The glioblastoma biomechanical landscape: A systematic review of magnetic resonance elastography (MRE) of brain tumors and healthy brain

Abstract

Diagnosis of a glioblastoma (GBM) brain tumor is associated with very poor prognosis. Currently, few preclinical models used to identify new therapies address the soft brain tissue environment and GBM mechanoresponses, which are implicated in disease progression. Understanding the GBM biomechanical landscape is critical to deriving improved preclinical models and magnetic resonance elastography (MRE) holds promise to address this gap. Due to technical and feasibility issues for MRE of patient tumors at scale, most studies only report on small cohorts of patients, thus limiting the conclusions that may be drawn from individual studies. To thus gain a better overview, we have undertaken a systematic review and meta-analysis of the reported tissue viscoelastic property values from studies of both healthy brain and brain tumors, with a particular focus on delineating measurements relative to MRE transducer vibration frequency. Based on these analyses, healthy white matter consistently appears stiffer than gray matter. Further, analyses of pooled healthy brain tissue measurements vs human GBM suggested that, overall, the GBM has the same stiffness as the surrounding healthy tissue. This contrasted with mouse models of GBM, where the tumors appear softer than brain tissue. The limited number of studies of human GBM in situ is a caveat to these conclusions and MRE analyses of larger GBM patient cohorts are urgently needed. Meanwhile, the information from this analysis can be used to guide engineering of improved preclinical models with features that recapitulate the in vivo brain tissue environment.

I. INTRODUCTION

Only 1 in every 10 patients will still be alive 5 years after diagnosis of a glioblastoma (GBM) brain tumor.¹ The failure to translate promising preclinical findings into enhanced patient survival has, in part, been attributed to the shortcomings of preclinical models.² It is increasingly appreciated that mechano-regulation of cell behavior^3,4 may play a critical role in GBM disease progression and therapy response,^3,5,6 with studies revealing mechanosensitive GBM behavior.^7–10 Yet, most preclinical analyses use rigid tissue culture dishes that do not reflect the soft brain tissue environment.^2,11,12

Detection of cancers by palpation is a clue to the stiffer matrix that characterizes many solid tumors.¹³ The stiff tumor environment has been shown to stimulate tumor proliferation, invasion, and response to therapy.^13,14 Whether GBMs are stiffer than the surrounding brain parenchyma remains an open question with conflicting reports in the literature.^3,15–18 A limitation of the reports that GBMs are stiffer than healthy brain is the measurement of tissue properties in ex vivo tissue sections,¹⁵ lacking the confinement imposed by the skull, blood flow, and cerebrospinal fluid, which alter the brain's mechanical properties. By contrast, magnetic resonance elastography (MRE) is a noninvasive imaging technique for measuring tissue viscoelastic properties in situ.¹⁹ Critically, MRE data suggest that GBM may be softer than the surrounding healthy tissue.^3,16–18 However, individual studies of brain tumor MRE only report on small patient cohorts as the reality of treatment for this devastating disease has been a barrier to analyzing large cohorts of patients.²⁰

Brain is a viscoelastic tissue, combining both elastic and viscous characteristics that can be quantified by MRE. The biomechanical features are described by the Storage Modulus G′ that represents the elastic component of the tissue; the Loss Modulus G″ that represents that viscous component; and a combination of the storage and loss moduli | G∗ | to describe the overall viscoelastic behavior.^3,21,22 A previous systematic review focusing on the potential for MRE to guide surgical options for brain cancer concluded that the majority of GBMs may be softer than healthy brain tissue.²³ This systematic review and meta-analysis aims to delineate the mechanical parameters required for building physiologically relevant preclinical models for brain cancer, with a particular focus on whether GBMs are softer than healthy brain tissue. To achieve this, we have quantitatively synthesized MRE data to define the values of |G^*|, G′, and G″ in both healthy brain tissue and GBM across human and murine studies. MRE studies of other primary tumors have been included, to help position the GBM mechanical features in the spectrum of brain tumors. By integrating data from normal and tumor-bearing brain tissue, our goal is to provide a comprehensive biomechanical profile of the brain and GBM microenvironment. This will support the development of models that better recapitulate the viscoelastic landscape of the human brain and improve the translational fidelity of GBM research.

Poor correlation between preclinical and patient efficacy data in GBM treatment has been attributed in part to limited correlation between preclinical efficacy and subsequent activity in Phase I clinical trials, highlighting the need for improved preclinical models.² It is hoped that this information will inform the development of preclinical models that truly reflect the brain parenchyma/GBM environment and consequently improve translation of preclinical development into successful therapies.

