**The Role of Daily Adaptive Stereotactic MR-Guided Radiotherapy for Renal Cell Cancer**

**Shyama U. Tetar 1,**†**, Omar Bohoudi 1,**†**, Suresh Senan 1, Miguel A. Palacios 1, Swie S. Oei 1, Antoinet M. van der Wel 1, Berend J. Slotman 1, R. Jeroen A. van Moorselaar 2, Frank J. Lagerwaard <sup>1</sup> and Anna M. E. Bruynzeel 1,\***


Received: 24 July 2020; Accepted: 22 September 2020; Published: 25 September 2020

**Simple Summary:** Standard treatment for localized renal cell carcinoma (RCC) is surgery. Stereotactic radiotherapy given in a few high dose fractions is a promising treatment for this indication and could be an alternative option for patients unsuitable for surgery. Stereotactic MR-guided radiotherapy (MRgRT) is clinically implemented as a new technique for precise treatment delivery of abdominal tumors, like RCC. In this study, we evaluated the clinical impact of stereotactic MRgRT given in five fractions of 8 Gy and routine plan re-optimization for 36 patients with large primary RCCs. Our evaluation showed good oncological results with minimal side-effects. Even in this group with large tumors, daily plan re-optimization was only needed in a minority of patients who can be identified upfront. This is a favorable result since online MRgRT plan adaptation is a time-consuming procedure. In these patients, MRgRT delivery will be faster, and these patients could be candidates for even less fractions per treatment.

**Abstract:** Novel magnetic-resonance-guided radiotherapy (MRgRT) permits real-time soft-tissue visualization, respiratory-gated delivery with minimal safety margins, and time-consuming daily plan re-optimisation. We report on early clinical outcomes of MRgRT and routine plan re-optimization for large primary renal cell cancer (RCC). Thirty-six patients were treated with MRgRT in 40 Gy/5 fractions. Prior to each fraction, re-contouring of tumor and normal organs on a pretreatment MR-scan allowed daily plan re-optimization. Treatment-induced toxicity and radiological responses were scored, which was followed by an offline analysis to evaluate the need for such daily re-optimization in 180 fractions. Mean age and tumor diameter were 78.1 years and 5.6 cm, respectively. All patients completed MRgRT with an average fraction duration of 45 min. Local control (LC) and overall survival rates at one year were 95.2% and 91.2%. No grade ≥3 toxicity was reported. Plans without re-optimization met institutional radiotherapy constraints in 83.9% of 180 fractions. Thus, daily plan re-optimization was required for only a minority of patients, who can be identified upfront by a higher volume of normal organs receiving 25 Gy in baseline plans. In conclusion, stereotactic MRgRT for large primary RCC showed low toxicity and high LC, while daily plan re-optimization was required only in a minority of patients.

**Keywords:** MR-guided; radiotherapy; MRgRT; stereotactic ablative radiotherapy; stereotactic ablative radiation therapy (SABR); renal cell cancer; RCC; online adaptive

#### **1. Introduction**

A radical or partial nephrectomy is the preferred standard curative treatment for localized renal cell carcinoma (RCC) [1–4]. Ablative local treatment, such as radiofrequency ablation (RFA), cryoablation (CA), or microwave ablation (MWA), is an alternative in elderly patients who present with a high surgical risk due to several comorbidities [3]. Radiotherapy does not have a prominent role in current international and national guidelines in treating primary RCC [1–4]. In recent years, stereotactic ablative radiation therapy (SABR) has been evaluated in several smaller retrospective and prospective studies [5–14], usually in RCC patients unsuitable for surgery. Outcomes of a multi-institutional pool from nine institutions, utilizing either single or multi-fractionated treatment in 223 patients, have been reported by the International Radiosurgery Oncology Consortium for Kidney (IROCK) [15]. SABR for RCC was found to be well tolerated, achieved local control (LC) rates exceeding 95% at four years of follow-up and grade ≥3 toxicity rates of 1.3%, and had an average decrease in glomerular filtration rate of 5.5 mL per minute. The majority of the tumors in this pooled analysis was ≤4 cm and clinical data for larger tumors is limited. A retrospective analysis of a subgroup of 95 patients with tumors >4 cm was recently published [16], but with the exception of these data, clinical outcomes on cT1b-T2 RCC SABR are scarce. Due to the inherent limitations to a pooled analyses, the Trans-Tasman Radiation Oncology Group (TROG) and the Australian and New Zealand Urogenital and Prostate Cancer Trials Group (ANZUP) have initiated a prospective, multi-institutional phase II study in 70 patients with biopsy-confirmed medical inoperable RCC patients [17]. Full accrual has recently been completed, and the data of this trial are eagerly awaited.

