1 INTRODUCTION

Through the characteristic Bragg peak, proton beam therapy spares patients the “exit” radiation dose associated with X-ray treatments, such as three-dimensional (3D) conformal photon therapy and intensity-modulated photon therapy (IMRT). Proton therapy confers numerous purported advantages, including the ability to escalate the radiation dose in tumors positioned near critical organs and reduce the overall volume of normal tissue exposed to low- and intermediate-radiation doses. The latter is particularly appealing in pediatric radiotherapy, where the potential for decades-long survivorship, inherent radiosensitivity of developing organs, and high rate of germline mutations in childhood cancer patients drives a constant pursuit to reduce or eliminate unnecessary radiation exposure. However, to date, the vast majority of pediatric patients worldwide have been treated using a passive modulation proton technique that employs a scattering foil to produce a clinically useful field size. This scattering foil is the source of neutrons that can expose the total body of a child to the damage of potent secondary ionization, estimated in doses distant from the field edge 10 times higher than those characteristic of IMRT.¹ Such provocative models are an understandable source of both ongoing concern^{2, 3} and controversy,^{4, 5} calling for empiric validation in the real-world clinical setting. Therefore, we performed a comprehensive review of second tumors in pediatric patients treated with double-scattered proton therapy over the 13-year history of the University of Florida Health Proton Therapy Institute (UFHPTI).

2 METHODS

3 RESULTS

Patient and treatment characteristics of the 1713 patients are detailed in Table 1. The 5- and 10-year overall survival rates were 83.5% and 82.2%, respectively. The median follow-up was 3.3 years (range 0.01–12.8) and included 6587 total person-years in follow-up. Overall, 549 patients had ≥5 years of follow-up. For this 549-patient subset, the follow-up duration was 7.1 years (range 5–12.8).

3.1 Second solid tumor events

Overall, 11 patients developed second solid tumors resulting in a 5- and 10-year cumulative incidence of 0.8% (95% CI, 0.4–1.9%) and 3.1% (95% CI, 1.5–6.2%), respectively (Figure 1A). Table 2 contains second solid tumor details. The SIR was 13.5 and the AER was 1.5/1000 person-years. The median time to development of a second solid tumor was 66 months following irradiation (range 17–102 months). Eight of 11 patients had malignant tumors. The risk of developing a second solid tumor was 1.7% if irradiated at age ≤5 years versus 0.1% if older (p < .0005). The risk for patients with a known tumor predisposition syndrome developing a second tumor was 10.8% versus 0.4% in those without (p < .001). The second tumor risk was most pronounced in patients with Li–Fraumeni (33.3% vs. 0.5%; p < .001). Excluding patients with known tumor predisposition syndromes, the 5- and 10-year risks of second solid tumor development were 0.6% (95% CI, 0.2–1.7%) and 1.7% (95% CI, 0.7–4.0%), respectively (Figure 1B). Two of the four patients with known tumor predisposition syndromes developed multiple second tumors. Excluding the patients with known tumor predisposition syndrome, all of the malignant second tumors occurred in the high-dose region (Appendix SA1). In this subset of patients without a tumor predisposition syndrome, age ≤5 years remained a significant risk factor (p < .01). Examining the subset of 549 patients with at least 5-year follow-up reveals a 10-year second tumor cumulative incidence of 2.3% (95% CI, 0.9–5.7%). For this cohort, the SIR was 10.0 and AER was 1.2/1000 person-years. Among these patients, age ≤5 years (p < .05), tumor predisposition syndrome (p < .01), and the use of craniospinal radiation (p < .05) correlated with the development of a second solid tumor. When excluding the patients with known tumor predisposition syndrome from those with 5-year follow-up (n = 538), the 10-year second tumor cumulative incidence was 1.1% (95% CI, 0.3–3.6%). Overall and across the subset analyses, sex, histology, radiation dose, chemotherapy, and alkylating chemotherapy were not significantly associated with the development of a second solid tumor. The number of events precluded a valid multivariate analysis.

TABLE 2. Second solid tumors

Case #	Age at RT (years)	Primary tumor	Interval (months)	Secondary tumor	Dose at site of secondary tumor (Gy)	Genetic syndrome?	Chemotherapy?
1	5	Medulloblastoma	66	High-grade glioma	54	No	Yes
2	4	Ependymoma	81	High-grade glioma	12.6–59.4	No	No
3	5	Craniopharyngioma	68	High-grade glioma	10.8–54	No	No
4	2	Ependymoma	96	High-grade sarcoma	10.8–54	No	Yes
5	3	Ewing sarcoma	56	High-grade sarcoma	54	No	Yes
6	2	ATRT	56	Schwannoma	10.8–50.4	No	Yes
7	7	Ewing sarcoma	44	Inflammatory myofibroblastic tumor	12–21	No	Yes
8	1	S-PNET	102	High-grade bone sarcoma	35–54	ATM	Yes
9	5	Medulloblastoma	93, 122	Low-grade glioma, choroidal melanoma	36, 36	NF-1	Yes
10	3	Rhabdomyosarcoma	17	Giant cell tumor	0–41.4	p53	Yes
11	2a	Rhabdomyosarcoma	41, 66	Medulloblastoma, high-grade soft tissue sarcoma	0, 50.4	p53	Yes

• Abbreviations: ATM, ataxia-telangiectasia mutation; ATRT, atypical teratoid rhabdoid tumor; RT, radiation therapy; S-PNET, supratentorial primitive neuroectodermal tumors.
• ^a Same patient as in Table 3.

