Proton Therapy and Limited Surgery for Paediatric and Adolescent Patients with Craniopharyngioma: A Phase 2 Study

Abstract

Background

Proton therapy reduces exposure of normal brain compared to photon therapy in patients with craniopharyngioma which might reduce cognitive deficits associated with radiotherapy. Because there are known physical differences between the two modalities, we aimed to estimate progression-free (PFS) and overall survival (OS) distributions for pediatric and adolescent patients with craniopharyngioma treated with limited surgery and proton therapy while monitoring for excessive central nervous system toxicity.

Methods

Children, adolescents, and young adults diagnosed with craniopharyngioma were treated between 2011 and 2016 using passively scattered proton beams, 54Gy (RBE), and a 0·5cm clinical target volume margin. Surgery was individualized prior to proton therapy and included no surgery, single procedures of catheter and Ommaya reservoir placement via burr hole or craniotomy, endoscopic resection, transsphenoidal resection, craniotomy, or multiple procedure types. After completing treatment, patients were evaluated clinically and by neuroimaging for tumor progression and evidence of necrosis, vasculopathy, permanent neurologic deficits, vision loss, and endocrinopathy. Neurocognitive tests were administered at baseline and annually through five years. Outcomes were compared to a historical cohort treated with surgery and photon therapy. The coprimary endpoints were progression-free and overall survival distributions. Progression was defined as an increase in tumor dimensions on successive imaging evaluations more than 2 years after treatment. The study used the CTCAE Version 4·0 for toxicity and performance reporting. Grade 3 and 4 toxicities were reported to the institutional review board.

This was a single arm study that was not designed as a non-inferiority trial. A 3-year PFS of 94.6% was chosen from a historical photon trial. The study was designed with a binary endpoint to have 91% statistical power (1-sided test: α=0.05) to detect a drop in the 3-year success rate from 94.6% to 86%. After 3 years of follow-up for all patients, 91 out of 94 participants (96.8%) did not have an event. The one-sided binomial test showed that there was no evidence to suggest that the 3-year PFS was less than 94.6%. The trial was registered with the US National Library of Medicine (ClinicalTrials.gov NCT01419067).

Findings

Between Aug 22, 2011, and Jan 19, 2016, 94 patients were enrolled and treated with proton therapy of whom 46 were female and 55 were male. The median age was 9 years (IQR 6–13) at the time of radiotherapy. With a median follow-up of 7·52 years (IQR 6.28–8.53), 3-year PFS was 96.8% (95% CI: 90.4–99.0%; p=0·89), with progression occurring in three of 94 patients. No deaths occurred at 3 years, such that overall survival was 100%. No deaths occurred as of data cutoff (02/02/2022). There was no difference in five-year PFS and cumulative incidence (CI) of severe complications, change in vision, or endocrinopathy comparing patients treated with proton therapy to those treated with photon therapy (photon patient values in parentheses): CI of necrosis 2·13% ± 1·50% (1·98% ± 1·39%), severe vasculopathy 4·49% ± 2·21% (4·99% ± 2·19%), permanent neurologic conditions 3·19% ± 1·82% (2·97% ± 1·70%), and change from normal to impaired vision 7·41% ± 3·60% (6·04% ± 3·42%). There was no difference in the 5-year incidence (95% confidence interval) of central hypothyroidism 68.89% (0.5278, 0.8047) vs. 78.95% (0.6149, 0.8915), adrenal insufficiency 50.98% (0.3640, 0.6382) vs. 57.14% (0.3882,0.7183), or hypogonadotropic hypogonadism 43.43% (0.3348,0.5296) vs. 42.86% (0.3208, 0.5318) comparing patients treated using proton therapy to those treated using photon therapy. The most common grade 3–4 adverse events were headache (six [6%] of 94 patients), seizure (five [5%]), and vascular disorders (six [6%]).

Interpretation

Proton therapy did not improve survival outcomes in paediatric and adolescent patients with craniopharyngioma and severe complication rates were similar between the proton therapy cohort and an historical photon cohort; however, cognitive outcomes with proton therapy were improved over photon therapy.

Funding

American Lebanese Syrian Associated Charities (ALSAC), American Cancer Society, Cancer Center Support Grant No. CA21765-23 from the US National Cancer Institute, and Research to Prevent Blindness

Introduction

Craniopharyngioma is an intracranial tumor that presents in children with wide-ranging symptoms and catastrophic effects. Surgery and radiation therapy are the mainstays of treatment. Each modality may be uniquely tailored to successfully address this locally aggressive midline tumor intimately associated with the visual pathways, hypothalamic-pituitary axis, and central components of the cerebral vasculature.¹

For newly diagnosed patients, decisions to use radical or limited surgery and definitive irradiation are driven by institutional preferences and experiences. The low incidence of this tumor limits treatment experience for most centers and impacts the ability to establish a standard of care.^2,3

Most children treated in North America for craniopharyngioma receive fractionated external beam radiation therapy at the time of initial diagnosis or progression after radical resection. The use of unsealed radioactive sources or chemotherapy to treat tumor cysts, single-fraction radiosurgery to treat limited volume residual tumor, and hypofractionated radiation therapy regimens have been selectively applied to small cohorts or in the setting of recurrence after conventional irradiation.^4,5

