Baseline total brain volume predicts changes in quality of life and overall survival after cranial radiotherapy in older patients with glioblastoma: Results from the prospective BRITER study

Abstract

Background

Short-course partial brain radiotherapy ± chemotherapy for older patients with GBM extends survival but there is no validated evidence for prediction of individual risk of acute radiotherapy-related side effects.

Methods

This prospective multicentre observational trial recruited patients with newly diagnosed GBM aged ≥65 planned for cranial radiotherapy. Baseline MRI scans were analyzed for markers of brain resilience including relative total brain volume (ratio of cerebrospinal fluid (CSF) volume to total intracranial volume (TIV)) and their relationship to change in quality of life (QoL).

Results

126 patients enrolled: mean age 72 years (range 65-83). 77% had debulking surgery. 79% received radiotherapy with concurrent TMZ, and 21% received palliative radiotherapy alone. The median OS was 10.7 months. After accounting for age, sex, treatment, and baseline MoCA score, there was a relationship between baseline CSF:TIV and change in QoL score at 8 weeks post treatment. For each unit point of increase in CSF:TIV, there was a corresponding decrease in QoL score of 1.72 (95% CI −3.24 to −0.19 P = .027). 35 participants were too unwell to complete questionnaires or had died by the 8 week follow-up visit. In this subgroup, post hoc logistic regression showed baseline CSF:TIV was related to the risk of non-attendance (OR 1.35, 95% CI 1.01 to 1.80, P = .042). Cox regression models showed baseline CSF:TIV was associated with worsened OS (HR 1.41, 95% CI 1.19 to 1.66, P < .001).

Conclusions

This study provides evidence to support the use of an imaging biomarker to help assess the risk:benefit ratio for radiotherapy.

Glioblastoma (GBM) is the commonest primary malignant brain tumor among the adult population with approximately 2000 new cases diagnosed in the UK every year. Incidence peaks in the 7th and 8th decades of life and as the global population ages, rates are increasing. Outcomes from this disease remain extremely poor with median life expectancy around 12–18 months, dropping to 3–6 months in the older population.¹ The reasons for this are multifactorial, including more aggressive tumor biology in the older age group, decreased tolerance to treatment-related side effects and the tendency for clinicians to undertreat elderly patients.^2,3

Given the poor prognosis, treatment must be balanced against side effects and potential for worsening quality of life. In patients aged 65 or over there is a lack of consensus on standard of care. Radiotherapy provides a survival advantage over best supportive care⁴ however the optimal dose of radiotherapy is yet to be established.^5,6 A Phase III trial randomized GBM patients aged 60 and above to standard radiotherapy (60 Gray (Gy) in 30 fractions), hypofractionated radiotherapy (34 Gy in 10 fractions), or temozolomide (TMZ) chemotherapy alone. Patients receiving TMZ or hypofractionated radiotherapy experienced better survival outcomes than those treated with standard radiotherapy.⁷ In patients receiving chemotherapy alone, longer survival was observed in those whose tumors exhibited MGMT gene promoter methylation, as replicated in the NOA_08 trial.⁸ Published evidence has shown a survival benefit from adding concomitant and adjuvant TMZ to a hypofractionated radiotherapy regime of 40 Gy in 15 fractions in patients aged over 65, again with greater effect seen in those with MGMT promoter methylation.⁹ Hence there is now evidence to support the use of concomitant chemoradiotherapy or chemotherapy or radiotherapy as single agents amongst elderly GBM patients and an increasing interest in using MGMT promoter methylation status as a biomarker. However there remains a paucity of data surrounding the clinical and radiological basis by which individual patients are assessed for treatment.

There is evidence to suggest that radiotherapy can stabilize or improve functional ability for some older patients with GBM^10,11 as well as providing a survival advantage. However clinical experience shows that the severity of side effects experienced and their impact on quality of life varies widely within this patient cohort. Short term side effects from radiotherapy include fatigue, headache, cognitive defects, nausea, weakness and a need for increased steroid doses. Longer term side effects include persistent cognitive and memory defects, long term fatigue, and hormonal imbalances^,12,13. Radiation causes an inflammatory response within brain tissue as well as disrupting the blood brain barrier. It affects the vasculature of the brain with endothelial cell damage leading to microvascular dilatation, thickening of the vessel wall, and increased risk of microbleeds, and ischemic strokes in the months to follow¹⁴.

