Volumetric Brain Assessment of Long-Term Head and Neck Cancer Survivors

Abstract

Background:

Radiation therapy (RT) for locally advanced head and neck cancer (HNC) often exposes subcortical brain structures to radiation. We performed this study to assess region-specific brain volumetrics in a population of long term HNC survivors.

Methods and Materials:

Forty HNC survivors were enrolled at a mean of 6.4 years from completion of RT. Patients underwent a research MRI protocol that included a 3D T1- weighted whole-brain scan on a 3 Tesla MRI scanner. Voxel based morphometry was performed using the Computational Anatomy Toolbox with the Neuromorphometrics atlas. Healthy controls from the Human Connectome Project were used as a comparison cohort. Study participants also completed a comprehensive neurocognitive assessment.

Results:

The final study cohort consisted of 38 participants after excluding 2 participants due to image quality. HNC survivors displayed widespread reduction in gray matter (GM) brain region volumes that included bilateral medial frontal cortex, temporal lobe, hippocampus, supplemental motor area, and cerebellum. Greater radiation exposure was associated with reduced GM volume in the left ventral diencephalon (r=−0.512, p=0.003). Associations between cognition and regional GM volumes were identified for motor coordination and bilateral cerebellum (left, r=0.444, p=0.009; right, r=0.372, p=0.030), confrontation naming and left amygdala (r=0.382, p=0.026), verbal memory and bilateral thalamus (left, r=0.435, p=0.010; right, r=0.424, p=0.012), right amygdala (r=0.339, p=0.050), and right putamen (r=0.364, p=0.034).

Conclusions:

Reductions in GM were observed within this cohort of primarily non-nasopharyngeal HNC survivors as compared to a control sample. GM volumes were associated with performance in multiple cognitive domains. Results of this exploratory study support the need for investigation of anatomic brain changes as an important translational corollary to cognitive problems among HNC survivors.

INTRODUCTION

Radiation therapy (RT) is an important component of therapy for most patients with locally or regionally advanced head and neck cancer (HNC).¹ Benefits of RT for HNC include the opportunity for organ preservation when used as primary treatment and improved local control when administered after surgery. The benefit of RT is unfortunately tempered by a range of late adverse effects, and neurocognitive problems experienced by HNC survivors are increasingly recognized. The impact of RT for HNC on neurocognitive functioning was first appreciated among patients treated for nasopharyngeal carcinoma, where radiation exposure to the temporal lobes is common.^2,3 More recently, neurocognitive decline has also been observed in patients with non-nasopharyngeal HNC treated with chemoradiotherapy.⁴ Though a growing body of work now describes the impact of neurocognitive problems in HNC survivors, the underlying pathophysiology and contributing factors have not been fully explored.

Prior studies in adult brain tumor survivors have consistently observed reduced cortical gray matter (GM) after RT exposure, with volumetric changes observed even at lower dose levels.^5–9 The changes in cortical gray matter volume have also been associated with neurocognitive deficits. By comparison, the brain radiation exposure during RT for HNC occurs primarily within subcortical structures, though the temporal lobes may receive exposure when the tumor is located more superiorly, as with nasopharyngeal carcinoma.¹⁰ HNC patients also have normal brain anatomy uninvolved by tumor. Whether non-nasopharyngeal HNC survivors exhibit volumetric brain changes similar to those observed in other populations has not yet been explored. The primary purpose of this study was to describe brain volumes using a voxel-based morphometry (VBM) approach in a population of long-term HNC survivors and to explore the relationship between brain volumes and radiation dose. We also sought to compare brain volumes in this HNC population with a reference population and assess for a relationship between brain volumes and neurocognitive functioning.