II. METHODS

A. Literature search

A systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.²⁴ The protocol for this systematic review was registered with PROSPERO (International Prospective Register of Systematic Reviews); Registration No.: CRD42024567910. Ovid MEDLINE, Ovid Embase, PubMed, Web of Science, and Google Scholar were searched on 11 July 2024. Combined database-specific subject headings with free text terms in the title, abstract, and keyword fields for “Brain,” “Tumour,” “Cancer,” “MRE,” and “Magnetic Resonance Elastography” were used. Articles with the term “review” in the title and keyword field were excluded. The search strategy was built using text and index terms and run in all databases from inception up to 11 July 2024. No database filters were used. A full description of the search strategy is listed in supplementary data 1 in Ref. 25. Covidence (Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia. Available at www.covidence.org.), a web-based collaboration software platform that streamlines the production of systematic reviews and other reviews of the literature, was used to download all database results and retrieve full text. Two reviewers independently evaluated and selected studies using Covidence, which facilitated blinded screening of titles, abstracts, and full text articles. Discrepancies were resolved through consensus. Reasons for exclusion at the full text stage were recorded.²⁵

B. Inclusion criteria

Studies for inclusion were selected on the following inclusion criteria: (i) observational studies involving human or murine subjects, encompassing retrospective cohort studies, case series and case reports; (ii) articles published up until 11 July 2024; (iii) full text articles that have undergone peer review; (iv) MRE studies on entirely normal healthy human brain with and without comparison to human subjects with brain tumors OR MRE in animal models reporting on results of tumors in the brain with and without comparison to healthy controls were included. The mouse studies were included based on the expectation of a limited dataset from human patients.

C. Exclusion criteria

Excluded studies were as follows:(i) review articles, systematic reviews/meta-analyses, conference abstracts, theses, articles posted on pre-print servers, (ii) studies lacking quantitative data, (iii) articles not authored in English, (iv) studies only reporting on data obtained from non-magnetic resonance elastography techniques, and (v) studies lacking a statement of ethics.

D. Data extraction

Data extracted from the selected studies included the study characteristics (first author, year of publication, sample size), clinical details (patient age, sex, and tumor type where applicable), MRE parameters (imaging resolution, analysis algorithm applied, transducer frequency, location of measurement), and outcomes (estimate and modality of viscoelastic property).²⁵ To collect all relevant data, manuscript appendixes/supplementary material were also reviewed. Data were summarized as mean estimates of G′, G′′, |G^*|, and “shear stiffness” with standard deviation estimates also collected. However, since the term “shear stiffness” has been used inconsistently in previous literature, we treated it separately in our analysis. Some studies define shear stiffness as the magnitude of the complex shear modulus (|G∗|), while others calculate it as density × wave speed².²² Where only graphical data were available, PlotDigitizer (https://plotdigitizer.com) was used to extract the data. This is a Java-supported program that allows the extraction of raw data from X-Y-type scatter or line plots.^26,27 Where data were not presented as mean and standard deviation, previously described methods were used to obtain estimates of the mean and standard deviation of the data presented in the studies.^28,29

