Proton Therapy Mediates Dose Reductions to Brain Structures Associated With Cognition in Children With Medulloblastoma

Abstract

Purpose:

Emerging evidence suggests proton radiation therapy may offer cognitive sparing advantages over photon radiation therapy, yet dosimetry has not been compared previously. The purpose of this study was to examine dosimetric correlates of cognitive outcomes in children with medulloblastoma treated with proton versus photon radiation therapy.

Methods and Materials:

In this retrospective, bi-institutional study, dosimetric and cognitive data from 75 patients (39 photon and 36 proton) were analyzed. Doses to brain structures were compared between treatment modalities. Linear mixed-effects models were used to create models of global IQ and cognitive domain scores.

Results:

The mean dose and dose to 40% of the brain (D40) were 2.7 and 4.1 Gy less among proton-treated patients compared with photon-treated patients (P = .03 and .007, respectively). Mean doses to the left and right hippocampi were 11.2 Gy lower among proton-treated patients (P < .001 for both). Mean doses to the left and right temporal lobes were 6.9 and 7.1 Gy lower with proton treatment, respectively (P < .001 for both). Models of cognition found statistically significant associations between higher mean brain dose and reduced verbal comprehension, increased right temporal lobe D40 with reduced perceptual reasoning, and greater left temporal mean dose with reduced working memory. Higher brain D40 was associated with reduced processing speed and global IQ scores.

Conclusions:

Proton therapy reduces doses to normal brain structures compared with photon treatment. This leads to reduced cognitive decline after radiation therapy across multiple intellectual endpoints. Proton therapy should be offered to children receiving radiation for medulloblastoma. © 2024 Elsevier Inc. All rights reserved.

Introduction

Treatment for children with medulloblastoma includes surgery followed by radiation therapy (RT) to the craniospinal axis, as well as a boost to the tumor bed and chemotherapy. Radiation may be delivered using protons or photons, depending on available technology. Considerable effort has been made to decrease potential long-term effects from treatment for pediatric malignancies. Although treatment for medulloblastoma still includes treatment of the whole brain and spine, the boost volume has been reduced from the entire posterior fossa to the surgical bed with a margin.¹ Smaller volumes of irradiated brain are known to result in better intellectual capacity posttreatment,^2,3 which is one of the main advantages of proton therapy compared with photon irradiation.

Until recently, protons were being used in pediatrics based solely on first principles, to reduce the risk of secondary effects from nontarget dose (such as cognitive change, endocrinopathy, cardiomyopathy, secondary malignancy).⁴ Dosimetric studies also have shown hypothetical dosimetric improvements with protons versus photons, including less dose to supratentorial brain and temporal lobes.⁵ Recent work has shown an improvement in cognitive outcomes among pediatric patients with medulloblastoma treated with protons compared with a cohort treated with photons,⁶ suggesting that proton RT should be standard of care.⁷

The mechanism behind the cognitive gain observed in treatment of patients with medulloblastoma treated with proton therapy—who all receive craniospinal irradiation (CSI), regardless of radiation modality—is subject to some controversy.⁸ Debates in the clinical literature have argued about whether proton boost provides a clinically significant benefit with respect to sparing normal brain tissue.^9,10

The association of proton therapy with cognitive performance has already been demonstrated.⁶ However, the goal of the present study was to elucidate the mechanistic reason behind the superiority of proton therapy over photon therapy in the treatment of children with medulloblastoma. We hypothesized that proton therapy results in less dose to normal brain structures from the boost portion of RT, leading to improvement in cognitive scores.

Methods and Materials

This was a retrospective study of children treated with RT at 2 institutions for medulloblastoma. Eligible patients were identified from institutional databases at The Hospital for Sick Children (photon RT only, Toronto, Ontario, Canada) and Texas Children’s Hospital (proton RT only, Houston, TX) after approvals from the research ethics and institutional review boards, respectively. Waiver of consent was obtained for this minimal risk, retrospective study. Eligibility criteria were previously published⁶; in summary, this study included pediatric medulloblastoma patients from both centers treated following the SJMB03 (NCT00085202) or SJMB12 (NCT01878617) protocols between 2007 and 2018, with available cognitive evaluations. In the present study, patients must have had available digital radiation dosimetry data in RT-DICOM format.

