https://orcid.org/0000-0001-8836-0235
Abstract
Central nervous system (CNS) injury following brain-directed radiotherapy remains a major challenge. Proton radiotherapy (PRT) minimizes radiation to healthy brain, potentially limiting sequelae. We characterized CNS radiotoxicity, including radiation-induced leukoencephalopathy (RIL), brain tissue necrosis (TN), and cerebral microbleeds (CMB), in glioma patients treated with PRT or photons (XRT).
Thirty-four patients (19 male; median age 39.6 years) with WHO grade 2-3 gliomas treated with partial cranial radiotherapy (XRT [n = 17] vs PRT[n = 17]) were identified and matched by demographic/clinical criteria. Radiotoxicity was assessed longitudinally for 3 years post-radiotherapy via serial analysis of T2/FLAIR- (for RIL), contrast-enhanced T1- (for TN), and susceptibility (for CMB)-weighted MRI sequences. RIL was rated at whole-brain and hemispheric levels using a novel Fazekas scale-informed scoring system.
The scoring system proved reliable (ICC > 0.85). Both groups developed moderate-to-severe RIL (62%[XRT]; 71%[PRT]) within 3 years; however, XRT was associated with persistent RIL increases in the contralesional hemisphere, whereas contralesional hemispheric RIL plateaued with PRT at 1-year post-radiotherapy (t = 2.180; P = .037). TN rates were greater with PRT (6%[XRT] vs 18%[PRT]; P = ns). CMB prevalence (76%[XRT]; 71%[PRT]) and burden (mean #CMB: 4.0[XRT]; 4.2[PRT]) were similar; however, XRT correlated with greater contralesional hemispheric CMB burden (27%[XRT]; 17%[PRT]; X2 = 4.986; P = .026), whereas PRT-specific CMB clustered at the radiation field margin (X2 = 14.7; P = .002).
CNS radiotoxicity is common and progressive in glioma patients. Injury patterns suggest radiation modality-specificity as RIL, TN, and CMB exhibit unique spatiotemporal differences following XRT vs PRT, likely reflecting underlying dosimetric and radiobiological differences. Familiarity with such injury patterns is essential to improve patient management. Prospective studies are needed to validate these findings and assess their impacts on neurocognitive function.
This study suggests that delayed CNS radiotoxicity, including radiation-induced leukoencephalopathy (RIL), tissue necrosis (TN), and cerebral microbleeds (CMB), is prevalent and progressive after cranial irradiation regardless of radiation modality. However, radiation-associated brain injury patterns suggest unique differences between protons and photons: photons potentially cause greater RIL in contralesional hemispheres, while protons may increase CMB and possibly TN risk at the distal radiation field. These findings propose new hypotheses about the radiotoxicity profiles of protons and photons, likely reflecting their dosimetric and radiobiological differences. These insights may inform prevention, diagnosis, and management of CNS radiotoxicity and improve clinical decision-making in oncology. Larger prospective studies are needed to corroborate these results and assess putative impacts on neurocognitive function.
Introduction
Cranial radiotherapy is integral to brain-directed antineoplastic therapy but central nervous system (CNS) radiotoxicity remains an unresolved challenge.1,2 With multimodal therapy and improved survivorship for patients with CNS malignancies, addressing the growing burden of treatment-related sequelae is imperative.3 Late CNS radiotoxicities are common and include cognitive dysfunction,2,4 leukoencephalopathy,5-7 brain atrophy,8,9 brain tissue necrosis,10-12 cerebrovascular complications,7,13-15 endocrinopathies,16,17 and secondary CNS malignancies.18 These sequelae are largely progressive and irreversible, causing permanent neurological disability, diminishing quality of life, and even impacting survival.1 Adequate clinical management is frequently hindered by diagnostic difficulties and poor treatment options.1 Emerging prevention strategies focus on optimizing existing radiation techniques to spare organs-at-risk from radiation exposure.19
The dosimetric and potential radiobiologic advantages of proton- (PRT) over photon (XRT) beam radiotherapy have prompted increased use of PRT for pediatric and adult CNS tumors.20,21 While associated improvements in long-term neurological outcomes are well-recognized among pediatric patient populations,20 the purported benefit of PRT among adults with CNS malignancies remains under active investigation.22-24 Early studies in adult patients suggest toxicity profile advantages of protons in mitigating radiation-associated cognitive dysfunction25-27 and brain volume atrophy,9,25,28 by virtue of reduced exposure to non-target brain regions such as the contralateral hemisphere. However, there has been concern of higher rates of focal toxicity such as brain tissue necrosis with PRT given greater relative biological effectiveness (RBE) at or near the target.29,30 We therefore aimed to characterize and compare the long-term burden and pattern of 3 common CNS radiotoxicity manifestations—radiation-induced leukoencephalopathy (RIL), treatment-induced brain tissue necrosis (TN), and cerebral microbleeds (CMB)—in 2 matched cohorts of adult patients with glioma irradiated with either XRT or PRT.
Patients and methods
We conducted a retrospective case-matched analysis of demographic, clinical, radiographic, and histopathological data from 34 patients treated with XRT or PRT-based partial cranial radiotherapy for a glioma (WHO grades 2-3) diagnosis at the Massachusetts General Hospital (MGH) between Novemer 1998 and Octoter 2017. Patient data were identified from the institutional database of the MGH Neuro-Oncology Division. This study received approval by the MGH Institutional Review Board for all activities.
Eligibility and group allocation
All 34 patients were treated at MGH and met the following criteria: (1) tissue-based diagnosis of a glioma WHO grades 2-3 according to the WHO 2021 classification (tumors originally signed out as “oligoastrocytoma,” as per WHO 2016 classification, were re-classified based on available data, whenever possible), (2) age ≥18 years at time of diagnosis, (3) either XRT or PRT-based partial cranial radiotherapy following surgery or tumor biopsy, (4) documented clinical follow-up with 3 years of progression-free survival (PFS), (5) availability of axial T1 contrast-enhanced, T2 fluid-attenuated inversion recovery (T2/FLAIR), and T2* susceptibility-weighted (SWI, SWAN) MRI sequences as part of cancer surveillance at baseline (6 weeks post-radiotherapy), 1 year, and 2 years post-radiotherapy. In all cases, gliomas only involved one hemisphere and were treated with tumor-targeted partial cranial radiotherapy, enabling evaluation of both focal and global effects on the brain.
PRT subjects were identified from ongoing longitudinal single-arm outcome studies conducted at MGH (NCT01358058, NCT03286335). XRT subjects were identified from an institutional database and individually matched to PRT subjects using an 11-tiered criterion: age, sex, tumor type/location/ laterality, mutational status (isocitrate dehydrogenase [IDH] 1 mutation; 1p19q deletion), concurrent/adjuvant chemotherapy, and radiation dose/fractions. Over 80% (n = 28/34) of subjects were individually matched using all 11 variables, and efforts were made to identify the closest match for the remaining 6 subjects and confirmed with group statistics.
Variables
Demographic, clinical, therapeutic, and outcome parameters were collected for each patient. Collected radiographic variables of interest specific to the manifestation and dynamics of CNS radiotoxicity (RIL, CMB, and TN) included: time of onset, anatomic location, spatial correlation to the main field of prior irradiation, quantification of lesion burden, and spatiotemporal evolution over time for up to 3 years post-radiotherapy.
Quantification of radiation-induced leukoencephalopathy
A novel scoring system was designed to longitudinally quantify RIL extent based on presence of white matter signal hyperintensity (WMH) on axial T2/FLAIR-weighted MRI sequences. RIL quantification was based on extent of WMH present around (1) the tumor resection cavity (RC-WMH), (2) periventricular white matter (P-WMH), and (3) deep white matter (D-WMH), whereby grading of WMH severity (grades 1-3) was based on a modified Fazekas scale31 (Figure 1A). As per literature consensus, P-WMH and D-WMH were differentiated via the “10 mm rule” (ie, P-WMH was defined as WMH lesions ≤10 mm from the ventricle wall).32,33
(A) Illustrative case examples depicting the evolution of white matter hyperintensity (WMH) severity (grades 1-3). Representative axial T2/FLAIR-weighted brain MRIs demonstrate evolving hyperintensities in periventricular (P-WMH; top panel), deep (D-WMH; middle panel), and resection cavity (RC-WMH; bottom panel) white matter, corresponding to the 3 categories captured in the RIL scoring system. (B) Summary of novel RIL scoring system, depicting calculation of RIL score and corresponding RIL grade, at both hemispheric and global (= whole-brain) levels.
RIL and disease progression were differentiated by clinical-radiographic follow up. In those patients who eventually developed disease progression, a case-by-case radiographic evaluation was conducted to differentiate the onset of recurrence-associated T2/FLAIR changes from pre-existing RIL-related hyperintensities. Patients in whom radiographic differentiation was not reliably possible were excluded from the analysis of RIL-related variables.
In each patient, RIL burden was quantified at baseline (6 weeks post-radiotherapy) and at 1-, 2-, and (for n = 28/34 patients with available imaging) 3-years post-radiotherapy at both global (whole brain) and individual hemispheric levels. RIL scoring was carried out for both P-WMH and D-WMH categories. The category of RC-WMH was added for patients with resection cavities. Scores across categories were summed and normalized into a combined “RIL score” (Figure 1B). As a proxy for injury burden, the RIL score of each patient was translated into a RIL grade (ie, classifying RIL as either absent [0], mild [1], moderate [2], or severe [3]) at both global and hemispheric levels at each time point (Figure 1B). Analysis of RIL burden was conducted by comparing absolute RIL grade values longitudinally within and between groups. Analysis of RIL injury dynamics was conducted within and between groups by comparing the average percent change in RIL score between 1- and 3 years post-radiotherapy. In the subset of patients (n = 3 XRT and n = 3 PRT) for whom data were not available at the 3-year time point, the last available time point (ie, 2 years post-radiotherapy) was chosen for analysis.
Radiographic analysis
Radiotoxicity (RIL, CMB, and TN) was characterized longitudinally in each patient during periods of PFS up to 3 years post-radiotherapy via serial MRI analysis using standard medical imaging software (eUnity v5.10.2.489, Client Outlook Inc.).
For RIL, radiographic analysis of axial T2/FLAIR-weighted MRI sequences with subsequent quantification of RIL burden as per the newly designed RIL scoring system was conducted by a trained investigator (S.F.W.). Interrater reliability of RIL scoring was determined through additional analysis of a data subset (5 subjects over 4 time points each; n = 20 MRIs) by a second investigator (J.D.) blinded to treatment condition. Intraclass correlation coefficient (ICC) identified strong interrater reliability (ICC 0.85) of the new scoring method (Supplementary Table).