METHODS AND MATERIALS

Inclusion Criteria and Regulatory Approval:

This cross-sectional study enrolled survivors of cancers of the head and neck region who received radiation therapy as a component of their treatment. Participants were recruited from a larger cohort who took part in a cognitive outcomes study with this MRI sub-study as an optional component. Eligible participants were required to have received a dose of at least 50 Gray (Gy) and had survived at least 2 years from the time of RT completion. The study population was limited to patients >2 years from RT in order to be most generalizable to long-term survivors. HNC was defined as any malignant tumor of the upper aerodigestive track, cervical lymph nodes, or base of skull; locally advanced skin cancers of the head and neck region were also allowed if the radiation targets extended to include the neck lymphatics or base of skull. Patients were excluded from the study if they had a neurocognitive impairment from a known neurologic condition (such as prior stroke), if magnetic resonance imaging (MRI) was deemed unsafe, or if implanted material was expected to cause severe MRI artifact. Potentially eligible patients were identified by screening the electronic medical record. Patients were approached for study participation at routine clinic visits, or by telephone if they were no longer following at our facility. This MRI sub-study was closed after it reached its accrual goal of 40 participants. This study was reviewed and approved the institutional review board at the University of Alabama at Birmingham.

MRI acquisition:

Participating HNC survivors underwent a one-time research MRI protocol that included a 3D T1-weighted MPRAGE (magnetization prepared rapid acquisition gradient echo) whole-brain scan on a 3 Tesla MRI scanner (Siemens Prisma, Siemens, Erlangen, Germany) with the following parameters: repetition time (TR)=2400 MS, echo time (TE)=2.22 ms, flip angle (α) 9 degrees, field of view=256mm × 256 mm × 170mm, acquisition matrix=256 × 256, slice thickness=0.80 mm.

Voxel-based morphometry (VBM):

VBM analyses were conducted using Statistical Parametric Mapping version 12 (SPM12; The Welcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, London, UK) with the Computational Anatomy Toolbox (CAT12; Jena University Hospital, Departments of Psychiatry and Neurology) and MATLAB version 9.5 (The MathWorks, MA, USA). Raw images were converted from Digital Imaging and Communications in Medicine (DICOM) to Neuroimaging Informatics Technology Initiative (NIFTI) format. The NIFTI images were spatially normalized to Montreal Neurologic Institute (MNI) template space and then segmented into GM, white matter (WM), and cerebral spinal fluid (CSF) tissue classes according to the DARTEL approach, and modulated by the Jacobian determinants of the deformations in order to preserve volumes in native space.^11,12 An isotropic Gaussian kernel of 6 mm full width at half maximum (FWHM) was applied to the modulated tissue maps. Total intracranial volume (TIV) and the native space volumes of GM, WM, and CSF maps were estimated as well.¹³ TIV was included in all region of interest (ROI) analyses as a covariate, given TIV is highly correlated with ROIs. Age, sex, and time since radiation were also included as covariates for within-group analyses of the head and neck cancer survivors’ group. For ROI analyses, regional tissue volumes were estimated in different regions based on the probabilistic atlases (Neuromorphometrics Inc. Somerville, MA, USA). All analyses were corrected for multiple comparisons utilizing a Holm-Bonferroni corrected p<0.05 in addition to a 100-voxel extent threshold.

For between-group VBM analysis, healthy control data for 40 individuals (23 male, 17 female) was obtained from the Human Connectome Project (HCP) database (https://ida.loni.usc.edu/). HCP is the result of efforts of co-investigators from the University of Southern California, Martinos Center for Biomedical Imaging at Massachusetts General Hospital (MGH), Washington University, and the University of Minnesota, and provides preprocessed MPRAGE data that can be analyzed using the same VBM pipeline as described above. To assess morphometric differences in the brain of the head and neck cancer survivor’s cohort compared to control cohort, general linear model (GLM) analyses were conducted using t-tests within SPM12. Sex, scanner type, and total intracranial volume (TIV) were included as covariates of no interest to control for confounding linear effects.