All patients underwent resection via craniotomy and received RT and chemotherapy. Patients received RT to the whole brain and spine, followed by tumor bed boost (0.5 cm or 1 cm Clinical Tumour Volume expansion on the Gross Tumour Volume, defined as tumor bed and residual tumor). Patients with the smaller boost margin were considered separately in a subgroup analysis. All photon treated patients were treated with three dimensional conformal radiation therapy (3DCRT) for the CSI component and intensity modulated RT for the boost component.

Brain structures (right and left frontal lobes, right and left temporal lobes, right and left hippocampi, brain, supratentorial brain) were manually segmented using a shared contouring atlas (Appendix E1) in RayStation (RaySearch Laboratories). Composite dosimetric parameters including D40 (dose to 40% of the structure), D50 (dose to 50% of the structure), and mean dose were calculated for all structures. Intelligence scores were evaluated at baseline and prospectively followed for all included patients. We evaluated scores that represented global IQ (ie, Wechsler/Stanford-Binet Full Scale IQ score; Wood-cock-Johnson General Intellectual Ability score), as well as verbal comprehension index (VCI), perceptual reasoning index (PRI), working memory index (WMI), and processing speed index (PSI). All intellectual scores are standardized (population mean = 100, standard deviation = 15), with lower scores indicating worse functioning.

Patient characteristics, dosimetry data, and baseline cognitive outcomes were summarized using descriptive statistics and compared between RT modalities using independent t test for continuous variables and the χ² or Fisher exact test for categorical variables. Cognitive outcomes (VCI, PRI, WMI, PSI, global IQ) were each analyzed using linear mixed effects models with a random intercept and modelled as follows, based on prior work⁶: Cognitive outcome score = ß₀ (intercept) + ß₁ dose parameter (Gy) × Time since RT (years) + ß₂ age at RT (years) × Time since RT (years) + ß₃ posterior fossa syndrome (yes/no). Models were graphically presented using the predicted values of cognitive outcomes versus dose parameter and posterior fossa model terms, with overlaid data points by RT modality. Dose structures were selected for inclusion in presented models based on strength of association with the cognitive endpoint and statistical significance. We also adjusted P values for multiple comparisons to different brain structures using the false discovery rate (FDR) method, which are reported as Q values. All analyses were done using R version 4.2.2.

Results

A total of 75 patients were identified meeting eligibility criteria (36 treated with proton RT, 39 treated with photon RT). Baseline characteristics of both cohorts are reported in Table 1. More proton patients underwent ventriculoperitoneal (VP) shunt placement; 31% of proton patients had VP shunt compared with 10% of photon patients (P = .04). There were 33% of proton patients who received high-dose CSI (>23.4 Gy) compared with 15% of photon patients, although this difference was not statistically significant. The total RT prescription was slightly lower among proton patients, compared with photon patients (mean, 5462 vs 5608 cGy; P < .001). Median times from start of RT to first (baseline) and last psychological evaluations were 0.2 and 4.8 years, respectively. The median and mean follow-up times (from start of RT to last psychological evaluation) for each of the compared cohorts are as follows: 4.1 and 4.7 years (respectively) for patients treated with proton RT, and 5.2 and 4.7 years (respectively) for patients treated with photon RT. Baseline cognitive scores did not differ significantly between groups and are reported in Appendix E2.

Table 1.