E. Data analysis

As factors such as transducer frequency and brain subregion influence the measured viscoelastic properties,^3,22 the data were segmented by these two factors and a subgroup analysis was performed. For analysis, the white matter subgroup incorporated only studies which reported values for “white matter” and did not include values for regions which could be considered as white matter such as values reported for corpus callosum or corona radiata. Likewise, gray matter subgroup analysis incorporated only studies reporting on values for “gray matter” and not for regions which could be considered as gray matter such as values reported for thalamic structures. The methods described by Neyeloff et al. were used to estimate the population mean of G′ and G′′, |G^*| and the shear stiffness, μ, for each of the brain regions at each reported transducer frequency as a continuous mean with 95% confidence intervals. Meta-analysis was undertaken in Excel using a previously described method.³⁰ Extracted data were entered into the Excel spreadsheet as continuous data of mean and standard deviation. Variance was obtained by squaring the standard deviation, and study weights were calculated as the inverse of the variance. Weighted effect sizes and additional terms required for heterogeneity analysis (e.g., weighted squares) were computed and summed across studies. Either fixed-effects or random-effects models were used. Where random-effects modeling could not be used due low heterogenicity (e.g., in a single study sample), a fixed-effects model was used to estimate the mean and errors of the population.³⁰ The test for intra-study heterogeneity was measured using the Cochran Q test (Q) and I². Q is calculated as the weighted sum of squared differences between the effects of the individual study and the pooled effect between the studies.³⁰ I² is expressed as a percentage of total variability in a set of effect sizes due to heterogeneity, with higher values indicating more heterogeneity between studies.^30,31 Sensitivity analyses were performed to assess the robustness of results when the number of studies for each subgroup analysis >3 and I² > 50%. A leave-one-out method was used to assess the impact of each study on the overall reported viscoelastic property using a Brown–Forsythe ANOVA test with alpha =0.5. Linear regression analysis and power-law slope estimates were used to compare white and gray matter using Graphpad Prism. G′ and G′′ as well as |G^*| and the shear stiffness of various brain tumors were compared to the normal brain using a Z-test for linear regression and extra sum-of-squares F test for power-law slope estimates with α = 0.05. Calculated values are shown in supplementary data 4 and 5 in Ref. 25, respectively. Notably, there was no significant difference in the power-law slopes²⁵ and since the MRE data captured in this review span only 1–2 frequency octaves, thus constraining model parameter estimation, simpler linear or exponential fits, or frequency ratios, are used here as per previous reports.³² Shear modulus phase angles were estimated for each tissue type (whole brain, white matter, and gray matter) at each transducer frequency and plotted against the corresponding MRE-induced shear frequency.²⁵ Linear regression was performed to assess the dependency of phase angle on frequency. F-tests were used to evaluate whether regression slopes significantly differed from zero and to compare intercepts between tissue types using GraphPad Prism with α = 0.05.

III. RESULTS

A total of 2203 studies were identified comprising studies from the following sources: Ovid MEDLINE (1000), (511), (408), (212), and (72). From these, 866 duplicate records were removed by Covidence. Then, 1337 abstracts were screened and 1129 were removed for failing to meet all inclusion criteria or fulfilling an exclusion criterion (Fig. 1). Following full text review of 202 potential candidate studies, a total of 90 were finally included for systematic review. Reasons for exclusion of studies at the full text review stage are outlined in supplementary data 2 in Ref. 25.

FIG. 1.

PRISMA Flow Diagram describing article screening and selection process. Two reviewers (TJ and GON) independently undertook blinded screening of titles abstracts and full text articles. PROSPERO Registration No.: CRD42024567910.

A. White matter is stiffer than gray matter

Following data extraction and compilation,²⁵ linear regression analysis was performed to compare G′, G″, μ, and | G∗ | of the gray and white matter relative to transducer vibration frequency (Fig. 2 and Table I). There were no differences between gray and white matter in the slopes of the curves for G′ (0.024 vs 0.018 kPa/Hz, P = 0.15), G″ (0.016 vs 0.015 kPa/Hz, P = 0.93), and | G∗ | (0.035 vs 0.036 kPa/Hz, P = 0.51); however, in each case, there was a significant difference in the y-intercepts G′ (0.51 vs 0.97, P = 0.0003), G″ (0.076 vs 0.165, P = 0.0005), and | G∗ | (0.088 vs 0.133, P = 0.0076), suggesting that white matter is stiffer than gray matter. In the case of shear stiffness, there was a significant difference in the slope of curves (0.053 vs 0.210 kPa/Hz, P < 0.0001), suggesting that gray matter may have a higher shear stiffness at lower frequencies and a lower shear stiffness at higher frequencies compared to white matter. However, when the single study that reported shear stiffness values at 100 Hz is removed using Grubb's test for outliers, the slopes of the linear regression analysis curves are not significantly different (0.024 vs 0.030 kPa/Hz, P = 0.53). Significant difference in the intercepts (1.289 vs 4.251; P = 0.0369) supports the conclusion that white matter is stiffer than gray matter.

FIG. 2.

Population estimates for storage, loss, shear as well as complex shear modulus of human whole brain, white matter, and gray matter were determined by meta-analysis of published MRE studies in human subjects. Population estimates were plotted against the induction shear frequency used in MRE. Data presented as mean ± SEM.

TABLE I.

Population estimates of storage, loss, complex shear as well as shear stiffness of gray and white matter in human brain. Estimates of values G′, G′′, | G∗ |, and the shear stiffness obtained through meta-analysis of published studies presented for each reported transducer frequency. Number of studies and total number of subjects for each condition are also listed. The test for intra-study heterogeneity was measured using the Cochran Q test (Q) and I², n = number, Hz = Hertz, kPa = kilopascals.