Craniopharyngioma has several features that make it amenable to advanced methods of targeting, localization, and external beam delivery. These features are deep, central intracranial locations and in general, distinct borders. When treating children, normal tissue sparing, achieved by reducing target volume margins or the use of advanced methods, has been a priority while total dose and fractionation regimens have remained largely unchanged.⁶

We designed a clinical trial to estimate the progression-free and overall survival distributions for children and young adults with craniopharyngioma treated with limited surgery and proton therapy using a 0·5cm clinical target volume margin while monitoring for excessive central nervous system necrosis, clinically significant vasculopathy, and permanent neurological conditions or deficits.⁷

The trial was designed based on our experience testing reduced target volume margins and the use of photon therapy.⁸ When our models suggested that proton therapy, which reduces the volume of normal tissues exposed to intermediate or low doses,⁹ might improve cognitive outcomes,¹⁰ the current trial was proposed. Although proton therapy was considered the logical next step for children with craniopharyngioma, when the protocol was designed there were no clear guidelines for treatment planning and beam delivery (target volume margins, dose, and fractionation), and the radiobiological differences between photon and proton beams required special consideration and monitoring for unanticipated side effects.¹¹

Research in Context

Evidence before this study

We searched PubMed on July 1, 2011, for clinical trials published between 1988 and 2011 in the English language using different combinations of the terms “craniopharyngioma”, “radiotherapy”, “pediatric”, and “outcomes”; we also searched https://clinicaltrials.gov/ using the same terms for clinical trials with posted results for children age < 18 years treated for craniopharyngioma using proton radiotherapy. Our search found no published data describing outcomes for children and adolescents treated using proton radiotherapy and the estimated benefit of proton radiotherapy when compared to radiotherapy administered using photons.

In a phase 2 study, we estimated the rate of tumor control in young patients with craniopharyngioma using passively scattered proton beams and 0.5cm clinical target volume margin. Proton therapy start dates were September 19, 2011 to February 22, 2016. We compared progression free and overall survival to a historic cohort treated using photon beams. We monitored patients for severe complications and change in cognition. The latter was an exploratory aim to determine whether there is a benefit associated with proton radiotherapy and reduced normal tissue exposure.

Added value of this study

The lack of difference in tumor control rates and the incidence of severe complications between modalities was expected. The benefit of proton beam therapy over photon beam therapy is, to our knowledge, one of the first prospective studies, involving pediatric and adolescent patients with a single tumor type, to demonstrate an advantage with long-term follow-up. The study provides a model approach to the evaluate of newer methods of radiation therapy in the treatment of childhood brain tumors and highlights the need for rigorous and long-term follow-up.

Implications of all the available evidence

These results may be practice changing if they convince caregivers to recommend proton beam therapy over radical surgery or the referral of patients for proton beam therapy instead of radiation therapy using photons. Healthcare systems have invested substantial resources to acquire access to advanced radiation therapy methods hoping for convincing evidence of a benefit of proton beam therapy. Public health care agencies promoting universal access and health equity should weigh the available evidence on the benefits, risks, and cost effectiveness of new and established treatments to support referral of patients with rare conditions such as craniopharyngioma for treatment with proton therapy.

Methods

Eligibility

Patients diagnosed with craniopharyngioma by histology, intraoperative assessment, or neuroimaging were eligible. Patients previously treated with radiation therapy or intracystic therapies were not eligible. There was no minimum performance level. Age at the time of enrollment was > 12 months and < 21 years. The cost of proton treatment was covered by private or public health insurance or St. Jude Children’s Research Hospital. No patient or family were required to pay for care or ancillary clinical or research-related expenses. Protocol related expenses were entirely supported by St. Jude Children’s Research Hospital.

Surgery

There was no limit to the number of surgical procedures performed prior to proton therapy. Many initial operations were completed by the referring teams. The first tumor-directed surgery was performed at the enrolling institutions in 20 patients and a total of 33 patients had a tumor-directed surgery at the enrolling institutions prior to irradiation. Patients were grouped according to surgery type: none, transsphenoidal, closed (bur hole) placement of catheter and Ommaya reservoir, open (craniotomy) placement of catheter and Ommaya reservoir, endoscopic via bur hole, craniotomy, or multiple approaches.

Proton Therapy

Conformal proton therapy was administered using passive scattering methods with apertures and compensators. The gross tumor volume (GTV) was defined as the postoperative tumor bed and residual tumor, the clinical target volume (CTV) included an anatomically defined margin of 0·5cm surrounding the GTV, and the planning target volume (PTV) was a geometric margin of 0·3–0·5cm surrounding the CTV. The CTV was meant to include subclinical microscopic disease. The PTV was meant to account for variation in daily treatment, beam uncertainties, and aperture design. Proton-specific uncertainties were accounted for in the design of each proton beam. For recording and reporting purposes, a generic PTV was constructed from the CTV by a uniform expansion equal to the lateral setup margins. The prescribed total dose was 54CGE using conventional fractionation of 1·8CGE per day. Weekly non-contrast MR imaging was performed to monitor for changes in the target volume that would require replanning.