Certain risk factors for radiation toxicity have been identified, namely radiation dose, dose per fraction, and age, but more accurate tools for predicting which patients are more likely to suffer side effects are urgently needed. MRI changes after radiotherapy, including radiation induced atrophy, have been shown to correlate with neurocognitive deficits¹⁵ and MRI can detect microhemorrhages and other ischemic changes prior to treatment which may indicate a “vulnerable” brain.¹⁶ Pilot work by our group has shown a correlation between the degree of baseline cerebral atrophy and survival outcomes in older patients with GBM.¹⁷

The aim of this study was to examine whether pre-treatment MRI parameters relate to changes in QoL measures in elderly patients in the acute toxicity period (7–9 weeks) after receiving cranial radiotherapy for a GBM. Secondary outcome measures included overall survival, radiotherapy-related toxicity as measured by the rate of CTCAE Grade 3–5 events, and change in Montreal Cognitive Assessment (MoCA) score.

Materials and Methods

Participants

Eligible participants were those aged ≥ 65 years with a new diagnosis of GBM who were planned to receive radiotherapy. Diagnosis could be achieved by an MRI scan showing changes consistent with high grade glioma confirmed by a neuroradiologist. Subsequent histological verification was preferred but not mandated. Treatment was determined by the local physician and included cranial radiotherapy, with or without chemotherapy. Ethical approval was granted by the NHS Research Ethics Committee and the Health Research Authority.

Questionnaires

Patients provided informed written consent, underwent baseline data collection, and completed a MoCa screening questionnaire¹⁸ as well as the EORTC QLQ C30 QoL (C30) questionnaire (version 3.0) with the brain (BN20) and elderly (ELD14) subsets of questions added. Repeat questionnaires were performed at 3–5 weeks (time point 2) and 7–9 weeks (time point 3) after completion of radiotherapy. All C30, BN20, and ELD14 scales range from 0 to 100. Subscale scores for C30 were calculated using the qlqc30 command in Stata 17.0.¹⁹ C30 summary score was calculated as the mean of 13 subscales (global health and financial difficulties subscales were not included) after reverse scoring of the symptom subscales.^20,21 Greater C30 summary score indicates better QoL. BN20 and ELD14 subscale scores were calculated using the module QoLR.²²

MRI Variables

Baseline MRI scans were performed including the following sequences:

• Axial T2 4 mm slice thickness.

• Volumetric T1 pre contrast (1 mm or below).

• Post contrast volumetric T1 (1 mm or below).

• DWI/ADC (4 mm slice thickness).

• Susceptibility weighted imaging (acquired volumetrically typically in the region of 1.5 mm slice thickness but varying with scanner setup).

• 3D volumetric inversion recovery or MP-RAGE (additional sequence—to allow accurate quantification of cortical thickness/volumes using Freesurfer software in addition to scoring methods of atrophy, must be acquired at 1mm or below).

• Axial T2* gradient echo (4mm slice thickness).

Analysis of the following variables was performed using Freesurfer software:

• Total brain volume (ratio of cerebral spinal fluid volume to total intracranial volume—CSF:TIV).

• Contralateral medial temporal lobe volume (mm³).

• Volume of T1 white matter hypointensities in contralateral hemisphere (mm³).

• Number of white matter microhemorrhages seen on T2* or SWI in the contralateral hemisphere (range 0–7).

• Fazekas scale assessing white matter changes on T2 weighted imaging calculated in contralateral hemisphere (range 0–3).

Radiotherapy planning target volumes (cc) were acquired from planning software.

Patients with midline tumors and callosal involvement were excluded from the volumetric analysis.

In order to adjust for changes in mass effect and cerebral edema, whole brain volumetric assessment was not performed and volumetric assessment was made only on the whole contralateral, normal appearing, tumor free hemisphere (excluding the corpus callosum). The segmentation analysis allows for anatomical distortion due to midline shift. Post processing segmentation results were checked qualitatively for accurate segmentation of anatomical areas and failed volumetric analysis cases were excluded.

The measured estimated total intracranial volume from which we derived the CSF:TIV ratio is an estimate of the total volume within the calvarium, calculated only on the single side examined (the normal appearing contralateral hemisphere to the tumor). It is an estimate of the total volume within the skull vault. The presence of the tumor did not affect this as it was determined by skull vault size.