Baseline characteristics of patients, stratified by treatment modality

Characteristic	All patients N = 75		Photon n = 39		Proton n = 36		Comparison P value
	No.	%	No.	%	No.	%
Sex							.79
Male	52	69	26	67	26	72
Female	23	31	13	33	10	28
VP shunt	15	20	4	10	11	31	.04
Posterior fossa syndrome	38	51	19	49	19	53	.90
CSI dose, cGy							.07
1500	1	1	0	0	1	3
2340	56	75	33	85	23	64
3240	1	1	0	0	1	3
3600	17	23	6	15	11	31
	Mean (SD)	Range (min-max)	Mean (SD)	Range (min-max)	Mean (SD)	Range (min-max)	P value
Age at diagnosis, y	8.7 (3.0)	3.6–15.3	8.3 (3.2)	3.6–15.3	9.0 (2.8)	3.7–14.4	.34
Age at RT, y	8.7 (3.0)	3.6–15.4	8.4 (3.2)	3.6–15.4	9.1 (2.8)	3.8–14.5	.33
CSI dose, cGy	2626 (550)	1500–3600	2534 (461)	2340–3600	2727 (623)	1500–3600	.14
Total RT dose, cGy	5538 (126)	5100–5940	5608 (97)	5580–5940	5462 (108)	5100–5580	<.001
Maximal tumor diameter (cm)	4.6 (1.0)	2.0–7.4	4.6 (1.1)	2.0–7.4	4.6 (0.9)	3.0–6.6	.85
Number of psychological evaluations	4.1 (2.5)	1–11	3.8 (2.1)	1–8	4.3 (2.8)	1–11	.43

Abbreviations: CSI = craniospinal irradiation; RT = radiation therapy; VP = Ventriculoperitoneal; Sex = refers to sex assigned at birth.

Doses to normal brain tissues are reported in Table 2. There was a statistically significant reduction in doses to the brain (mean and D40), supratentorial brain (mean), left and right temporal lobes, and left and right hippocampi among patients treated with proton therapy, compared with photon radiation. The whole brain D40 was 4.1 Gy less among proton-treated patients (P = .007), and the left and right hippocampi received 11.2 Gy less among proton-treated patients (P < .001 for both). The magnitude of difference in doses to brain structures exceeded the average 1.46 Gy higher total prescription dose among photon-treated patients, compared with proton-treated patients (Table 1). A visual representation of doses to all structures is provided in Figure 1.

Table 2.

Comparison of mean composite doses to normal brain tissues, stratified by treatment modality

Dose	All patients (N = 75)	Photon (n = 39)	Proton (n = 36)	P value
Whole brain
Mean	33.4 (5.2)	34.7 (4.6)	32.0 (5.6)	.03
D50	28.9 (6.2)	29.6 (6.0)	28.2 (6.4)	.33
D40	30.9 (6.6)	32.9 (6.2)	28.8 (6.4)	.007
Left frontal
Mean	28.3 (5.9)	28.6 (5.4)	28.1 (6.4)	.71
D50	27.7 (6.0)	27.7 (5.6)	27.8 (6.4)	.92
D40	28.3 (6.2)	28.6 (6.0)	27.9 (6.5)	.65
Right frontal
Mean	28.2 (5.8)	28.4 (5.4)	28.1 (6.3)	.79
D50	27.9 (6.6)	28.1 (6.9)	27.8 (6.3)	.85
D40	27.9 (6.1)	28.0 (5.9)	27.8 (6.3)	.92
Left temporal
Mean	34.7 (6.9)	38.0 (5.4)	31.1 (6.7)	<.001
D50	33.8 (8.0)	37.9 (6.0)	29.3 (7.6)	<.001
D40	35.0 (8.2)	39.6 (5.9)	30.2 (7.6)	<.001
Right temporal
Mean	34.5 (6.9)	37.9 (5.5)	30.8 (6.4)	<.001
D50	33.5 (7.9)	37.6 (6.2)	29.0 (7.1)	<.001
D40	34.8 (8.2)	39.4 (6.2)	29.7 (7.2)	<.001
Left hippocampus
Mean	41.2 (9.1)	46.5 (7.6)	35.3 (6.8)	<.001
D50	41.0 (9.9)	47.1 (7.9)	34.5 (7.3)	<.001
D40	42.4 (9.5)	47.9 (7.7)	36.5 (7.6)	<.001
Right hippocampus
Mean	40.0 (9.0)	45.3 (7.4)	34.1 (6.6)	<.001
D50	39.8 (9.7)	45.8 (7.9)	33.3 (7.0)	<.001
D40	40.9 (9.5)	46.3 (8.1)	35.1 (7.2)	<.001
Supratentorial brain
Mean	31.2 (5.4)	32.5 (4.8)	29.7 (5.8)	.03
D50	28.3 (6.0)	28.6 (5.8)	27.8 (6.3)	.57
D40	29.4 (6.20)	30.4 (5.9)	28.3 (6.4)	.13

