Subcortical alterations following pediatric posterior fossa brain tumor treatment: A longitudinal MRI study

Abstract

Background

Increasing survival rates in pediatric brain tumor patients highlight the importance of understanding treatment-related neurodevelopmental consequences. The posterior fossa is a primary site for many of these tumors, which are treated with surgery, chemotherapy, and/or radiation therapy that could impact brain function and structure. This retrospective longitudinal study examines changes in subcortical brain structures of pediatric posterior fossa tumor patients over five years post-diagnosis and compares patients who were treated for low- and high-grade tumors.

Methods

We analyzed 558 T1-weighted (T1w) brain MR images of 57 pediatric posterior fossa tumor patients (mean age at diagnosis = 9.65 (SD 4.06), 44 % female). There were 39 patients with a low-grade tumor treated with surgery and/or chemotherapy and 18 patients with a high-grade tumor treated with surgery, chemotherapy, and radiation treatment. Volumes of the globus pallidus, caudate, putamen, nucleus accumbens, amygdala, thalamus and hippocampus were calculated using SPM CAT12 (Statistical Parametric Mapping Computational Anatomy Toolbox). Linear mixed effect models were estimated to study the association between volumetric changes and time for these 7 subcortical structures.

Results

Significant interactions between time and group were shown in the hippocampus (β = −0.06, SE = 0.02, p < .001) and globus pallidus (β = 0.01, SE = 0.00, p = .016). Patients treated for low-grade tumors exhibited increasing hippocampal volume over time, while those treated for high-grade tumors experienced a decline. In the globus pallidus, volume decreased over time for the low-grade tumor group but remained stable for the high-grade tumor group. Patients with a history of hydrocephalus had smaller thalamic (β = −0.40, SE = 0.14, p = .004) and hippocampal (β = −0.31, SE = 0.12, p = .012) volumes. Younger age at diagnosis was associated with smaller putamen and thalamus, and male sex was associated with larger volumes of the putamen, nucleus accumbens and amygdala.

Conclusions

These findings suggest that children who received treatment for low- or high-grade tumors had different subcortical volumes over time, which may be related to treatment, complications, or demographic factors such as age and sex. These results suggest that while brain tumor treatment primarily aims to cure patients, it may inevitably affect neural development and may contribute to a range of cognitive, behavioral, and emotional long-term deficits. This emphasizes the need for long-term monitoring and prospective longitudinal studies into structural and functional brain changes after treatment for pediatric brain tumor.

Highlights

• •Hippocampus and globus pallidus development differs between low- and high-grade tumor patients.
• •Subcortical development may be impacted by neuro-oncological treatment, hydrocephalus, age, and sex.
• •Longitudinal neuroimaging reveals distinct subcortical trajectories within five years after diagnosis.

1. Introduction

Advances in neuro-oncological treatment have resulted in higher survival rates for pediatric brain tumor patients. Despite this, many treatment modalities and complications contribute to the acute and chronic neuropsychological impairments observed in survivors (Gheysen et al., 2018). More than 50 % of pediatric brain tumor survivors could develop neuropsychological impairments after treatment, such as memory and attention deficits that impact daily functioning (Gheysen et al., 2018; Wagner et al., 2020). To guide future treatment practices, understanding of the long-term sequalae of pediatric brain tumors is important.

Besides the tumor itself, neuro-oncological treatments (i.e. surgery, chemotherapy, and radiation therapy) are known to impact the developing brain, with varying degrees of impact depending on the treatment. Based on structural and functional neuroimaging studies, it appears that surgical resection as a sole procedure for posterior fossa tumors may cause damage to cerebellar pathways and their connection to the deep grey matter and supratentorial regions, affecting cognitive, associative, and behavioral functions (Beez et al., 2021; Heitzer et al., 2020; Moberget et al., 2015; Phillips et al., 2021; Oyefiade et al., 2021)Besides surgery, adjuvant chemotherapy has an additional effect on the neuroarchitecture of the brain; for example, chemotherapy is associated with lower grey- and white matter volume and lower fractional anisotropy in various areas of the brain in long-term survivors (Kesler et al., 2021). Furthermore, chemotherapy is associated with poorer neurocognitive functioning (Kesler et al., 2021; Kundu et al., 2017; Nelson et al., 2020), which is hypothesized to be related to the changes in underlying neuroarchitecture including white matter. Finally, the addition of cranial radiation therapy is particularly neurotoxic, especially in younger children and with larger doses or volumes, as in cranio-spinal irradiation (CSI). This heightened toxicity in young children is primarily due to the vulnerability of their developing brains, which are more susceptible to radiation-induced damage. Radiation can disrupt neurogenesis, myelination, and vascular development, leading to significant long-term deficits in cognitive and neurobehavioral functions (Gheysen et al., 2018; Grill et al., 1999; Oyefiade et al., 2021; Palmer et al., 2020; Tsang et al., 2020; Yahya and Manan, 2020).