Technical challenges in renal SABR include the management of intra-fractional motion, and potential solutions using an internal target volume-approach, fiducial-assisted robotic SABR or abdominal compression [18] have been described. Magnetic-resonance (MR)-guided radiotherapy (MRgRT) has been considered a promising option because of its improved visualization of kidney tumors in relation to critical adjacent organs such as a small bowel, duodenum, and stomach and the opportunity of real-time tumor tracking and automated gated delivery [18,19]. MRgRT also facilitates daily plan re-optimization as a means to reduce organs at risk (OAR) doses when abdominal organs are near the primary tumor. Furthermore, MRgRT is an outpatient treatment for which no invasive procedures or anesthesia is required. However, to the best of our knowledge, clinical data on MR-guided SABR for localized RCC have not been reported.

Stereotactic MRgRT with routine daily plan adaptation was clinically implemented at our center in 2016 for a variety of clinical indications. The aim of the current paper is to describe our technique, early clinical outcomes, and the role of daily plan adaptation in MRgRT for patients with primary large RCC.

#### **2. Materials and Methods**

Data from all patients treated with MRgRT on the MRIdian-system (ViewRay Inc., Mountain View, CA, USA) at the Amsterdam University Medical Centers are collected within a prospective institutional review board approved database. Between May 2016 and February 2020, a total of 51 patients were treated for a primary RCC (*n* = 36), local recurrences (*n* = 5), renal metastases from other primary tumors (*n* = 3), or a diagnosis of urothelial carcinoma (*n* = 7). This analysis is restricted to the remaining 36 patients who were treated for primary RCC.

All patients underwent stereotactic adaptive MRgRT delivered to a dose of 40 Gy in five fractions in a two-week period. Implanted fiducials were not required, and the adaptive workflow was similar to that which had been described previously for pancreatic tumors [20]. Briefly, for simulation, both a MR-scan (0.35T True-FISP, TR/TE: 3.37 ms/1.45 ms, FA: 60◦, 17-s with 1.6 mm × 1.6 mm × 3.0 mm resolution) and computed tomography (CT)-scan (slice thickness of 2 mm) are acquired during a shallow-inspiration breath-hold. Geometric accuracy of the MRIdian system is < 0.1 cm in a sphere of 10 cm radius around the isocenter, and <0.15 cm in a sphere of 17.5 cm radius. Every patient was brought as close to the isocenter as possible for each fraction, and the maximum distance from the tumor or any other critical structure to the isocenter was always below 10 cm. Geometric accuracy was assessed with two different dedicated phantoms for spatial integrity measurements. Contouring of the primary tumor (also called gross tumor volume; GTV) and OAR is performed on breath-hold MR-images with the aid of diagnostic imaging, generally contrast-enhanced CT scans. The PTV (planning target volume) is derived from the GTV plus an isotropic 3-mm margin. A co-planar baseline plan consisting of between 30 and 42 intensity modulated radiotherapy (IMRT)-segments is generated, using the MRIdian treatment planning software. Dose calculation was executed with a VMC and EGSnrc code-based Monte-Carlo algorithm (statistical uncertainty of 1% and a grid size of 0.3 cm × 0.3 cm × 0.3 cm) using the deformed electron density map from the simulation CT scan. Institutional target coverage and OAR constraints are summarized in Table 1.

