Radiation oncology is a field constantly seeking to improve patient outcomes by bringing novel and sophisticated treatment paradigms into the standard of care. Technological innovation in radiation oncology often emphasizes enhancing precision and accuracy in planning and treatment delivery by developing processes that widen the therapeutic ratio by allowing for dose-escalation to tumor tissues while simultaneously reducing dose to surrounding normal tissues. MRI-guided radiation therapy (MRgRT) is a rapidly emerging technology that seeks to achieve these goals through a variety of mechanisms, including taking advantage of superior soft-tissue contrast compared with the current standard of care CT-guided radiation therapy (CTgRT).1 MRgRT further leverages MRI capabilities to observe motion and deformation of tumor targets and organs at risk in real-time. This not only allows for beam-on gating, whereby radiation is only delivered to the target structure when it falls within a predefined gating margin,2 but also the innovative and paradigm-shifting possibility of real-time, on-table adaptive treatment delivery3 which can be augmented through integration of advanced imaging techniques such as functional imaging.4

Given the clinical promise of MRgRT, there has been significant work verifying and validating the use of MRgRT in the United States, Europe, and Asia in recent years.5,6 Although still early in its development, MRgRT appears to be capable of significantly disrupting the current process of care in radiation oncology which has been relatively stable for decades. More than just a linear accelerator combined with an MRI device, this technology has been acknowledged to offer more than the sum of its parts, largely due to the real-time imaging features, adaptive planning possibilities of MRgRT, and functional imaging serving as a real-time biomarker for treatment response. In total, MRgRT allows radiation oncologists to operate more akin to surgeons who observe, modify and actively manage their treatments with real-time feedback. The rapidly developing and transformative treatment delivery modality of MRgRT thus invites several important health policy and healthcare economic considerations which will be discussed herein.

When a new medical technology emerges that disrupts the routine process of care, it is incumbent for stakeholders to evaluate how the new technology will fit into the existing system from the perspective of delivery of care and value. Incorporation of a new technology, in this case MRgRT, requires that it demonstrates value within the healthcare system. This is particularly true in radiation oncology, which has often been historically singled out for scrutiny given the typically high upfront costs in spite of eventually proving to be high-value care.7,8 As proposed by Steinberg et al.,9 value within the field of radiation oncology is influenced by 4 key components: cost, outcomes, structure, and process. Cost and outcomes represent recognizable and intuitive components of value,10 while the other components of structure and process derive from a separate yet related model for assessing quality in healthcare.11 Structure refers to the larger organization of care delivery as well as the facilities where medical care is being provided, including the equipment, staff, and institutions, and emphasizes technologically current and safe environments. Structure can also extend to include insurance coverage and reimbursement models.12 Process, on the other hand, relates to the components of healthcare delivery that originate from the patient perspective, emphasizing patient-centered care as well as the technical delivery of care including factors such as physician expertise and physics quality assurance methods. This model can be summarized and expressed as an equation which includes all 4 components, in which value = (outcomes + structure + process)/cost, and where the numerator can also be considered as “quality.” Through this equation, there are multiple ways in which value can be enhanced, such as by increasing the numerator of quality and/or decreasing the denominator of cost. However, it is worth noting that value can also be enhanced even in the context of increasing costs through proportionally greater increases in quality, and it is ultimately the overall ratio of quality/cost that is of importance when evaluating value. Thus, a new treatment or technology, in this case MRgRT, can still demonstrate value in radiation oncology despite potentially increased costs, by offsetting these costs with substantial improvements in the structure, process, and outcomes of care.