One method for examining sequelae is to evaluate structural or functional changes in the brain after treatment. Neuroimaging studies on the effect of treatment in posterior fossa tumors have focused mostly on structural findings, including white matter structure throughout the brain as well as grey matter volume such as the hippocampus (Ailion et al., 2017; Kesler et al., 2021; Nagel et al., 2004; Riggs et al., 2014; Zureick et al., 2018). These studies found that higher treatment intensity, including cranial radiation, is related to changes in grey- and white matter structure, smaller hippocampal volume, and cognitive decline. Consequently, these findings have guided research into hippocampal sparing in radiation treatment (Nagel et al., 2004; Tsai et al., 2015). Moreover, some evidence suggests that adult survivors have preserved memory and improved quality of life when the hippocampus is avoided during radiation (Gondi et al., 2010, 2014). However, the influence of treatment on other subcortical structures, i.e., amygdala, thalamus, and basal ganglia have not yet been investigated. These subcortical structures are known to play an important role in neuropsychological development, including cognitive, motor, and emotional functioning (Dima et al., 2022; Wierenga et al., 2014). Therefore, it is important to study all subcortical structures, in addition to the hippocampus, to help understand the impact of treatment on brain development in pediatric brain tumor patients. Furthermore, the impact of treatment on brain development is often described for long-term survivors using cross-sectional designs. However, the first years after diagnosis might be a crucial time window to study the early changes in the brain in relation to oncological treatment in the pediatric posterior fossa brain tumor population.

The aim of this retrospective longitudinal study was to investigate subcortical brain volumes over time in the first five years after diagnosis in patients with a tumor in the posterior fossa region. We investigated the differences between patient groups, based on tumor grade and therefore the intensity of the treatment that they received. We hypothesized that patients with high-grade tumors (treated with surgery, chemotherapy, and radiation therapy) would have greater changes over time when compared to the patients with a low-grade tumor (treated with surgery with or without chemotherapy). We also examined demographic factors such as age at diagnosis and sex at birth and hydrocephalus as a clinical factor. Age and sex are known demographic factors in the literature that may be related to the development of subcortical volumes (Wierenga et al., 2014). Also, hydrocephalus is commonly experienced in children with posterior fossa tumors and may also impact brain development (Uh et al., 2020). These results will help to define the potential changes that occur in subcortical structures following within five years after treatment for pediatric brain tumors and thereby may help with treatment planning in the future.

2. Material and methods

2.1. Participants

2.1.1. Patients

We included all posterior fossa tumor patients who were diagnosed at the Princess Máxima Center from June 2018 until October 2020 and who signed informed consent for retrospective studies (n = 102) (Fig. 1). Exclusion criteria were age at diagnosis <4 years, diffuse intrinsic pontine glioma/diffuse midline glioma diagnosis, passing away within 1 year after diagnosis, less than two MR images within the first year after diagnosis, and radiation therapy because of relapse. This resulted in 62 participants that were eligible for inclusion. However, scans of 5 participants were of unusable data/too low quality for further data processing, which resulted in 57 patients in the sample. For analyses, we grouped patients into those who were treated for a low-grade tumor (WHO grade 1 and 2) (n = 39) and those who were treated for a high-grade tumor (WHO grade 4) (n = 18). We included all T1w MR images up to five years after diagnosis, until September 2023. Neurologic assessment of all included patients was performed, including diagnosis of post-operative pediatric cerebellar mutism syndrome (ppCMS). In our study, ppCMS was defined as a transient syndrome occurring after posterior fossa surgery, characterized by delayed-onset mutism, emotional lability, and motor speech deficits. This definition aligns with established diagnostic criteria reported in the literature (Gudrunardottir et al., 2016). Obstructive hydrocephalus was noted based on the potential treatment given (i.e., external ventricular drain, shunt, ventriculostomy) and/or radiological diagnosis. All patients, except for one, were hospitalized in the Princess Máxima Center for their neurosurgical treatment.

Fig. 1.

– Flowchart of patient enrollment.

2.2. Procedure

2.2.1. MR image acquisition and processing

There were 558 T1w MR images from all included patients. 3D T1w MR images without gadolinium enhancement were made on a 1.5T Philips Ingenia (repetition time (TR) = 8.25ms; echo time (TE) = 3.78ms; flip angle = 8°; number of slices = 230; voxel size: 1.00x0.96 × 0.96 mm) and 3T Philips Ingenia Elition X scanner (TR = 8.22ms; TE = 3.76ms; flip angle = 8°; number of slices = 230; voxel size: 1.00x0.48 × 0.48 mm) (Philips Medical System, Best, The Netherlands). T1w MR images were obtained for diagnosis, treatment planning, and standard monitoring.