As previously reported¹², the clinical target volume margins were 1.0cm for the first 25 photon patients and ≤ 0.5cm for the remaining 76 photon patients. The PTV margins for photon patients ranged from 0.5 cm for the earliest patients to 0.3 cm for those treated using image-guidance.

Clinical Evaluation and Follow-up

Patients were evaluated by a multidisciplinary team at baseline. A physical exam was timed to match imaging assessments which were performed every 3 months dated from the start of treatment through 12 months followed by semi-annual evaluations until year 5, and annually thereafter. MR angiography was performed at the time of baseline and annual surveillance imaging evaluation. Severe stenosis was further evaluated with cerebral angiography. Additional treatment with acetylsalicylic acid was allowed at the discretion of the treating physician. The trial was a collaborative effort between St. Jude Children’s Research Hospital (SJCRH) and the University of Florida Health Proton Therapy Institute (UFHPTI). Baseline and follow-up protocol evaluations were performed at SJCRH. Proton therapy was administered at UFHPTI. Information about ophthalmology and endocrinology are included in the supplement.

Imaging Response and Evaluation

Progression was defined as an increase (>25%) in perpendicular tumor dimensions on two or more successive imaging evaluations two-three years after treatment. The study used the CTCAE Version 4·0 for toxicity and performance reporting. Grade 3 and 4 toxicities were reported to the institutional review board.

Neuropsychology

Neuropsychological assessment was performed at baseline and repeated annually for five years. Intellectual functioning was assessed using the age-appropriate Wechsler scale, with derivation of a full-scale IQ (FSIQ).^13,14 Adaptive functioning, or self-care skills, was assessed using parent report on the Adaptive Behavior Assessment System, Second Edition.¹⁵ The photon comparison group also participated in annual neuropsychological assessments that included the age-appropriate Wechsler scale,^16,17 with derivation of FSIQ or estimated IQ (EIQ) depending on time point; FSIQ and EIQ were combined to maximize data for analysis. In the photon group, parents also provided report of adaptive functioning using the Vineland Adaptive Behavior Scales.¹⁸ All measures are considered gold-standard for the field and scores are age-standardized using large, representative normative samples. All age-standarized scores have a mean of 100 and standard deviation of 15; higher scores indicate better performance.

Outcomes

The primary objective of the RT2CR protocol was to estimate the progression-free (PFS) and overall survival (OS) distributions for children and youth with craniopharyngioma treated with limited surgery and proton therapy using a 5mm clinical target volume margin while monitoring for excessive central nervous system necrosis, clinically significant vasculopathy and permanent neurological conditions or deficits. The secondary objective was to estimate the cumulative incidence of cystic intervention and the event-free survival (EFS) distribution; and compare the distributions of PFS, EFS and OS to those of a historic photon therapy cohort. The cumulative incidence of cystic intervention will be reported separately. The trial was also designed to estimate the distributions of progression-free survival and overall survival for children and youth with craniopharyngioma treated only with primary surgical resection and to compare these distributions with the distributions observed for patients treated with limited surgery and proton therapy. Seven subjects were enrolled on this stratum, and they were not included in the current analysis. Exploratory objectives included potential associations of clinical and treatment factors with the incidence and severity of neurological, endocrine, and cognitive deficits. Descriptively comparing findings for patients treated with protons to those treatment with photon therapy. These outcomes are included in the present analysis. The trial included research exploring sleep, fatigue and quality of life, physical performance and movement, the incidence and severity of structural, functional, and vascular effects of normal brain using magnetic resonance and positron emission tomography imaging methods, growth factor and cytokine responses, and exploratory genetic analyses to better understand the biology that underlies craniopharyngioma, treatment response, and various side effects. Analysis of longitudinal outcomes for the exploratory aims was ongoing at the time of submission and will be reported separately. A complete list of the primary, secondary, and exploratory outcomes is in the supplemental appendix.

Biostatistics

Long-term disease control and outcome data were available for comparison from a cohort of 101 pediatric and adolescent patients with craniopharyngioma enrolled on or treated according to a phase II single institutional protocol^12,19 after 1998. The 5 year progression-free survival (PFS), 5 year overall survival (OS) and their 95% confidence interval (CI) were estimated using the Kaplan-Meier method. Log rank test was used to compare survival distributions between two protocols. A one-sided binomial test was used to test whether the 3-year PFS rate in proton group was lower compared with photon group. For the survival analysis, we included all patients in the proton cohort who had both proton therapy and limited surgery, excluding any who only had surgery. The cumulative incidences of necrosis, vasculopathy, and permanent neurologic deficits unrelated to necrosis or vasculopathy were calculated while disease progression or death were treated as competing events if they occurred prior to the events of interest. The estimate and its 95% CI were calculated for the cumulative incidence at year 5. Gray’s method was used to compare the cumulative incidences between the two protocols. The estimated cumulative incidence with standard error were reported. The linear mixed model was used to investigate the changes of neurocognitive scores over time. CSF shunting was used as a proxy for hydrocephalus when evaluating the impact of clinical covariates on neurocognitive scores. The chisquare test was used to investigate the associations of categorical variables. The t-test was used to compare continuous variables. All the analyses mentioned above are pre-specified in the protocol. In addition, the post-hoc analyses included: (1) Estimate of HR and its confidence interval between two protocols for PFS was calculated using Cox regression method. (2) Fine Gray method was used to estimate the hazard ratio and its 95% CI for necrosis and vasculapathy between two protocols. All analyses were conducted using SAS version 9.4 (SAS Institute Inc).

