Geriatric assessment of glioblastoma patients is feasible and may provide useful prognostic information

Abstract

Background

Glioblastoma (GBM) is the most common and most lethal primary brain tumor in adults. Clinical trials in older patients with GBM have explored the use of single and multimodality treatment regimens with modest survival benefits; however, trial criteria are commonly based on chronological age and do not reflect the heterogeneity of this cohort. Geriatric assessment (GA) techniques predict survival and treatment tolerance in other tumor sites and thus may objectively guide the decision-making process, but data are lacking in the neuro-oncology cohort.

Methods

We performed a prospective, multicenter feasibility study involving patients age 65 years or older with newly diagnosed GBM. A modified GA was undertaken in the outpatient setting prior to starting treatment. Feasibility was determined primarily by recruitment rate, alongside data completeness, impact on clinic time, and acceptability to patients and staff. Factors associated with survival were explored using Cox regression models.

Results

Fifty patients were recruited within a prespecified time period with a recruitment rate of 82% (target 80%). Data completeness was greater than 80% in all except one assessment. Median overall survival was 9.5 months (95% confidence interval [CI] 5.0-14.0 months). Among the GA screening factors analyzed, a baseline impaired Montreal Cognitive Assessment (hazard ratio [HR] = 2.7, 95% CI 1.128-6.530) and impairment in instrumental activities of daily living (HR = 2.9 95% CI 0.983-8.541) were associated with poorer survival.

Conclusion

In the first study of this kind among elderly GBM patients, we have shown that undertaking a neurologically focused GA screen is feasible and may provide useful prognostic information.

Glioblastoma (GBM) is the most common and most lethal primary brain tumor in adults. Incidence increases with age, peaking in the seventh and eighth decades of life.¹ Prognosis is poor, with average survival of around 18 months for younger patients, but only 3-6 months in those older than 70 years.¹ Treatment in younger patients is based on the Stupp protocol of maximal surgical resection followed by 6 weeks of fractionated radiotherapy with concurrent temozolomide (TMZ) chemotherapy, then 6 months of adjuvant TMZ.² Intensive multimodality treatments have increasing morbidity with increasing age and thus are not routinely offered to elderly patients. Clinical trials in the older age group have explored the use of surgery followed by single-agent chemotherapy, radiotherapy, or combined chemoradiotherapy (CRT) regimens with modest survival benefits. Data from these trials support the use of shorter, hypofractionated radiotherapy regimens rather than extended 6-week protocols, but also highlight the adverse effects of treatment on older patients with GBM.^3–5 Clinical studies have identified molecular biomarkers, including O6-methylguanine DNA methyltransferase (MGMT) methylation status, that can help inform chemotherapy-related treatment decisions,^6,7 but there are very few clinical data on which to base treatment decisions concerning whether to deliver or withhold radiotherapy. Methods or tools for patient selection have not been properly evaluated, and patients recruited into clinical trials usually have better prognostic factors and are more motivated than the general population affected by this disease and therefore do not always faithfully represent this heterogeneous population. Hence there is a need for more research to develop and validate clinical biomarkers that can facilitate a personalized approach to older patients with GBM, with treatment options focusing not just on overall survival (OS), but incorporating quality of life and treatment toxicity endpoints that are crucial in this patient group.

The physiological and functional ability of older adults is often poorly represented by their chronological age, and this number alone should not be used as a decision-making tool.⁸ Oncological treatment decisions should instead be based on a theory of biological age, a more heterogeneous concept that is reflected not by a single number but by the results of a comprehensive, holistic assessment process. Geriatric assessment (GA) tools that aim to give an estimate of biological age are available that evaluate multiple domains, including socioeconomic factors, functional status, cognition, and psychological health. The use of GA tools has been shown to predict morbidity and mortality from cancer in other tumor types^9,10 and thus may objectively guide the treatment decision-making process, but as yet this approach has not been specifically developed for or validated within the neuro-oncology population.

To address this deficiency we undertook the Geriatric assessment for OLDEr patients with Glioblastoma within Neuro-oncology clinics (GOLDEN) study, a feasibility study conducted across 3 United Kingdom cancer centers that assessed whether a neuro-oncology–focused GA could be implemented within busy outpatient clinics.

Materials and Methods

Patient Selection and Assessment

All patients recruited to this multicenter study provided appropriate consent. Patients were recruited from the Royal Marsden National Health Service (NHS) Trust, the Sussex Cancer Centre, and the Beatson West of Scotland Cancer Centre. Ethical approval was granted by the NHS research ethics committee and the Health Research Authority. Eligible patients had been diagnosed with a GBM either by histological confirmation or expert radiological opinion within the setting of a neuro-oncology multidisciplinary team. Patients were age 65 years or older and had been referred to a neuro-oncology outpatient clinic for discussion of nonsurgical oncological treatment options. Patients considered eligible were approached in the neuro-oncology outpatient clinic by a specialist nurse. Participants were excluded if there was uncertainty over their diagnosis, they lacked capacity, or they were unable to complete the questionnaires even with the help of a carer or member of the trial team.

A modified GA tool was developed by the research team specifically for this study. This covered the domains suggested by the European Organisation for Research and Treatment of Cancer minimal dataset¹¹ and used previously validated questionnaires but focused particularly on neurological symptoms. These comprised the Lawton and Brody instrumental activities of daily living (IADLs),¹² the Hospital Anxiety and Depression Scale (HADS),¹³ the Geriatric 8 screening questionnaire (G8),¹⁴ the Montreal Cognitive Assessment tool (MoCA),¹⁵ the Trail Making Test B (TMTB),¹⁶ Eastern Cooperative Oncology Group (ECOG) performance status (PS), and the Charlson Comorbidity Index (CCI).¹⁷ Additional data collected included date of birth, sex, social status, mobility, medications, details of tumor location, and focality and treatment received. If patients had undergone surgery, molecular data were collected, including isocitrate dehydrogenase 1 (IDH1) mutation and MGMT promoter methylation status. Treatment decisions were made by the treating consultant and patients and families based on local policies and patient preferences.