2.2.2. MR image processing pipeline

The images were processed with the FMRIB Software Library (FSL 6.0) (Jenkinson et al., 2002; Jenkinson and Smith, 2001) and Statistical Parametric Mapping (SPM12, version 7771) (Gaser et al., 2024; Penny et al., 2011), using the Computational Anatomy Toolbox (CAT12, version 12.8 (1931)) (Gaser et al., 2024). First, voxel sizes from both the 1.5T and 3T scans were rescaled to 1.00x1.00 × 1.00 mm³ isotropic using FSL FLIRT. Volumes were estimated on the individual participant level with the segmentation pipeline using the default settings (Gaser et al., 2024). The SPM CAT12 pipeline consists of several steps of preprocessing and segmentation, modulation and smoothing and quality control. Images were first spatially normalized to MNI space, followed by segmentation into grey matter, white matter, and cerebrospinal fluid compartments. CAT12 then uses the AAL3 atlas for the identification of subcortical regions, including the hippocampus (Rolls et al., 2020; Tzourio-Mazoyer et al., 2002). The segmented regions were adapted to account for differences in individual brain sizes and smoothed to enhance the signal-to-noise ratio. To ensure the reliability of the results, all processed images were subjected to quality control measures, including checks for scanner artifacts, misregistration, or segmentation errors. Moreover, all segmentations with an image quality rating (IQR) score above 2 were visually inspected to identify any inaccuracies. This IQR is a weighted average score of three image quality measures, and it is calculated by the CAT12 toolbox. Five patients were excluded because of processing pipeline failure (as described above) (Rolls et al., 2020). There were seven subcortical structures examined in this study: hippocampal formation, amygdala, thalamus, globus pallidus, caudate, putamen, and nucleus accumbens (see Fig. 2 (Larivière et al., 2021)). For all structures, the volumes of both hemispheres were highly correlated, therefore mean values across hemispheres were used in the analysis. To further strengthen our quality control, we manually reviewed the segmentations of all subcortical structures in collaboration with a board-certified neuroradiologist. This manual review was performed in a randomly selected 5 % of the sample (n = 28 scans). Segmentations were rated as optimal, mild, moderate, or severe.

Fig. 2.

– Segmented subcortical regions

Note: There were seven subcortical structures that were examined in this project: hippocampal formation, amygdala, caudate, nucleus accumbens, globus pallidus, putamen, and thalamus. In rows A and B, examples of real data with volume masks are superimposed over the participant's native-space image. Row A is from a patient without hydrocephalus, Row C is a patient with hydrocephalus. Row C: View is from medial to lateral.

2.3. Statistical analysis

Statistical tests were performed in R software environment (RCoreTeam, 2024), and all assumptions for statistical models were checked where applicable. Participant characteristics were descriptively reported. Independent sample t-tests were employed to examine differences between low- and high-grade tumor groups in scan interval between first and last MRI scan, age at diagnosis, and number of T1w MR images. Chi-square tests were used to test differences in sex, hydrocephalus diagnosis, and ppCMS diagnosis.

Linear mixed models were used to account for the longitudinal nature of the data (lme4 package 1.1–35.1 (Bates et al., 2015)). We estimated a model for each subcortical structure and included an interaction between time and group, corrected for age at diagnosis (centered around the grand mean (M = 9.23)), sex assigned at birth (female or male), history of hydrocephalus at any timepoint (yes/no), and MRI field strength (1.5T or 3T). The signal-to-noise ratio differences between field strengths can potentially influence the measured volumes and therefore this was included as an additional factor in the models. Other factors, including ppCMS, were not included in modeling because of statistical limitations due to the small sample size of low-grade tumor patients with ppCMS. Participant ID was incorporated in the models a random factor with random intercepts and random slopes. To explore whether variation in craniospinal irradiation (CSI) dose might influence subcortical volume trajectories, we performed an exploratory subgroup analysis comparing patients who received 23.4 Gy CSI versus ≥36 Gy CSI.

3. Results

3.1. Sample characteristics

Demographics and medical characteristics are shown in Table 1. Patients were between 4 and 18 years at diagnosis. Low-grade tumors consisted of pilocytic astrocytoma, and the ‘other’ category were low grade glioma not otherwise specified, pleomorphic xanthoastrocytoma (WHO grade II), papillary glioneuronal tumor (WHO grade I), teratoma, and angiocentric glioma (WHO grade I). The majority (95 %) received surgery only and one patient received adjuvant chemotherapy. High-grade tumors consisted of medulloblastoma and treatment included surgery, chemotherapy, and cranio-spinal proton beam radiation, except for one patient who received cranio-spinal photon radiation. T-tests between groups showed that there were no differences between low- and high-grade tumors in the interval between first and last MRI scan, age at diagnosis, and number of surgeries (p > .05), although there was a significant difference for the number of T1w MR images (t = −6.90, df = 55, p < .001). Chi-square tests showed that there was a significant difference between patient groups for ppCMS incidence (χ² = 5.03, df = 1, p = .025), but not for hydrocephalus (χ² = 2.71, df = 1, p = .099).