Radiation Dosimetry:

Radiation dosimetry was calculated using Varian Eclipse software (Varian Medical Systems, Palo Alto, CA, USA). Since Eclipse software functions in DICOM space, the ROIs defined by the CAT12 toolbox could not be utilized for dosimetry calculation because CAT12 performs image deformation to the MNI space. We therefore utilized SynthSeg software as an alternative approach to segment brain ROIs in native space.¹⁴ SynthSeg utilizes an artificial intelligence model to segment bilateral cerebral and cerebellar GM, WM, thalamus, caudate, putamen, pallidum, hippocampus, amygdala, accumbens, ventral diencephalon, lateral ventricles as well as the brainstem, 3^rd ventricle, and 4^th ventricle. The MRI data set with SynthSeg ROIs were imported as DICOM objects into Eclipse software and coregistered to the treatment planning computed tomography (CT) scan using skull based rigid registration. The mean radiation dose to each ROI was calculated from each subject’s radiation treatment plan. Partial correlation analyses, controlling for age, sex, and time since radiation, were conducted for each ROI (FDR corrected for multiple comparisons) to determine association of radiation dosimetry with GM volume.

Neurocognitive Assessment:

Patients completed a neurocognitive assessment battery to assess general intelligence (Weschler Abbreviated Scale of Intelligence¹⁵), executive functioning (Flanker Inhibitory Control and Attention Test¹⁶, Dimensional Change Card Sort Task¹⁷), working memory (List Sorting Working Memory Test¹⁸), verbal memory (Weschler Memory Scale Logical Memory and Verbal Paired Associates subtests¹⁹), confrontation naming (Boston Naming Test²⁰), and motor coordination (Grooved Pegboard Test²¹). Neurocognitive assessments were completed by trained research staff under the supervision of a board-certified neuropsychologist. Raw scores were transformed to T-scores (mean = 50, SD = 10) based on normative sample groups published in each assessment’s manual. Partial correlation analyses, controlling for age, sex, and time since radiation, were conducted to determine association of neurocognitive performance with GM volume.

RESULTS

MRI acquisition was completed for a total of 40 HNC survivors. Two participants’ MPRAGE sequences were not suitable for VBM analysis due to quality of scans acquired and the final study cohort therefore consisted of 38 participants whose demographic, disease, and treatment information is presented as Table 1. The median time from RT to study MR acquisition was 4.2 years (interquartile range: 2.6 – 10.1 years).

Table 1.

Demographics and disease characteristics

		Participants (N=38)	Approached but did not participate (N=26)	Not approached (N=158)
Age (years)	Mean (±SD)	60.6 (±9.3)	66.5 (±8.2)	63.4 (±10.5)
Years from RT	Mean (±SD)	6.4 (±4.9)	7.2 (±6.6)	5.0 (±3.5)
Sex	Female	9 (23.7%)	7 (26.9%)	45 (28.5%)
	Male	29 (76.3%)	19 (73.1%)	113 (71.5%)
T-Stage	T0-T2	22 (57.9%)	18 (69.2%)	90 (57.0%)
	T3-T4	16 (42.1%)	8 (30.8%)	68 (43.0%)
	N0	6 (15.8%)	9 (34.6%)	41 (25.9%)
	N1	12 (31.6%)	7 (26.9%)	32 (20.3%)
N-Stage	N2	15 (39.5%)	8 (30.8%)	72 (45.6%)
	N3	4 (10.5%)	1 (3.8%)	6 (3.8%)
	N/A1	1 (2.6%)	0	0
	Missing	0	1 (3.8%)	7 (4.4%)
Primary site	Nasopharynx	4 (10.5%)	2 (7.7%)	8 (5.1%)
	Oropharynx	22 (57.9%)	12 (46.2%)	77 (48.7%)
	Oral cavity	2 (5.3%)	4 (15.4%)	21 (13.9%)
	Larynx	4 (10.5%)	1 (3.8%)	23 (14.6%)
	Hypopharynx	0	0	1 (0.6%)
	Paranasal sinuses	1 (2.6%)	0	2 (1.2%)
	Salivary gland	1 (2.6%)	4 (15.4%)	1 (0.6%)
	Unknown primary or other1	4 (10.5%)	3 (11.5%)	14 (8.9%)
Surgery	Yes	20 (52.6%)	14 (53.8%)	97 (70.3%)
	No	18 (47.4%)	12 (46.2%)	61 (29.7%)
	Cisplatin	23 (60.5%)	11 (42.3%)	70 (44.3%)
	Carboplatin	1 (2.6%)	3 (11.5%)	10 (6.3%)
Chemotherapy	Cetuximab	3 (7.9%)	3 (11.5%)	7 (4.4%)
	Other	0	1 (3.8%)	5 (3.2%)
	None	11 (28.9%)	8 (30.8%)	66 (41.8%)
Radiotherapy Technique 2,3	3D-CRT	2 (5.2%)	-	-
	IMRT	36 (94.8%)	-	-
Tumor directed dose	>66 Gy	26 (68.4%)	-	-
	60–66 Gy	11 (28.9%)	-	-
	<60 Gy	1 (2.6%)	-	-
	Unknown	0	-	-