The study originally included 2 strata with small sample sizes: proton therapy (105 patients), and observation (25 patients) after radical surgery (gross-total resection). Only the proton therapy group was included in this manuscript because the observation strata did not accrue the planned number of patients and meaningful comparisons are not possible. The institutional review board allowed for 10 ineligible patients for a total of 140 patients; however, there were no ineligible patients. The original statistical design developed prior to the activation of the study in 2011 used available data from 93 photon patients. The statistical design was based on a binary endpoint to have 91% statistical power (1-sided test: α=0.05) to detect a drop in the 3-year PFS rate from 94.6% to 86%. The design led to a sample size of 101 patients. The study was stopped early after enrolling 94 proton therapy patients when a newer method of proton therapy became available. A power analysis was performed using 94 subjects instead of the planned 101 subjects. The revised power analysis showed 89.1% instead of 91% power to detect the 3-year PFS rate difference. This study was not designed as a non-inferiority trial. This study was designed as a single-arm study and the hypothesis was H0: p=0.946 vs. Ha: p<0.946, where p is the 3-year PFS. The 94.6% was chosen based on the 3-year PFS from a historical photon trial. The study was designed based on this binary endpoint to have 91% statistical power (1-sided test: α=0.05) to detect a drop in the 3-year success rate from 94.6% to 86%. In the current proton therapy trial, in year 3, 91 out of 94 participants (96.8%) did not have any event. The one-sided binomial test showed that there was no statistical evidence to support that the3-year PFS was less than 94.6% (p=0.89).

Role of the Funding Source

The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.

Results

Between August 22, 2011 and January 19, 2016, 94 patients were enrolled and started proton therapy during the time interval September 19, 2011 to February 22, 2016. Nine patients experienced tumor progression with median time to progression of 3·09 years (IQR: 2.05 to 5.06 years). The median follow-up was 7·52 years (IQR: 6.28 years to 8.53 years) for those who did not experience tumor progression. Outcomes for the proton therapy patients were compared to a cohort of 101 patients treated using photon therapy. The median follow-up with (IQR) was 13.25 years (9.43, 16.65) in photon therapy. The proton and photon groups were comparable based on sex, race, and age at the time of treatment; those treated with photon therapy were more likely to have a CSF shunt (P=0·0002) and higher mean doses to the supratentorial brain and temporal lobe volumes (P<0·0001; Table 1). Those treated using photon therapy had more tumor directed surgery procedures (P=0.0019). Additional information about surgery is provided in the supplemental appendix. Comparing proton and photon (parenthesis) groups by surgery type, 140 (197) tumor directed procedures were performed prior to proton therapy. Craniotomy alone was performed once in 30 (23) or more than once in 9 (13) patients. Placement of a catheter and Ommaya reservoir was performed as a closed (bur hole) procedure alone and once in 6 (14) patients and twice in one (3) patient. Placement of a catheter and Ommaya reservoir alone and once was performed via craniotomy in 9 (4) patients. Transsphenoidal resection was performed alone and once in 9 (2) patients and twice (2) in 2 patients. Endoscopic resection was performed alone and once in 5 (7) patients. Nineteen (30) patients underwent multiple types of tumor-directed surgery, and 4 (5) patients did not have any type of tumor directed surgery. All patients completed proton therapy and received the protocol-specified dose of 54 Gy RBE. The median number of elapsed days was 42 (range 39–49). Cyst expansion required re-planning in 8 cases, cyst drainage in 4 cases, and both in 2 cases.

Table 1.

Patient and treatment characteristics for photon and proton therapy cohorts.

		Photon Cohort	Proton Cohort	P-value
	Patients	101	94
	Follow-up (years)	14.70(12.18–17.27	7.52 (6.28–8.53)	··
	Age at RT (years)	8·98 (6.31–13.02)	9·.39 (6.39–13.38)	0·5224
Sex	Female	46 (45.54%	49 (52.13%)	0·3914
	Male	55 (54.46%)	45 (47.87%)
Race	Asian	1 (0.99%)	2 (2.13%)	<0·0001*
	Black	22 (21.78%)	16 (17.02%)
	Hispanic	2 (1.98%)	··
	Other	··	14 (14.90%)
	White	76 (75.25%)	62 (65.95%)
CSF Shunt	No	74 (73.27%)	88 (93.62%)	0·0002
	Yes	27 (26.73%)	6 (6.38%)
Tumor Directed Procedures	Total	197†	140‡	0·0019
Surgery	Multiple Procedures	46	31	0·1610
	Single Procedures	50	59
	No Procedures	5	4
Mean Dose (Gy)	Brain	16·89 (14.46–18.79)	8·79 (6.74–10.61)	<0·0001
	Temporal Left	18·70 (13.87–21.52)	7·99 (5.60–10.62)	<0·0001
	Temporal Right	18·51 (14.17–22.49)	7·99 (5.60–10.76)	<0·0001

^*When all patients are included, association of race by protocol is significant (P<0·0001. When the major categories Black and White are compared, association of race by protocol is not significant (P=0·8543).