Table 1.

– Demographics and medical characteristics for patients.

Variable	Low-grade tumor (n = 39)	High-grade tumor (n = 18)	p-value
Sex at birth (Female)	20 (51 %)	5 (28 %)	0.09645b
Age at diagnosis (years)	9.77 (3.65)	9.39 (4.95)	0.7729a
Range	4–16	4–18
Tumor type (%)
Medulloblastoma	0 (0 %)	18 (100 %)
Pilocytic Astrocytoma	34 (87 %)	0 (0 %)
Other	5 (13 %)	0 (0 %)
Number of surgeries Mean (SD)	1.44 (0.88)	1.78 (0.81)	0.1582a
Radiation
Radiation	0 %	18 (100 %)
Prescribed dose to tumor bed
54 Gy	N/A	15 (83 %)
55.8 Gy	N/A	3 (17 %)
CSI dose
23.4 Gy	N/A	11 (61 %)
36 Gy	N/A	5 (28 %)
39.6 Gy	N/A	2(11 %)
Anesthesia during RT	N/A	10 (56 %)
Chemotherapy (%)	1 (2.56 %)	18 (100 %)
Hydrocephalus (%)	24 (62 %)	15 (83 %)	0.0999b
ppCMS (%)	5 (13 %)	7 (39 %)	<0.025b
MRI sessions	7.85 (2.87)	14.00 (3.65)	<0.001a
Scan interval between first and last MRI scan in years	3.35 (0.77)	3.75 (0.99)	0.1458a

CSI = cranio-spinal irradiation, ppCMS = post-operative pediatric cerebellar mutism syndrome.

¹ n (%).

^at-test.

^bChi-square test.

3.2. Difference in subcortical volumes between patients with low-grade tumors and patients with high-grade tumors

We investigated whether there was a significant difference in the volumetric change over time of seven subcortical volumes between patients with low-grade tumors and patients with high-grade tumors. Estimated average volumes of the structures at the time of diagnosis are presented in Table 1 as the intercept. Linear mixed model analysis showed interaction effects between time and patient group in two subcortical regions, including the hippocampus and globus pallidus (Table 2, Fig. 3, Appendix Table 1). Specifically, hippocampal volume showed a significant increase over time for patients who were treated for a low-grade tumor (β = 0.03, SE = 0.01, p = .001), while it decreased over time for patients who were treated for a high-grade tumor (β = −0.06, SE = 0.02, p < .001). This indicates a shift from a positive to a negative slope for the patients with a high-grade tumor (initial slope of 0.03 becoming −0.03). Patients diagnosed with hydrocephalus showed smaller average hippocampal volume compared to those without this diagnosis (β = −0.31, SE = 0.12, p = .012). Female (vs. male) sex (β = −0.27, SE = 0.11, p = .018) was associated with smaller hippocampal volume and 3T (vs. 1.5) MRI field strength (β = 0.17, SE = 0.02, p < .001) and was associated with larger hippocampal volume.

Table 2.

Results linear mixed effect models of patients treated for low-vs. high-grade posterior fossa tumors.

	Hippocampus		Globus pallidus		Caudate		Putamen		Nucleus accumbens		Amygdala		Thalamus
Predictors	Estimate	CI	Estimate	CI	Estimate	CI	Estimate	CI	Estimate	CI	Estimate	CI	Estimate	CI
(Intercept)	4.54∗∗∗	4.29–4.79	0.56∗∗∗	0.51–0.60	4.02∗∗∗	3.73–4.30	5.47∗∗∗	5.17–5.77	0.95∗∗∗	0.88–1.02	1.46∗∗∗	1.38–1.54	4.65∗∗∗	4.36–4.94
Time	0.03∗∗	0.01–0.05	−0.01∗∗∗	−0.02–−0.01	−0.02	−0.05–0.00	0	−0.02–0.03	0.01	−0.00–0.02	0.01	−0.00–0.02	0	−0.03–0.04
Group [High grade]	−0.09	−0.33–0.16	−0.05∗	−0.11–−0.00	−0.2	−0.49–0.09	−0.04	−0.35–0.27	−0.02	−0.09–0.05	−0.01	−0.10–0.08	−0.12	−0.41–0.17
Age at Diagnosis	0.01	−0.01–0.04	−0.01∗∗∗	−0.02–−0.01	−0.02	−0.05–0.01	−0.03∗	−0.07–−0.00	−0.01	−0.01–0.00	0	−0.00–0.01	−0.05∗∗	−0.08–−0.02
MRI Strength [3T]	0.17∗∗∗	0.13–0.21	−0.04∗∗∗	−0.06–−0.03	0.03	−0.01–0.08	−0.1∗∗∗	−0.14–−0.06	0	−0.01–0.01	0	−0.02–0.01	0.32∗∗∗	0.26–0.38
Sex [Female]	−0.27∗∗	−0.49–−0.05	−0.04∗	−0.08–−0.00	−0.11	−0.36–0.15	−0.52∗∗∗	−0.78–−0.25	−0.06∗	−0.12–−0.00	−0.15∗∗∗	−0.21–−0.08	−0.2	−0.45–0.06
Hydrocephalus [Hydrocephalus]	−0.31∗∗	−0.55–−0.07	−0.03	−0.07–0.01	−0.22	−0.50–0.05	−0.27	−0.56–0.02	−0.05	−0.12–0.01	−0.05	−0.12–0.02	−0.4∗∗	−0.68–−0.13
Time x Group [High grade]	−0.06∗∗∗	−0.09–−0.03	0.01∗∗	0.00–0.02	0	−0.04–0.04	−0.01	−0.05–0.03	−0.01	−0.03–0.00	−0.01	−0.03–0.00	−0.02	−0.07–0.04
Marginal R2	0.22		0.345		0.096		0.239		0.123		0.197		0.223