Sample Size

The original sample size calculation for the study was based on the predicted association between MRI variables and participants achieving a ≥10 point improvement in EORTC QLQ questionnaire score. However, the number of participants expected to achieve such a change was low (5%) and attrition was greater than expected, greatly limiting the modeling possibilities using logistic regression. While developing the Statistical Analysis Plan (after recruitment was complete but prior to any analyses taking place), the decision was made to change the analyses to more efficiently model associations between MRI variables and continuous EORTC QLQ questionnaire score using multiple linear mixed effects regression, with random effects for the participant to allow data at all time points to be included in the models, and fixed effects for other variables potentially related to the outcome. After an extension due to COVID-19, 128 participants were recruited for the study. The analyses should be considered exploratory, since although the number of participants was sufficient to fit the models outlined in the SAP, power calculations were not performed for these models.

Statistical Analysis

Mixed effects multiple linear regression models with continuous C30 score as the outcome were fitted for all participants with a C30 score at time point 2 or 3 and complete data for the fixed effects listed below. Models included random effects for participants to account for repeated measurements and fixed effects for 1 of the 6 MRI variables per model, age, gender, C30 summary score at baseline, treatment received, baseline MoCA, and time point. Separate models were fitted for each MRI variable as collinearity was expected. Multivariable Cox regression models were fitted for time to death, with the same fixed effects as for the primary outcome.

Results

Baseline Characteristics and Treatment

Hundred and twenty eight patients were analyzed of whom 117 had baseline data available (see Figure 1). The mean age of participants was 72 years (standard deviation 4.2, range 65–83); 67% were men, 33% women. 78% of participants lived with someone, and 22% lived alone. 74% of participants had ECOG PS 0–1, 23% ECOG PS 2, and 3% ECOG PS 3. 36% of participants had a normal (≥26) baseline MoCA score. 62% of patients were taking dexamethasone at their baseline assessment (dose range 1–8 mg daily).

Figure 1.

Consort diagram.

77% of participants had undergone debulking surgery, 19% biopsy, and 4% had a radiological diagnosis. 79% received radiotherapy with concurrent TMZ (47% 40 Gy/15 fractions, 53% 60 Gy/30 fractions), 21% received palliative radiotherapy alone (25% 30 Gy/6 fractions, 75% 40 Gy/15 fractions).

Primary Outcome Measure

The median quality of life questionnaire C30 score was 74.4 (IQR 63.9–81.3) at baseline and 73.7 (IQR 65.4–81.9) 7–9 weeks after radiotherapy treatment. Mixed effects multiple linear regression models showed evidence of an association between change in C30 summary score and 2 baseline MRI variables—CSF:TIV ratio and number of white matter microhemorrhages. For both, increases in the MRI variable were associated with decreases in C30 summary score, corresponding to worsening quality of life (see Table 1).

Table 1.

Showing Associations between each MRI variable and change in QLQ C30 quality of life summary score

MRI variable	Coefficient	95% CI	P-value	No. participants	No. observations
CSF:TIV	−1.72	−3.24 to −0.19	.027	56	91
Medial temporal lobe volume	0.34	−2.85 to 3.53	.834	57	92
WM hypointensities	0.00	−0.88 to 0.87	.993	58	93
No. of white matter microhemorrhages
0	Reference			51	85
1+	−7.83	−15.04 to −0.61	.033	11	17
Radiotherapy PTV	−2.19	−4.99 to 0.61	.125	71	117
Fazekas score
0	Reference			20	32
1	−5.38	−11.04 to 0.28	.062	36	59
2+	−2.64	−10.65 to 5.37	.518	10	18

MRI variables were scaled to avoid excessively small or large coefficients. CSF: total intracranial volume × 10,000 (coefficient represents a 0.0001-unit change); medial temporal volume/1000 (coefficient represents a 1000-unit change); WM hypointensities/1000 (coefficient represents a 1000-unit change); radiotherapy PTV/100 (coefficient represents a 100-unit change). White matter microhemorrhages has been grouped as 0 or 1 + due to few instances of multiple microhemorrhages and Fazekas score has been grouped as 0, 1, or 2 + as few participants scored above 2.