For CMB, axial T2* susceptibility-weighted images generated from gradient-echo pulse MRI sequences (ie, SWI or GRE [Siemens]; GRE or SWAN [General Electric]) were analyzed to classify CMB presence, number, and anatomical distribution.
CMB were defined as small (2-10 mm) round hypointense lesions on gradient-echo T2*-weighted MRI sequences, not detectable on T1-, T2, or T2/FLAIR sequences.34,35 CMB anatomical distribution was stratified into ipsi- vs contralesional hemisphere, and lobar (cortical/subcortical hemispheric white matter), deep territory (basal ganglia, thalamus, internal/external capsule, periventricular white matter), and infratentorial (brainstem, cerebellum) regions. Radiographic analysis of CMB was conducted by a trained investigator (S.F.W.); strong interrater reliability (ICC > 0.9) was determined through additional analysis of sample data (5 subjects over 3 time points each; n = 15 MRIs) carried out by a second blinded investigator (J.D.).
For TN, axial T1 contrast-enhanced MRI sequences were analyzed and contextualized with clinical, radiographic, and, whenever available, histopathological reports confirming TN diagnosis. In accordance with literature consensus, de novo contrast-enhancing lesions manifesting ≥6 months post-radiotherapy and a clinical-radiographic or histopathological TN diagnosis were included in the analysis.1,10
Correlation of radiotoxicity with radiation dose distribution plans
Whenever available, patient radiotherapy dose distribution plans were spatially correlated to each CMB and/or TN lesion to extrapolate radiation dose exposure at individual lesion levels. This lesion-to-radiation field (RF)-correlation was conducted digitally using MiM medical imaging software (Version 7.0.5; MiM Software Inc.), whereby CT-based radiotherapy dose distribution plans were spatially correlated to representative MRI sequences displaying CMB/TN lesions. When digital radiotherapy plans were unavailable (n = 3 patients), analog dose distribution plans were used for correlative spatial analysis. In addition to extrapolated Gy dose values, CMB/TN lesions were classified as localized either to the main RF (ie, ≥90% of prescribed target Gy dose), RF margin (ie, 70-89% of prescribed target Gy dose), or outside main RF (ie, <70% of prescribed target Gy dose).
Statistical analysis
Data analyses were conducted in SPSS 27 (IBM SPSS Statistics for Windows, Version 27.0. Armonk, NY: IBM Corp.) Descriptive statistics were calculated for all demographic, diagnostic and treatment data to characterize the 2 study groups (PRT and XRT) included in the study. Non-parametric analyses were used to compare demographic, diagnostic, and treatment characteristics between the 2 groups. Measures of age, sex, tumor type, tumor location, laterality of tumor, IDH mutation status, 1p19q co-deletion status, concurrent treatment, adjuvant chemotherapy, total Gy dose, and number of fractions were analyzed using the Mann–Whitney U test of independence. Changes in RIL were measured as follows: the first portion of the analysis evaluated descriptive statistics to establish group trends. For comparisons within groups, a repeated-measures ANOVA was performed to evaluate changes in RIL burden over time, ie, from baseline up until 3 years post-radiotherapy. For intergroup comparisons, paired t-tests were conducted to longitudinally assess differences in RIL burden and overall injury dynamics (ie, % change between one- and 3-years post-radiotherapy) at global and hemispheric levels, and a chi-square test of independence was performed to examine differences in 3-year cumulative incidence of RIL grades. For CMB, intergroup comparisons were conducted with chi-square tests for categorical variables (eg, prevalence) and with independent samples t-tests for quantitative variables (eg, counts of CMBs). Intraclass correlation was used to examine the percentage of variance between the raters on radiographic analyses. Given the overall goal of this retrospective analysis to generate hypotheses, we retained a P-value of .05 for consideration of significance, acknowledging the potential for type 1 error given multiple comparisons.
Results
Patient characteristics and treatment specifics
Patient characteristics and treatment specifics are summarized in Table 1. Briefly, patients (56% males) had a median age of 39.9 years at time of diagnosis. Approximately two-thirds (68%) of tumors were WHO grade 2 gliomas; pathologies included astrocytoma (74%), oligodendroglioma (21%), and low-grade astrocytoma not otherwise specified (6%). Tumors were generally located in the frontal (50%) or temporal (32%) lobes and predominantly (82%) isocitrate dehydrogenase (IDH) 1 gene mutant. Baseline patient performance status at time of radiotherapy was high (median Karnofsky performance status [KPS] prior to radiotherapy: 90/100); ≥1 cardiovascular comorbidities were present in approximately two-thirds of patients (71% [XRT] vs 53% [PRT]; P = ns). Most patients had undergone surgical resection (77%) prior to radiotherapy and a small subset (15%) chemotherapy prior to surgery. As part of tumor-directed therapy following resection (or biopsy), all patients underwent partial cranial radiotherapy, either in the setting of newly diagnosed (n = 22) or recurrent (n = 12) glioma, followed by adjuvant chemotherapy (68%) during the 3-year assessment period. Patients treated with cranial XRT (n = 17) received a median dose of 59.4 Gy (33 × 1.8 Gy) (range, 54-59.4 Gy). In all patients treated with PRT (n = 17), passive scattered protons were used with a median dose of 54 Gy(RBE) (30 × 1.8 Gy) (range 54-59.4 Gy). Median progression-free survival (PFS) post-radiotherapy was 5 years (4.7 [XRT] vs 5.1 [PRT]), and recurrence rates post-radiotherapy were 58.8% for XRT and 29.4% for PRT, with a median follow-up time of 9.8 and 5.5 years, respectively. There were no significant differences between matched XRT and PRT groups on any of the demographic variables, tumor characteristics, and other treatment/clinical variables.
Summary of patient characteristics, treatment specifics, and clinical outcome.
| Patient characteristics | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| No of patients included | 34 | 17 | 17 | |
| Demographics | ||||
| % sex ratio (m/f) | 56 / 44 | 59 / 41 | 53 / 47 | X2 = 0.119, P = .730 |
| Median age at diagnosis (y) | 39.6 | 38.6 | 39.6 | t = −0.115, P = .126 |
| Tumor specifics | ||||
| Intracranial location | X2 = 4.877, P = .431 | |||
| % Left/right | 56 / 44 | 47 / 53 | 65 / 35 | X2 = 1.074, P = .300 |
| % Frontal (N) | 50.0 (17) | 47.1 (8) | 52.9 (9) | |
| % Temporal (N) | 32.4 (11) | 41.2 (7) | 23.5 (4) | |
| % Insular (N) | 5.9 (2) | 0 (0) | 11.8 (2) | |
| % Thalamic (N) | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| % Parietal (N) | 2.9 (1) | 0 (0) | 5.9 (1) | |
| % Cerebellopontine angle (N) | 2.9 (1) | 5.9 (1) | 0 (0) | |
| WHO grade % (N) | X2 = 0.582, P = .748 | |||
| 2 | 64.7 (22) | 58.8 (10) | 70.6 (12) | |
| 3 | 29.4 (10) | 35.3 (6) | 23.5 (4) | |
| Low grade, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Histopathology | X2 = 0.183, P = .913 | |||
| % Astrocytoma* (N) | 73.5 (25) | 76.5 (13) | 70.6 (12) | |
| % Oligodendroglioma (N) | 20.6 (7) | 17.6 (3) | 23.5 (4) | |
| % Low grade astrocytoma, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Molecular-genetic profile | ||||
| % IDH1 mutant (N) | 82.4 (28) | 76.5 (13) | 88.2 (15) | X2 = 0.81, P = .368 |