Note: ∗p ≤ .05, ∗∗p ≤ .01, p ≤ .001.

Fig. 3.

– Results linear mixed effect models of patients treated for low-vs. high-grade posterior fossa tumors

Note: estimated linear effect models for low-grade tumors (n = 39) and high-grade tumors (n = 18); each dot represents a different MR image and the shaded area represents the confidence interval for the estimated slopes. An asterisk (∗) indicates a significant group by time interaction from the linear mixed models (p < .05).

Volume of the globus pallidus decreased over time for patients treated for a low-grade tumor (β = −0.01, SE = 0.00, p < .001) and remained stable over time for patients treated for a high-grade tumor (β = 0.01, SE = 0.00, p = .016). This results in a shift from a negative to a neutral slope (initial slope of −0.01 becoming 0.00). At baseline, patients treated for a high-grade tumor had a smaller volume of the globus pallidus compared to patients treated for a low-grade tumor (β = −0.05, SE = 0.03, p = .034). Moreover, female sex (vs. male; β = −0.04, SE = 0.02, p = .047) was associated with smaller volume of the globus pallidus. In contrast, 3T MRI field strength (vs. 1.5; β = −0.04, SE = 0.01, p < .001) and younger age at diagnosis (β = −0.01, SE = 0.00, p < .001) were related to smaller volume of the globus pallidus.

The other subcortical structures did not show significant group by time interactions. Younger age at diagnosis was associated with smaller volume of the putamen (β = −0.03, SE = 0.02, p = .037) and thalamus (β = −0.05, SE = 0.02, p = .001). Female sex (vs. male) was associated with smaller volumes of the putamen (β = −0.52, SE = 0.13, p < .001), nucleus accumbens (β = −0.06, SE = 0.03, p = .036) and amygdala (β = −0.15, SE = 0.03, p < .001). Patients diagnosed with hydrocephalus showed smaller average volume of the thalamus compared to those without this diagnosis (β = −0.4, SE = 0.14, p = .004). Finally, 3T MRI field strength (vs. 1.5T) was associated with larger volume of the thalamus (β = 0.32, SE = 0.03, p < .001) and smaller volumes of the putamen (β = −0.1, SE = 0.02, p < .001).

3.3. Difference in subcortical volumes between patients with 23.4 Gy CSI and ≥36 Gy CSI

With the high-grade tumor group, there were 11 patients who received 23.4 Gy and 7 patients who received ≥36 Gy (Table 1). Despite small sample sizes in each radiotherapy group, we performed an exploratory analysis and examined the differences between those who received 23.4 Gy and 36 Gy CSI. There were no consistent associations between CSI dose and subcortical volumes. Full model results are reported in Appendix Table 2.

4. Discussion

This study investigated the trajectory of subcortical volumes in pediatric posterior fossa tumor patients in the first five years after diagnosis. We assessed the differences between patient groups based on low- and high-grade tumor classification, and thereby the intensity of treatment that they received. We also examined the demographic and clinical variables that may be associated with subcortical volume in these patients.

