**4. Discussion**

To the best of our knowledge, this is the first series of patients treated for primary RCC using MRgRT with routine daily plan re-optimization. We applied a commonly used fractionation scheme of 40 Gy in five fractions [18,23,24] in an overall treatment time of two weeks. Only a single patient reported nausea as acute toxicity, and no grade ≥ 2 late toxicity was observed. Despite the inclusion of large tumors, mostly T1b and T2, which had a mean tumor diameter of 5.6 cm and were generally unsuitable for other local therapies, we observed an LC rate of 95.2%. Our local response scoring has been according to the RECIST 1.1 criteria, and 83.3% had stable disease. In addition, 11.1% had partial remission, while 5.6% showed local progression. Fast tumor size regression is uncommon after SABR as previously reported by Sun and colleagues [11]. This preponderance of stable disease is in accordance with their paper. Both LC and OS are reported to be poorer for larger primary RCC than for the smaller lesions [25,26]. Despite this observation, our LC rate is within the high range of what was reported in recent systematic reviews, meta-analyses, and pooled analyses of SABR for primary RCC [15,24,27].

MRgRT with daily plan re-optimization was feasible with an average fraction duration of 45 min, even in poorer condition patients with multiple co-existing diseases. Despite this prolonged treatment duration, all patients were able to complete treatment, which indicates good tolerability. Our fractionation scheme of 40 Gy in five fractions is commonly used and seems safe without severe toxicity. With a mean interval of well over one year, the mean decline in eGFR in our study was only 6.0 (SD <sup>±</sup> 9.8) mL/min/1.73 m2. This value corresponds well with the mean decline in eGFR of 5.5 (SD <sup>±</sup> 13.3) mL/min/1.73 m2 that was described in previous SABR studies [15,28]. This limited decline in renal function in our patients with relatively large RCC may well be the result of this gated approach with small mobility boundaries, instead of using internal target volumes incorporating full tumor motion.

MRgRT also offers the advantage of using plan re-optimization for each delivered fraction at the cost of additional time. Our offline analysis showed that daily plan re-optimization was required in only 16% of fractions in which the predicted plan failed to meet the predetermined high-dose OAR constraints or target coverage objectives. Decision tree analysis showed that patients for whom daily plan re-optimization is not required can be identified upfront on the basis of a V25Gy of the combined OAR of less than 0.5 cc in the baseline plan. It is, however, unlikely that an isolated single fraction violating high OAR dose or target constraints will be clinically relevant, and three out of five insufficient predicted plans were seen in only 14% of patients. Performing MRgRT without plan re-optimization indicates that the re-contouring, plan adaptation, and plan quality assurance phases can be omitted, which would enable respiratory-gated MRgRT fractions to be completed in 30 min. Furthermore, when plan adaptation is redundant, this indicates that the presence of the radiation oncologist at the MR Linac is not necessary. As a result of our analysis, we are currently introducing the found V25Gy selection criterion in clinical practice.

The main limitation of our study is the relative short and unstructured patient follow-up. The limited number of RCC patients reflects the limited role of SABR in current international treatment guidelines, as only patients unsuitable for or refusing other local treatments are referred for curative radiation therapy. Another limitation includes the absence of pathology in half of our patients. Incomplete pathology confirmation is partly inherent to our patient population with generally frail elderly patients, which is unsuitable for other treatment modalities. Moreover, in a number of patients, a diagnostic biopsy was considered contra-indicated because of anticoagulant use or the anatomical location of the tumor. All patients had been discussed in a multidisciplinary tumor board with access to all available diagnostic imaging. Contrast enhanced multi-phasic CT has a high sensitivity and specificity for characterization and detection of RCC [3,29] and this specific imaging was available for all patients without pathological confirmation.

Prior to the MRgRT era, the need for radiologists to implant fiducial markers has also been an obstacle for referral for SABR. Our data show that MRgRT can be a valid alternative in patients unsuitable for the more commonly used local treatments, because of patient vitality or tumor size. The only contra-indication for MRgRT is having MR-incompatible devices. The main advantage of MRgRT is that it is an outpatient, non-invasive treatment for which not even the placement of fiducial markers is necessary. Whether MRgRT can also be considered as an alternative to partial nephrectomy or cryotherapy needs to be addressed in a prospective randomized study, which should also evaluate quality of life and cost-effectiveness. With regard to the favorable outcome in the data on SABR literature as well as the current analysis on MRgRT, a more prominent role of SABR in the treatment guidelines for RCC appears warranted.