As described above, cost is a key component in determining the value of a treatment or technology. In addition, both provider and patient perspectives on cost are important. While this is true for all new treatments, it is of particular importance in radiation oncology where upfront technology costs can be significant.13 Such is the case for MRgRT, where the device cost is typically higher than state of the art CT-based treatment delivery devices, and there are supplemental construction costs associated with installing MRI shielding.14 Moreover, setting in place the additional care delivery infrastructure such as increased medical physics oversight and staffing cost required for these devices, which may have lower throughput compared to standard CT-based treatment devices, all need to be taken into consideration.15 However, these apparent expenditures are perhaps less impactful than initially anticipated when one considers other cost savings and increased utility that can be associated with MRgRT. For example, when compared with CTgRT delivery in time-driven activity-based costing analyses, it was estimated that MRgRT was only incrementally more expensive for the delivery of stereotactic body radiation therapy (SBRT) for hepatocellular carcinoma14 and in absolute dollars, just $1,497 (in 2021 United States Dollars) more expensive per course of SBRT for prostate cancer16 from the health system perspective. Furthermore, while the upfront device costs seem expensive on the surface, the machine costs are amortized over time, and much of the long-term costs associated with MRgRT actually derive from associated staffing and personnel needs.17,18 Ultimately though, cost associated with new technology is often ephemeral in nature, and there are multiple advantages of MRgRT that may actually invite cost-savings and improve value, as will be discussed. Given the upfront investment required to implement MRgRT programs, early acquisition and use of MRgRT remains primarily in larger, well-resourced medical facilities, and widespread adoption may initially be limited by the inability of smaller facilities to acquire the necessary equipment and infrastructure.19, 20, 21 At the same time, MRgRT has also demonstrated its ability to successfully integrate into the workflows of some varied US healthcare practice settings—ranging from large academic centers22 to community practices23 to universal access healthcare systems such as the Veterans Health Administration.24 This suggests that even the high upfront costs are surmountable in a number of practice environments.

Early economic analyses of MRgRT, which mainly examine prostate cancer and focus on cost from the provider perspective, highlight a few important points regarding the economic implications of this technology.15 As it currently stands and as noted above, MRgRT is associated with incrementally increased costs compared with CTgRT. However, these studies identify multiple practicable ways by which MRgRT can become (or already is) cost-effective compared with CTgRT (Fig. 1). One of the more obvious ways to demonstrate the cost-effectiveness of MRgRT is to show that it improves outcomes, such as radiation-associated toxicities or disease control. Improvements in these metrics would result in decreased costs for patients and the healthcare system overall and would thus successfully engage the value proposition in healthcare and radiation oncology.9

A valid concern is that, in its current rendition, treatment times are relatively long with MRgRT, yielding decreased patient throughput with this technology.25 This phenomenon holds true even when not performing real-time on-table adaptive treatment and is thought to be related to issues of machine functionality and capability, such as the intensity modulated radiation therapy (IMRT) treatment delivery modality being a step-and-shoot technique rather than the more efficient volumetric arc therapy technique,14 as well as the need to further optimize adaptive workflows. Longer treatment times and workflow inefficiency thus impact the value proposition for MRgRT at this point in time. It should be pointed out, however, that full implementation of a MRgRT-only workflow (ie, MRI simulation only) is likely to be associated with reduced cost and improved efficiency over the current process of care.26,27 In the current fee-for-service model, included in the value equation is the number of fractions per treatment course. To the extent that MRgRT enhances our ability to safely deliver ultra-hypofractionated treatments, further cost savings from wide MRgRT implementation are likely to ensue. While some published studies found MRgRT 5 fraction SBRT to be more costly than CT-based treatment,15 further reducing the number of fractions delivered is another possible method for improving its cost-effectiveness.14 Such fraction-reducing treatment paradigms are currently under prospective evaluation in prostate cancer28,29 as well as other disease sites.30, 31, 32, 33 The underlying rationale for studies such as these is sound, given that MRgRT has already demonstrated incremental improvements in both patient-and physician-reported toxicity profiles of ultra-hypofractionated radiation therapy.34 Moreover, ultra-hypofractionated treatment courses may mitigate patient financial toxicity through less missed work, reduced caregiver burden, and fewer travel-related expenses, among other components,35, 36, 37, 38 and shorter treatment courses may even have positive environmental impacts as well.39,40 Furthermore, there are factors specific to the CTgRT process, such as the need for the procedural insertion of fiducial markers for certain disease sites,41 which are not required with MRgRT, thereby creating additional opportunities for cost savings and improvements in patient outcomes. Finally, MRgRT most often requires both CT and MRI simulations for planning in its current state.42 When the treatment planning process can omit CT simulation, which is potentially the case with synthetic CTs,16,43, 44, 45 the costs of MRgRT can be improved even further.