4 DISCUSSION

Similar to general age models, there is strong evidence supporting a linear dose–response for second solid tumors following radiation in children.¹² This concept has been invoked by numerous proton therapy proponents who expect less carcinogenesis in children treated with proton therapy.^13-15 However, a landmark article by Eric Hall published in 2006 indicated that the neutrons generated by the scattering foil used in passive-scattered radiation devices results in doses distant to the field edge that are 10 times greater than those characteristic of IMRT and even higher when compared to 3D conformal photon radiation.¹ Furthermore, due to the smaller physical size of children undergoing radiation, the absolute risk of secondary neoplasms may be more pronounced in younger patients. The conclusions by Hall have subsequently been challenged from both physics and radiobiology perspectives,^{5, 16} but clinical evidence is lacking on both sides of the issue. An important barrier to resolution is the lack of suitable comparison data for second tumors following photon radiation. Many existing large cohort reports are based on survivorship clinic patients, excluding patients with less than 5 years of follow-up. Yet the relative risk of developing a subsequent malignancy after radiation is highest in the first 10 years after radiation for the primary cancer diagnosis,^{6, 17-19} and, like ours, multiple series report solid tumors occurring within 3 years following radiotherapy.^{6, 19, 20} Some reports exclude patients with cancer predisposition syndromes. For example, Berger et al. reported a cumulative incidence rate of 1.0% at 5 years, but excluded patients with neurofibromatosis and retinoblastoma.¹⁹ Other studies exclude “nonmalignant second neoplasms” or “nonglioma CNS neoplasms” from rate estimates.^{6, 7, 17, 18, 21, 22} Other reports generate SIRs using mixed therapy cohorts, including patients who did not receive radiotherapy.^{7, 18, 19, 23} Furthermore, the SIR in any given series may vary as a function of the irradiated primary tumor, where patients with CNS tumors, sarcoma, and Hodgkin lymphoma may have a two to three-times higher rate of developing a second malignancy compared to other types of tumors.^{6, 10, 19, 23} As a result of the wide variation in methodology, SIR estimates following photon radiation range broadly from 3.6 to 22.9.^{6, 7, 10, 17-19, 23}

It is widely recognized that certain cancer predisposition syndromes, such as neurofibromatosis, Li–Fraumeni, retinoblastoma, and ataxia telangiectasia, increase sensitivity to ionizing photon radiation. The magnitude of elevated risk is uncertain as one cannot determine if, for any given event, radiation actually caused the second neoplasms so common in these patients. Furthermore, no meaningful clinical data exist on the relative impact of neutron scatter in children with cancer predisposition syndromes. Nonetheless, it is hoped that when these children do require therapeutic radiation, the risk of second tumors will be reduced if proton therapy is utilized. Given that 90% of children in our cohort with known tumor predisposition syndromes did not develop a second tumor, our early data are encouraging, especially considering the actual rate would be lower as some of the documented second tumors would likely occur in the absence of any radiation. Our report might serve as a useful foundation for future photon and proton radiation studies intended to address this knowledge gap. Beyond patients with a cancer predisposition syndrome, our observations also confirmed the elevated risk of second tumors in any patient irradiated during early childhood. When radiation cannot be delayed, however, double-scattered proton therapy seems to impart a fairly low risk in the first decade of follow-up: among our 1134 children older than 5 years at the time of treatment, the incidence of any second neoplasm was 0.3%.

Important limitations of our data must be acknowledged. Patients were screened for known predisposition syndromes using standard methods, but the true incidence of cancer predisposition genes in children may exceed 8%.²⁸ Family history is an incomplete predictor in most patients and, therefore, the true number of children with cancer predisposition (and impact thereof) may be underestimated in our analysis. Moreover, despite >500 patients with over 5 years of follow-up, the latency of some second tumors requires decades of follow-up to truly characterize the risk faced by survivors. It is unclear how second cancer latency correlates with tissue dose: the lower doses afforded by proton therapy may simply delay, rather than eliminate, second neoplasm risks, although the limited existing clinical information shows no correlation between dose and latency.²⁰ Given the high rate of participation in our prospective outcome study, we are hopeful that this patient cohort will continue to contribute to our future analyses of both incidence and timing of these second tumors. Our study is also susceptible to another limitation common to long-term research: evolving technology. Nearly all new proton therapy devices today have replaced passive modulation with a scanning beam. In theory, this will reduce the rate of second tumors even further.¹ This limitation does not negate the value of our findings however, as thousands of survivors worldwide have been treated with double-scattered proton therapy and would benefit from more accurate risk estimates. Moreover, some scanning proton systems still require patient-specific apertures that generate neutron scatter²⁹ and other heavy particles that are on the horizon, like carbon, generate secondary neutrons.³⁰ Therefore, the data herein will remain useful and offer an opportunity to refine radiobiologic models and inform risk–benefit determination in the future.

CONFLICT OF INTEREST

In the last 36 months, Raymond B. Mailhot Vega and Julie A. Bradley were awarded unrestricted educational grants from IBA.