		Gray						White
Modulus	Frequency (Hz)	Studies (n)	Subjects (n)	Mean (kPa)	95% CI (kPa)	Q	I2	Studies (n)	Subjects (n)	Mean (kPa)	95% CI (kPa)	Q	I2
Storage	40	2	72	1.41	1.29–1.53	8	88.16	2	72	1.56	1.46–1.65	6	82.15
	50	3	27	1.88	1.57–2.19	57	96.51	6	50	2.07	1.78–2.36	290	98.28
	60	2	72	2.01	1.89–2.13	14	92.80	2	72	2.10	2.03–2.18	6	83.59
	80	3	80	2.36	2.31–2.4	2	18.21	3	80	2.43	2.39–2.47	1	0.00
	90	2	51	2.73	2.01–3.45	184	99.46	2	51	2.53	2.21–2.86	35	97.18
Loss	40	2	72	0.71	0.52–0.89	62	98.39	2	72	0.68	72–0.68	0	0.00
	50	3	27	1.1	1–1.21	7	70.69	4	34	1.23	34–1.23	185	98.38
	60	2	72	0.96	0.89–1.04	17	94.14	2	72	1.05	72–1.05	22	95.46
	80	3	80	1.15	1.1–1.19	13	84.29	3	80	1.18	80–1.18	0	0.00
	90	2	51	1.72	0.38–3.05	207	99.52	2	51	1.69	51–1.69	230	99.57
Complex shear	20	0	0					1	42	0.78	0.75–0.8	0
	30	1	44	1.08	44–1.08	0		2	77	1.09	1.03–1.15	0
	40	2	70	1.54	70–1.54	0		2	70	1.62	1.43–1.8	0
	60	2	70	2.3	70–2.3	0		2	70	2.33	2.3–2.37	0
	80	1	26	2.74	26–2.74	0		1	26	2.76	2.68–2.84	0
	Multi-frequency	1	5	1.03	5–1.03	0		4	78	1.4	1.16–1.63	105	97.13
Shear stiffness	40	1	10	2.24	10–2.24	0		1	10	3.36	10–3.36	0
	50	4	41	2.53	41–2.53	241	98.75	6	54	2.64	54–2.64	389	98.71
	60	2	14	2.74	14–2.74	44	97.74	3	42	2.53	42–2.53	2376	99.92
	100	1	25	5.22	25–5.22	0		1	25	13.6	25–13.6	0

B. GBM stiffness is equivalent to normal brain in human studies

To date, a limited number of studies have been performed using MRE to investigate the mechanical characteristics of brain tumors and only a few specifically focus on GBM (six studies) in comparison to other tumor types, including gliomas excluding GBM (four studies), meningioma (eight studies), metastatic B-cell lymphoma (one study), and other tumors that have metastasized to the brain (three studies). In total, 11 studies comprising a total of 142 subjects investigated G′ (two studies), G′′ (two studies), and | G∗ | (seven studies) (Table II). No significant difference in G′ between GBM tumors and whole normal brain data, collected across included studies, measured at 50 Hz was detected using a two-tailed unpaired t-test (P = 0.4147). Similarly, G′′ at 50 Hz was not significantly different between the two (P = 0.8555). Other conditions in which G′, G′′, and | G∗ | were obtained for GBM could not be compared with normal brain due to the lack of comparator values in the published literature. | G∗ | of gliomas other than GBM was also not significantly different from normal whole brain measurements at 60 Hz (P= 0.0689). In contrast, the complex shear modulus of meningiomas was significantly higher than that of the normal whole brain at 60 Hz (P < 0.0001). When analyzing the |G^*| values obtained from multi-frequency analysis, compared to normal analysis of the whole brain, GBM was not significantly different (P = 0.5151), nor were other gliomas (P = 0.8909), meningioma (P = 0.3266), or metastatic tumors (P =0.8225).

TABLE II.

Population estimates of viscoelastic properties in human brain studies with tumors. Estimates of values G′, G′′, | G∗ |, and the shear stiffness obtained through meta-analysis of published studies presented for each reported transducer frequency. Selected estimates of human whole brain viscoelastic properties presented for comparison. The test for intra-study heterogeneity was measured using the Cochran Q test (Q) and I², n = number, Hz = Hertz, kPa = kilopascals.