RT=Radiation Therapy

^†96 photon therapy patients

^‡90 proton therapy patients

Progression-free and Overall Survival

The three-year and five-year progression-survivals were, respectively, 96.81% (95% CI: 90.43%, 98.96%) and 93·62% (95% CI: 86.34%, 97.08%). The three-year and five-year overall survivals were: 100% and 100%. There were no deaths on the study. A one-sided binomial test result showed that there was no statistical evidence of a decline in the 3-year progression free rate (p=0.89). Thus, the primary end point for 3-year PFS was not met and there was no statistical evidence to show that the 3-year PFS rate using proton therapy was lower compared to photon therapy. Similarly, there was no difference when comparing progression-free and overall survivals at five years. The hazard ratio for progression-free survival was HR 0.538 (95% CI: 0.243, 1.189). The incidence of tumor progression was lower with proton therapy, although the difference was not statistically significant. The number of events for progression-free survival at 3 and 5 years was 4 of 101 (3.96%) and 10 of 101 (9.90%) for photon and 3 of 94 (3.19%) and 6 of 94 (6.38%) for proton. The number of events for progression-free survival over the study period for photon therapy was 20 of 101 (19.80%) and 9 of 94 (9.57%) for proton therapy. The five-year progression-free and overall survivals were 90·00% (95% CI: 82.21%, 94.49%) and 98·02% (95% CI: 92.31%, 99.50%) for the photon cohort. (Figure 1.)

Figure 1.

Progression-free survival (upper) and overall survival (lower), comparing pediatric and adolescent patients with craniopharyngioma treated with passively scattered proton therapy to photon conformal radiation therapy.

Necrosis

Two proton patients experienced CNS necrosis 3.9 and 7.4 months after the initiation of proton therapy. Both had imaging evidence of cerebral ischemia after their initial surgery, and both were treated successfully for necrosis with hyperbaric oxygen therapy. The cumulative incidence of necrosis at five years was 2·13% ± 1·50% for patients treated with proton therapy. There was no difference when comparing these patients to those treated with photon therapy where the cumulative incidence at five years was 1·98% ± 1·39%. Two patients in the photon cohort developed necrosis at 4.7 and 5.8 months. (Figure 2.) The incidence of radiation necrosis was higher with proton therapy, although the difference was not statistically significant. Two (2%) patients in the proton cohort and 2 (2%) in the photon cohort developed necrosis.

Figure 2.

Cumulative incidence of cumulative incidence of necrosis (upper left), vasculopathy (upper right), severe vasculopathy (middle left), and permanent neurologic conditions unrelated to necrosis or vasculopathy (middle right and change from normal vision to abnormal vision (lower left) in pediatric and adolescent patients with craniopharyngioma treated with passively scattered proton therapy compared to photon conformal radiation therapy, corrected for change in visual field assessment from confrontational testing to automated static perimetry.

Vasculopathy

Five patients had pre-existing vasculopathy. Two of these patients developed severe vasculopathy and required revascularization. After proton therapy, seven patients develop vasculopathy. Four cases were characterized as severe and three were characterized as moderate vasculopathy. Severe cases had severe stenosis based on MRA findings that were confirmed by cerebral angiography. These patients showed no evidence of compensation of perfusion deficits. All were treated with revascularization surgery. Moderate cases had severe stenosis based on MRA findings that were confirmed by cerebral angiography; however, they had evidence of compensation of perfusion deficits on MRI perfusion studies. All were treated with low-dose (81mg daily) acetylsalicylic acid therapy at the discretion of the treating physician. Finally, two patients who had MRA evidence of severe stenosis who underwent MR perfusion and did not show deficits. These patients did not have cerebral angiography. The cumulative incidence of vasculopathy at five years was 7·87% ± 2·87% and the cumulative incidence of severe vasculopathy at five years was 4·49% ± 2·21%. This was similar to the five-year cumulative incidence of severe vasculopathy in our comparative photon cohort 4·99% ± 2·19%. (Figure 2.) The time to severe vasculopathy was 24.0, 25.9, 36.9, and 43.73 months in the proton cohort and 42.27, 44.3 and 51.0 months in the photon cohort. Four (4%) patients in the proton cohort and 5 (5%) in the photon cohort developed severe vasculopathy.