Also fundamental to a treatment's value is how it is incorporated into healthcare insurance coverage and reimbursement policies, and indeed, changes in our billing and coding paradigms are critical for making MRgRT cost-effective as well. Such an undertaking must balance the often-competing interests of payers, patients, the centers investing in MRgRT capabilities, as well as the healthcare system overall. Despite the significant innovations offered by MRgRT, the current billing and coding nomenclature neither distinguishes nor does it differentially reimburse for MRgRT or the process by which it is used in adaptive treatment delivery.46,47 The reflex may be to simply plug these novel treatment delivery processes into the pre-existing adaptive billing lexicon. However, the current coding structure for adaptive radiation therapy is based on legacy adaptive radiation planning and thus does not sufficiently account for the radically different processes utilized in MRgRT adaptive planning. This is especially important to consider under the proposed radiation oncology-alternative payment model (RO-APM),48 where adaptive MRgRT may be susceptible to reduced reimbursement relative to its infrastructure costs in the setting of bundled payments.49 Continued thought and discussions are needed on how to best account for the different processes introduced by MRgRT in the billing and coding nomenclature in order to maximize the benefit of this technology for all financial stakeholders.

The ability of a treatment or technology to improve oncologic outcomes, such as toxicity or disease control, is a crucial component of value in radiation oncology and healthcare as a whole. Multiple studies have started to report on the clinical outcomes of MRgRT in a variety of cancer sites.50 While these data are mostly retrospective in nature, there exist some high-quality prospective studies which begin to confirm the theoretical benefits of MRgRT with respect to cancer outcomes.

The recently published MIRAGE clinical trial34 randomized men with clinically localized prostate cancer to either CTgRT or MRgRT and sought to evaluate whether reductions in isotropic planning target volume (PTV) margins from 4 mm to 2 mm, enabled by MRgRT, would reduce the risk of acute genitourinary (GU) and gastrointestinal (GI) toxicity from SBRT. Indeed, the incidence of physician-reported acute GU (24.4% versus 43.4%, 56.2% relative reduction) and GI (0.0% versus 10.5%, 100% relative reduction) toxicities were improved in the MRgRT arm, as were patient-reported toxicity outcomes. In the economic analyses discussed above, relative reductions in toxicity of 7-54% were needed for MRgRT to become cost-effective compared with CTgRT,51,52 a threshold which was met and indeed exceeded by the results of this trial. Additionally in this trial, patients in the CTgRT arm required fiducial placement whereas those in the MRgRT arm did not, adding further cost savings to the healthcare system otherwise unaccounted for with MRgRT. However, this trial also demonstrated some of the economic concerns with MRgRT, including the need for 2 simulation scans (CT and MRI) in the healthcare system where this single institution study was carried out and increased post-imaging treatment delivery times with MRI versus CT guidance (median 1133 seconds versus 232 seconds). Nevertheless, a rigorously demonstrated decrease in radiation-associated toxicity should not be overlooked as it not only improves patient health-related quality of life, but it also reduces acute hospital encounters related to treatment toxicity which can be associated with high costs, procedures, and overall burden for patients, providers, and the healthcare system. Drawing parallels between MRgRT and other historical innovations in radiation oncology, namely IMRT, recall that the widespread adoption of this novel technology was primarily spurred by its associated improvements in normal tissue sparing and toxicity,53, 54, 55 and perhaps an analogous implementation threshold would be worthwhile for MRgRT.

While the MIRAGE trial demonstrated the toxicity benefits that MRgRT can provide without daily adaptation, the SMART trial further demonstrated the potential advantages of MRgRT by incorporating adaptive planning.56 In this multi-center, single-arm phase II trial of MRgRT in borderline resectable or locally advanced pancreatic cancer, patients were treated to an unprecedented dose of 50 Gy in 5 fractions, and on-table adaptive re-planning was used if the original radiation plans recomputed onto the daily anatomy would not meet treatment constraints. This resulted in adaptive planning being performed in 93.1% of fractions. Overall, there were low rates of acute grade 3 or higher GI toxicity probably (2.2%) and definitely (0%) related to radiation therapy at the augmented dose. Furthermore, this study also reported noteworthy oncologic outcomes including 1 year local control (82.9%), distant progression free survival (50.6%), and overall survival (65.0%). Previous research of MRgRT in inoperable pancreatic cancer found that dose-escalation above a biologically effective dose10 (BED10) of 70 Gy, which was primarily achievable by adaptive MRgRT, was associated with improved overall survival without an increase in toxicity.57 In fact, patients treated with dose-escalated adaptive MRgRT experienced less toxicity than patients treated at a lower BED10 without adaptive radiation therapy. The SMART trial dosing regimen equates to a BED10 of 100 Gy, further highlighting the clinical advantage of on-table adaptive planning enabled by MRgRT in this study. Importantly, the primary study objective of demonstrating <15.8% acute grade 3 or higher GI toxicity definitely related to radiation therapy was met, indicating that additional prospective evaluation of this technique should be performed.