There was insufficient evidence to conclude that Fazekas score was associated with C30 score. There was no association noted between baseline MoCA score, gender, age, planning target volume, and change in quality of life. As it was not possible to calculate validated summary scores for the BN20 and ELD-13 QoL subscale questionnaires, these questionnaires were not included in the modeling analyses. In our analyses, there was no evidence to indicate that the treatment group was associated with any of the primary outcomes. Models for each primary outcome with/without the treatment group variable included were compared using Likelihood ratio tests. P-values ranged from .149 to .427, indicating that the inclusion of the treatment group variable did not improve the fit of the models. However, treatment group was included in the main models as this was defined a priori.

Secondary Outcomes Measures

4.8% of participants experienced recorded CTCAE grade 3–4 fatigue at 3–5 weeks and 1.6% at 7–9 weeks after the end of radiotherapy. 2.6% of participants experienced grade 3–4 seizures at 3–5 weeks and 1.8% at 7–9 weeks after radiotherapy finished. No other CTCAE Grade 3–5 toxicities were recorded.

Mixed effects multiple linear regression models were fitted for MoCA total score (continuous), in the same way as for the primary outcome. There was no evidence of association between MoCA total score and any of the MRI variables however a high proportion of participants had impaired MoCA scores at baseline (64.0%).

Median overall survival (OS) was 10.7 months (IQR 4.9–13.6). There was strong evidence for an association between CSF:TIV and hazard of death. Baseline CSF:TIV was associated with worsened OS (HR 1.41, 95% CI 1.19 to 1.66, P < .001). The hazard ratio (95% CI) for each 0.0001-unit increase was 1.41 (1.19 to 1.66). See Table 2 and Figures 2 and 3.

Table 2.

Adjusted Hazard Ratio for death for each MRI Variable.

MRI variable	Hazard ratio	95% CI	P-value	No. participants
CSF:TIV	1.41	1.12 to 1.66	<.001	88
Medial temporal lobe volume	1.03	0.68 to 1.56	.899	88
WM-hypointensities	1.00	0.94 to 1.07	.949	89
No. of white matter microhemorrhages				94
0	Reference			79
1+	1.50	0.77 to 2.90	.235	15
Radiotherapy PTV	1.05	0.86 to 1.28	.647	107
Fazekas score				101
0	Reference			23
1	2.17	1.09 to 4.31	.027	56
2+	1.18	0.53 to 2.61	.683	22

For the Fazekas Score, the Hazard Ratio (95% CI) for the 1 Group Compared to the 0 Group was 2.17 (1.09 to 4.31). There was no evidence for a difference in the 2 group vs 0 group comparison, but this group was much smaller. A likelihood ratios test was used to assess the overall effect of the Fazekas score categorical variable by comparing models with and without it included. The inclusion of the variable improved the model, indicating it is associated with the hazard of death (P = .029).

Figure 2.

Kaplan–Meier plot of OS in participants with below median CSF:TIV ratio (ie, less cortical atrophy) and above median CSF:TIV ratio (ie, greater degree of cortical atrophy).

Figure 3.

Freesurfer segmented brain overlays for coronal views of (left) Patient with high atrophy score CSF:total (contralateral hemi) brain volume 1.12⁻³ (OS = 80 days) and (right) Patient with low atrophy score CSF:total (contralateral hemi) brain volume 2.52⁻⁴ (OS = 995 days, alive at end of study). The cerebellum and brain stem have been removed from the overlays. Medial temporal lobe volume segmentation is depicted in light gray.

There was some evidence for an association between Fazekas score and hazard of death although this was less strong than for global atrophy (Table 2). There was no association found between gender, age, treatment received or planning target volume, and overall survival. It must be noted that the number of patients who did not receive surgery was very small and therefore the impact of surgical management cannot be fully assessed in this cohort.

Post hoc Analysis

Following analysis, it was clear that missing quality of life outcome data complicated interpretation of the results, as a substantial number of participants were not included in the modeling due to missing C30 summary scores at follow-up time points. A high attrition rate was expected due to the COVID-19 pandemic leading to the redeployment of research staff as well as to the frailty of this patient cohort. Data were therefore collected on the reasons for participants’ inability to complete their follow-up assessments. A post hoc analysis was proposed to model associations between MRI variables and being too unwell to provide data at follow-up.