Permanent neurological conditions not related to vasculopathy or necrosis

The three-year cumulative incidence of permanent neurological conditions was 3·19% ± 1·82% for patients treated with proton therapy compared to 2·97% ± 1·70% for those treated with photon therapy. Permanent neurological conditions in the proton therapy group included paresthesia, basal ganglia syndrome and dystonia. (Figure 2.) Three (3%) patients in the proton cohort and 3 (3%) in the photon cohort developed permanent neurologic conditions.

Vision

A total of 818 ophthalmology evaluations were available for 94 patients through the time of analysis. There were 54 patients with normal visual acuity and visual field at baseline. At the last follow-up, any visual field deficit was observed in ten patients and a decline in visual acuity to the level of monocular impairment was observed in one patient. The cumulative incidence of decline was 18·52 ± 5·34% at three years and 20·86 ± 5·69% at five years. When excluding decline discovered when transitioning from confrontational to Humphrey visual field testing in younger patients, and temporally associated tumor progression, the cumulative incidence was 7·41 ± 3·60% at both three- and five-year time points. (Figure 2.) Optic atrophy was not associated with change in vision. Amongst the 40 patients with abnormal vision prior to proton therapy, the cumulative incidence of improvement was 25·00% ± 6·95% at three years and 27·50% ± 7·17% at -five years. Visual acuity remained largely stable or improved for those with normal or impaired vision at the start of treatment. The cumulative incidence of decline was 6·04% ± 3·42% for the photon cohort. Four (7%) of 54 patients with normal vision in the proton cohort and 3 (6%) of 50 in the photon cohort with normal vision developed visual impairment with long-term follow-up.

Endocrinopathy

Amongst the 94 proton therapy patients, 9 (9.5%) had diabetes insipidus at diagnosis and 45 (47.9%) acquired diabetes insipidus at surgery. None of the remaining 40 (42.6%) patients developed diabetes insipidus. At the time of evaluation for proton therapy, growth hormone deficiency was present in 70 of 94 (74.5%), central hypothyroidism in 56 of 94 (59.6%), and central adrenal insufficiency in 59 of 94 (62.8%). Amongst the 101 photon therapy patients, 7 (6.9%) had diabetes insipidus at diagnosis and 47 (46.5%) acquired diabetes insipidus at surgery. None of the remaining 47 (46.5%) patients developed diabetes insipidus. When evaluated before photon therapy, growth hormone deficiency was present in 30 of 36 (83.3%) patients based on provocative testing, 21 of 54 (38.8%) based on clinical factors and screening laboratory evaluation, 9 were not tested because of logistical or medical reasons, and 2 were previously prescribed growth hormone replacement. Growth hormone deficiency was confirmed or suspected in 53 of 92 (57.6%) patients. Central hypothyroidism was present in 61 of 101 (60.3%) patients and central adrenal insufficiency was present in 53 of 101 (52.5%). The 3 and 5-year cumulative incidence of hypothyroidism, adrenal insufficiency, and hypogonadism was not significantly different comparing proton and photon patients (supplemental appendix).

Neuropsychology

Statistically significant differences (points/year, P-value) in longitudinal scores were observed with decreasing values for IQ (−1·0939, P=0·0070) and adaptive behavior (−1·4849, P=0·0303) in those treated with photon therapy compared to those treated with proton therapy. (Figure 3.) The estimated difference was 4·58 points over five years for IQ, and 7·34 points over five years for adaptive behavior. When the dosimetry information was combined for all patients, the mean dose to the individual temporal lobes had a significant impact on longitudinal change (points/Gy/year) in IQ. For mean dose to temporal left lobe, the decreasing rate over time was −0·0661/Gy/year (p=0·0142) after taking shunt and age at RT into account. For mean dose to temporal right lobe, the decreasing rate over time was −0·0738 /Gy/year (p=0·0078) after taking shunt and age at RT into account. We included CSF shunting as a covariate in our analysis to represent a severe form of hydrocephalus. Baseline IQ scores were 10.8 points higher for those with no CSF shunt (p=0.0037) when age and protocol were included in the model of change in IQ over time. Baseline values (±SD) for IQ based on the model were 103.41(3.33) for no shunt versus 92.61(4.46) for shunt.

Figure 3.

Estimated mean values (± SD) and modeled curves for longitudinal IQ (upper) and Adaptive Behavior (lower) scores comparing pediatric and adolescent patients with craniopharyngioma treated with passively scattered proton therapy to those treated with photon conformal and intensity-modulated radiation therapy. The differences in the curves were statistically significant IQ (1·0939 points/year, P=0·007) and Adaptive Behavior (1·4849 points/year, P=0·0303).

Common Toxicity Reporting

The grade 1–2 events recorded (number of patients, %) for more than 10% of patients were headache (17, 18.09%) and respiratory disorders (13, 13.83%). Grade 3 events occurring more than one patient were gastrointestinal (3, 3.19%), vomiting (2, 2.13%), central nervous system necrosis (2, 2.13%), headache (6, 6.38%), seizure 5 (5.32%), other central nervous system disorders ((3, 3.19%), vascular disorders (6, 6.38%), thromboembolic event (2, 2.13%), and surgery or medical procedures (2, 2.13%). Five patients experienced grade 5 events including sepsis (1, 1.06%), hyponatremia (1, 1.06%), hypernatremia (2, 2.13%), and eye disorders (1, 1.06%). In total, 10 serious adverse events (sepsis, hypernatremia, hyponatremia, headache, paresthesia, thromboembolic event) were reported in 5 patients. There were no treatment or drug-related deaths. CTCAE V 4·0 adverse events are summarized in Table 2.