4.1. Subcortical volume differences between patients with low- and high-grade brain tumors

Our main hypothesis was partly supported. The models showed that the volumes of the hippocampus and globus pallidus had different changes over time between patients treated for low- and high-grade tumors. This effect was not found for the other subcortical structures. Patients with high-grade tumors had treatment that included surgery, chemotherapy, and CSI therapy, per standard protocol for medulloblastoma patients. In contrast, all but one low-grade tumor patients received surgery only. Our results therefore suggest a potential effect of tumor type, chemotherapy and/or cranial radiation therapy on the trajectory of hippocampal volumetric growth, consistent with previous studies in adults (Nagtegaal et al., 2021) and children (Nagel et al., 2004). It has also been shown that higher radiation dose to the hippocampus is related to smaller volumes and poorer memory functioning (Acharya et al., 2019; Gondi et al., 2013, 2014; Tsai et al., 2015). Notably, the susceptibility of the hippocampus to radiation-induced injury has influenced treatment protocols to optimize sparing of these regions (Tsai et al., 2015). Therefore, we replicate earlier findings of susceptibility of the hippocampus to oncological treatment.

Our findings imply that globus pallidus volumes have different changes over time when comparing low- and high-grade tumors. Our results suggest that the volume of globus pallidus was significantly smaller for patients with high-grade tumors (vs. low-grade tumors) at the first time point, which then deviated over time. For low-grade patients, the volume of the globus pallidus decreased over time, whereas this was stable in the high-grade patients. This indicates that there is already a difference between the volumes of the globus pallidus at the time of diagnosis and the maturational trajectories differed between patient groups. Therefore, the globus pallidus may be impacted by tumor type and grade from an early time point, but also treatment and complications over time. The clinical implications of this finding are not yet known, but the globus pallidus is involved in motor control and regulating voluntary movement by modulating the activity of the thalamus (Javed and Cascella, 2020). The globus pallidus is also connected to frontal and limbic systems through its efferent connections and therefore may be involved with cognitive and emotional functioning (Javed and Cascella, 2020). Future research (e.g., with high-resolution imaging and neurocognitive testing) will be needed to further examine these relationships and clinical implications.

4.2. Risk factors for subcortical volume differences and its clinical implications

Differences in brain tumor treatment groups may be related to differences in treatment intensity, as noted above. In addition, previous research showed that greater risk for cognitive impairment is related to higher intensity of therapies (Aarsen et al., 2004; Hanzlik et al., 2015; Hardy et al., 2008). Our results were shown in a cohort of medulloblastoma patients who (mostly) received proton cranial-spinal radiation compared to a cohort of pilocytic astrocytoma and other low-grade tumors treated with surgery and/or chemotherapy. Previous retrospective studies on proton cranial radiation have shown better neurocognitive outcomes for patients treated with proton radiation therapy compared to patients who were treated with photon therapy (Lassaletta et al., 2023). This included better neurocognitive outcomes for domains such as intelligence, verbal comprehension, perceptual reasoning indices, visual motor integration, and verbal memory; the findings for other domains including attention or processing speed were inconsistent. The current study shows that radiation may be associated with changes in subcortical structures over time, although further investigation is needed to understand the clinical implications of these changes.

Hydrocephalus was included as an additional predictor, since this is previously shown to be a risk factor for brain development and cognitive problems (Duffner, 2010; Hardy et al., 2008). The thalamus and hippocampus had smaller volumes in association with hydrocephalus, which may be partially explained by its location next to the ventricles. The increased pressure and volume from the enlarged ventricles due to hydrocephalus can compress nearby brain structures, including the thalamus and hippocampus. This can lead to atrophy (shrinkage) and reduced volume of these regions. Recognizing the association between hydrocephalus and smaller thalamic and hippocampal volumes underscores the importance of early diagnosis and intervention. For example, shunting procedures or other treatments are already conducted to manage cerebrospinal fluid (CSF) levels to reduce pressure on brain structures. Also, previous research has shown the association between history of hydrocephalus and poorer cognitive performance in posterior fossa patients (Aarsen et al., 2004; Hardy et al., 2008; Moxon-Emre et al., 2014). Our findings suggest that treated hydrocephalus may also impact the volumes of subcortical structures which suggests that follow-up or additional neurocognitive assessment with subsequent interventions may be needed for these patients. It is important to note that the course of hydrocephalus usually differs between low- and high-grade tumors. For example, in low grade tumors, hydrocephalus may exist for a longer period due to the gradual growth of the lesion; whereas, in high grade tumors, there is rapid tumor growth and thus this can lead to quick onset of symptoms due to hydrocephalus. (Hardy et al., 2008). Suggestions for future research in this regard are discussed below.