The SMART trial highlights a unique and important strength of MRgRT compared with the current radiation treatment process—notably, the capability for on-table adaptive radiation therapy in real-time, which allows for target dose escalation while minimizing the risk of injury to surrounding normal tissues. The increased use of MRgRT in recent years has been accompanied by an increase in the percentage of on-table adapted fractions with this technology.5,6 The logistics of this process has been described in multiple studies,2,3,50,56 and implementation of this multifaceted workflow is important to consider (Fig. 2). What should also be considered is the amount of time required to complete adaptive treatments, as this has been identified as a cost concern for MRgRT.3,14,16 The economic evaluations described above did not include adaptive planning in their total treatment time, which has been documented to range from 50-100 min,3,25 although one did note that that this would increase healthcare system costs by only $529 per adaptive treatment.14 Additionally, increased staffing and resources are needed for this process, as are technologies and systems for ensuring sufficient plan evaluation and safe treatment delivery.58 However, mechanisms to improve protracted treatment times and inefficiencies are being actively studied, such as through the deployment of artificial intelligence and auto-segmentation59,60 and deep learning.61

Moving forward, the use of MRgRT and real-time adaptive planning is currently being studied and implemented in a variety of disease sites and clinical situations, including central nervous system tumors,50,62 head and neck cancers,50 and breast and lung neoplasms50,62,63 as well as re-irradiation64 and single fraction metastatic cases,65,66 and may even be expanded to other situations where real-time, high-resolution soft tissue imaging is crucial, such as palliative celiac plexus irradiation67, 68, 69 (Fig. 3). Additionally, MRgRT capabilities will continue to progress as well. Beyond motion management and adaptive planning, MRI-specific functional imaging represents a future method through which MRgRT can continue to demonstrate its value.62,70,71 For example, diffusion-weighted imaging has been identified as a potential biomarker for treatment response and could be incorporated into the MRgRT treatment process for dose escalation of non-responsive disease or dose de-escalation of rapidly responding disease.4 The expansion of MRgRT to additional clinical situations and the continued development and improvement of the adaptive treatment process will further support the economic justification of this promising technology.

As an emerging technology, a sufficient body of evidence demonstrating the benefit of MRgRT is needed to successfully argue for its incorporation into standard of care of radiation paradigms. A recent survey investigation highlighted the potential challenges associated with widespread adoption of MRgRT based on interviews of personnel involved with MRgRT, revealing that lack of current evidence of clinical benefit was a primary concern surrounding its implementation.19 Certainly, additional evidence, including not only traditional objective oncologic outcomes but also more thorough cost-effectiveness analyses72,73 and subjective measures like patient-reported outcomes,74 would help to further support the MRgRT value proposition. However, emerging technology also frequently outpaces the evidence supporting its use, and thus our field is in the position where we must consider and respect the need for evidence prior to widespread implementation of MRgRT while simultaneously respecting the drive to enable our patients to experience the intuitive but untested benefits of MRgRT. Additionally, it should be acknowledged that technology trials are challenging to conduct,75 and it is often infeasible to perform a randomized controlled trial for every new treatment paradigm. While some endpoints require robust prospective data and extended time periods to measure, such as disease-specific and overall survival, there are other relevant endpoints, such as acute toxicity, which can be determined in a shorter timeframe. Thus, the widespread use of MRgRT need not be delayed until longer-term endpoints are met, similar to the intuition around and the adoption of IMRT.76,77 It should also be noted that level I evidence is similarly not reasonable nor a required threshold for the creation of a new Current Procedural Terminology (CPT) code to, for example, capture the work and practice expense of MR-guided adaptive radiation therapy. Therefore, we must carefully consider the evidence threshold required for widespread adoption of MRgRT and associated processes of care. Fortunately, efforts are already ongoing to ensure the appropriate accumulation of data for the introduction of MRgRT.78

While the advantages of MRgRT with respect to cancer outcomes make it appealing from the value perspective, additional value may come from the disruptive workflows inherent to MRI simulation and adaptive processes of care. These aspects relate to the structure and process components of the value equation describe above. Regarding the MRgRT simulation process, being able to omit CT simulation improves value from the patient and system perspectives through reduced appointments, radiation exposure, and overall healthcare congestion, among other factors. Instrumental to this process is the use of synthetic CTs, in which MRI simulation images are converted to synthetic CT data (eg, Hounsfield Units) necessary for dose calculation and treatment planning, which are actively being studied and validated.43, 44, 45 Additionally, use of both CT and MRI simulations for treatment planning, which is increasingly the case for disease sites where soft tissue resolution is paramount such as central nervous system, head and neck, and gastrointestinal tumors,42 can introduce errors during the image registration process,79 introducing yet another opportunity for improvement in outcomes with MRI-only workflows.