**Table 1.** Dose prescription for institutional target coverage and normal tissue constraints. The constraints represent the cut-off doses for radiotherapy planning with the aim of dose sparing in the surrounding organs (contralateral kidney, liver, duodenum, bowel, and stomach) while, at the same time, aiming to achieve a high dose in the tumor with the margin, which is represented as the planning target volume. Organs at risk are only re-contoured within 2 cm of the tumor and, for an adaptive setting, only the dose in these structures are optimized.


We perform routine plan re-optimization using the daily pre-SABR breath-hold MR-imaging acquired in the treatment position. After rigid registration on the GTV, OAR contours are propagated to the repeat MR using deformable image registration. The ViewRay deformable image registration algorithm uses an intensity-based algorithm, which minimizes a cost function that measures the similarity between the images including a regularization term in order to obtain smoother deformation fields and prevent sharp discontinuities. The GTV and OAR contours are checked and adjusted where needed within a 2-cm distance of the PTV by the attending radiation oncologist. Next, the baseline IMRT plan is recalculated on the new anatomy ("predicted plan"), and subsequently re-optimized using the target and OAR optimization objectives of the baseline plan ("re-optimized plan"). Plan re-optimization prioritizes avoiding high doses to OARs, even when this is at the cost of decreased PTV coverage. Both the predicted and re-optimized plans are reviewed, and the re-optimized plan is selected for the actual delivery.

MRgRT delivery is performed using respiratory gating during subsequent breath-hold periods in shallow inspiration. The tracking structure for gating is either the primary tumor, or the kidney itself on a single sagittal plane (Figure 1), depending on the visibility on this sagittal plane. Gating is augmented by visual and/or auditory feedback provided to patients during treatment [21]. Visual feedback is performed with the aid of an in-room MR compatible monitor on which both the tracking structure (GTV or kidney) and the gating boundary (3 mm), generally corresponding to the PTV, is projected in real-time. The 2D MR images during treatment were acquired with a True FISP sequence with the MRIdian (0.35 T) at a frequency of four frames-per-second (TR: 2.1 ms, TE: 0.91 ms, FA: 60◦). FOV was 0.35 cm × 0.35 cm and the slice thickness was 0.7 cm. Due to the low magnetic field and low FA, "real-time" MR images of the patient were performed without interruption during the beam-on time. A previous analysis showed a treatment duty cycle efficiency between 67% and 87% for upper abdominal tumors [22].

**Figure 1.** Sagittal plane for tumor tracking: either (**a**) tracking on gross tumor volume (green) or (**b**) tracking on the whole kidney (orange). A boundary of 3 mm (red) for gated delivery.

Baseline patient and tumor characteristics and follow-up data including LC, renal function, and toxicity were collected. Acute and late toxicity was scored using the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. Follow-up imaging was assessed by a CT-scan or ultrasound, and the tumor response was classified according to RECIST 1.1. criteria.

An offline analysis was performed to evaluate the need for daily plan re-optimization in MRgRT for RCC in a total of 180 fractions. For this purpose, predicted and re-optimized plans were analyzed for adherence with planning target objectives and OAR constraints, i.e., a V38Gy of the GTV ≥ 90%, and V33Gy ≤ 1 cc for stomach, duodenum, and bowel. Re-optimization was defined as "needed" when the predicted plan violated the above-mentioned GTV and/or OAR constraints, which was subsequently corrected by re-optimization. In contrast, plan re-optimization was defined as "redundant" when predicted plans already complied with the planning objectives. In addition, the value of plan re-optimization was analyzed on a patient level by studying the number of fractions per patient that were considered suboptimal.