| % 1p19q co-deleted (N) | 37.0 (10/27) | 38.5 (5/13) | 35.7 (5/14) | X2 = 0.02, P = .883 |
| % p53 mutated (N) | 37.0 (10/27) | 54.5 (6/11) | 25.0 (4/16) | X2 = 6.367, P = .041 |
| % ATRX loss | 42.1 (8/19) | 33.3 (2/6) | 46.2 (6/13) | X2 = 6.085, P = .048 |
| % MGMT promoter methylated (N) | 50.0 (6/12) | 42.9 (3/7) | 60.0 (3/5) | X2 = 1.238, P = .538 |
| Clinical status | ||||
| % w/cardiovascular comorbidities | 61.8 | 70.6 (12) | 52.9 (9) | X2 = 3.84, P = 0.147 |
| % w/recurrence prior to RT (N) | 35 (12) | 24 (4) | 47 (8) | X2 = 2.06, P = .151 |
| Median KPS baseline pre-RT | 90 | 90 | 90 | t = 1.343, P = .165 |
| Median KPS 2 years post-RT | 90 | 95 | 90 | t = −0.026, P = .475 |
| Treatment details | ||||
| Extent of surgical resection | X2 = 5.727, P = .126 | |||
| % GTR (N) | 8.8 (3) | 5.9 (1) | 11.8 (2) | |
| % NTR (N) | 17.6 (6) | 0 (0) | 35.3 (6) | |
| % STR (N) | 32.4 (11) | 29.4 (5) | 35.3 (6) | |
| % PR (N) | 17.6 (6) | 17.6 (3) | 17.6 (3) | |
| % Biopsy only (N) | 23.5 (8) | 41.2 (7) | 5.9 (1) | X2 = 5.1, P = .024 |
| Radiotherapy regimen | n/a | |||
| % Photons (N) | 50 (17) | 100 (17) | N/A | |
| % IMRT (N) | 29.4 (5/17) | N/A | ||
| % VMAT (N) | 17.7 (3/17) | N/A | ||
| % 3D CRT (N) | 5.9 (1/17) | N/A | ||
| % IFRT, not otherwise specified | 47.1 (8/17) | |||
| % Protons (N) | 50 (17) | N/A | 100 (17) | |
| Median dose in Gy (range) | 54 (9) | 59.4 (9) | 54(RBE) (5.4) | t = 1.92, P = .065 |
| Median no. fractions (range) | 30 (5) | 33 (5) | 30 (3) | t = 1.41, P = .167 |
| Median fraction size in Gy | 1.8 | 1.8 | 1.8(RBE) | t = 1.00, P = .325 |
| Systemic treatment | ||||
| % w/ neoadjuvant Ctx (N) | 14.7 (5) | 17.6 (3) | 11.8 (2) | X2 = 5.034, P = .412 |
| % w/ concurrent Ctx (N) | 32.4 (11) | 35.5 (6) | 29.4 (5) | X2 = 0.134, P = .714 |
| % w/ adjuvant Ctx (N) | 67.6 (23) | 58.8 (10) | 76.5 (13) | X2 = 1.209, P = 0.271 |
| Clinical outcome | ||||
| Median PFS post-RT in years (SD) | 5.0 (3.69) | 4.7 (4.95) | 5.1 (1.26) | P = .134, CI −5.52 −8.3 |
| % w/ recurrence post-RT (N) | 44.1 (15) | 58.8 (10) | 29.4 (5) | X2 = 2.43, P = .119 |
| Patient characteristics | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| No of patients included | 34 | 17 | 17 | |
| Demographics | ||||
| % sex ratio (m/f) | 56 / 44 | 59 / 41 | 53 / 47 | X2 = 0.119, P = .730 |
| Median age at diagnosis (y) | 39.6 | 38.6 | 39.6 | t = −0.115, P = .126 |
| Tumor specifics | ||||
| Intracranial location | X2 = 4.877, P = .431 | |||
| % Left/right | 56 / 44 | 47 / 53 | 65 / 35 | X2 = 1.074, P = .300 |
| % Frontal (N) | 50.0 (17) | 47.1 (8) | 52.9 (9) | |
| % Temporal (N) | 32.4 (11) | 41.2 (7) | 23.5 (4) | |
| % Insular (N) | 5.9 (2) | 0 (0) | 11.8 (2) | |
| % Thalamic (N) | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| % Parietal (N) | 2.9 (1) | 0 (0) | 5.9 (1) | |
| % Cerebellopontine angle (N) | 2.9 (1) | 5.9 (1) | 0 (0) | |
| WHO grade % (N) | X2 = 0.582, P = .748 | |||
| 2 | 64.7 (22) | 58.8 (10) | 70.6 (12) | |
| 3 | 29.4 (10) | 35.3 (6) | 23.5 (4) | |
| Low grade, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Histopathology | X2 = 0.183, P = .913 | |||
| % Astrocytoma* (N) | 73.5 (25) | 76.5 (13) | 70.6 (12) | |
| % Oligodendroglioma (N) | 20.6 (7) | 17.6 (3) | 23.5 (4) | |
| % Low grade astrocytoma, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Molecular-genetic profile | ||||
| % IDH1 mutant (N) | 82.4 (28) | 76.5 (13) | 88.2 (15) | X2 = 0.81, P = .368 |
| % 1p19q co-deleted (N) | 37.0 (10/27) | 38.5 (5/13) | 35.7 (5/14) | X2 = 0.02, P = .883 |
| % p53 mutated (N) | 37.0 (10/27) | 54.5 (6/11) | 25.0 (4/16) | X2 = 6.367, P = .041 |
| % ATRX loss | 42.1 (8/19) | 33.3 (2/6) | 46.2 (6/13) | X2 = 6.085, P = .048 |
| % MGMT promoter methylated (N) | 50.0 (6/12) | 42.9 (3/7) | 60.0 (3/5) | X2 = 1.238, P = .538 |
| Clinical status | ||||
| % w/cardiovascular comorbidities | 61.8 | 70.6 (12) | 52.9 (9) | X2 = 3.84, P = 0.147 |
| % w/recurrence prior to RT (N) | 35 (12) | 24 (4) | 47 (8) | X2 = 2.06, P = .151 |
| Median KPS baseline pre-RT | 90 | 90 | 90 | t = 1.343, P = .165 |
| Median KPS 2 years post-RT | 90 | 95 | 90 | t = −0.026, P = .475 |
| Treatment details | ||||
| Extent of surgical resection | X2 = 5.727, P = .126 | |||
| % GTR (N) | 8.8 (3) | 5.9 (1) | 11.8 (2) | |
| % NTR (N) | 17.6 (6) | 0 (0) | 35.3 (6) | |
| % STR (N) | 32.4 (11) | 29.4 (5) | 35.3 (6) | |
| % PR (N) | 17.6 (6) | 17.6 (3) | 17.6 (3) | |
| % Biopsy only (N) | 23.5 (8) | 41.2 (7) | 5.9 (1) | X2 = 5.1, P = .024 |
| Radiotherapy regimen | n/a | |||
| % Photons (N) | 50 (17) | 100 (17) | N/A | |
| % IMRT (N) | 29.4 (5/17) | N/A | ||
| % VMAT (N) | 17.7 (3/17) | N/A | ||
| % 3D CRT (N) | 5.9 (1/17) | N/A | ||
| % IFRT, not otherwise specified | 47.1 (8/17) | |||
| % Protons (N) | 50 (17) | N/A | 100 (17) | |
| Median dose in Gy (range) | 54 (9) | 59.4 (9) | 54(RBE) (5.4) | t = 1.92, P = .065 |
| Median no. fractions (range) | 30 (5) | 33 (5) | 30 (3) | t = 1.41, P = .167 |
| Median fraction size in Gy | 1.8 | 1.8 | 1.8(RBE) | t = 1.00, P = .325 |
| Systemic treatment | ||||
| % w/ neoadjuvant Ctx (N) | 14.7 (5) | 17.6 (3) | 11.8 (2) | X2 = 5.034, P = .412 |
| % w/ concurrent Ctx (N) | 32.4 (11) | 35.5 (6) | 29.4 (5) | X2 = 0.134, P = .714 |
| % w/ adjuvant Ctx (N) | 67.6 (23) | 58.8 (10) | 76.5 (13) | X2 = 1.209, P = 0.271 |
| Clinical outcome | ||||
| Median PFS post-RT in years (SD) | 5.0 (3.69) | 4.7 (4.95) | 5.1 (1.26) | P = .134, CI −5.52 −8.3 |
| % w/ recurrence post-RT (N) | 44.1 (15) | 58.8 (10) | 29.4 (5) | X2 = 2.43, P = .119 |
*The majority (n = 21) of tumors were classified as “Astrocytoma, IDH mutant” as per WHO 2021 classification. In the remainder (n = 4), imaging, histopathology, and disease course were consistent with grade 2/3 astrocytoma although detailed molecular information were not available given the retrospective nature of the study.
Abbreviations: ATRX, alpha-thalassemia/mental retardation, X-linked; Ctx, chemotherapy; f, female; GTR, gross total resection; IDH1, isocitrate dehydrogenase 1; IFRT, involved-field radiation therapy; IMRT, intensity-modulated radiation therapy; KPS, Karnofsky Performance Satus Scale; m, male; MGMT, Methylguanine methyltransferase; NTR, near total resection; PFS, progression-free survival; PR, partial resection; RT, radiotherapy; STR, subtotal resection; VMAT, volumetric modulated arc therapy; WHO, World Health Organization; yrs, = years.
Summary of patient characteristics, treatment specifics, and clinical outcome.
| Patient characteristics | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| No of patients included | 34 | 17 | 17 | |
| Demographics | ||||
| % sex ratio (m/f) | 56 / 44 | 59 / 41 | 53 / 47 | X2 = 0.119, P = .730 |
| Median age at diagnosis (y) | 39.6 | 38.6 | 39.6 | t = −0.115, P = .126 |
| Tumor specifics | ||||
| Intracranial location | X2 = 4.877, P = .431 | |||
| % Left/right | 56 / 44 | 47 / 53 | 65 / 35 | X2 = 1.074, P = .300 |
| % Frontal (N) | 50.0 (17) | 47.1 (8) | 52.9 (9) | |
| % Temporal (N) | 32.4 (11) | 41.2 (7) | 23.5 (4) | |
| % Insular (N) | 5.9 (2) | 0 (0) | 11.8 (2) | |
| % Thalamic (N) | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| % Parietal (N) | 2.9 (1) | 0 (0) | 5.9 (1) | |
| % Cerebellopontine angle (N) | 2.9 (1) | 5.9 (1) | 0 (0) | |
| WHO grade % (N) | X2 = 0.582, P = .748 | |||
| 2 | 64.7 (22) | 58.8 (10) | 70.6 (12) | |
| 3 | 29.4 (10) | 35.3 (6) | 23.5 (4) | |
| Low grade, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Histopathology | X2 = 0.183, P = .913 | |||