Tumor	Modality	Frequency (Hz)	Studies (n)	Participants (n)	Value (kPa)	95% CI (kPa)	Q	I2
B-cell lymphoma	Complex shear modulus	45	1	1	1.4
Glioblastoma	Complex shear modulus	45	1	3	1.24	0.89–1.59	0.00
		30–60	4	45	1.29	1.19–1.37	4.36	31.21
	Loss	50	1	10	0.66	0.62–0.70	0.00
		30–60	1	3	0.63	0.48–0.77	0.00
	Storage	50	1	10	1.4	1.32–1.48	0.00
		30–60	1	3	1.09	0.91–1.26	0.00
Gliomas other than glioblastoma	Complex shear modulus	45	1	7	1.3	1.11–1.49	0.00
		60	1	18	2.2	1.88–2.52	0.00
		30–60	2	4	1.46	1.26–1.65	0.47	0.00
	Loss	30–60	1	1	0.77	0.00– 1.54	0.00
	Storage	30–60	1	1	0.76	0.00– 1.67	0.00
Meningioma	Complex shear modulus	45	1	2	2.09	2.02–2.17	0.00
		60	1	18	3.12	2.55–3.69	0.00
		30–60	3	19	1.62	1.57–1.68	1.15	0.00
	Loss	30–60	1	3	1.19	0.99–1.39	0.00
	Storage	30–60	1	3	1.19	1.03–1.35	0.00
Metastasis	Complex shear modulus	45	1	3	1.25	0.97–1.54	0.00
		30–60	2	6	1.84	0.95–1.62	0.02	0.00
	Loss	30–60	1	1	0.61	0.00–1.24	0.00
	Storage	30–60	1	1	1.06	0.66–1.47	0.00
Normal human whole brain	Complex shear modulus	60	3	82	2.37	82–2.37	0
		Multi-frequency	7	127	1.39	127–1.39	247	97.57
	Loss	50	5	181	0.73	0.56–0.9	659.04	99.39
	Storage	50	7	197	1.80	1.58–2.01	516	98.84

C. GBM are softer than the surrounding brain in mouse studies

Given the small number of studies of human brain tumors, studies in mouse models were also included. Five studies used MRE to measure the viscoelastic properties of GBM brain tumors in animal models, including one study with healthy controls and four studies without healthy controls. Data were available from 52 GBM orthotopic xenografts and six healthy control mice collectively (Table III). G′ (one study), G′′ (two studies), and |G^*| (two studies) were investigated at frequencies of 900 (one study), 1000 (three studies), and 1800 Hz (one study). G′ of GBM at 1000 Hz (one study) indicated that the tumor was softer than healthy mouse brain (one study) (P = 0.0043), a result mirrored in G′′ results calculated from data using 1000 Hz transducer vibration frequency (one study) (P = 0.0001).

TABLE III.

Population estimated of viscoelastic properties in mouse brain tumor MRE studies. Estimates of values G′, G′′, and | G∗ | obtained through meta-analysis of published studies of mouse MRE studies with brain tumor with and without comparison to normal mouse brain, presented for each reported transducer frequency. Number of studies and total number of animals for each condition are also listed. The test for intra-study heterogeneity was measured using the Cochran Q test (Q) and I², n = number, Hz = Hertz, kPa = kilopascals.

Tumor	Modality	Frequency (Hz)	Studies (n)	Animals (n)	Value (kPa)	95% CI (kPa)	Q	I2
Glioblastoma	Complex shear	900	1	20	5.34		0.00
		1000	2	10	5.71	4.60–6.83	13.22	92.43
	Loss	1000	1	20	2.51	2.19–2.84	0.00
		1800	1	12	3.24	2.21–4.27	0.00
	Storage	1000	1	20	4.25	3.70–4.80	0.00
Healthy control	Loss	1000	1	6	4.36	4.22–4.50	0.00
	Storage	1000	1	6	5.89	5.75–6.03	0.00

IV. DISCUSSION

We sought to methodically interrogate the published literature using meta-analysis to better define the mechanical environment of healthy brain and GBM. Analyses of the combined studies suggest limited difference in tissue viscoelastic properties between GBM and healthy brain in humans. Conversely, in mouse models, engrafted human GBM tumors appear softer than the surrounding tissue. While there are caveats to extrapolating these findings to all human and mouse GBMs, the results highlight the benefit of this systematic review for future GBM model development. Most notably, in no case were GBM tumors shown to be globally stiffer than the surrounding brain tissue by MRE.