Thirty five patients were too unwell to complete questionnaires or had died by the 8 week follow-up visit. Post hoc logistic regression showed that baseline CSF:TIV ratio was related to risk of non-attendance (OR 1.35, 95% CI 1.01 to 1.80, P = .042).

Discussion

Active treatment decision making amongst the population of older patients with glioblastoma remains difficult. Although evidence supports the use of chemotherapy, radiotherapy, or combined concurrent chemo-radiotherapy, deciding with the patient on the correct form of therapy for the individual must incorporate consideration of a wider clinical assessment and requires a broader evidence base.

This study population was a relatively fit group of older patients with glioblastoma. 97% had ECOG performance status 0–2 at baseline and 41% were planned for treatment with 6 weeks of radical concurrent chemo-radiotherapy indicating that the treating physician felt they had a good baseline functional ability. Despite this, 31% of patients enrolled and treated actively had died or were too unwell to attend follow-up only 8 weeks following the end of their treatment. This is a strong indicator that our choice of patient for treatment requires significant improvement.

Comprehensive geriatric assessments (CGAs) have been shown to provide prognostic and predictive information for patients with cancer and to assess fitness more accurately than performance status.²³ Data in the neuro-oncology cohort is sparse, however retrospective work by Lombardi et al has shown that a baseline CGA score can predict survival amongst older GBM patients.²⁴ In their work, 47% of patients classified as “frail” by CGA still had KPS of 70–100 and were treated with combined chemoradiotherapy. There is evidence that a neuro-oncology focused geriatric assessment is feasible to implement within the outpatient setting and can help empower members of clinical teams in their clinical treatment decision making.²⁵

There were a number of limitations to our study. Due to the aim of including radiologically diagnosed patients, we did not account for MGMT promoter methylation status, a known prognostic and predictive biomarker in this patient cohort. Future studies will include molecular biomarkers alongside imaging biomarkers. We did not collect data on the tumor location within the brain which may have an effect on QoL. In order to reflect a real-world population our cohort was heterogenous with different treatment regimes incorporated. Our follow-up data collection was limited by the Covid-19 pandemic and therefore our initial statistical analysis plan had to be altered, leading our data to become less robust and more exploratory. Our study was designed to use baseline imaging data to help guide treatment decisions for the patient at that point. We did not therefore collect data on follow-up imaging to assess for radiation induced changes on subsequent MRI scans. Future work will involve assessing the degree of radiation induced damage as assessed by imaging criteria, and the correlation with clinical and QoL changes.

Our study provides preliminary data to support the hypothesis that baseline MR imaging, routinely performed during diagnosis for the majority of patients, can be utilized to quantify the degree of global cortical atrophy and used to predict the severity of radiotherapy induced acute toxicity and its effect on quality of life. Previous studies have shown radiotherapy can cause cortical thinning and atrophy in longer term follow-up.²⁶ We hypothesize that a higher degree of baseline cortical atrophy predisposes these patients to accelerated radiotherapy induced side effects. We propose that gold standard assessment of older GBM patients for treatment decisions should incorporate clinical and cognitive features with a neuro-oncology focused CGA; biological features including molecular markers; and radiological features including the degree of cortical atrophy. Multimodality assessments enable a more personalized treatment approach, helping manage patient and carer expectations from treatment of this difficult and devastating disease.

Conflict of interest statement

C.L., J.B., C.J., I.C., S.M., A.C., A.W., and M.R. have no conflicts of interest to declare. J.G. declares speaker disclosures honoraria and consultation fees for Roche, Pfizer, and ONCOTYPE. G.T. has provided consultancy for translational brain imaging in neurodegenerative disorders, but this is unrelated to this study and has not been remunerated.

Funding

The study was funded by The Sussex Cancer Fund and Brainstrust UK.

Authorship statement

The authors confirm their contribution to the paper as follows: study conception and design: C.L., S.M., A.C., J.B., A.W., G.H.; data collection: C.L., I.C., J.B., J.G., A.C.; analysis and interpretation of results: C.L., S.M., M.R., C.J., J.B.; draft manuscript preparation: C.L., J.B., S.M., C.J., A.C. All authors reviewed the results and approved the final version of the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author, upon reasonable request.