However, there are multiple potential advantages of MRI simulation in and of itself compared with CT simulation, which invite additional opportunities for innovation and improving value. For example, the superior soft tissue visualization with MRI allows for improved target delineation such that MRI simulation-based target delineation notably results in smaller clinical target volumes (CTVs) in prostate80 and cervical cancer.81 Furthermore, when combined with online adaptive planning in lung cancer, MRI-based planning resulted in smaller PTVs than would have been generated from an internal target volume (ITV) approach.82 Additionally, a recent small retrospective study found that the use of MRI simulation was associated with improved local control in nasopharyngeal cancer, particularly in stage T4 disease, compared with PET/CT simulation, which was proposed to be the result of full visualization of disease extent with MRI.83 MRI-only workflows have been developed84,85 (Fig. 4), and the feasibility and safety of a same-day MRI-only simulation with an adaptive MRgRT workflow was recently demonstrated in palliative radiotherapy cases with encouraging results.86 The improved value to the patient provided by this accelerated process is apparent in the palliative setting, where symptom control and minimizing treatment length are emphasized, though this expedited workflow would certainly benefit all patients, especially those receiving a single fraction of radiation or travelling from great distances. Further in the future, with the real-time motion management and plan adaptation capabilities of MRgRT, the concept of the PTV may be rendered obsolete, and future treatments might be planned based solely on the gross tumor volume (GTV) or CTV, which would allow for further reduction in dose delivered to organs at risk. Beyond the standard anatomic considerations at the time of treatment planning, MRI simulation also unlocks functional tumor-specific considerations as well, which can be interrogated with MRI-based functional imaging, as noted previously. Given the rapidly increased use of MRI-simulation in recent years, the American Association of Physicists in Medicine (AAPM) have published a task group report providing recommendations on the safe and optimal use of MRI simulation,87 which is a key aspect of the structure component in the value framework described above. These advantages, as well as the more complicated process of care, validate the notion that MRI-only simulation should not just be conflated with standard simulation nomenclature. Instead, its work and practice expense should be valued for what are already established equipment costs and process of care differences.

Also crucial to this discussion of incorporating MRgRT into the standard of care in radiation oncology is the realization that the process of care for MRI-based radiation therapy, particularly MRI adaptive radiation therapy, is fundamentally a very different radiation treatment delivery process than that currently utilized for CT-based radiation therapy. The ability for on-table adaptive planning and re-planning with MRgRT represents the most unique aspects of this technology.88 While adaptive radiation treatment is not a new concept in radiation oncology,89 it historically has been limited due to the logistics of CTgRT. Specifically, legacy approaches to adaptive radiation therapy have not been performed on-table in real-time.50 Instead, they are done off-line prior to the next treatment, and they cannot address target or organ at risk deformation. MRgRT, in contrast, allows for a feasible and superior mechanism for performing adaptive radiation therapy, specifically while the patient is on the treatment table, and can be repeated daily. While the data supporting adaptive radiation therapy continue to accumulate, our field must remain open to the plasticity that may accompany the radical shifts in the process of care that MRgRT affords.

Peering into the future, one might imagine a treatment process where simulation as we have come to know it is not performed at all, and treatment is simply delivered de novo based on the daily anatomy. A simulation-free radiation treatment process has been explored in palliative cases, though this process relied on diagnostic images for planning.90 In the future, we might be able to rely on treatment machine-generated MRI images alone. These images will require adequate soft-tissue contrast during the on-table acquisition period, which is currently under evaluation.91 Omission of traditional simulation scans might thus represent a potential avenue to enhance accuracy and quality of care, particularly from the patient perspective as this would reduce appointments, costs, and time from consult to treatment, which might be of particular benefit for patients experiencing symptoms from their tumor such as pain or bleeding.