[1] 14w Radiation-Induced Cerebral Microbleeds in Pediatric Patients With Brain Tumors Treated With Proton Radiation Therapy

[1] 307w This study investigates radiation-induced cerebral microbleeds (CMBs) in pediatric patients with primary brain tumors treated with proton radiation therapy. This complication of radiation therapy has not been studied significantly after proton radiation therapy in the pediatric patient population, particularly with regard to timing, incidence, and associated risk factors. CMBs have been associated with decreased neurocognitive function, indicating potential clinical significance. This research Purpose: Proton beam radiation therapy (PBT) has been increasingly used to treat pediatric brain tumors; however, limited information exists regarding radiation-induced cerebral microbleeds (CMBs) among these patients. The purpose of this study was to evaluate the incidence, risk factors, and imaging appearance of CMBs in pediatric patients with brain tumors treated with PBT. Materials and Methods: A retrospective study was performed of 100 pediatric patients with primary brain tumors treated with PBT. CMBs were diagnosed by examination of serial magnetic resonance imaging scans, including susceptibility-weighted imaging. Radiation therapy plans were analyzed to determine doses to individual CMBs. Clinical records were used to determine risk factors associated with the development of CMBs in these patients. Results:The mean age at time of PBTwas 8.1 years. The median follow-up duration was 57 months. The median time to development of CMBs was 8 months (mean, 11 months; range, 3-28 months). The percentage of patients with CMBs was 43%, 66%, 80%, 81%, 83%, and 81% at 1 year, 2 years, 3 years, 4 years, 5 years, and >5 years from completion of proton radiation therapy. Most of the CMBs (87%) were found in areas of brain exposed to 30 Gy. Risk factors included maximum radiation therapy dose (P Z .001), percentage and volume of brain exposed to 30 Gy (P Z .0004, P Z .0005), and patient age at time of PBT (P Z .0004). Chemotherapy was not a significant risk factor (P Z .35). No CMBs required surgical intervention.

[1] 123w Radiation therapy is one of the primary therapies used to treat pediatric brain tumors. Adverse effects of radiation therapy on the brain include radiation necrosis, atrophy, gliosis, telangiectasia, microhemorrhages, cavernous malformations, and large vessel vasculopathy. Magnetic resonance imaging (MRI) scans performed after cranial radiation therapy frequently detect small parenchymal lesions that demonstrate susceptibility artifact and have been termed radiation-induced cerebral microbleeds (CMBs). CMBs have been found to be associated with worse executive function in pediatric brain tumor survivors treated with radiation therapy and among patients with nasopharyngeal carcinoma treated with radiation therapy (1,2). Among patients without history of cranial radiation therapy, CMBs also have been associated with cognitive impairment in Alzheimer disease, stroke, vascular dementia, small vessel ischemic disease, and increasing age (3)(4)(5)(6)(7)(8)(9)(10).

[2] 192w Compared with conventional (photon) radiation therapy, proton beam therapy (PBT) offers the advantages of the absence of an exit dose, a highly conformal dose distribution, and a reduced radiation dose to adjacent normal tissue (11). Therefore, potential benefits of PBT in patients with pediatric brain tumors may include reduction of the negative long-term effects of radiation, such as cognitive deficits, endocrine abnormalities, vascular abnormalities, and secondary malignancies (12). PBT increasingly has been used to treat pediatric brain tumors, including craniopharyngiomas, ependymomas, germinomas, and medulloblastomas; however, limited information exists regarding CMBs after PBT (13)(14)(15)(16). Although CMBs can be diagnosed on histopathology, brain MRI is the preferred method for detection of CMBs. Gradient echo (GRE) imaging and susceptibility-weighted imaging (SWI) MRI sequences are necessary for detection of CMBs because these MRI sequences have been developed to increase the conspicuity of the blood products that cause CMBs. SWI represents the most current and advanced MRI sequence commercially available for detection of CMBs and has been shown to be much more sensitive than GRE for detection of CMBs (17)(18)(19). Therefore, clinical research investigating the formation of CMBs should include SWI for the best possible imaging assessment.

[3] 31w The purpose of this research was to evaluate the incidence, imaging appearance, and risk factors for radiationinduced CMBs among pediatric patients with brain tumors treated with PBT using MRI with SWI.

[1] 54w A total of 100 pediatric patients met inclusion criteria, and 26 patients were excluded from the study. Mean age at time of PBT was 8.1 years (range, 0.75-18 years), and the male:female ratio was 63:37. Median follow-up duration was 57 months (mean, 52 months; range, 7-116 months). Patient characteristics are listed in Table 1.