| % Astrocytoma* (N) | 73.5 (25) | 76.5 (13) | 70.6 (12) | |
| % Oligodendroglioma (N) | 20.6 (7) | 17.6 (3) | 23.5 (4) | |
| % Low grade astrocytoma, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Molecular-genetic profile | ||||
| % IDH1 mutant (N) | 82.4 (28) | 76.5 (13) | 88.2 (15) | X2 = 0.81, P = .368 |
| % 1p19q co-deleted (N) | 37.0 (10/27) | 38.5 (5/13) | 35.7 (5/14) | X2 = 0.02, P = .883 |
| % p53 mutated (N) | 37.0 (10/27) | 54.5 (6/11) | 25.0 (4/16) | X2 = 6.367, P = .041 |
| % ATRX loss | 42.1 (8/19) | 33.3 (2/6) | 46.2 (6/13) | X2 = 6.085, P = .048 |
| % MGMT promoter methylated (N) | 50.0 (6/12) | 42.9 (3/7) | 60.0 (3/5) | X2 = 1.238, P = .538 |
| Clinical status | ||||
| % w/cardiovascular comorbidities | 61.8 | 70.6 (12) | 52.9 (9) | X2 = 3.84, P = 0.147 |
| % w/recurrence prior to RT (N) | 35 (12) | 24 (4) | 47 (8) | X2 = 2.06, P = .151 |
| Median KPS baseline pre-RT | 90 | 90 | 90 | t = 1.343, P = .165 |
| Median KPS 2 years post-RT | 90 | 95 | 90 | t = −0.026, P = .475 |
| Treatment details | ||||
| Extent of surgical resection | X2 = 5.727, P = .126 | |||
| % GTR (N) | 8.8 (3) | 5.9 (1) | 11.8 (2) | |
| % NTR (N) | 17.6 (6) | 0 (0) | 35.3 (6) | |
| % STR (N) | 32.4 (11) | 29.4 (5) | 35.3 (6) | |
| % PR (N) | 17.6 (6) | 17.6 (3) | 17.6 (3) | |
| % Biopsy only (N) | 23.5 (8) | 41.2 (7) | 5.9 (1) | X2 = 5.1, P = .024 |
| Radiotherapy regimen | n/a | |||
| % Photons (N) | 50 (17) | 100 (17) | N/A | |
| % IMRT (N) | 29.4 (5/17) | N/A | ||
| % VMAT (N) | 17.7 (3/17) | N/A | ||
| % 3D CRT (N) | 5.9 (1/17) | N/A | ||
| % IFRT, not otherwise specified | 47.1 (8/17) | |||
| % Protons (N) | 50 (17) | N/A | 100 (17) | |
| Median dose in Gy (range) | 54 (9) | 59.4 (9) | 54(RBE) (5.4) | t = 1.92, P = .065 |
| Median no. fractions (range) | 30 (5) | 33 (5) | 30 (3) | t = 1.41, P = .167 |
| Median fraction size in Gy | 1.8 | 1.8 | 1.8(RBE) | t = 1.00, P = .325 |
| Systemic treatment | ||||
| % w/ neoadjuvant Ctx (N) | 14.7 (5) | 17.6 (3) | 11.8 (2) | X2 = 5.034, P = .412 |
| % w/ concurrent Ctx (N) | 32.4 (11) | 35.5 (6) | 29.4 (5) | X2 = 0.134, P = .714 |
| % w/ adjuvant Ctx (N) | 67.6 (23) | 58.8 (10) | 76.5 (13) | X2 = 1.209, P = 0.271 |
| Clinical outcome | ||||
| Median PFS post-RT in years (SD) | 5.0 (3.69) | 4.7 (4.95) | 5.1 (1.26) | P = .134, CI −5.52 −8.3 |
| % w/ recurrence post-RT (N) | 44.1 (15) | 58.8 (10) | 29.4 (5) | X2 = 2.43, P = .119 |
| Patient characteristics | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| No of patients included | 34 | 17 | 17 | |
| Demographics | ||||
| % sex ratio (m/f) | 56 / 44 | 59 / 41 | 53 / 47 | X2 = 0.119, P = .730 |
| Median age at diagnosis (y) | 39.6 | 38.6 | 39.6 | t = −0.115, P = .126 |
| Tumor specifics | ||||
| Intracranial location | X2 = 4.877, P = .431 | |||
| % Left/right | 56 / 44 | 47 / 53 | 65 / 35 | X2 = 1.074, P = .300 |
| % Frontal (N) | 50.0 (17) | 47.1 (8) | 52.9 (9) | |
| % Temporal (N) | 32.4 (11) | 41.2 (7) | 23.5 (4) | |
| % Insular (N) | 5.9 (2) | 0 (0) | 11.8 (2) | |
| % Thalamic (N) | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| % Parietal (N) | 2.9 (1) | 0 (0) | 5.9 (1) | |
| % Cerebellopontine angle (N) | 2.9 (1) | 5.9 (1) | 0 (0) | |
| WHO grade % (N) | X2 = 0.582, P = .748 | |||
| 2 | 64.7 (22) | 58.8 (10) | 70.6 (12) | |
| 3 | 29.4 (10) | 35.3 (6) | 23.5 (4) | |
| Low grade, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Histopathology | X2 = 0.183, P = .913 | |||
| % Astrocytoma* (N) | 73.5 (25) | 76.5 (13) | 70.6 (12) | |
| % Oligodendroglioma (N) | 20.6 (7) | 17.6 (3) | 23.5 (4) | |
| % Low grade astrocytoma, not otherwise specified | 5.9 (2) | 5.9 (1) | 5.9 (1) | |
| Molecular-genetic profile | ||||
| % IDH1 mutant (N) | 82.4 (28) | 76.5 (13) | 88.2 (15) | X2 = 0.81, P = .368 |
| % 1p19q co-deleted (N) | 37.0 (10/27) | 38.5 (5/13) | 35.7 (5/14) | X2 = 0.02, P = .883 |
| % p53 mutated (N) | 37.0 (10/27) | 54.5 (6/11) | 25.0 (4/16) | X2 = 6.367, P = .041 |
| % ATRX loss | 42.1 (8/19) | 33.3 (2/6) | 46.2 (6/13) | X2 = 6.085, P = .048 |
| % MGMT promoter methylated (N) | 50.0 (6/12) | 42.9 (3/7) | 60.0 (3/5) | X2 = 1.238, P = .538 |
| Clinical status | ||||
| % w/cardiovascular comorbidities | 61.8 | 70.6 (12) | 52.9 (9) | X2 = 3.84, P = 0.147 |
| % w/recurrence prior to RT (N) | 35 (12) | 24 (4) | 47 (8) | X2 = 2.06, P = .151 |
| Median KPS baseline pre-RT | 90 | 90 | 90 | t = 1.343, P = .165 |
| Median KPS 2 years post-RT | 90 | 95 | 90 | t = −0.026, P = .475 |
| Treatment details | ||||
| Extent of surgical resection | X2 = 5.727, P = .126 | |||
| % GTR (N) | 8.8 (3) | 5.9 (1) | 11.8 (2) | |
| % NTR (N) | 17.6 (6) | 0 (0) | 35.3 (6) | |
| % STR (N) | 32.4 (11) | 29.4 (5) | 35.3 (6) | |
| % PR (N) | 17.6 (6) | 17.6 (3) | 17.6 (3) | |
| % Biopsy only (N) | 23.5 (8) | 41.2 (7) | 5.9 (1) | X2 = 5.1, P = .024 |
| Radiotherapy regimen | n/a | |||
| % Photons (N) | 50 (17) | 100 (17) | N/A | |
| % IMRT (N) | 29.4 (5/17) | N/A | ||
| % VMAT (N) | 17.7 (3/17) | N/A | ||
| % 3D CRT (N) | 5.9 (1/17) | N/A | ||
| % IFRT, not otherwise specified | 47.1 (8/17) | |||
| % Protons (N) | 50 (17) | N/A | 100 (17) | |
| Median dose in Gy (range) | 54 (9) | 59.4 (9) | 54(RBE) (5.4) | t = 1.92, P = .065 |
| Median no. fractions (range) | 30 (5) | 33 (5) | 30 (3) | t = 1.41, P = .167 |
| Median fraction size in Gy | 1.8 | 1.8 | 1.8(RBE) | t = 1.00, P = .325 |
| Systemic treatment | ||||
| % w/ neoadjuvant Ctx (N) | 14.7 (5) | 17.6 (3) | 11.8 (2) | X2 = 5.034, P = .412 |
| % w/ concurrent Ctx (N) | 32.4 (11) | 35.5 (6) | 29.4 (5) | X2 = 0.134, P = .714 |
| % w/ adjuvant Ctx (N) | 67.6 (23) | 58.8 (10) | 76.5 (13) | X2 = 1.209, P = 0.271 |
| Clinical outcome | ||||
| Median PFS post-RT in years (SD) | 5.0 (3.69) | 4.7 (4.95) | 5.1 (1.26) | P = .134, CI −5.52 −8.3 |
| % w/ recurrence post-RT (N) | 44.1 (15) | 58.8 (10) | 29.4 (5) | X2 = 2.43, P = .119 |
*The majority (n = 21) of tumors were classified as “Astrocytoma, IDH mutant” as per WHO 2021 classification. In the remainder (n = 4), imaging, histopathology, and disease course were consistent with grade 2/3 astrocytoma although detailed molecular information were not available given the retrospective nature of the study.
Abbreviations: ATRX, alpha-thalassemia/mental retardation, X-linked; Ctx, chemotherapy; f, female; GTR, gross total resection; IDH1, isocitrate dehydrogenase 1; IFRT, involved-field radiation therapy; IMRT, intensity-modulated radiation therapy; KPS, Karnofsky Performance Satus Scale; m, male; MGMT, Methylguanine methyltransferase; NTR, near total resection; PFS, progression-free survival; PR, partial resection; RT, radiotherapy; STR, subtotal resection; VMAT, volumetric modulated arc therapy; WHO, World Health Organization; yrs, = years.
Radiation-induced leukoencephalopathy: injury burden and dynamics
The novel RIL scoring system was reliable (ICC 0.85; Supplementary Table). Longitudinal analysis of RIL burden identified a significant increase in global RIL in both XRT (F[3, 57] = 8.63, P < .001) and PRT (F[3, 61] = 4.69, P < .005) groups over time (Figure 2A). Most patients (62% [XRT] and 72% [PRT]) developed moderate or severe RIL within 3 years, with the ipsilesional hemisphere more severely affected (Table 2; Figure 2C and E). Intergroup comparison indicated, on average, slightly greater absolute RIL grade (driven by greater ipsilesional RIL) with PRT at all time points, although this observation was only significant at baseline (average RIL grade: 0.8 [XRT] vs 1.1 [PRT], t = 2.379, P = 0.024) (Figure 2A).
Summary of CNS radiotoxicity.