Table 2.

Adverse events grade 1 or 2 occurring in ≥10% of patients and recorded grade 3, 4, and 5 events by patient.

94 patients Treated by Proton Therapy		Grade*
		2	3	4
EVENTCATEGORY	EVENTDESCRIPTION	N(%)	N(%)	N(%)
Blood and lymphatic system disorders	Anemia		1(1%)
	Blood and lymphatic system disorders		1(1%)
Endocrine disorders	Adrenal insufficiency		1(1%)
	Endocrine disorders		1(1%)
Eye disorders	Eye disorders			1(1%)
Gastrointestinal disorders	Gastritis		1(1%)
	Gastrointestinal disorders		3(3%)
	Vomiting		2(2%)
Infections and infestations	Catheter related infection		1(1%)
	Device related infection		1(1%)
	Sepsis			1(1%)
Metabolism and nutrition disorders	Dehydration		1(1%)
	Hypernatremia			2(2%)
	Hyponatremia			1(1%)
Nervous system disorders	Central nervous system necrosis		2(2%)
	Dysarthria		1(1%)
	Dysphasia		1(1%)
	Headache	17(18%)
	Headache		6(6%)
	Hydrocephalus		1(1%)
	Hypersomnia		1(1%)
	Nervous system disorders		3(3%)
	Oculomotor nerve disorder		1(1%)
	Paresthesia		1(1%)
	Seizure		5(5%)
	Stroke		1(1%)
Psychiatric disorders	Psychiatric disorders		1(1%)
	Psychosis		1(1%)
Respiratory, thoracic and mediastinal disorders	Respiratory, thoracic and mediastinal disorders	13(14%)
	Respiratory, thoracic and mediastinal disorders		1(1%)
Surgical and medical procedures	Surgical and medical procedures		2(2%)
Vascular disorders	Thromboembolic event		2(2%)
	Vascular disorders		6(6%)

^*CTCAE v 4.0, no patient experienced a grade 5 event.

Discussion

This study documents disease control in pediatric and adolescent patients with craniopharyngioma during the first five years after radiation therapy using passively scattered protons and demonstrates similar tumor control rates when compared to radiation therapy using photons. There was little doubt about the equivalence between the two radiation modalities when the study was initiated. Passively scattered proton therapy provides a relatively uniform dose distribution across the targeted volume and the use of a planning target volume mitigated treatment delivery uncertainties. Answering this fundamental question was required based on the applied target volume definitions and the limited reported experience using proton therapy in children with this tumor. Because of the small number of patients with progression after proton therapy, further follow-up is required to determine if there is an association between clinical and treatment factors and outcomes. Our experience showed that race and permanent CSF shunting impacted progression-free survival after photon therapy and the critical need to monitor these patients with frequent imaging during treatment.²⁰

Nearly all the patients treated on this study had tumor-directed surgery prior to proton therapy. The low incidence of craniopharyngioma, and often acute presentation, explain why the first tumor directed surgery was performed at the referring institutions in most cases. This study using first-generation proton therapy showed contemporary trends in the surgical management of craniopharyngioma. Patients were treated with a variety of surgical approaches: craniotomy, multiple surgical approaches, endoscopic resection, transnasal, and open and closed Ommaya catheter placements. Additional studies will be required to determine if the type of surgical approach ultimately affects tumor control or functional outcome after irradiation. The trend toward using hypothalamus-sparing approaches is apparent in many series including Madsen et al.²¹ who showed endonasal resection results in less injury to children and a higher rate of gross-total resection when compared to surgery performed using open procedures. They showed that rates of ischemia were higher in patients treated using open procedures when matched for initial tumor size, and that body mass index was uniformly higher in the open surgery group.

Considering that there is no difference in disease control when comparing patients treated with radical surgery to those treated with more limited surgery and irradiation,³ we have been selective when considering radical surgery and favor the use of limited surgery and irradiation. Rock et al.²² showed that more than 30% of craniopharyngioma patients have post-operative complications and 15% have major complications defined as “single or multi-organ dysfunction that would require intermediate care or intensive care unit management.” Hypertension and length of surgery were found to be risk factors when determining the rate of severe complications. The incidence and severity of post-operative complications appear to be reduced in the modern treatment era. Fouda et al.²³ evaluated the impact of contemporary treatment approaches. They observed reductions in the incidence of visual complications, panhypopituitarism and diabetes insipidus, cognitive impairment, and obesity in the modern era. Tan et al.²⁴ showed no change in overall rates of recurrence, hypothalamic obesity, hypothalamic damage, or vision loss comparing patients treated prior to 2000 to those after 2000; however, they did find a lower incidence of diabetes insipidus and panhypopituitarism amongst patients treated with partial resection and limited surgery when compared to complete resection.