[2] 199w The median time to development of CMBs was 8 months (mean, 11 months; range, 3-28 months). The percentage of patients with CMBs was 43% (16 of 37), 66% (27 of 41), 80% (20 of 25), 81% (26 of 32), 83% (35 of 42), and 81% (29 of 36) at 1 year, 2 years, 3 years, 4 years, 5 years, and >5 years, respectively, from completion of proton radiation therapy (Fig. 1). The median number and range of CMBs per patient was 6 (0-49) at 1 year, 6 (0-63) at 2 years, 6 (0-72) at 3 years, 7 (0-130) at 4 years, 8 (0-136) at 5 years, and 11 (0-136) at >5 years from completion of proton radiation therapy (Fig. 2). The median size of CMBs was 0.2 cm (range, 0.1-0.5 cm). CMBs were detected in all areas of the brain, specifically the brain stem, cerebellum, thalami, basal ganglia, and cerebral hemispheres. No CMBs demonstrated resolution on follow-up imaging. No CMBs required surgical intervention. Four patients (4%) demonstrated imaging appearance consistent with a cavernous malformation, which developed at a median time of 46 months (range, 14-72 months). Representative examples of CMBs and cavernomas are shown in Figures 3 and 4.

[3] 128w Evaluation of radiation dose to individual CMBs demonstrated that 87% of CMBs occurred in areas of the brain exposed to 30 Gy (32.8% for 30-40 Gy, 18.8% for 40-50 Gy, 35.7% for >50 Gy) and 13% in areas exposed to <30 Gy (4.1% for <10 Gy, 0.6% for 10-20 Gy, and 7.9% for 20-30 Gy). There were statistically significant differences among patients without CMBs and patients with CMBs in the percentage of brain exposed to 30 Gy (10.1% vs 42.1%, P Z .0004) and volume of brain exposed to 30 Gy (133 cm 3 vs 601 cm 3 , P Z .0005). There was a statistically significant positive correlation between the number of CMBs and percentage of brain exposed to 30 Gy (r Z 0.39, P Z .0002).

[4] 73w There was a statistically significant difference in mean age at time of PBT between patients with and without CMBs (6.9 vs 10.6 years, P Z .0004). There was a statistically significant difference in maximum radiation therapy dose between patients with and without CMBs (55.9 vs 51.6 Gy, P Z .001). Chemotherapy was not a statistically significant risk factor (P Z .35) for development of CMBs between patients treated with chemotherapy versus without chemotherapy.

[1] 351w In this study, approximately 80% of pediatric patients with brain tumor treated with PBT developed CMBs, indicating a high frequency in this patient population. In comparison, the incidence of CMBs in pediatric patients with primary brain tumors treated with conventional/photon radiation therapy has been reported to occur over a wide range of frequencies, approximately 3% to 80% (1,(21)(22)(23). The high incidence found in our study is similar to a study by Passos et al of 100 childhood primary central nervous system tumors treated with photon radiation therapy, where CMBs were identified in 80.6% of patients (23). However, Passos et al also reported cavernous malformations in 52.8% of patients, which is different from the 4% in our study (23). A potential explanation for differences in cavernous malformations may be a difference in follow-up duration, approximately 4.5 years in this study compared with 11 years in Passos et al (23). The wide range in incidence of CMBs and cavernous malformations after photon radiation therapy is also dependent on whether GRE, SWI, or neither was used to detect them. Careful review of the research study methods is necessary because studies reporting CMBs using GRE for detection will tend to demonstrate a lower incidence compared with studies using SWI. For example, Burn et al ( 1) reported an incidence of 3.4% but did not use GRE or SWI for detection of CMBs, whereas Roddy et al ( 21) reported an incidence of 48.8% but used a combination of GRE and SWI imaging. Undoubtedly, the use of SWI to detect CMBs in this study results in an incidence that is at the high end of the range reported with photon radiation therapy. We believe that using SWI, rather than GRE, to detect CMBs in our study is a major advantage because it represents the most sensitive of MRI technology currently clinically available and allows the best imaging depiction of what is present histopathologically. If MRI technology advances from SWI to an even greater level of sensitivity for detection of CMBs, future studies would be necessary to determine whether the incidence is even greater in these patients.

[2] 109w CMBs appeared at a median of 8 months after completion of PBT. Similar to our results, Peters et al demonstrated that, in children treated with photon radiation therapy, CMBs may appear as early as 5 months and with much more variable rates of formation and faster rates than are observed in adults (24). Tanino et al demonstrated a wider time range to development of CMBs of 3 months to 9 years (mean, 33 months), but this study included children and adults with a mean age of 49 years (range, 13-78 years) (25). These results suggest that development of CMBs is affected by patient age at time of radiation therapy.