| Radiotoxicity type | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| Radiation-induced leukoencephalopathy (RIL) | ||||
| No of patients included | 33/34 | 16/17 | 17/17 | |
| Average RIL Grade (G0–3) | ||||
| Baseline (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.0 | 0.8 | 1.1 | t = 2.379, P = .024 |
| Hemispheric, IL/CL | 1.0 / 0.7 | 0.8 / 0.7 | 1.1 / 0.8 | t = 2.379, P = .024/t = 0.484, P = .632 |
| 1 year (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.6 | 1.4 | 1.7 | |
| Hemispheric, IL/CL | 1.6 / 1.0 | 1.4 / 0.9 | 1.7 / 1.1 | |
| 2 years (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.8 | 1.6 | 1.9 | |
| Hemispheric, IL/CL | 1.8 / 1.2 | 1.9 / 1.2 | 1.9 / 1.2 | |
| 3 years (N) | 27/34 | 13/16 | 14/17 | |
| Global | 1.9 | 1.8 | 2.1 | t = 1.212, P = .237 |
| Hemispheric, IL/CL | 2.0 / 1.2 | 1.8 / 1.2 | 2.1 / 1.3 | t = 1.212, P = .237/t = 0.214, P = .833 |
| 3-year cumulative incidence by grade | X2 = 2.478; P = .3 | |||
| % absent [G0] (N) | 0.0 (0/27) | 0.0 (0) | 0.0 (0) | |
| % mild RIL [G1] (N) | 33.3 (9/27) | 38.5 (5/13) | 28.6 (4/14) | |
| % moderate RIL [G2] (N) | 37.0 (10/27) | 46.2 (6/13) | 28.6 (4/14) | |
| % severe RIL [G3] (N) | 29.6(8/27) | 15.4 (2/13) | 42.9 (6/14) | |
| Average injury dynamics (% change from baseline) | ||||
| % Global RIL change yr 1-3 | 31.8% | 31.3% | 29.4% | t = 0.826, P = .912 |
| % IL RIL change yr 1-3 | 34.8% | 31.3% | 38.2% | t = −0.389, P = .700 |
| % CL RIL change yr 1-3 | 22.2% | 43.8% | 7.8% | t = 2.180; P = .037 |
| Cerebral microbleeds (CMB) | ||||
| No of patients included | 34 | 17 | 17 | |
| All-time total CMB identified | 362 | 213 | 149 | |
| All-time cumulative prevalence, % (N) | 91.2 (31) | 88.2 (15) | 94.1 (16) | X2 = 0.3656; P = .56 |
| All-time #CMB/patient—mean (SD) | 10.6 (12.5) | 12.5 (15.1) | 8.7 (9.4) | t = 0.869; P = .391 |
| 3-year #CMB/patient—mean (SD) | 4.1 (4.6) | 4.0 (4.5) | 4.2 (4.9) | t = −.106; P = .91 |
| 3-year cumulative prevalence, % (N) | 73.5 (25) | 76.5 (13) | 70.5 (12) | X2 = 0.1511; P = .70 |
| Median latency to first CMB (range) | 17 (1-55) | 13 (1-51) | 17.5 (4-55) | t = −1.36; P = .186 |
| Anatomic distribution | X2 = 4.980; P = .08 | |||
| % lobar (N) | 63.0 (228) | 61.0 (130) | 65.8 (98) | |
| % deep territory (N) | 33.7 (122) | 33.8 (72) | 33.6 (50) | |
| % infratentorial (N) | 3.0 (11) | 4.7 (10) | 0.7 (1) | |
| Laterality | X2 = 4.986; P = .026 | |||
| % IL (N) | 77.3 (280) | 73.2 (156) | 83.2 (124) | |
| % CL (N) | 22.7 (82) | 26.8 (57) | 16.8 (25) | |
| Radiation field correlation [% of target Gy dose] | X2 = 12.604; P = .002 | |||
| % main field [≥ 90%] (N) | 57.3 (192/335) | 53.2 (99/186) | 62.4 (93/149) | |
| % margin [70%-89%] (N) | 20.6 (69/335) | 16.1 (30/186) | 26.2 (39/149) | |
| % outside main field [<70%] (N) | 22.1 (74/335) | 30.6 (57/186) | 11.4 (17/149) | |
| Median Gy dose received (range) | 50.6 (8.0–62.9) | 51.6 (8.0–62.9) | 50.0 (6.0–56.6) RBE | |
| Brain tissue necrosis (TN) | ||||
| No. of patients included | 34 | 17 | 17 | |
| % 3-year cumulative incidence (N) | 11.8 (4/34) | 5.9 (1/17) | 17.6 (3/17) | X2 = 1.1333; P = .29 |
| Median onset latency (range) | 18 (10–30) | 30 (n/a) | 12 (10–24) | n/a |
| Median duration until resolution (range) | 31.5 (9–48) | 21 (n/a) | 42 (9–48) | n/a |
| % symptomatic (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| % w/ treatment (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| N radiographic dx/biopsy | 3 / 1 | 1 / 0 | 2 / 1 | n/a |
| Median no./patient (range) | 1 (0) | 1 (n/a) | 1 (0) | n/a |
| Median maximum diameter, mm (range) | 6 (2–30) | 2 (n/a) | 7 (5–30) | n/a |
| % periventricular location (N) | 50.0 (2/4) | 100.0 (1/1) | 33.3 (1/3) | n/a |
| % in main radiation field [≥90%] (N) | 100.0 (4/4) | 100.0 (1/1) | 100.0 (3/3) | n/a |
| Radiotoxicity type | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| Radiation-induced leukoencephalopathy (RIL) | ||||
| No of patients included | 33/34 | 16/17 | 17/17 | |
| Average RIL Grade (G0–3) | ||||
| Baseline (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.0 | 0.8 | 1.1 | t = 2.379, P = .024 |
| Hemispheric, IL/CL | 1.0 / 0.7 | 0.8 / 0.7 | 1.1 / 0.8 | t = 2.379, P = .024/t = 0.484, P = .632 |
| 1 year (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.6 | 1.4 | 1.7 | |
| Hemispheric, IL/CL | 1.6 / 1.0 | 1.4 / 0.9 | 1.7 / 1.1 | |
| 2 years (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.8 | 1.6 | 1.9 | |
| Hemispheric, IL/CL | 1.8 / 1.2 | 1.9 / 1.2 | 1.9 / 1.2 | |
| 3 years (N) | 27/34 | 13/16 | 14/17 | |
| Global | 1.9 | 1.8 | 2.1 | t = 1.212, P = .237 |
| Hemispheric, IL/CL | 2.0 / 1.2 | 1.8 / 1.2 | 2.1 / 1.3 | t = 1.212, P = .237/t = 0.214, P = .833 |
| 3-year cumulative incidence by grade | X2 = 2.478; P = .3 | |||
| % absent [G0] (N) | 0.0 (0/27) | 0.0 (0) | 0.0 (0) | |
| % mild RIL [G1] (N) | 33.3 (9/27) | 38.5 (5/13) | 28.6 (4/14) | |
| % moderate RIL [G2] (N) | 37.0 (10/27) | 46.2 (6/13) | 28.6 (4/14) | |
| % severe RIL [G3] (N) | 29.6(8/27) | 15.4 (2/13) | 42.9 (6/14) | |
| Average injury dynamics (% change from baseline) | ||||
| % Global RIL change yr 1-3 | 31.8% | 31.3% | 29.4% | t = 0.826, P = .912 |
| % IL RIL change yr 1-3 | 34.8% | 31.3% | 38.2% | t = −0.389, P = .700 |
| % CL RIL change yr 1-3 | 22.2% | 43.8% | 7.8% | t = 2.180; P = .037 |
| Cerebral microbleeds (CMB) | ||||
| No of patients included | 34 | 17 | 17 | |
| All-time total CMB identified | 362 | 213 | 149 | |
| All-time cumulative prevalence, % (N) | 91.2 (31) | 88.2 (15) | 94.1 (16) | X2 = 0.3656; P = .56 |
| All-time #CMB/patient—mean (SD) | 10.6 (12.5) | 12.5 (15.1) | 8.7 (9.4) | t = 0.869; P = .391 |
| 3-year #CMB/patient—mean (SD) | 4.1 (4.6) | 4.0 (4.5) | 4.2 (4.9) | t = −.106; P = .91 |
| 3-year cumulative prevalence, % (N) | 73.5 (25) | 76.5 (13) | 70.5 (12) | X2 = 0.1511; P = .70 |
| Median latency to first CMB (range) | 17 (1-55) | 13 (1-51) | 17.5 (4-55) | t = −1.36; P = .186 |
| Anatomic distribution | X2 = 4.980; P = .08 | |||
| % lobar (N) | 63.0 (228) | 61.0 (130) | 65.8 (98) | |
| % deep territory (N) | 33.7 (122) | 33.8 (72) | 33.6 (50) | |
| % infratentorial (N) | 3.0 (11) | 4.7 (10) | 0.7 (1) | |
| Laterality | X2 = 4.986; P = .026 | |||
| % IL (N) | 77.3 (280) | 73.2 (156) | 83.2 (124) | |
| % CL (N) | 22.7 (82) | 26.8 (57) | 16.8 (25) | |
| Radiation field correlation [% of target Gy dose] | X2 = 12.604; P = .002 | |||
| % main field [≥ 90%] (N) | 57.3 (192/335) | 53.2 (99/186) | 62.4 (93/149) | |
| % margin [70%-89%] (N) | 20.6 (69/335) | 16.1 (30/186) | 26.2 (39/149) | |
| % outside main field [<70%] (N) | 22.1 (74/335) | 30.6 (57/186) | 11.4 (17/149) | |
| Median Gy dose received (range) | 50.6 (8.0–62.9) | 51.6 (8.0–62.9) | 50.0 (6.0–56.6) RBE | |
| Brain tissue necrosis (TN) | ||||
| No. of patients included | 34 | 17 | 17 | |
| % 3-year cumulative incidence (N) | 11.8 (4/34) | 5.9 (1/17) | 17.6 (3/17) | X2 = 1.1333; P = .29 |
| Median onset latency (range) | 18 (10–30) | 30 (n/a) | 12 (10–24) | n/a |
| Median duration until resolution (range) | 31.5 (9–48) | 21 (n/a) | 42 (9–48) | n/a |
| % symptomatic (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| % w/ treatment (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| N radiographic dx/biopsy | 3 / 1 | 1 / 0 | 2 / 1 | n/a |
| Median no./patient (range) | 1 (0) | 1 (n/a) | 1 (0) | n/a |
| Median maximum diameter, mm (range) | 6 (2–30) | 2 (n/a) | 7 (5–30) | n/a |
| % periventricular location (N) | 50.0 (2/4) | 100.0 (1/1) | 33.3 (1/3) | n/a |
| % in main radiation field [≥90%] (N) | 100.0 (4/4) | 100.0 (1/1) | 100.0 (3/3) | n/a |
Abbreviations: CL, contralesional; CMB, cerebral microbleeds; dx, diagnosis; IL, ipsilesional; RIL, radiation-induced leukoencephalopathy; SD, standard deviation; TN, tissue necrosis; yr, year.