Data are presented as gray (SD).

Abbreviations: D40 = dose to 40% of the brain structure; D50 = dose to 50% of the brain structure.

Fig. 1.

Composite doses received by brain structures, stratified by radiation modality. Medians are shown with a thick horizontal line, and means are shown with a closed circle. The boxplots represent first and third quartiles (IQR), whiskers denote values 1.5 × IQR away from the first and third quartiles, and outliers are shown with open circles. Abbreviations: PRI = perceptual reasoning index; VCI = verbal comprehension index; WMI = working memory index.

To ensure a balanced comparison between groups, a subgroup analysis of patients excluding those who received a boost CTV of 0.5 cm or 15 Gy CSI was created; these exclusion criteria removed 14 proton-treated patients (who were treated on SJMB12). This subgroup analysis is presented in Appendix E3; proton-treated patients continued to receive less dose to the brain (mean and D40), supratentorial brain, left and right temporal lobes, and left and right hippocampi compared with photon-treated patients.

Subgroup analysis was also completed on those patients who received lower-dose CSI (23.4 Gy). This included 33 photon patients and 23 proton patients. This confirmed statistically significant decrease in doses to brain structures within the proton group compared with photon-treated children (Appendix E4).

Models of cognition and associations with dose to brain structures are presented in Table 3. Higher dose to the whole brain led to reduced VCI. Reduced PRI was associated with increasing right temporal lobe D40 dose. WMI scores were negatively correlated with left temporal mean dose, which persisted after correction for multiple comparisons. Increased dose to 40% of the brain was associated with reduced PSI and global IQ scores, the former of which persisted after correction for multiple comparisons. In addition, age at RT was associated with PRI, WMI and PSI. The presence of posterior fossa syndrome led to declines in PRI, WMI, PSI and global IQ. A visual representation of estimated cognitive endpoints by dose to brain structures is presented in Figure 2, overlaying actual cognitive scores and corresponding doses to brain structures to provide a visual guide to the dose differences between patients who receive photon versus proton RT.

Table 3.

Models of cognitive endpoints

Endpoint	Variable	Coefficient	SE	P value	Q value
VCI	ß0 (intercept)	96.134	2.329	<.001
	ß1 (whole brain mean × time since RT)	−0.037	0.017	.033	0.12
	ß2 (age at RT × time since RT)	0.093	0.066	.159
	ß3 (posterior fossa syndrome)	−3.279	3.183	.306
PRI	ß0 (intercept)	96.664	2.873	<.001
	ß1 (right temporal lobe D40 × time since RT)	−0.038	0.019	.048	0.094
	ß2 (age at RT × time since RT)	0.161	0.070	.024
	ß3 (posterior fossa syndrome)	−9.364	3.970	.021
WMI	ß0 (intercept)	99.917	2.318	<.001
	ß1 (left temporal lobe mean dose × time since RT)	−0.103	0.021	<.001	<0.001
	ß2 (age at RT × time since RT)	0.256	0.075	.001
	ß3 (posterior fossa syndrome)	−9.533	3.145	.003
PSI	ß0 (intercept)	91.967	2.349	<.001
	ß1 (brain D40 × time since RT)	−0.107	0.021	<.001	<0.001
	ß2 (age at RT × time since RT)	0.248	0.071	.001
	ß3 (posterior fossa syndrome)	−15.812	3.217	<.001
Global IQ	ß0 (intercept)	95.295	2.743	<.001
	ß1 (brain D40 × time since RT)	−0.036	0.017	.030	0.18
	ß2 (age at RT × time since RT)	0.072	0.0573	.211
	ß3 (posterior fossa syndrome)	−8.214	.803	.034

The coefficients and equation are represented here: Cognitive outcome score = ß₀ (intercept) + ß₁ dose parameter (Gy) × Time since RT (years) + ß₂ age at RT (years) × Time since RT (years) + ß₃ posterior fossa syndrome (yes/no).