[3] 265w In this study, we have identified volume of brain exposed to 30 Gy, higher maximum PBT dose, and younger age at time of PBT as risk factors for the development of CMBs. These risk factors likely in part explain why some patients do not develop CMBs. Our results demonstrated that 87% of CMBs occurred in areas exposed to 30 Gy, which is similar to the 91% occurring in areas receiving >30 Gy reported by Roddy et al (21), who evaluated patients treated with photon radiation therapy. Our results are not completely similar, however, to those of Tanino et al (25); in a smaller study of 34 pediatric and adult patients treated with photon radiation therapy, they reported that all CMBs were in areas exposed to >25 Gy. The percentage of patients with CMBs stabilized at 3 years after completion of PBT, indicating if a patient will develop CMBs, it will most likely occur within 3 years of completion of proton radiation therapy. However, the number of CMBs per patient continued to increase over time. Similar to these findings, Lupo et al demonstrated that the rate of CMBs increased after 2 years after radiation therapy for adult patients with high-grade gliomas (26). We also demonstrate that younger age at time of PBT was a risk factor for development of CMBs. Similarly, Passos et al demonstrated that age at time of radiation therapy was a risk factor for development of CMBs (23). Lastly, in this study, chemotherapy was not associated with development of CMBs, similar to the findings of the previous study by Passos et al (23).

[4] 425w This study has a few important limitations. First, pathologic correlation to confirm these lesions represent CMBs was not performed, and it is our presumption that these CMBs are secondary to radiation therapy. It is not standard clinical care to biopsy these areas, and prior research has demonstrated that CMBs are not seen in patients treated only with chemotherapy (23,27,28). One potential explanation for the association between radiation therapy and CMBs is that vascular endothelial growth factor is increased after radiation therapy (29-31). There are rare genetic causes of multiple cavernomas in which patients have numerous cavernomas, although such causes are unlikely in our patient population. Genetic testing was not performed in these patients because none had multiple cavernomas before PBT, which would be expected with a familial cavernoma syndrome, and because a high percentage of patients (w80%) develop CMBs after PBT, indicating that a rare genetic syndrome etiology is unlikely. However, it is possible that unknown genetic differences among these patients may predispose some patients to greater susceptibility to CMB formation after exposure to radiation therapy. Although CMBs can also be caused by systemic hypertension, amyloid angiopathy, or diffuse axonal injury, our patient population does not support these as potential etiologies, which is why we believe the CMBs in these patients are related to PBT. Second, our criteria for including patients with SWI imaging resulted in some patients who could not be evaluated at every time interval from completion of PBT. SWI has been replacing GRE at many institutions and represents the most sensitive MRI technique for the detection of hemorrhage currently available for clinical use. Several factors can affect the image quality and sensitivity of the SWI sequence, including echo time, flip angle, slice thickness, and magnetic field strength; however, the optimal parameters have yet to be determined (4). Because this was a retrospective study, we could not control for patients with SWI sequence performed at either 1.5T or 3T. We chose to include all patients with SWI performed at 1.5T or 3T and acknowledge this choice could affect our results. Lastly, we did not assess cognitive function in these patients. Because of the retrospective nature of this study, not all patients had neuropsychology testing, and the neuropsychology testing was neither performed at a consistent time after PBT nor standardized to provide an adequate analysis. We anticipate a similar negative impact on cognitive function associated with CMBs, which has been previously reported (1,2). Future research in correlating CMBs with neuropsychology testing and longer follow-up duration to assess development of cavernous malformations would be useful.

[1] 68w CMBs are seen with high frequency and develop in the first few years after PBT for brain tumors in pediatric patients. Risk factors for development of CMBs include younger age at time of PBT, percentage and volume of brain exposed to 30 Gy, and higher maximum radiation therapy dose. These findings demonstrate similarities with CMBs that develop in pediatric patients with brain tumor treated with photon radiation therapy.