¹One patient with MALT lymphoma of the head and neck who received >50 Gy.

²All treatments delivered in daily fraction sizes between 180 cGy and 212 cGy per day.

³Radiation treatment plan information not available for all non-participants.

The brain regions that demonstrated significant less volume (all p<0.001) in the head and neck cancer survivors when compared to the control comparison cohort are shown in Figure 1 and Table 2. Generally, the head and neck cancer survivors displayed widespread reduction in GM brain region volumes that included bilateral medial frontal cortex, temporal lobe, hippocampus, supplemental motor area, and cerebellum. WM, TIV, and CSF had no significant structural difference between the two groups. Additionally, there were no brain regions that demonstrated lower volume in the control cohort compared to patient cohort.

Figure 1.

Brain regions demonstrating reduced gray matter (p < 0.0001) in study cohort of head and neck cancer survivors compared to control cohort.

Table 2.

List of brain regions with significant GM reductions compared to control cohort.

	Left	Right
P<0.00000 1	Ventral diencephalon Gyrus rectus Subcallosal area Medial orbital gyrus Thalamus Proper Basal Forebrain Supramarginal gyrus Medial frontal cortex Cerebellum Exterior Anterior cingulate gyrus Anterior orbital gyrus	Ventral diencephalon Medial orbital gyrus Basal Forebrain Subcallosal area Gyrus rectus Thalamus Proper Medial frontal cortex Middle temporal gyrus Planum polare Superior frontal gyrus medial segment Anterior orbital gyrus
	Posterior orbital gyrus	Inferior temporal gyrus
P<0.0001	Middle temporal gyrus Precentral gyrus Lingual gyrus Superior frontal gyrus medial segment Planum polare Cuneus Precuneus Central operculum Superior temporal gyrus Posterior insula Planum temporale Anterior insula Middle cingulate gyrus Calcarine cortex Parietal operculum	Central operculum Calcarine cortex Middle cingulate gyrus Anterior cingulate gyrus Posterior orbital gyrus Precentral gyrus Temporal pole Cerebellum Exterior Parahippocampal gyrus Frontal pole

Table 3 provides a description of the mean radiation doses to the SynthSeg brain ROIs averaged across the cohort. Greater radiation exposure was associated with reduced GM volume in the left ventral diencephalon ROI, includes the hypothalamus, mammillary body, subthalamic nuclei, substantia nigra, red nucleus, lateral geniculate nucleus (LGN), and medial geniculate nucleus (MGN), r=−0.512, p=0.003 (FDR multiple comparisons correction of left sided brain regions, p=0.045). No other regions survived multiple comparison corrections. Step-wise regression assessed additional effects of Cisplatin exposure was not significant (p = 0.39), indicating radiation dose was the best predictor of GM volume for the left ventral diencephalon ROI.

Table 3.

Mean radiation doses to SynthSeg ROIs, averaged across the study cohort.