Abbreviations: PRI = perceptual reasoning index; PSI = processing speed index; RT = radiation therapy; VCI = verbal comprehension index; WMI = working memory index.

Fig. 2.

Modeled (estimated) cognitive index values at 3 years postradiation (y-axis), based on dose (Gy, x-axis) to brain structures most associated with cognition. Estimated cognitive scores (colored lines) are stratified by the presence or absence of posterior fossa syndrome. Overlaid on the graphs is a scatter plot of delivered dose to each individual patient and their actual test score values. By inspection, one can see that the filled circles predominate the left side of the x-axis, reflecting that patients who had proton therapy had less dose to brain structures overall. Abbreviation: RT = radiation therapy.

Discussion

Our previous work demonstrated a clear association of proton therapy with improved cognitive outcomes in children treated for medulloblastoma.⁶ In the present study, we elucidate the mechanism behind this benefit of proton therapy by evaluating dosimetry to treated patients. We have shown that proton therapy reduces dose to normal brain structures across composite RT plans because of the more focused nature of the proton boost compared with photon treatment. We further demonstrate that reduction in doses to normal brain structures is associated with less cognitive decline across intellectual endpoints. Statistically significant associations of dose with VCI, PRI and global IQ were lost after adjustment for multiple comparisons, although there remained a strong negative association between left temporal lobe mean dose and working memory (WMI), as well as dose to 40% of the brain (brain D40) and PSI.

Using the modeled coefficients in Table 3, we estimated the detriment to cognitive function at 3 and 5 years after RT for the dosimetric parameters that remained significant after correction for multiple comparisons (Table 4). For example, an increase in D40 of 10 Gy to the whole brain would be expected to result in a decrease in PSI of 3.2 points at 2 years after RT and 5.4 points at 5 years. This data may help clinicians describe the actual benefit to patients from these dose reductions.

Table 4.

Estimated detriment to cognitive scores by varying doses to brain structures

Brain structure	Dose metric	Cognitive endpoint	Estimated change in cognitive endpoint at 3 y post-RT			Estimated change in cognitive endpoint at 5 y post-RT
			5 Gy increase	10 Gy increase	15 Gy increase	5 Gy increase	10 Gy increase	15 Gy increase
Left temporal lobe	Mean dose	WMI	−1.5 points	−3.1 points	−4.6 points	−2.2 points	−2.9 points	−7.7 points
Brain	D40	PSI	−1.6 points	−3.2 points	−4.8 points	−2.7 points	−5.4 points	−8.0 points

Intellectual scores are standardized to a population mean of 100 and a standard deviation of 15. Lower scores indicate worse functioning.

Abbreviations: PSI = processing speed index; RT = radiation therapy; WMI = working memory index.

This mechanistic study shows a dosimetric benefit to proton RT compared with photon RT when using craniospinal irradiation followed by RT boost, and a strong correlation of that difference to cognitive outcomes. Although consensus opinion supports proton therapy referral for children with brain tumors including medulloblastoma,^7,11 there remain many jurisdictions where access to proton therapy is challenging due to insurance delays,¹² denials,¹³ geography,¹⁴ or access.¹⁵ This study helps reduce the hesitation of many radiation oncologists regarding the potential dosimetric and cognitive benefits of proton therapy for children with medulloblastoma, notwithstanding the fact that the CSI component of treatment is similar between radiation modalities.¹⁰