Younger age at diagnosis was also significant predictor of smaller volumes. In healthy children, it is known that subcortical structures continue to show maturational changes from childhood into adolescence. The subcortical structures first show increasing volume, and they reach their largest volume at different ages. According to Wierenga et al. (2014), the age at peak volume for both sexes is around 17.3 years old for the hippocampus and 17.2 years for the globus pallidus. By the end of the adolescence period, the volume of all subcortical structures is declining, as reported by multiple studies (Frangou et al., 2022; Wierenga et al., 2014). On average, our patient cohort was 9 years old at diagnosis, and thus, smaller volumes at younger ages may be consistent with normal healthy development. However, age at diagnosis is a commonly known risk factor for neurocognitive decline in brain tumor survivors (Wauters et al., 2021). The early years of a child's life are crucial for brain development and if a posterior fossa brain tumor is diagnosed at a younger age, the necessary treatments and the tumor itself may disrupt the normal developmental processes of the brain. Our results suggest that brain tumor patients may have a different maturational pattern in subcortical volumes, and part of this finding may be related to younger age of the patient. However, the observed decline in globus pallidus volume over time in the non-irradiated low-grade tumor group is not consistent with normal subcortical development, which typically shows increasing volume until mid-adolescence (Wierenga et al., 2014). This suggests that, even in the absence of radiotherapy, other factors such as the tumor itself, surgical intervention, or indirect treatment effects may have interfered with the expected maturational trajectory of the globus pallidus. These findings point to a potentially altered developmental pattern in this subcortical structure among children with posterior fossa tumors, which warrants further investigation. As discussed below in the strengths and limitations section, interpretation of globus pallidus volume should be made cautiously due to its small size and the limitations of standard MRI resolution in reliably distinguishing its substructures. Finally, the result that male sex was associated with larger volume in five out of seven structures is consistent with previous studies (Herting et al., 2018; Wierenga et al., 2014).

4.3. Strengths and limitations

The study has several strengths, providing new insights into brain development in pediatric brain tumor patients. Given the rarity of pediatric brain tumors, the sample size and number of T1w MR images for patients is large. Also, since children with a brain tumor are scanned regularly, this provides a unique dataset of MR images with on average more than 10 scans per patient over time. Therefore, this study is novel in charting the developmental trajectory of subcortical structures in pediatric patients with posterior fossa tumors.

The timing of the study, from diagnosis to a period extending up to five years post-diagnosis, investigates a crucial phase in patient recovery and development. It offers a detailed examination of the potential alterations taking place in a time frame that is essential for evaluating therapeutic interventions and determining long-term prognoses. Unlike most research focused on white matter and the hippocampus, this analysis of all subcortical structures offers a more comprehensive view on the developmental trajectory of the brain and its sensitivity to cancer and subsequent treatment in these young patients, thereby contributing to the understanding of treatment impacts on brain development.

Pediatric post-operative cerebellar mutism syndrome (ppCMS) is a condition arising from disruptions to posterior fossa structures and has been associated with involvement of the dentato-rubro-thalamo-cortical tract (DRTCT), as evidenced in regions such as the inferior olivary nuclei, red nuclei, and periaqueductal grey area (e.g. (McAfee et al., 2022; Obdeijn et al., 2024; Patay et al., 2014). These findings suggest that posterior fossa tumors and related pathophysiological processes, including ppCMS, may produce downstream effects on structures outside of the cerebellum. For instance, the amygdalae, known for their role in emotional regulation and stress responses, have demonstrated hyperconnectivity in patients with ppCMS (McAfee et al., 2023), though structural changes remain unexplored. There was a high prevalence of ppCMS in the high-grade tumor group (39 % vs. 13 % in low-grade tumor group). There could be additional potential factors playing a role in the high percentage of patients having ppCMS. For example, this could be tumor location. Since it is a small sample, there is also a possibility of sampling error to that may lead to this high percentage. It was statistically not feasible to disentangle the potential effects of ppCMS from tumor group. While we did not include ppCMS as a predictor in our modeling due to statistical constraints, we recognize the importance of its implications for future research. For example, future studies could explore the subcortical mechanisms of ppCMS, including its interaction with radiation therapy, which could provide valuable insights into the broader neurodevelopmental impact of pediatric posterior fossa tumors.

The globus pallidus had the smallest mean volume of the analyzed subcortical structures, therefore the volume changes of the globus pallidus as described in our series with distinction between high-grade and low-grade patients should be cautiously addressed. It was not possible to examine the internal and external segments of the globus pallidus in the current MRI processing software. Given the structural and functional differences between the internal and external segments of the globus pallidus, it is suggested for future research to examine these segments using high-resolution imaging. This may provide further information on the structure-function relationships within the globus pallidus, but also other relatively small regions of the brain. Additionally, MRI field strength is a significant predictor in most models, which could be due to the differences in signal-to-noise ratio and partial volume effects. Not only the vulnerability of the globus pallidus should be investigated further, but also the effect of MRI field strength. Modern harmonization tools could be advantageous in dealing with different MR scanners in hospital settings (Zhao et al., 2024). Our irradiated cohort were all children with a medulloblastoma. The biggest differences in dose distribution are based on the standard CSI (23.4 Gy) or high dose CSI (36 Gy or 39.6 Gy) craniospinal part of the treatment. However, the retrospective design of the study limited analysis of individual voxel-wise dose maps, which limits precise dose-volume correlations.