Region of Interest	Median (Range) Mean Radiation Dose in Gy
	Left	Right
Cerebral white matter	1.6 (0.3 – 12.9)	1.4 (0.3 – 12.8)
Cerebral cortex	2.1 (0.4 – 13.2)	1.9 (0.3 – 13.2)
Cerebellar white matter	16.2 (1.2 – 51.8)	14.7 (1.1 – 44.1)
Cerebellar cortex	20.1 (1.5 – 51.9)	17.4 (1.4 – 44.8)
Thalamus	1.7 (0.4 – 18.6)	1.6 (0.4 – 19.8)
Caudate	1.2 (0.3 – 27.4)	1.3 (0.3 – 24.8)
Putamen	1.6 (0.3 – 14.5)	1.6 (0.3 – 13.8)
Pallidum	1.7 (0.4 – 19.1)	1.8 (0.3 – 18.6)
Hippocampus	2.6 (0.5 – 27.9)	2.8 (0.5 – 23.6)
Amygdala	2.6 (0.5 – 38.4)	2.7 (0.5 – 33.1)
Accumbens area	1.8 (0.4 – 27.4)	1.8 (0.3 – 27.8)
Ventral diencephalon	2.4 (0.5 – 22.4)	2.4 (0.5 – 20.5)

The results of the neurocognitive battery are summarized in Figure 2. Reduced GM volume across several brain regions was associated with poorer neurocognitive functioning: motor coordination and bilateral cerebellum (left, r=0.444, p=0.009; right, r=0.372, p=0.030), confrontation naming and left amygdala (r=0.382, p=0.026), verbal memory and bilateral thalamus (left, r=0.435, p=0.010; right, r=0.424, p=0.012), right amygdala (r=0.339, p=0.050), and right putamen (r=0.364, p=0.034).

Figure 2.

Bot plot of neurocognitive battery T-scores. Scores falling below 40 (red line) were considered impaired.

DISCUSSION

Results of this study support the need for investigation of anatomic brain changes as an important translational corollary to cognitive problems that are becoming increasing recognized as common in HNC survivors. In this cross-sectional study we observed that long-term HNC survivors had multiple brain regions with reduced gray matter as compared to a control cohort sample but no difference in white matter. The anatomic distribution of most ROIs with reduced gray matter was consistent with the typical radiation distribution associated with RT for HNC, and dose was associated with gray matter volume in the ventral diencephalon.

VBM quantifies the relative amount of gray matter present within regions of interest throughout the brain.²² VBM has been extensively applied to a wide range of primary neurologic conditions such as Alzheimer disease, Parkinson disease, and multiple sclerosis where gray matter loss has been associated with adverse patient outcomes including cognitive decline, memory impairment, and accelerated aging.^23–25 Studies utilizing VBM to quantify gray matter in cancer survivors who received RT are comparatively scarce. Voon et al. recently performed a review of MRI correlate studies in HNC patients and found that prior studies focused exclusively on Asian (endemic) nasopharyngeal cancer survivors.³ Three studies assessed brain volume, two of which observed volume reduction within the temporal lobes.^26,27 Lv et al. paid special attention to hippocampal changes and observed that nasopharyngeal cancer survivors developed dose-dependent volume loss within hippocampal subfields.²⁸

Our study population consisted primarily of non-nasopharyngeal HNC survivors whose brain radiation exposure was typically limited to subcortical ROIs. We observed GM volume reductions primarily within subcortical ROIs which are at risk of exposure to moderate doses of radiation with modern RT techniques.¹⁰ Prior studies in the setting of cranial radiation have observed that subcortical GM loss after radiation exposure is dose-dependent and without a clear threshold below which no GM loss occurs.⁷ We also observed reduced GM within a limited number of cortical ROIs which are not exposed to significant radiation doses. Investigating the factors which contributed to GM changes outside of regions exposed to radiation is an important future direction, as many HNC patients are exposed to cisplatin which has been associated with both cognitive changes and GM loss within the prefrontal cortex.²⁹ Reduced cerebral blood flow has also been associated with cortical atrophy; the carotid arteries are almost always exposed to high radiation doses during HNC treatment.^30–33

