tumor; a = +7 for bladder and a = +7 for rectum) [4,5]. For a better inter-patient comparability gEUDis were normalized to the prescribed dose. The percentage deviation of gEUD for the CTVi (%gEUDCTVi) and organs-at-risk at fraction i was determined for ART and IGRT in comparison to the reference-plan. The accumulated gEUD-values were examined over all fractions per patient. In this study, all adapted-plans, deformed images and structures were analyzed by an experienced team of medical physicists and expert radiation oncologists.

Normalization
All dosimetric characteristics for a deformed structure in the synthetic-CT (sCT) from dose fraction i of a series were given as the percentage deviation from the respective value for the undeformed structure in the planning-CT. All scheduled dose distributions were recalculated using the synthetic-CTs. Therefore, for example, the percentage deviation of Dmin for the CTVi at fraction i is  [6]. When both processes have random components, they affect the efficacy of a treatment series [6].
For patients with tumors unsuitable for brachytherapy there is a need of high precision external beam therapy to mimic brachytherapy [7][8][9]. For such dose escalated schedules the smallest possible PTV-margins below 5 mm must be aimed for. Online ART can realize here its full potential as shown in the present study. During ART image-quality must be sufficient for exact target-volume-delineation. Image-quality of the CBCT in this study was sufficient to delineate organs-at-risk and its anatomic relation with the residual tumor. Its boundaries, especially in lateral direction within the parametria and towards the pelvic wall, were precisely known from the planning MRDG-PreBT-MR. In some cases, surgical clips helped to increase precision of CTVdelineation. If bladder or rectal wall were infiltrated, the infiltrated zones could be accurately detected on the high-quality CBCT. Our experience shows, that for online ART of gynecologic cancers a current MRI-study available for review of residual tumor on daily CBCT and on current diagnostic-CT is comparable with the precision of MRIguided-brachytherapy. The same was found by others for image-guided-adaptivebrachytherapy [1,10]. According to the IBS-GEC ESTRO-ABS-recommendations for CT-guided brachytherapy a diagnostic-MR prior to therapy, a pre-brachytherapy-MR within 1 week prior to brachytherapy and a CT with the applicator in place are comparable with the gold standard of MRI-guided approach for brachytherapy of cervical cancer [1]. Likewise, Mahantshetty et al. report that CT-based target and organ-at-risk delineation using MRI at diagnosis and real-time transrectal-ultrasound information during BT seems comparable with the gold standard MRI-guided-adaptivebrachytherapy for cervical cancer [11].
Only very limited data on ART is available at this time. Yock et al. used a software emulator to retrospectively analyze standard large pelvis radiotherapy for patients with cervical cancers using a PTV-margin of 5mm [12]. They found that Dmin for the CTV of the primary and the nodal regions could be improved by adaptive radiotherapy by in average 0.25+0.30 Gy per fraction, where the paired-differences of the scheduled with the adaptive-plan became significant [12]. Lakomy et al recently described their initial experience with online adaptive radiotherapy using an MR-linac in 10 patients with gynecologic cancer, treated with 90 fractions [13]. The treatment with online plan adaptation took in median 42 min. They used a PTV-margin of 3-5 mm and 17 of the fractions were treated with online-adapted plans. The mean accumulated dose of the GTV showed a decline above -5% in comparison to the reference plan [13].
Hadi et al. used a 0.35 Tesla MR-Linac for an online-adaptive boost with a PTV-margin of 5 mm for patients with gynecologic cancers ineligible for brachytherapy [14]. The median online treatment-time was with 77 min rather long. This treatment showed good tolerability. While these studies showed the feasibility of online adaptive radiotherapy, the on-coach times have to be shortened sufficiently to prevent intra-fraction organ shifts. In addition, though, MRI is considered the imaging modality of choice for gynecologic cancers, nowadays not 0.35 Tesla, but 1.5 and 3T-MRI is considered as minimum and optimum field strengths for sufficient diagnostic accuracy.
The median on-coach treatment-times in the present study with a CBCT-based adaptive treatment approach were 33 min using VMAT-optimization. Using a 12-field IGRT-optimization these times can be further shortened by approx. 8 min. The second CBCT for final verification, embedded within a MRDG-PreBT-MR/CTBT environment [1], combined with fast treatment delivery allows a nearly continuous intra-fraction monitoring.

eFigure 1A.
Linear-Quadratic Dependence of the %gEUDCTVi on %Dmin for the CTVi (%DminCTVi). The quadratic term was significantly smaller than 0 (p < 0.0001, t-test). Data points for the scheduled (red) and adapted plans (blue) follow the same relation, while the scatter of the %gEUD values of the adapted plans were much smaller than that of the scheduled plan.

eFigure 1B.
Dependence of %gEUD on the %D99 for the CTVi. Only the linear term (p < 0.0001, t-test) but not the intercept or the quadratic term became significant. The relation was close to a 1:1 relation with a slope of 1.173 +/-0.035.