Furthermore, the low-grade and high-grade tumor groups in this study had a significantly different number of MR images and frequency of ppCMS diagnosis. Medulloblastoma patients made up the high-grade tumor group in this study, and due to the higher grade of this tumor and risk for relapse, they have more MRI scans than low-grade tumors. (Pettersson et al., 2022). This type of complexity and overlap between variables is a common factor in pediatric brain tumor research, and thus it is suggested to examine these interactions in a larger prospective study.

Another limitation of this study is the quantification of hydrocephalus status. Hydrocephalus was defined as yes or no when there was a radiological diagnosis or surgery (e.g. shunt placement). Potentially, the use of total CSF volume could be used as a proxy for hydrocephalus severity. While the total CSF volume measure of SPM CAT 12 provides a general measure of cerebrospinal fluid distribution, it does not specifically isolate ventricular CSF, which is most directly impacted by hydrocephalus. This approach may dilute the relationship between hydrocephalus severity and subcortical volume changes due to contributions from extraventricular CSF, such as in the sulci. Future studies would benefit from employing more refined measures of ventricular CSF volume, which could be achieved through the development and implementation of automated segmentation tools specifically designed for ventricular CSF quantification (e.g. (Quon et al., 2021). Such tools could provide more accurate and reliable measures of hydrocephalus severity and allow for dynamic tracking of ventricular changes over time, particularly in relation to treatment interventions such as shunting or third ventriculostomy. These advancements would enable a more precise evaluation of the relationship between hydrocephalus and subcortical brain structure volumes, offering deeper insights into the impact of this condition on neurodevelopment and aiding in the optimization of treatment strategies.

Also, the inclusion of patients with ependymoma to the sample would have been advantageous in examining the effect of focal radiation treatment. The patients with ependymoma, however, were excluded because of their young age in this sample. Healthy control data were also not available in this study, and thus comparisons to healthy development was not possible. As our aim was to compare between low-grade and high-grade tumor groups, a healthy control group was not included in this current study. In future research, we suggest including a comparison with healthy peers to examine the amount of deviation from typical development.

To strengthen the methodological rigor of our study, we implemented additional quality control and exploratory analyses. First, we conducted a structured visual inspection of subcortical segmentations in a randomly selected 5 % of scans, rated in collaboration with a board-certified neuroradiologist. This review confirmed generally good anatomical accuracy, with consistent deviations observed in the caudate and hippocampus in the context of hydrocephalus, likely due to displacement of the ventricular system. These effects were consistent across time and were accounted for in our statistical models by including hydrocephalus, MRI field strength, age, and sex as covariates. Additionally, appendix Fig. 1 shows hippocampal volume (cm³) over time since first scan, stratified by the different variables. Together, these additional steps enhance the robustness of our findings and demonstrate the longitudinal stability and group-level interpretability of the derived subcortical volumes.

4.4. Conclusions and future directions

In conclusion, the trajectory of subcortical volumes of posterior fossa patients changes over time, with the hippocampus and globus pallidus showing differences over time according to tumor grade and treatment intensity. Hydrocephalus was also associated with smaller hippocampal and thalamic volumes. These results suggest that although brain tumor treatment aims to cure the patient, it may inevitably affect neural development and may contribute to a range of cognitive, behavioral, and emotional late and long-term deficits (Oyefiade et al., 2021). Future studies should investigate the relationship with neurocognitive outcomes, especially proton radiation therapy and its relation to the volumetric development of subcortical structures, considering different RT doses. Another direction of future research is to examine the sensitivity of the globus pallidus and the other subcortical structures following treatment in a prospective manner using high-resolution imaging.

Ethical approval

Patient data were acquired following approval by the Biobank and Data Access Committee of the Princess Máxima Center for pediatric oncology (PMCLAB2019.084). Written informed consent was obtained from parents for children under the age of 12 years. For children aged 12–16 years, parental consent was obtained in addition to individual consent. For participants aged 16 years and over, written informed consent was only obtained from the individual.

CRediT authorship contribution statement

Anne E.M. Leenders: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Bruno M.de Brito Robalo: Writing – review & editing, Investigation, Data curation. John H. Maduro: Writing – review & editing, Conceptualization. Marta Fiocco: Writing – review & editing, Methodology, Formal analysis. Maarten Lequin: Writing – review & editing, Supervision, Conceptualization. Eelco Hoving: Writing – review & editing, Supervision, Conceptualization. Marita Partanen: Writing – review & editing, Writing – original draft, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Formal analysis, Conceptualization.

Funding sources

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of competing interest

The authors have no financial/personal interest or belief that could impact their objectivity.

Appendix A. Supplementary data

The following are the Supplementary data to this article.

Data availability

The data that has been used is confidential.