Summary of CNS radiotoxicity.
| Radiotoxicity type | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| Radiation-induced leukoencephalopathy (RIL) | ||||
| No of patients included | 33/34 | 16/17 | 17/17 | |
| Average RIL Grade (G0–3) | ||||
| Baseline (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.0 | 0.8 | 1.1 | t = 2.379, P = .024 |
| Hemispheric, IL/CL | 1.0 / 0.7 | 0.8 / 0.7 | 1.1 / 0.8 | t = 2.379, P = .024/t = 0.484, P = .632 |
| 1 year (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.6 | 1.4 | 1.7 | |
| Hemispheric, IL/CL | 1.6 / 1.0 | 1.4 / 0.9 | 1.7 / 1.1 | |
| 2 years (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.8 | 1.6 | 1.9 | |
| Hemispheric, IL/CL | 1.8 / 1.2 | 1.9 / 1.2 | 1.9 / 1.2 | |
| 3 years (N) | 27/34 | 13/16 | 14/17 | |
| Global | 1.9 | 1.8 | 2.1 | t = 1.212, P = .237 |
| Hemispheric, IL/CL | 2.0 / 1.2 | 1.8 / 1.2 | 2.1 / 1.3 | t = 1.212, P = .237/t = 0.214, P = .833 |
| 3-year cumulative incidence by grade | X2 = 2.478; P = .3 | |||
| % absent [G0] (N) | 0.0 (0/27) | 0.0 (0) | 0.0 (0) | |
| % mild RIL [G1] (N) | 33.3 (9/27) | 38.5 (5/13) | 28.6 (4/14) | |
| % moderate RIL [G2] (N) | 37.0 (10/27) | 46.2 (6/13) | 28.6 (4/14) | |
| % severe RIL [G3] (N) | 29.6(8/27) | 15.4 (2/13) | 42.9 (6/14) | |
| Average injury dynamics (% change from baseline) | ||||
| % Global RIL change yr 1-3 | 31.8% | 31.3% | 29.4% | t = 0.826, P = .912 |
| % IL RIL change yr 1-3 | 34.8% | 31.3% | 38.2% | t = −0.389, P = .700 |
| % CL RIL change yr 1-3 | 22.2% | 43.8% | 7.8% | t = 2.180; P = .037 |
| Cerebral microbleeds (CMB) | ||||
| No of patients included | 34 | 17 | 17 | |
| All-time total CMB identified | 362 | 213 | 149 | |
| All-time cumulative prevalence, % (N) | 91.2 (31) | 88.2 (15) | 94.1 (16) | X2 = 0.3656; P = .56 |
| All-time #CMB/patient—mean (SD) | 10.6 (12.5) | 12.5 (15.1) | 8.7 (9.4) | t = 0.869; P = .391 |
| 3-year #CMB/patient—mean (SD) | 4.1 (4.6) | 4.0 (4.5) | 4.2 (4.9) | t = −.106; P = .91 |
| 3-year cumulative prevalence, % (N) | 73.5 (25) | 76.5 (13) | 70.5 (12) | X2 = 0.1511; P = .70 |
| Median latency to first CMB (range) | 17 (1-55) | 13 (1-51) | 17.5 (4-55) | t = −1.36; P = .186 |
| Anatomic distribution | X2 = 4.980; P = .08 | |||
| % lobar (N) | 63.0 (228) | 61.0 (130) | 65.8 (98) | |
| % deep territory (N) | 33.7 (122) | 33.8 (72) | 33.6 (50) | |
| % infratentorial (N) | 3.0 (11) | 4.7 (10) | 0.7 (1) | |
| Laterality | X2 = 4.986; P = .026 | |||
| % IL (N) | 77.3 (280) | 73.2 (156) | 83.2 (124) | |
| % CL (N) | 22.7 (82) | 26.8 (57) | 16.8 (25) | |
| Radiation field correlation [% of target Gy dose] | X2 = 12.604; P = .002 | |||
| % main field [≥ 90%] (N) | 57.3 (192/335) | 53.2 (99/186) | 62.4 (93/149) | |
| % margin [70%-89%] (N) | 20.6 (69/335) | 16.1 (30/186) | 26.2 (39/149) | |
| % outside main field [<70%] (N) | 22.1 (74/335) | 30.6 (57/186) | 11.4 (17/149) | |
| Median Gy dose received (range) | 50.6 (8.0–62.9) | 51.6 (8.0–62.9) | 50.0 (6.0–56.6) RBE | |
| Brain tissue necrosis (TN) | ||||
| No. of patients included | 34 | 17 | 17 | |
| % 3-year cumulative incidence (N) | 11.8 (4/34) | 5.9 (1/17) | 17.6 (3/17) | X2 = 1.1333; P = .29 |
| Median onset latency (range) | 18 (10–30) | 30 (n/a) | 12 (10–24) | n/a |
| Median duration until resolution (range) | 31.5 (9–48) | 21 (n/a) | 42 (9–48) | n/a |
| % symptomatic (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| % w/ treatment (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| N radiographic dx/biopsy | 3 / 1 | 1 / 0 | 2 / 1 | n/a |
| Median no./patient (range) | 1 (0) | 1 (n/a) | 1 (0) | n/a |
| Median maximum diameter, mm (range) | 6 (2–30) | 2 (n/a) | 7 (5–30) | n/a |
| % periventricular location (N) | 50.0 (2/4) | 100.0 (1/1) | 33.3 (1/3) | n/a |
| % in main radiation field [≥90%] (N) | 100.0 (4/4) | 100.0 (1/1) | 100.0 (3/3) | n/a |
| Radiotoxicity type | Total cohort | XRT | PRT | P value for difference between groups |
|---|---|---|---|---|
| Radiation-induced leukoencephalopathy (RIL) | ||||
| No of patients included | 33/34 | 16/17 | 17/17 | |
| Average RIL Grade (G0–3) | ||||
| Baseline (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.0 | 0.8 | 1.1 | t = 2.379, P = .024 |
| Hemispheric, IL/CL | 1.0 / 0.7 | 0.8 / 0.7 | 1.1 / 0.8 | t = 2.379, P = .024/t = 0.484, P = .632 |
| 1 year (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.6 | 1.4 | 1.7 | |
| Hemispheric, IL/CL | 1.6 / 1.0 | 1.4 / 0.9 | 1.7 / 1.1 | |
| 2 years (N) | 33/34 | 16/17 | 17/17 | |
| Global | 1.8 | 1.6 | 1.9 | |
| Hemispheric, IL/CL | 1.8 / 1.2 | 1.9 / 1.2 | 1.9 / 1.2 | |
| 3 years (N) | 27/34 | 13/16 | 14/17 | |
| Global | 1.9 | 1.8 | 2.1 | t = 1.212, P = .237 |
| Hemispheric, IL/CL | 2.0 / 1.2 | 1.8 / 1.2 | 2.1 / 1.3 | t = 1.212, P = .237/t = 0.214, P = .833 |
| 3-year cumulative incidence by grade | X2 = 2.478; P = .3 | |||
| % absent [G0] (N) | 0.0 (0/27) | 0.0 (0) | 0.0 (0) | |
| % mild RIL [G1] (N) | 33.3 (9/27) | 38.5 (5/13) | 28.6 (4/14) | |
| % moderate RIL [G2] (N) | 37.0 (10/27) | 46.2 (6/13) | 28.6 (4/14) | |
| % severe RIL [G3] (N) | 29.6(8/27) | 15.4 (2/13) | 42.9 (6/14) | |
| Average injury dynamics (% change from baseline) | ||||
| % Global RIL change yr 1-3 | 31.8% | 31.3% | 29.4% | t = 0.826, P = .912 |
| % IL RIL change yr 1-3 | 34.8% | 31.3% | 38.2% | t = −0.389, P = .700 |
| % CL RIL change yr 1-3 | 22.2% | 43.8% | 7.8% | t = 2.180; P = .037 |
| Cerebral microbleeds (CMB) | ||||
| No of patients included | 34 | 17 | 17 | |
| All-time total CMB identified | 362 | 213 | 149 | |
| All-time cumulative prevalence, % (N) | 91.2 (31) | 88.2 (15) | 94.1 (16) | X2 = 0.3656; P = .56 |
| All-time #CMB/patient—mean (SD) | 10.6 (12.5) | 12.5 (15.1) | 8.7 (9.4) | t = 0.869; P = .391 |
| 3-year #CMB/patient—mean (SD) | 4.1 (4.6) | 4.0 (4.5) | 4.2 (4.9) | t = −.106; P = .91 |
| 3-year cumulative prevalence, % (N) | 73.5 (25) | 76.5 (13) | 70.5 (12) | X2 = 0.1511; P = .70 |
| Median latency to first CMB (range) | 17 (1-55) | 13 (1-51) | 17.5 (4-55) | t = −1.36; P = .186 |
| Anatomic distribution | X2 = 4.980; P = .08 | |||
| % lobar (N) | 63.0 (228) | 61.0 (130) | 65.8 (98) | |
| % deep territory (N) | 33.7 (122) | 33.8 (72) | 33.6 (50) | |
| % infratentorial (N) | 3.0 (11) | 4.7 (10) | 0.7 (1) | |
| Laterality | X2 = 4.986; P = .026 | |||
| % IL (N) | 77.3 (280) | 73.2 (156) | 83.2 (124) | |
| % CL (N) | 22.7 (82) | 26.8 (57) | 16.8 (25) | |
| Radiation field correlation [% of target Gy dose] | X2 = 12.604; P = .002 | |||
| % main field [≥ 90%] (N) | 57.3 (192/335) | 53.2 (99/186) | 62.4 (93/149) | |
| % margin [70%-89%] (N) | 20.6 (69/335) | 16.1 (30/186) | 26.2 (39/149) | |
| % outside main field [<70%] (N) | 22.1 (74/335) | 30.6 (57/186) | 11.4 (17/149) | |
| Median Gy dose received (range) | 50.6 (8.0–62.9) | 51.6 (8.0–62.9) | 50.0 (6.0–56.6) RBE | |
| Brain tissue necrosis (TN) | ||||
| No. of patients included | 34 | 17 | 17 | |
| % 3-year cumulative incidence (N) | 11.8 (4/34) | 5.9 (1/17) | 17.6 (3/17) | X2 = 1.1333; P = .29 |
| Median onset latency (range) | 18 (10–30) | 30 (n/a) | 12 (10–24) | n/a |
| Median duration until resolution (range) | 31.5 (9–48) | 21 (n/a) | 42 (9–48) | n/a |
| % symptomatic (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| % w/ treatment (N) | 25.0 (1/4) | 0.0 (0/1) | 33.3 (1/3) | n/a |
| N radiographic dx/biopsy | 3 / 1 | 1 / 0 | 2 / 1 | n/a |
| Median no./patient (range) | 1 (0) | 1 (n/a) | 1 (0) | n/a |
| Median maximum diameter, mm (range) | 6 (2–30) | 2 (n/a) | 7 (5–30) | n/a |
| % periventricular location (N) | 50.0 (2/4) | 100.0 (1/1) | 33.3 (1/3) | n/a |
| % in main radiation field [≥90%] (N) | 100.0 (4/4) | 100.0 (1/1) | 100.0 (3/3) | n/a |
Abbreviations: CL, contralesional; CMB, cerebral microbleeds; dx, diagnosis; IL, ipsilesional; RIL, radiation-induced leukoencephalopathy; SD, standard deviation; TN, tissue necrosis; yr, year.