When dose reductions to normal tissues were assessed by modality, the degree of effectiveness of proton therapy over advanced photon methods was found to depend on tumor location.²⁵ For those with suprasellar tumors, proton therapy was more likely to be superior to photon therapy by reducing exposure of the subventricular zones and hippocampi. Although our goal is to eventually evaluate these patients for their long-term cognitive outcomes, addressing the issue of severe complications of proton therapy and preserving the prescribed dose of 54CGE is of immediate importance. Indeed, the promise of proton therapy for suprasellar tumors largely hinges on improved cognitive outcomes. In a series of children treated using proton therapy, those given focal irradiation, including young children and those with craniopharyngioma, appeared to be spared significant cognitive decline when compared to patients treated with craniospinal irradiation, although the follow-up was short in that series.²⁶ Current findings are consistent with initial reports in the literature that suggest reduced risk for declines in intelligence^27,28 and adaptive functioning²⁹ following proton therapy. The current study has the advantage of a large, prospectively followed sample, with minimal attrition, and cognitive surveillance for five years, as well as a well-matched photon comparison group. Further, mean radiation dose to the temporal lobes was related to decline in intellectual functioning, providing direct evidence that reduced dose to normal tissues is driving at least some of the cognitive benefit of proton therapy.

Necrosis was observed in two patients. Necrosis is an uncommon but expected complication of radiation therapy. In children with brain tumors treated with radiation therapy, the highest incidence is observed in those treated with high dose craniospinal irradiation and boost treatment to posterior fossa subsites.³⁰ Radiation necrosis predominantly affects white matter and is thought to occur after small artery injury and thrombotic occlusion. It results from increased tissue pressure from edema and vascular injury leading to infarction, damage to endothelial cells, and fibrinoid necrosis of small arteries and arterioles.³¹

Vasculopathy is common among patients with craniopharyngioma and is responsible for some of the devastating effects observed after radiation therapy. The incidence and time to onset and factors predictive of severe and life-threatening vasculopathy have not been studied systematically.³² Surgery is believed to be responsible for peri-operative vasospasm and ischemia whereas late events are largely attributable to radiation dose and volume. Boekhoff et al.³³ reported a 11% incidence of cerebral ischemia in surgically treated craniopharyngioma. In Multivariable analysis in their study showed hydrocephalus and gross-total resection as significant risk factors. Amongst irradiated patients, ischemia was present prior to irradiation in all 12 irradiated patients. Managing vasculopathy is often difficult because medical or surgical intervention is instituted or considered only after the process has become established. This series shows that the cumulative incidence of vasculopathy is similar comparing proton and photon modalities, however, longer-term evaluation is required. In the meantime, we seek new ways to identify risk factors for vasculopathy, improve imaging protocols to study vascular effects of irradiation, and streamline assessments for early intervention.

We found that 40 of 94 patients (43%) had visual impairment prior to irradiation. This value was similar or better than the proportion of patients with visual impairment at diagnosis (74% adult and 59% children) as reported by Wijnen et al.³⁴ Although the proportion of patients with any level of visual impairment increased to 54% in our series, this value remained lower than the long-term proportion of patients with visual acuity (63%) and visual field (66%) impairment noted in other series that combined irradiated and non-irradiated patients. The impact of surgery on visual outcomes was studied by Akinduro et al.³⁵ Their systematic review of adults treated with surgery showed that there was no difference in impaired or improved vision comparing patients treated with GTR to those with <GTR. The rate of improved vision was 10% regardless of surgical extent and the rates of improved vision were 42% for GTR compared to 38% for <GTR. They concluded that GTR was not necessary to achieve meaningful decompression in patients with visual impairment at diagnosis.

There is a high incidence of hypopituitarism and diabetes insipidus in patients treated for craniopharyngioma. Although detailed information about hypopituitarism and diabetes insipidus was available for both cohorts, it is not considered a worthwhile endpoint for this study given the significant proportion of patients who present with pre-existing deficits prior to radiotherapy. Further, diabetes insipidus is not considered a radiotherapy lesion because the incidence is low or non -existent. As shown in these series, diabetes insipidus most often arises after surgery and may occasionally be present at the time of diagnosis.

The current study has limitations. The use of a historic cohort is susceptible to the inherent differences encountered when performing and interpreting assessments, and external factors beyond the healthcare environment. The relative additional costs of proton therapy, including equipment, staffing, and access, compared with photon therapy are significant and affect the generalizability of our findings. Craniopharyngioma is a rare disease, and a randomized study would not be feasible. Both the proton study and photon study assessments were carried out under the supervision of the same follow-up team and were consistent across both studies.

Summary

There was no difference between proton beam therapy and photon therapy in terms of progression-free, event-free, or overall survival. The potential benefit of proton therapy in the treatment of craniopharyngioma is to reduce the volume of normal brain exposed to low doses. This is most relevant to critical structures not adjacent to the targeted volume. A reduction in the volume of normal brain exposed to radiation seems to reduce the cognitive effects of irradiation. Further improvements will require careful study, assessment of relevant domains, and long-term follow-up.