A secondary retrospective case-control study design was used to investigate demographic, clinical (e.g., treatment variables), and histopathological data in a total of 34 adult patients with lower grade glioma (WHO grades 2–3) treated with either XRT or PRT at the Massachusetts General Hospital (MGH) between 11/1998–10/2017. Patients treated with PRT were identified from ongoing longitudinal single-arm outcome studies at our institution (NCT01358058, NCT03286335), which included routine brain MRI and serial neuropsychological assessments. Patients treated with XRT were identified from an archival database from the Departments of Radiation Oncology and the Division of Neuro-Oncology at MGH. All PRT patients signed informed consent for participation in the outcome studies enumerated above. The XRT patient data was gathered under a Mass General Brigham IRB approved research protocol for secondary retrospective analyses of medical record data. All research procedures were conducted in accordance with the Declaration of Helsinki [52] As indicated above, patients treated with XRT had not completed neuropsychological assessment as part of their routine care, so only PRT patients were included in the detailed cognitive analyses.
Inclusion criteria for all patients were: (a) pathologically confirmed grade 2 or grade 3 glioma as defined by the World Health Organization (pathology records included the WHO classification criteria that were in place at the time of diagnosis and were re-classified to meet the 2021 classification based on available data), (b) tumor burden limited to one hemisphere (e.g., tumors did not cross midline to minimize brain volume changes associated with tumor growth or regression after treatment), (c) age 18 years or older, (d) partial cranial radiation therapy (PRT or XRT) delivered at MGH, (e) availability of contrast-enhanced axial brain magnetic resonance imaging scans (e.g., MPRAGE, BRAVO, or thick slice T1 weighted MRI) at baseline (within eight weeks after completion of radiation therapy) as well as at yearly intervals or more frequently (part of routine cancer surveillance screening), and (f) a minimum of two years progression free survival following RT.
To limit sampling bias, XRT patients were closely matched to “case-control” PRT patients using a carefully developed eleven-tiered set of criteria: age, sex, tumor type, tumor location, laterality of tumor, isocitrate dehydrogenase (IDH) 1 mutation status, 1p/19q co-deletion status, concurrent chemotherapy, adjuvant chemotherapy, total radiation dose, and number of radiation fractions. Participants were first matched on an individual level, followed by a group match for patients (n = 6) that did not match on all eleven variables.
Structural MRI scans were collected for patients at baseline, one year, and at two years after RT. When available, thin-slice (1mm) imaging (e.g., MPRAGE, BRAVO, and T1) was used for volume quantification; thick-slice (6mm) images were used for those cases without thin-slice imaging. As in our prior studies [20, 53], we measured volume change in the lateral ventricle in the hemisphere contralateral to the tumor as an index of diffuse cerebral volume loss. Ventricular segmentation was performed manually using Slicer (Version 4.10.2) to generate a 3D model of the contralesional ventricle, as illustrated in Fig. 1. The manual segmentation process was carried out by two trained research technicians following extensive training from board-certified neuro-oncologists (J.D. and E.G.) to ensure accurate identification of the limits of the ventricle. Each researcher identified the area of interest or ventricular space on a single axial slice using a “point and click” method to place a “seed” in the ventricular space and then used the “grow from seeds” function in Slicer to fill in all areas of the ventricle across slices. The program computed volume as the product of voxel size in the individual patient’s image space multiplied by the number of voxels labeled as included in the ventricle. Inter-rater reliability was evaluated by random selection of ventricular measurements for ten patients at each time point using the same method.
INSERT FIGURE 1 HERE
As noted above, PRT patients completed neurocognitive assessment at baseline (e.g., after the surgical resection and before start of radiotherapy) and follow-up (e.g., two years after baseline). All neuropsychological evaluations were carried out by a neuropsychologist (J.S.) with testing assistance from a psychometrist. The detailed neuropsychological evaluation included cognitive screening measures, attention/working memory, processing speed, executive functioning, language, memory, and measures of depression and anxiety (See Supplemental Table 1 for complete list of tests). Embedded within the assessment was the "clinical trial battery composite” (CBTC) which generates a composite score routinely used in large scale clinical trials of cognitive outcomes in patients with brain tumors [54, 55]. The CTBC is the mean of the Z-scores of the following six metrics: Controlled oral word association (COWA), Trail Making Test parts A and B (TMT-A and B), and total recall, delayed recall, and recognition discrimination scores from the Hopkins Verbal Learning Test-Revised (HVLT-R)[54] Normative data (e.g., conversion to z-scores using published test norms) for neuropsychological assessment allows for a standardized comparison of individual performance while controlling for demographic factors (e.g., age, education). In repeat or serial cognitive assessment, it is controversial whether discrete norms yield invalid or additional error [56, 57]; therefore, raw scores were used in analyses assessing change over time. Linear regression and correlation coefficients were calculated to examine the relationship between change in ventricular volumes and change in cognitive performance in patients treated with PRT. Non-parametric tests were used for binary variables indicating cognitive deterioration.
Statistical analyses were performed with the Statistical Package for the Social Sciences (SPSS-29). Descriptive statistics were calculated for demographic, diagnostic and treatment data for the two study groups (PRT and XRT) and compared with non-parametric analyses including the Mann-Whitney U-test of independence to evaluate their equivalence. Percent change in lateral ventricle volume between the two study groups was evaluated using a repeated-measures ANOVA with treatment as the between groups factor (XRT and PRT) and time (year 1 and 2 from baseline) as the within patients variable. Paired t-tests and simple linear regression analyses were used to test overall change from baseline to follow-up. Associations between lateral ventricle volume changes and measures of neurocognitive functioning, mood and anxiety were explored using correlations and linear regressions. To examine the relationship between ventricular volume and cognitive decline, the mean change in raw scores for each cognitive test was determined and those patients whose cognitive decline was 1 standard deviation or more below the mean change for the group were defined as “decliners” on that test. The number of cognitive tests showing a decline was tabulated for each individual and plotted into a simple linear regression. An alpha of .01 was used to control for multiple comparisons of the neurocognitive baseline and follow-up analyses and should be interpreted as exploratory.