A, C, E: Evolution of mean RIL grade in XRT (blue lines) and PRT (orange lines) groups at global (A), ipsilesional (C), and contralesional (E) hemispheric levels. In both groups, global and hemispheric RIL grade increases significantly from baseline up through 3-years post-radiotherapy. The PRT group exhibits slightly greater average RIL grade across all time points (finding only significant at baseline). B, D, F: Analysis of RIL injury dynamics (mean % change between 1- and 3-years post-radiotherapy) in XRT (blue lines) and PRT (orange lines) groups at global (B), ipsilesional (D), and contralesional (F) hemispheric levels, demonstrating an association of XRT with significantly greater RIL injury dynamics in the contralesional hemisphere. Mean and SEM are given.
Analysis of RIL injury dynamics, measured as average percent change between 1- and 3-years post-radiotherapy, indicated no significant intergroup differences at a global level (31% [XRT] vs 29% [PRT]) (Figure 2B). However, analysis of RIL injury dynamics at the hemispheric level suggested radiation modality-specific differences: XRT was associated with a significantly greater increase in contralesional hemispheric RIL at 3 years post-radiotherapy compared to PRT (44% [XRT] vs 8% [PRT], t = 2.175; P = .037) (Figure 2F). Notably, whereas PRT-associated RIL changes plateaued after 1 year post-radiotherapy, they continued to worsen significantly following XRT. This effect was not observed in ipsilesional hemispheres (Figure 2D).
Cerebral microbleeds: prevalence, burden, and distribution pattern
Longitudinal radiographic analysis of CMB proved reliable (ICC > 0.9). All-time cumulative prevalence of CMB was 91% (n = 31/34 patients; median follow-up period 5.0 years post-radiotherapy), and a total of 362 CMB (213 [XRT]; 149 [PRT]) were identified and included in the analysis. On average, patients developed 10.6 (SD, 12.5) CMB over the entire follow-up period. CMB typically manifested as numerous small round hypointense lesions on gradient-echo T2*-weighted MRI sequences, occurring in lobar (63%), deep territorial (34%), or rarely, infratentorial (3%) brain regions, largely in ipsilesional hemispheres (77%) within the main RF (57%) (Table 2). Median latency from radiotherapy completion to first CMB manifestation was 17 months (range 1-55). In all cases, individual CMB remained traceable longitudinally on serial MRIs, persisting throughout the entire follow-up period. At 3 years post-radiotherapy, the cumulative prevalence of CMB was 73.5% (n = 25/34), with an average number of 4.1 (SD, 4.6) lesions per patient (Table 2).
The intergroup analysis indicated that CMB 3-year cumulative prevalence, burden, and spatial distribution by anatomical location did not differ significantly between groups, although there was a trend toward greater infratentorial CMB burden (5%[XRT] vs 1%[PRT]; X2 = 4.980; P = .08) (Figure 3A-C). Moreover, CMB onset occurred earlier (median time to first CMB, 13 [XRT] vs 17.5 [PRT] months, P = ns) following XRT. Notably, XRT was associated with significantly greater CMB burden in remote, contralesional hemispheres (27% [XRT] vs 17% [PRT], X2 = 4.986; P = 0.026) (Table 2).
A–C: Intergroup comparison of CMB cumulative prevalence (A), mean number per patient (B), and anatomic distribution (C), demonstrating no significant differences between XRT (blue color) and PRT (orange color) groups. D: Correlative spatial analysis in a PRT-treated patient, depicting the CT-based radiotherapy dose distribution plan (left panel), an axial SWI MRI (middle panel) displaying CMB lesions (black arrows), and a corresponding overlay (right panel) which demonstrates preferential CMB enrichment along the radiation field margin. E–F: Schematic (E) and graphical (F) illustration of significant intergroup differences in CMB distribution (ie, in the main radiation field (RF), at the RF margin, or remote to the RF), suggesting PRT-associated CMB clustering at the RF margin and greater remote CMB development following XRT. G: T1 + contrast MRI (left) demonstrating a small nodular enhancement focus (white arrow) located in the left periventricular white matter (splenium), manifesting at 10 months post-PRT, consistent with TN (clinical course ruled out tumor given indolent nature of the lesion). Radiotherapy dose distribution overlay on axial computed tomography (right) demonstrates prior exposure of this region of interest to the main radiation field (56.2 Gy; orange line) and proximity to the distal fall-off region. H: Cumulative TN incidence following PRT vs XRT up until 30 years post-radiotherapy. Mean and SEM are given unless otherwise stated.
Finally, CMB-to-RF correlation identified significant intergroup differences (Figure 3D–F): PRT was associated with preferential CMB clustering at the RF margin (70-89% target dose) (16% [XRT] vs 26% [PRT]), whereas XRT resulted in greater remote CMB development outside the main RF (<70% target dose) (31% [XRT] vs 11% [PRT]) (X2 = 12.604; P = .002).
Brain tissue necrosis: incidence and distribution pattern
The cumulative incidence of brain tissue necrosis (TN) was 11.8% (n = 4) at 3 years post-radiotherapy (Table 2). Diagnosis of TN lesions was established through longitudinal clinical–radiographic follow up and, in one instance, histopathologically confirmed via stereotactic tissue biopsy. Among the 4 patients who developed brain tissue necrosis, 3 (75%) had IDH1-mutant, 1p19q co-deleted oligodendrogliomas.
Radiographically, TN lesions manifested as de novo nodular contrast-enhancing lesions on axial T1-weighted MRI sequences, measuring 2-30 mm in maximum diameter, located within the periventricular white matter (50%), resection cavity margin (25%), or subcortical white matter (25%). TN lesions occurred at a median of 18 months (range, 10-30) post-radiotherapy, were mostly (75%) asymptomatic, and persisted for a median of 31.5 months until radiographic resolution. One patient developed symptomatic, progressive TN at 10 months post-PRT, with subsequent biopsy confirmation.
Intergroup analysis pointed to radiation modality-specific differences, as TN development at 3 years post-radiotherapy was more common in PRT (17.6%; n = 3) than XRT (5.9%; n = 1) (Figure 3H), but this trend did not reach statistical significance (X2 = 1.1333; P = .29). TN manifestation occurred, on average, earlier following PRT (median onset 12 [PRT] vs 30 [XRT] months). Lesion-to-radiation field correlation identified that all (100%) TN lesions were located in the main RF (> 90% of target dose), in 2 instances (50%) with close proximity to the RF margin (Figure 3G).
Discussion
Delayed CNS injury after cranial radiotherapy remains a major challenge.1,2 Dose distribution advantages of protons over photons minimize low-dose radiation to healthy brain tissue, potentially limiting radiotoxic sequelae and improving treatment-related outcomes.20,21 Our findings suggest that delayed CNS radiotoxicity, in the form of RIL, CMB, and/or TN, is common and progressive in the first 3 years after photon or proton-based radiotherapy in adult patients with gliomas WHO grades 2-3. Most patients developed moderate-to-severe RIL (67%) and CMB (74%); a subset experienced TN (12%). Our figures are consistent with previous studies indicating similar CNS radiotoxicity rates in patients with brain tumors. For instance, Terziev et al (2021) reported a 42% 3-year cumulative RIL incidence in 81 patients with high-grade glioma.5 Another study identified that in patients irradiated to doses>20 Gy, the incidence of white matter changes detectable on MRI was nearly 100% 1 year post-radiotherapy.36 Analogously, all patients in our study developed at least mild RIL by the 2-year time point. Similarly, for CMB, prevalence rates between 67% and 100% have been documented in patients with CNS malignancies within the first years of radiotherapy.14,37,38 TN development can vary substantially depending on treatment variables, tumor histology, patient characteristics, and diagnostic criteria10; in cranially irradiated patients with glioma, cumulative rates range between 5% and 14%.11,39
Known risk factors of CNS radiotoxicity include variables related to the treatment (eg, radiation type, modality, dose, volume, fractionation schedule, systemic anti-neoplastic treatment), the patient (eg, age, cognitive reserve, comorbidities, smoking history, genetic predisposition), and the tumor (eg, molecular-genetic profile, histology, location).1 Several of these risk factors were prevalent among our patient cohort, including high rates of pre-existing cardiovascular comorbidities and substantial exposure to additional antineoplastic treatment (surgery, systemic therapies). While it is likely that chemotherapy contributed to some of the neurotoxic effects observed, systemic antineoplastic therapy exposure was well-balanced between groups. Moreover, among the 4 patients with TN, 3 (75%) were found to have IDH1-mutant, 1p/19q co-deleted oligodendrogliomas, ie, tumor-related variables associated with higher rates of radiotoxicities, including TN.40,41
Notably, intergroup analysis suggested radiation modality-specific injury patterns, as spatiotemporal radiographic features of RIL and CMB differed following XRT vs PRT (cf. Figure 4). These potentially distinct radiotoxicity profiles likely reflect underlying dosimetric and radiobiological differences between protons and photons. The purported dosimetric advantage of protons stems from the lack of exit dose thereby sparing normal tissues from low-dose radiation. The majority of proton dose deposition and maximum linear energy transfer occur at the distal end of the spread-out Bragg peak, where the RBE of protons may substantially increase, causing unwanted toxicities.42 Conversely, modern highly conformal photon-based techniques (eg, IMRT, VMAT) may increase normal tissue exposure to a low-dose radiation bath, potentially contributing to adverse effects.