The associations between regional GM volumes and neurocognitive functioning were consistent with known functional specialization within the brain. Specifically, the amygdala was found to be related to visual confrontational naming. The ventral temporal lobe is a key brain region in word retrieval, and neuroimaging studies in temporal lobe epilepsy and amyotrophic lateral sclerosis have shown that damage to connections between ventral temporal lobe and amygdala can negatively affect confrontational naming abilities.^34–36 Of particular interest, our results indicated that verbal memory performance was related to GM volumes in the thalamus, putamen, and amygdala. The anterior nuclei of the thalamus is specifically related to verbal memory processing, and damage to the mammillothalamic tract leads to significant verbal memory impairment.^37,38 The right putamen has also been implicated in verbal retrieval and memory, particularly in verbal fluency.³⁹ Collectively, these regions are consistent with our radiation exposure results as well, in which greater radiation exposure was associated with reduced GM volume in the left ventral diencephalon ROI. Lastly, it is not surprising that decreased volume of the cerebellum was significantly correlated with fine motor dexterity.⁴⁰

The association between cognition and GM volumes should also be interpreted in the context of the cognitive domains with lower T-scores. For instance, the relationship between cerebellar GM and motor coordination is particularly meaningful since cerebellar GM volumes were reduced compared to the control cohort and most participants’ Grooved Pegboard Test scores in the study were significantly impaired. Verbal memory was also impaired and correlated with GM volumes of subcortical ROIs that were, in turn, lower than controls. Though not conclusive of a causal relationship, these findings are suggestive that the GM changes and neurocognitive impairment are strongly related to one another. That these ROIs fall within areas of the brain that tends to have more radiation is noteworthy, but significantly more work is needed to characterize whether these changes are related to specific treatment exposures.

This study was exploratory in nature and has important limitations to recognize. The cross-sectional design was necessary for feasibility, but since the neuroimaging and neurocognitive assessments were only conducted at a single time point assessing for longitudinal changes was not possible. The sample size limited statistical power even though it was respectable as compared to other studies of HNC survivors and adequate size for neuroimaging analyses. The sample size also limited our ability to explore other factors, such as chemotherapy exposure and surgery, which could have directly impacted brain volumes or modified the effect of radiation. Participation bias is a potential concern that may limit external validity of this study’s findings. Since this study only enrolled survivors >2 years from RT completion, the results should not be generalized outside of this subset of patients with HNC. Regarding methodology, the control group from the HCP did not have HNC and age was not fully reported which is a potential confounder when assessing for difference in GM volumes. Because of these limitations, this study should primarily be viewed as hypothesis generating.

Despite its limitations, this study is among the first to assess GM changes in non-nasopharyngeal HNC survivors. Regional GM volume was correlated with neurocognitive performance, particularly among regions with reduced GM and cognitive domains with higher rates of impairment. When brain volumes in this cohort of HNC survivors were compared to healthy controls, reduction in GM predominantly occurred in regions of the brain where radiation exposure was expected. While this supports a possible relationship with radiation exposure the effects of chemotherapy exposure and surgical resection are will be important to consider moving forward. Given that long-term survival after treatment is increasingly common, the results of this study support a need for more comprehensive studies of brain changes in HNC survivors.

Highlights:

• In this study of 40 head and neck cancer survivors who received radiation at least 2 years prior and underwent a research brain MRI, widespread reductions in gray matter were observed when compared to a cohort of healthy controls.

• Reductions in gray matter were most prominent in regions of interest located in the inferior portion of the brain

• Within the study cohort of head and neck cancer survivors, radiation dose was inversely associated with gray matter volume within the left ventral diencephalon.

• Results of a neurocognitive battery identified associations between cognition and regional GM volumes for motor coordination, confrontation naming, and verbal memory.

Data Availability:

All data collected in this study is available for sharing by written request to the corresponding author.