There is ample evidence supporting proton therapy as an effective modality for the treatment of pediatric medulloblastoma, with respect to tumor control.^16–18 The present study also contributes to emerging evidence about the benefits of proton therapy with respect to cognition and reduction of late toxicities in this patient population.¹⁹ Specifically, Eaton et al evaluated intellectual functioning in children with medulloblastoma, using a matched case-control design.²⁰ In their study, children treated with proton RT had higher IQ and verbal and nonverbal scores, compared with those treated with x-rays. Processing speed and working memory did not differ by treatment modality. This contrasts with work by Kahalley et al, which found that working memory was better in children treated with proton therapy⁶; this difference may be due to a smaller data set in the Eaton et al study (n = 37, compared with n = 79 in Kahalley et al).

This is the first-known study that specifically compared dosimetry to brain structures and cognitive endpoints between proton and photon-treated cohorts of children. Our results are comparable with other studies that evaluated cognition in children treated with photon RT. We found that dose to the right temporal lobe was associated with perceptual reasoning, consistent with prior work in a photononly cohort of all brain tumor types.³ Using prospective patient data from a medulloblastoma study treated with photon therapy (SJMB03), Acharya et al found that corpus callosum and frontal white matter doses were associated with declines in processing speed.²¹ Merchant et al also demonstrated dose to brain structures, including temporal lobes, were correlated with IQ and reading/math scores.^22,23

There are some limitations to our present work. The included cohorts of patients were treated following study protocols, although some patients were not registered on study. As a result, a small number of patients received varying radiation doses, such as a total cumulative prescription of 5940 cGy. However, the magnitude of dosimetric benefit of proton therapy with respect to reducing dose to normal brain tissues is much larger in magnitude than the slightly higher mean prescription dose in the photon cohort. One patient in the proton cohort received 1500 cGy CSI, and some in the proton cohort were treated with a smaller boost CTV margin of 5 mm (compared with 10 mm). After those patients were excluded, the dosimetric benefits of proton therapy were retained. It was also not possible to control for other factors that could affect cognitive outcomes such as socioeconomic status, surgical complications, or parental education. It is reassuring that baseline cognitive scores did not differ between groups. A larger, prospective, controlled, comparative proton versus photon study would be able to confirm the dosimetric and cognitive benefits of proton therapy over x-ray treatment in children with medulloblastoma and provide a robust platform to detect and confirm associations between brain dosimetry and intellectual endpoints. A randomized trial would likely be challenging or impossible to accrue to, however, given the lack of equipoise on this point.⁷

The mechanism by which radiation adversely affects brain tissues is an area of active study. A recent study found worse white matter integrity and lower neuropsychological performance among pediatric brain tumor patients treated with proton versus photon RT.²⁴ Atrophy and loss of volume of brain structures may be caused by RT, which is subsequently associated with decreased cognitive function.²⁵ Radiation may also affect normal structural changes in brain cortex with development, which disrupts cognitive function.²⁶ White matter changes in brain structures are commonly observed post-RT, which are associated with decreased motor and processing speed.²⁷ There is an urgent need to find treatments to mitigate or reverse radiation-induced cognitive decline in children treated for brain tumors.^28–31 The Children’s Oncology Group is prospectively evaluating memantine in children (NCT04939597 or ACCL2031), with cognition at 12 months post-RT as a primary study endpoint. Hypothalamic-pituitary axis and hippocampus avoidance with intensity modulated proton therapy have also been trialed as a means of reducing dose to critical structures in medulloblastoma patients, although safety, efficacy, and toxicity of this novel CSI approach have not yet been established.³²

Conclusion

In this study of children treated with CSI plus tumor bed boost for medulloblastoma, we demonstrate that proton therapy is associated with reduced doses to normal brain structures, compared with photon (x-ray) therapy. This decrease in dose to brain structures is associated with reduced posttreatment cognitive change. This study elucidates the mechanism by which proton therapy leads to improved cognition in children treated for medulloblastoma and confirms that proton therapy should be offered as one of the standard options for children receiving RT for medulloblastoma.