No question: Proton therapy is safe

Based on radiobiological experiments and clinical reports, legitimate questions have arisen concerning the spatial accuracy of dose deposition and the biological impact of protons.¹ These have been the subject of lively debate among doctors who specialize in pediatric brain tumors. Therefore, I suspect many readers will eagerly turn to the paper by Upadhyay et al in seeking conclusive answers to the question, “Does proton therapy carry a higher risk of brainstem radiation necrosis than photon therapy?” ² But they will come away disappointed. The article from investigators at MD Anderson Cancer Center (MDACC) does not address this topic. Re-phrased, the more accurate question is, “When similar patients receive an identical dose of radiation to the brainstem, do those treated with proton therapy have a higher risk of symptomatic brainstem toxicity?” No study has adequately addressed this question, and the effort by Upadhyay et al is no exception. These are difficult investigations that at their core examine rare events in patients with a rare diagnosis treated with a rare technology. A conclusive study comparing the two radiation modalities would require hundreds of patients, both those treated with protons and those treated with photons over a contemporaneous period, with similar clinical features and thus similar baseline risk. The study would need to be carried out with modern treatment planning to assess the radiation dose to the structure at risk (ie, the brainstem) and not rely on analyses of nominal prescribed dose alone. Ideally, proton radiation treatment plans would utilize newer software that incorporates linear energy transfer (LET) or variable RBE-weighted doses. And patients would undergo similar follow-up surveillance, both in duration and in rigor, with events identified and graded using uniform and valid criteria. But in the absence of the ideal study, we are left with data that only tangentially address the question. The publications in the authors’ Table 5, while providing a succinct overview of recent literature on radiation-related brainstem injury, variously reflect all of these shortcomings.

Although Upadhyay et al do not provide a conclusive answer on the relative risk of proton therapy in brainstem injury, theirs remains a valuable study from numerous angles. The authors should be recognized for compiling the largest published series to date characterizing the incidence and risk factors for symptomatic brainstem injury (SBI). It is a representative sample of the most common pediatric brain tumors treated with radiation. It also contains a detailed description of treatment technique and further refines brainstem dose tolerance models. In this sense, the authors have provided a practical manuscript, with recommendations that can be applied immediately in proton therapy clinics worldwide.

Perhaps most reassuring is this: When delivered as described in this manuscript, SBI following proton therapy is rare, with the incidence of grade ≥3 toxicity at only 1.7%, and largely transient, with ultimate resolution of symptoms in 80% of patients.

The authors avoid many shortcomings of previous studies on this topic. Upadhyay et al are precise in capturing clinical features and brainstem dose, and they apply standardized event criteria and toxicity grading. They should be commended for appreciating the difference between clinically relevant toxicity in the form of SBI and transient asymptomatic radiographic changes (ARC), which have little to no discernable impact. The frequent conflation of SBI and ARC under the general term “necrosis” is more consequential than simply imprecise rhetoric. Combining these two entities has obscured prior analyses and ignores patient impact. The lack of subtly can cause distress for families and angst for regulatory bodies. While the MDACC data on ARC have serious limitations due to insufficient imaging follow-up (“a minimum of 1 post-treatment MRI at 3 months was considered adequate imaging”) and lack of comparative historic norms, the authors have made an important contribution by simply recognizing and isolating ARC from true SBI in the context of this analysis.

As these authors highlight, we still have a lot to learn about the clinical and treatment factors that predispose children to brainstem injury. For example, which aspects of high-dose chemotherapy exacerbate risk? What is the interplay of hydrocephalus and surgical trauma? What is more damaging, a moderate radiation dose to the whole brainstem or a very high dose to a small volume? Should we develop differential brainstem guidelines for girls and boys? Further reduction in the incidence of brainstem toxicity will likely require a broader understanding of multiple variables, and complex pathophysiology can easily get lost in singular discussions of protons vs photons. To their credit, Upadhyay et al leverage the size of the MDACC cohort to advance our knowledge of risk factors and, combined with other publications, we now possess a much clearer picture of risk factors than we did a decade ago (Table 1). This knowledge has a real impact, as evidenced in the reduction in SBI from 4.4% to 1.5% at MDACC.

Table 1.

Risk Factors for Brainstem Toxicity Proposed Over the Past Decade in Published Series

Series (year)	Brainstem Dose	Young Age	Chemotherapy	Sex	Hydrocephalus	Focal Radiotherapy	Tumor Histology	Surgical Trauma
Murphy et al3	X
Indelicato et al4	X	X
Nanda et al5								X
Gentile et al6	X		X
Bojaxhiu et al7			X		X		X
Upadhyay et al2	X	X	X	X		X	X

In many cases (as in children with ependymoma⁸ and chordoma⁹), reducing the dose to the brainstem often means reducing the dose to part or all of the tumor bed. The MDACC investigators acknowledge that “prescription doses were sometimes decreased if these [brainstem] constraints failed to be met. Another approach commonly used to achieve constraints was intentional under coverage of the target volume adjacent to the brainstem.” The authors critically indicate that this “underdosing” does not compromise local control. Taken with other reports of this nature, their observation is provocative and raises the question, have we been prescribing too high a dose for these tumors? The issue deserves further attention in disease-specific studies.

Finally, careful analysis and publication of outcome data remain underappreciated in pediatric radiotherapy. Technology evolves quickly and randomized trials are impractical, unethical, or both.¹⁰ The topic of radiation-related brainstem injury serves as an exemplar of how the field has rapidly adapted to improve safety, largely through unglamorous institutional registry studies performed outside of the NCI cooperative group framework. For this, we commend not only the study investigators but also MDACC and Texas Children’s Cancer Center administrations for allocating resources to this project. It is hoped that other centers—as well as larger funding bodies—follow their lead and recognize the practical utility of high-quality outcome data in pediatric radiotherapy.