Since MRE-derived mechanical measurements are influenced by the frequency of the applied shear waves, the results were segmented by the MRE transducer frequency. This was validated by examining G′, G′′, |G^*|, and μ across the whole brain, white matter, and gray matter, all of which increased with higher transducer frequencies (Fig. 2), aligning with earlier studies.^3,21,33–35 This frequency-dependent behavior is consistent with the viscoelastic nature of brain tissue, meaning that at lower frequencies (50–60 Hz, typical for human MRE), brain tissue behaves more viscously, which may mask subtle differences in the viscoelastic properties between GBM and surrounding tissue. In contrast, higher frequencies (e.g., 1000 Hz, used in small-animal MRE) emphasize elastic properties, which could make the differences in viscoelastic properties more pronounced, suggesting that findings in mice at 1000 Hz may not translate directly to human GBM at 50–60 Hz.

Comparison of white and gray matter in the studies of normal brain indicated that white matter is stiffer than gray matter, consistent with previously reported MRE^4,34,35 and rheometric measurements.^4,35 A caveat to these comparisons is that few studies use the same viscoelastic modality, transducer frequency, and region of interest. Moreover, 92/202 estimates examined were from single study samples.²⁵ A total of 11 studies reported the shear modulus of healthy human whole brain at 50 Hz, but even with this number of studies, there was a high degree of heterogeneity with an I² of 99.57. Similarly, while the estimation of the population mean shear modulus of the hippocampus at 50 Hz was derived from analysis of 236 participants, there was still a high degree of heterogeneity between studies with an I² of 94.15. Intra-study differences are likely to include patient population characteristics, study methodology, and analysis methods.^36,37 Sensitivity analysis confirmed that no single study had a qualitative influence on the pooled viscoelastic properties of tested subgroups. Analysis of the white and gray matter shear modulus phase angles revealed significant differences and suggest a higher viscous to elastic ratio for gray matter compared to white matter.²⁵ While the biological significance of this is yet to be fully determined, it potentially reflects differences in the tissue microstructure. The elongated cells that characterize white matter have been theorized to contribute to higher stiffness and lower relative viscosity as compared to the more amorphous cells that characterize the gray matter.^38,39 Finally, it is noted that one critical factor that is often overlooked in MRE studies is the anisotropic nature of the brain tissue, particularly in white matter, which is thought to exhibit direction-dependent mechanical properties. The analyzed studies did not consider anisotropy, which may influence the mechanical parameters.

GBMs are distinguished histologically by the presence of a focal necrotic core, surrounded by layers of tumor cells¹⁷ that move away from the necrotic core called pseudopalisades.^40,41 During the growth and expansion of GBM tumors, compressive forces are generated due to the confines of the cranium.³ The analyses described here suggest no discernible difference between brain and GBM, in contrast to the earlier finding that GBMs appear softer than the healthy brain tissue.²³ However, it is important to note some studies have identified intra- and intra-tumor heterogeneity in | G∗ | in patients with GBM, with some patients showing decreased and others increased values of | G∗ | compared to healthy brain.^18,42 Notably, both the previous study²³ and the present meta-analysis are similarly limited by a small GBM sample size (data and extraction and quantification from 5 and 11 studies, respectively). Our study compared GBM viscoelastic properties to pooled values of normal human brain, contrasting the earlier approach where GBM viscoelastic property values were compared directly with values of normal brain from the same patient.²³ Previous MRE measurements on bovine liver specimens have shown that static compressive pre-strain increase both G′ and G″ in a nonlinear manner. This is best described by an exponential model, with 10% linear compression resulting in a 47% overestimation of the true value of G′l.⁴³ It would stand to reason that in an enclosed space such as the skull, the mass effect caused by a tumor would induce pre-loading onto the healthy tissue, likely obscuring viscoelastic differences between healthy and tumor tissues. The value of the present systematic review therefore is the comparison of viscoelastic properties of tumor tissue with wholly normal brains which would not be subject to pre-loading. Our approach was motivated by the need to obtain a consensus picture of healthy brain to inform the engineering of preclinical models.