[1] 170w After institutional review board approval, a retrospective study was performed from January 2010 to January 2017 including pediatric patients aged 18 years with primary brain tumors who were treated with PBT. All patients either had a newly diagnosed primary brain tumor and were subsequently treated with PBT or had low-grade gliomas without prior treatment with radiation therapy and demonstrated tumor progression on chemotherapy necessitating treatment with PBT. Therefore, patients were excluded if there was any history of treatment with photon radiation therapy, including before, concurrent with, or after PBT. Patients were also excluded if there was more than 1 course of PBT. Patients without a minimum 3-month follow-up brain MRI after completion of PBT were excluded, including if death occurred before 3 months. Patients were also excluded if there was a clinical history of hypertension, known coagulopathy, or collagen vascular disease or other medical history that would predispose to intracranial hemorrhage. PBT treatment doses followed the standard of care in the United States at a children's oncology group treatment center.

[2] 71w Brain MRI consisted of imaging performed with 1.5T or 3T (Avanto and Verio, Siemens, Erlangen, Germany) MRI units with standard MRI sequences at our institution, including sagittal T1-weighted (T1W) magnetizationprepared rapid-acquisition gradient echo, axial T2weighted (T2W) turbo spin echo, axial fluid attenuated inversion recovery, axial diffusion-weighted imaging, axial SWI, coronal T1W turbo spin echo postcontrast with fat saturation, and axial 3-dimensional T1W magnetization prepared rapid acquisition gradient echo postcontrast pulse sequences.

[3] 211w Postcontrast imaging was performed in all patients after intravenous administration of 0.1 mmol/kg gadobenate dimeglumine (MultiHance; Bracco Diagnostics, Princeton, NJ). SWI was performed with the following parameters: 1.5T, flip angle 15, TR 49 ms, TE 40 ms, and 2 mm slice thickness; or 3.0T, flip angle 30, TR 27 ms, TE 20 ms, and 2 mm slice thickness. In most patients, brain MRI scans were performed approximately every 1 to 3 months for 2 to 3 years, then every 6 to 12 months with shorter or longer follow-up depending on clinical need to assess tumor stability, clinical symptoms, or other complications, including radiation necrosis. Patients were included if 1 or more of the follow-up brain MRI scans included SWI. Patients were excluded if none of the follow-up brain MRI scans included an SWI sequence. Because the SWI sequence was not available on all MRI scanners at our institution dating back to 2010, some patients did not have SWI on every followup brain MRI. The number of patients who had SWI available for review at each yearly interval was indicated in the results by the total number of patients at each time point. Only patients with SWI on every follow-up brain MRI were used to determine the time to development of CMBs.

[4] 172w A fellowship-trained, board-certified neuroradiologist (S.K.) with certificate of added qualification in neuroradiology independently evaluated all brain MRI scans of included patients. SWI sequences in conjunction with the other MRI sequences were evaluated for presence, number, and largest size of CMBs or cavernous malformation. Clinical data included patient age at time of PBT, patient sex, tumor pathology from surgical resection or biopsy, tumor location, total cranial radiation dose, chemotherapy, timing from completion of PBT to imaging diagnosis of CMBs, and surgical intervention for CMBs. CMBs were defined similarly to consensus MRI criteria described by Greenberg et al (20) as an intraparenchymal small (5 mm or less) round or ovoid hypointensity on SWI imaging that did not correspond to vessels or tumor, that did not border the surgical resection site or expected location of mineralization, and that was not hyperintense on T1W or T2W imaging (20). To distinguish CMBs from a cavernous malformation, a cavernous malformation was defined as a round or ovoid hypointensity on SWI that demonstrated hyperintense T1W and/or T2W appearance (20).

[5] 168w To determine the specific amount of radiation received by CMBs, radiation dosimetric treatment plans were either overlaid and aligned with SWI MRI sequence using Eclipse image registration software (v13.7, Varian Medical Systems) to match voxel intensity values or the locations of CMBs on SWI images were cross-referenced to locations on the treatment plan by side-by-side comparison. Overlaid MRI scans were first registered to the planning computed tomography scan using automatic match based on bony anatomy, after which the alignment was manually adjusted to correct for any errors. Dose information for individual CMBs were then recorded by adjusting the dose slider in 10 Gy increments. Based on a prior study indicating most CMBs occur at >30 Gy, the percentage and overall volume of brain receiving 30 Gy of total dose was calculated for each patient by using the treatment planning software, which determined total volumes of the contoured brain and the dose-volume histogram corresponding to the 30 Gy dose level (1). This information was then used for statistical analysis.

[6] 66w Statistical analysis of risk factors for CMBs including age at time of PBT, chemotherapy, percentage and volume of brain exposed to 30 Gy, and maximum proton radiation therapy dose was performed by using an unpaired t test, Fisher exact test, or Pearson correlation coefficient where appropriate. A P value <.05 was considered statistically significant. Statistics were performed using Graphpad Prism 7 Statistical Software (La Jolla, CA).