The spine is among the most frequent sites of metastasis, often presenting with symptoms that prioritize local ablative treatments. Conventional palliative radiotherapy for spinal metastases was the standard of care for a long time. However, it was associated with poor 1‑year local control rates and complete pain control, with rates ranging from 61% to 86% and 0% to 24%, respectively, for de novo metastases [10, 11]. Recent developments in the care of cancer patients have led to longer overall survival and necessitated improvements in local control rates and quality of life. Particularly in the oligometastatic setting, where survival rates are higher compared to those in polymetastatic disease, patients may benefit more from aggressive approaches [12]. Spinal SBRT has demonstrated superior local control and complete pain control rates ranging from 57% to 95% and 46% to 92%, respectively [10], and for single-fraction treatments, PTV Dmin > 15 Gy was shown to improve LC [13]. Consequently, for spinal oligometastases, the treatment paradigm has shifted from palliative conventional radiotherapy to SBRT in recent years.

As high doses with steep dose gradients are delivered with SBRT, protection of organs at risks, especially of the spinal cord, is essential. In recent years, spinal cord dose constraints have been well studied [10]. However, accurate delineation and interfractional imaging of the spinal cord remain challenging. The delineation of spinal cord is ideally done with the fusion of MRI and the simulation CT [14, 15], since MRI has better soft tissue contrast than CT [16]. The use of image-guided radiotherapy, preferably every fraction, is also generally recommended in SBRT for spinal bone metastases [17] to ensure target volume coverage as well as spinal cord sparing.

The MR-linac system allows for online adaptive planning during each treatment fraction. Moreover, it has a double-stack and double-focus multileaf collimator system which allows us to use a steep dose gradient [18]. Also, SMART permits direct visualization of the spinal cord, esophagus, sacral plexus, and other OARs before each fraction, thus enabling contouring of target volumes and organs at risk to evaluate the predicted plan and generate an online adaptive plan. Small setup positional changes can be compensated with adaptive planning in lieu of correcting patient position, which might be hard for patients who have pain due to vertebral metastases. MRI acquisition and on-table adaptive planning may result in longer treatment times with SMART than with cone-beam CT (CBCT)-based systems. Our median treatment time was 15.13 min, which is tolerable given that the repositioning would not be needed due to daily contouring and adaptive planning. Intrafractional motion may also affect the treatment. In their study, Oztek et al. showed that even intrafractional spinal cord motion may contribute to increased spinal cord doses [6]. Real-time tracking with the MR-linac can aid in detecting intrafractional motion and help to prevent both target misalignment and spinal cord overdosing. Our postoperative cases are also a demonstrative example of how 0.35‑T low-strength magnetic field MR images are not severely affected by metallic artifacts, which might be the case for higher magnetic fields which are more susceptible to field inhomogeneities that result in an obscured view of nearby soft tissue structures [19]. Although metallic-artifact-reduction techniques have been developed for CT, no standard technique was developed, and the existing techniques may cause secondary artifacts [20].

Yadav et al. conducted a dosimetric study comparing SBRT performed by MR-linac vs. TrueBeamTM STx-based volumetric modulated arc therapy (VMAT) plans [18]. They evaluated 10 previously treated metastatic lesions. The results showed that SMART plans had lower intermediate dose spillage; similar D98%, D2%, and D50%; and also similar homogeneity and conformity indexes as well as similar doses to the spinal cord, which led the authors to conclude that spinal SBRT delivered with an MR-guided system is comparable to TrueBeam VMAT in terms of target coverage, plan quality, and spinal cord sparing. However, the MR-linac has advantages of contouring the spinal cord and other OARs in every fraction, personalized adaptive online treatment planning, and continuous cine-MR tracking during treatment.

Choi et al. retrospectively generated three plans (MR-linac IMRT, MR-Co-60-IMRT, and VMAT) for 20 thoracic spine lesions with prescription doses of 18 Gy in a single fraction to compare dose–volume outcomes [21]. A notable result was that the average spinal cord volumes delineated based on CT and MRI images significantly varied. Also, while dose de-escalation had to be performed for spinal cord dose tolerances for VMAT and MR-Co-60-IMRT plans, 18 Gy in single fraction could be delivered without dose violation for MR-linac IMRT plans. Importantly, MR-linac IMRT plans resulted in the lowest spinal cord doses and the difference was significant (all p ≤ 0.003). Those results accentuate the importance of the MRI-based spinal cord contours as well as the superiority of MR-linac plans both in terms of deliverability of the intended doses and in terms of spinal cord protection.

An ongoing pilot clinical trial is investigating the feasibility of MRI-based simulation and SMART in spinal metastasis with a treatment scheme of five fractions delivered over 1–2 weeks (NCT03878485). Patients with previous radiotherapy to the target region will be excluded. The findings from an ongoing phase I/II clinical trial that utilizes SMART to treat all disease sites, including the spine, and evaluates the feasibility and efficacy of SMART are anticipated (NCT04115254).

Our retrospective results highlight the importance of the aforementioned studies. The study includes real-world data of patients treated with SMART for spinal metastases. Both oligometastatic and palliative irradiations were included, as well as reirradiations, which are not mentioned in the previous literature. We demonstrate that online adaptive plans which may compensate for target motion yield significantly better results in terms of target coverage compared to predicted plans. Spinal cord protection is also significantly improved, and this improvement is even more important for this patient group who are oligometastatic and who may require reirradiation for spinal metastases in the future.

Limitations of our study are its retrospective nature and the fact that some information might be missed, since the dataset was not designed for a prospective study and included very heterogenous patients with different pathologies. Increased follow-up time could yield more comprehensive results.