When the analysis was performed with pooled MRE data on other brain tumors, no differences in |G^*| of gliomas other than GBM at 60 Hz were found, compared to healthy control brains. In contrast, |G^*| of meningiomas at 60 Hz is significantly higher compared to the normal healthy brain. This echoes other studies showing increased meningioma stiffness is correlated with the intraoperative qualitative evaluation of stiffness by surgeons.^23,44 Tumor stiffness is a key characteristic that is considered in planning resection surgery, as softer tumors require a less invasive endoscopic procedure, while firm tumors may require more open surgical approaches.^23,45,46 Overall, the review highlights the paucity and inconsistencies of data on the mechanical properties of brain tumors, which has hindered the development of preclinical models of GBM that mimic the in vivo tumor microenvironment.

To expand the biomechanical characterization of GBM, we included analysis of mouse models of GBM.^47–50 We identified five MRE studies of orthotopic GBM mouse models, three reporting mouse GBM lines engrafted into immunocompetent mice^16,51,52 and two reporting orthotopic xenografts of human GMB lines. Despite the different GBM tumor sources, in all cases the GBMs in mouse models appear softer than normal mouse brain. This may reflect better experimental control in animal studies, resulting in lower intra- and inter-study variability.^{30,37,48–50,53} Alternatively, differences in tumor infiltration, vascularization, and extracellular matrix remodeling in xenografted mice could lead to an artificially softer tumor appearance compared to human GBM in situ. Moreover, the higher frequencies used in animal MRE studies may emphasize the elastic properties of the brain, which could make measurements of complex shear modulus differences more apparent as previously outlined. Yet another possibility is that the xenograft has a mass effect on the surrounding brain tissue, applying compressive forces which, if substantial, could alter the apparent brain viscoelastic properties due to the nonlinear behavior of brain tissue.⁵⁴ Notably, MRE analyses are hindered by the time it takes to scan mouse brains;¹⁷ clinical MRE in humans can be completed in 2 min with a single transducer driving frequency, while MRE in murine brains can exceed 30 min.⁵⁵ As a result, some mouse studies use reduced image resolution and number of transducer frequencies to minimize the total scan times.⁵⁶

A significant limitation to the ability to undertake MRE analyses of patients with GBM is the very poor prognosis for these patients. Current best practice dictates maximal safe surgical resection of the tumor as soon as practicable after diagnosis, followed by radiotherapy and chemotherapy. The window of opportunity for recruiting patients prior to treatment for MRE is therefore brief. Moreover, it is a time of heightened stress for patients dealing with a dire survival prognosis, often occurring after sudden symptoms that result in presentation at a hospital emergency department.⁵⁷ Thus, larger studies will require close cooperation between surgeons, radiation oncologists, the MRI team, and the investigators undertaking the study. Furthermore, this review highlights a lack of consistent reporting and study methodology of MRE studies in both patients and rodents, which restricts meaningful data comparison across studies.

To conclude, MRE has a distinct advantage over other methods of rheometric analysis, including the evaluation of tissues in vivo, higher resolution, and deeper penetration of tissue.^3,58–60 Although significant efforts have been made to model the mechanical properties of the brain, contradictory results and variations in study techniques have hindered progress and prevented meaningful comparison of data.^61,62 Furthermore, this review highlights the paucity of studies investigating GBM tumors along with a lack of longitudinal data over the treatment time course, information that could guide the development, testing, and delivery of novel therapies for GBM.^63,64 Importantly, in addition to further MRE studies it will be important to firmly establish the GBM mechano-environment characteristics via complementary technologies. Nonetheless, the review confirms that there is limited evidence from current MRE studies to suggest a global increase in stiffness in GBM vs healthy brain tissue, an important consideration in the design of novel preclinical laboratory assays for GBM.

AUTHOR DECLARATIONS

Conflict of Interest

The authors have no conflicts to disclose.

Author Contributions

Thuvarahan Jegathees: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Methodology (lead); Writing – original draft (equal); Writing – review & editing (equal). Lauriane Jugé: Conceptualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Eric Hau: Conceptualization (equal); Supervision (supporting); Writing – original draft (equal); Writing – review & editing (equal). Lynne E. Bilston: Conceptualization (equal); Supervision (supporting); Writing – original draft (equal); Writing – review & editing (equal). Geraldine M. O'Neill: Conceptualization (equal); Formal analysis (supporting); Project administration (lead); Resources (lead); Supervision (lead); Writing – original draft (equal); Writing – review & editing (equal).

DATA AVAILABILITY

The data that support the findings of this study are openly available in Zenodo at https://doi.org/10.5281/zenodo16863781, Ref. 25.