![Schematic illustration of observed radiation modality-specific CNS injury pattern, including differences in radiation-induced leukoencephalopathy (RIL; gray-shaded areas), cerebral microbleeds (CMB; black dots), and brain tissue necrosis (TN; delineated gray area). A: Proton dose distribution plan overlayed on axial CT in a patient with grade 3 astrocytoma, demonstrating highly conformal dose delivery around the target (tumor resection cavity [RC]) with relative sparing of the contralesional hemisphere but greater ipsilesional RIL severity as well as distinctive CMB enrichment and TN manifestation at the radiation field margin. B: Dosimetric comparison with a photon (VMAT) plan in the same patient illustrates contralesional hemisphere exposure to intermediate- and low-dose radiation, with comparatively greater RIL injury dynamics and CMB manifestation outside the main radiation field.](https://oup.silverchair-cdn.com/oup/backfile/Content_public/Journal/oncolo/29/12/10.1093_oncolo_oyae195/1/m_oyae195_fig4.jpeg?Expires=1747852232&Signature=h3QJM7DCqqLTjTqiO2q2scbPLU7Iu8iFqPL3vVQ8NQm61Dsq6oO5IXkmxbjTZCsM-PBmBbwZAQN8mIXKFdWcObBawwsi7fibo-FolAmW7uOV2amf8qBlwFVCo1AlDqqLJJ7UcflujCPwNJG6ChsgmbdFRuNBJJl89z7roz~WlgjyJpx1s0rgS7FzFQRbQ74eam-kmHpKjwFDawo~rVgWepVScBw~~WDwsswECxyFRnKYaEAmx38siG7RrnJIvJJxVd~ucrUeW-oGv3UVe7n2XTNipoMf4kxW0lOco9x~ZzfU4zg349v4LHDlzV6caZkco8HtiAO1Q1F9k-1ws5Jc9g__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
Schematic illustration of observed radiation modality-specific CNS injury pattern, including differences in radiation-induced leukoencephalopathy (RIL; gray-shaded areas), cerebral microbleeds (CMB; black dots), and brain tissue necrosis (TN; delineated gray area). A: Proton dose distribution plan overlayed on axial CT in a patient with grade 3 astrocytoma, demonstrating highly conformal dose delivery around the target (tumor resection cavity [RC]) with relative sparing of the contralesional hemisphere but greater ipsilesional RIL severity as well as distinctive CMB enrichment and TN manifestation at the radiation field margin. B: Dosimetric comparison with a photon (VMAT) plan in the same patient illustrates contralesional hemisphere exposure to intermediate- and low-dose radiation, with comparatively greater RIL injury dynamics and CMB manifestation outside the main radiation field.
In this study, PRT was associated with greater ipsilesional white matter hyperintensities at the post-radiotherapy baseline, preceding the delayed occurrence of RIL observed in these regions at subsequent time points. While the reasons are unknown, we speculate that the distinct radiobiological properties of protons—particularly RBE variations at the radiation field margin—may effectuate a greater degree of early focal inflammatory white matter changes in the tumor harboring hemisphere, relative to photons. However, some caution must be exercised when interpreting this finding at baseline, as the possibility of a higher extent of pre-existing white matter signal hyperintensities in the PRT group (ie, prior to radiotherapy) cannot be ruled out. Intra-cerebral variations in tissue response after proton therapy (ie, greater radiosensitivity of the periventricular region) constitute additional important considerations for treatment planning.43 Conversely, XRT appeared to result in greater RIL progression in the contralesional hemisphere over time, potentially reflecting “off-target” white matter injury from collateral radiation exposure beyond the intended target. RIL frequently co-occurs with progressive brain atrophy1,8; both conditions have been linked to radiation-induced cognitive impairment.2,44 Recent studies have investigated potential differences in the extent of radiation-induced brain atrophy following XRT vs PRT.9,25,28 Early evidence suggests improved gray and white matter volume preservation in the hemisphere contralateral to the tumor/target following proton radiotherapy, given superior dose conformality with greater normal tissue sparing, especially of the contralesional hemisphere.9,25,28 Whether the differential brain injury patterns associated with protons vs photons significantly impact neurocognitive outcomes is subject of ongoing investigations.45 Emerging cohort studies point to stable cognitive function and preserved quality of life in brain tumor patients up until 5 years after PRT.25,27,46
For CMB, prevalence, burden, and anatomic distribution pattern at 3 years post-treatment were similar across modalities. Consistent with previous studies suggesting predominant CMB formation in brain areas receiving >30 Gy,37,47 herein identified CMB received, on average, radiation doses of >50Gy and were mostly (57%) located in the main radiation field (ie, within 90% isodose line). Interestingly, photons were associated with greater contralesional hemispheric CMB burden, with a greater proportion of CMBs located outside the main radiation field, relative to protons. This injury pattern appears consistent with the proportionally greater contralesional RIL burden observed with photons in this study. Conversely, lesion-to-radiation field correlation identified distinctive clustering of CMB at the radiation field margin with protons. This observed preferential enrichment of CMB at the radiation field distal edge appears to be a proton-specific toxicity pattern given known RBE uncertainties in this region. Despite existing robust optimization methods to control for uncertainties in physical and biological dose distributions, unanticipated toxicities stemming from setup, organ motion, range, and RBE uncertainty could occur.29,48,49 For example, several studies have investigated a potential spatial correlation of high linear energy transfer regions with late post-radiation imaging changes, including TN.29,50,51 In a prospective study of 110 patients with low-grade glioma treated with PRT, Harrabi et al found that radiation-induced contrast-enhancing lesions (corresponding largely to asymptomatic TN cases) were almost exclusively seen at the proton beam distal end.29 To the best of our knowledge, our study is the first to suggest that cranial PRT is associated with a similar enrichment of CMB just distal to the main radiation field.
We further observed that PRT was associated with 3-fold higher TN rates at 3 years post-radiotherapy, compared to XRT. While this difference did not reach statistical significance, other studies have suggested similarly high TN rates with protons.29,30 In line with observations by Harrabi et al, all identified TN lesions occurred in maximum radiation dose areas and/or with close proximity to the distal edge of the main radiation field.
Our study has several strengths and limitations. CNS radiotoxicity was assessed objectively, holistically (including RIL, CMB, and TN), and longitudinally during periods of stable disease with PFS up to 3 years post-radiotherapy. This approach permitted robust spatiotemporal radiographic characterization of delayed radiogenic brain injury patterns. Of note, intergroup differences in follow-up time and disease recurrence (ie, longer follow up in the XRT cohort with consequently higher rates of recurrence in that group) were not of relevance to the imaging outcomes evaluated during the defined 3-year post-radiotherapy PFS interval. The development of a scoring system to quantify RIL in patients with (resected or unresected) glioma constitutes a novel and unique aspect of this study. Based on the widely established and validated Fazekas scale, this new instrument proved reliable for quantitative MRI-based analysis of treatment-related white matter changes in neuro-oncological patients. Important limitations of this study include its retrospective design and relatively small cohort size. Despite these constraints, XRT and PRT groups were meticulously matched to optimize intergroup comparability and minimize confounders. Another potential constraint relates to our 3-tiered stratification of the radiation field (ie, main field [≥90% of prescribed target Gy dose]—margin [70-89%]—outside main field [<70%]). As the inherent radiobiological differences between photons and photons translate to proportionally greater radiation field margin surface areas with photon plans (cf. Figure 3E), the herein observed PRT-associated enrichment of CMB in radiation field margins may actually be an underestimation. MRI scans in this study were clinically acquired at potentially different field strengths over time. Given the study’s retrospective design, inclusion of neurocognitive data to correlate injury patterns with functional outcomes was not possible. Finally, this study was designed to generate hypotheses based on our retrospective analyses, and stringent corrections for multiple comparisons were not used, leading to the possibility of type 1 error (ie, false-positive). In this context, we judged the possibility of type 2 error to be more detrimental. Future prospective research is needed and can use these methods to generate and test hypotheses. Indeed, several clinical trials are underway to assess neurocognitive phenotypes following PRT vs photon-based techniques.45,52
Conclusion
Our findings suggest a high burden of delayed and progressive CNS radiotoxicity, including RIL, CMB, and TN, in patients with gliomas WHO grades 2-3 following partial cranial radiotherapy. Our intergroup analysis points to distinct radiographic signatures associated with photon vs proton radiation-mediated brain injury. These distinct radiotoxicity profiles likely reflect underlying dosimetric and radiobiological differences between protons and photons. While photons are potentially associated with greater and sustained RIL and CMB burden in contralesional hemispheres, protons may elevate CMB and possibly TN risk at the distal radiation field. These novel findings may help inform the prevention, diagnosis, and management of CNS radiotoxicity and improve clinical decision-making in neuro-oncology. Further investigations, including prospective clinical trials with larger cohort sizes and longer follow-up periods, are warranted to validate these findings and determine the impact of radiation modality-specific structural sequelae on neurocognitive function.
Supplementary material
Supplementary material is available at The Oncologist online.
Acknowledgments
Portions of this study were presented by S.F.W. at the Society for Neuro-Oncology (SNO) 2022 Annual Meeting, in Tampa, FL, USA.
Author contributions
Conception/design: Sebastian F. Winter, Jorg Dietrich. Data acquisition/analysis: Sebastian F. Winter, Melissa M. Gardner, Katarina Nikolic, Philipp Karschnia, Eugene J. Vaios, Clemens Grassberger, Marc R. Bussière, Helen A. Shih, Michael W. Parsons. Data interpretation: Sebastian F. Winter, Melissa M. Gardner, Philipp Karschnia, Eugene J. Vaios, Clemens Grassberger, Marc R. Bussière, Katarina Nikolic, Thanakit Pongpitakmetha, Felix Ehret, David Kaul, Wolfgang Boehmerle, Matthias Endres, Helen A. Shih, Michael W. Parsons, Jorg Dietrich. Manuscript drafting/revision: Sebastian F. Winter, Melissa M. Gardner, Philipp Karschnia, Eugene J. Vaios, Clemens Grassberger, Marc R. Bussière, Katarina Nikolic, Thanakit Pongpitakmetha, Felix Ehret, David Kaul, Wolfgang Boehmerle, Matthias Endres, Helen A. Shih, Michael W. Parsons, Jorg Dietrich. Manuscript final approval: Sebastian F. Winter, Melissa M. Gardner, Philipp Karschnia, Eugene J. Vaios, Clemens Grassberger, Marc R. Bussière, Katarina Nikolic, Thanakit Pongpitakmetha, Felix Ehret, David Kaul, Wolfgang Boehmerle, Matthias Endres, Helen A. Shih, Michael W. Parsons, Jorg Dietrich.
Conflict of interest
The authors declare no Conflicts of Interest.
Funding
Berlin Institute of Health / Charité Junior Clinician Scientist Program Grant (to S.F.W.); Amy Gallagher foundation (to J.D.); Derrick Wong Family foundation (to J.D.); German Research Foundation, Germany’s Excellence Strategy (EXC-2049-390688087, to M.E.); Collaborative Research Center ReTune (TRR 295-424778381, to M.E.); Federal Ministry of Education and Research (BMBF) (to M.E.); German Center for Neurodegenerative Diseases (DZNE); (to M.E.); German Centre for Cardiovascular Research (DZHK) (to M.E.); European Union (to M.E.); Corona Foundation (to M.E.); Fondation Leducq (to M.E.); German Cancer Aid (Mildred-Scheel Postdoctoral Fellowship) (to F.E.); National Institutes of Health/National Cancer Institute (5R38-CA245204; to E.J.V.).
Data availability
Research data for this work are available upon request to the corresponding author by any qualified investigator.