Participants attended follow-up visits according to the usual schedule within their cancer center. If the patient had received active treatment, toxicity data were collected at each follow-up visit using CTCAE V4.0. Survival data were gathered and censoring performed after the final patient had completed the acute toxicity monitoring period.

Outcomes of Interest

The primary outcome measure for this feasibility study was recruitment rate. Secondary outcome measures included: (a) data completeness, calculated from questionnaires and trial pro formas (target completion rate 80%); (b) impact on clinic, assessed qualitatively via analysis of in-depth interviews with clinic staff; (c) acceptability of the trial to participants, assessed by a questionnaire given to participants after their clinic appointment; and (d) OS and toxicity.

Statistical Analysis

A formal sample size calculation was not performed for this feasibility study. Retrospective work performed by the authors of this study¹⁸ had shown around 50 patients per year across the 3 sites would meet the eligibility criteria. Sample size was therefore set at 50. Data were collated in Microsoft Excel and analyzed using SPSS. The primary outcome measure of recruitment rate was analyzed using descriptive statistics. A target recruitment rate was calculated by the number of patients recruited compared with the number who met the eligibility criteria over the set time period. This was prespecified at 80% based on the previous literature. Survival data were analyzed using Kaplan-Meier plots. Exploratory survival analyses were performed using Cox regression models to assess for effect size. These were reported using hazard ratios (HRs) and 95% confidence intervals (CIs). P values were not reported as the study was not powered to detect statistically significant changes in survival between different groups.

Feedback from participants gathered from the questionnaires was analyzed descriptively using Likert scales.

Results

Demographics and Treatment

Fifty patients were recruited within 14 months, meeting the study target. Table 1 shows baseline demographic and treatment data for the full cohort. Fifty-six percent of patients were male, consistent with the published literature.¹⁹ Whereas median ECOG PS was 1 (interquartile range [IQR] 1-2), 8% of patients were ECOG PS 3, indicating that PS alone was not a factor in determining recruitment to the study. All participants had IDH–wild-type GBM as expected in this age group. Of the 35 individuals in whom MGMT testing had been successfully undertaken, 31% showed promoter methylation.

Neurosurgical outcomes were categorized as none, biopsy, subtotal resection (STR, <90% resected), or gross total resection (GTR, >90% tumor resected) and were confirmed by neuroradiological assessment of postoperative MRI scans. Of the 41 patients who underwent surgery, 41% had a GTR. A wide range of first-line oncological treatments were delivered: radical radiotherapy (60 Gy in 30#) with concurrent and adjuvant TMZ to 28% of patients; short-course radiotherapy alone, 40 Gy in 15# to 22% and 30 Gy in 6# to 12%; and short-course CRT to 10%. Of those who underwent surgery, 6 patients had no further active treatment because of rapidly worsening functional state. Of the 9 patients who did not undergo surgery, 6 received either radiotherapy alone: 30 Gy in 6# (3 patients), 40 Gy in 15# (2 patients); or radical CRT (1 patient).

Of 36 participants who consented to receive radiotherapy, 34 (94%) completed the course as planned. All 5 participants receiving short-course CRT completed concomitant treatment but only 1 received adjuvant TMZ. Of 14 receiving radical CRT, 13 (93%) completed the concomitant phase.

Median OS for the whole cohort was 9.5 months (95% CI 5.0-14.0 months). Three-month survival was 85%, 6-month survival 63%, and 1-year survival 43%. Median OS in those receiving best supportive care only was 4.3 months (95% CI 2.5-6.1 months), consistent with national data.¹ Median OS for those receiving TMZ alone was 2.4 months (95% CI 2.3-2.4 months) and for radiotherapy alone was 8.1 months (95% CI 4.2-12.0 months). Median survival for those receiving combined CRT was not yet reached.

Recruitment Rate

Overall recruitment rate was 82% (95% CI 70%-91%), exceeding the predetermined target of 80% (Fig. 1). Thirty percent of participants were recruited from the Royal Marsden, 30% from the Beatson West of Scotland Cancer Centre, and 40% from the Sussex Cancer Centre.

Data Completeness

Data completeness was measured per questionnaire and per participant to detect whether difficulties with specific questionnaires or functional abilities of participants determined completeness. Completion rates for all GA questionnaires exceeded 80% apart from the TMTB (70%). There was no statistically significant difference among the 3 recruiting centers in terms of completion rates of individual questionnaires apart from the TMTB questionnaire, for which the completion rate was 100% in one center, 93% in the second center, and only 50% in the third. Ninety percent of participants completed at least 8 of the 9 questionnaires, and 54% of participants completed all of them, with no statistically significant difference in participant-related completion rates among centers.

Acceptability

An acceptability questionnaire was offered to participants either immediately after completion of assessments or at a subsequent clinic visit; 37 (74%) were completed. More than 85% of participants reported that they had time to understand the study and answer the questions and were treated with respect throughout. Eighty-one percent responded that they would take part in a similar study again. Interviews were performed with the multidisciplinary clinic staff involved with the study at each site. Qualitative analysis of these will be published in full elsewhere. Staff members reported that assessments did require more time in clinic but they considered this worthwhile given the extra information obtained. Examples of comments made about the assessments included that they provided objective measurements to empower their treatment decision making and that they would be keen to embed some of the assessments within their usual clinic practice.

GA

Based on the scoring systems validated in the published literature surrounding each GA questionnaire, the baseline parameters of the study cohort are displayed in Table 2. There are few data available to compare these baseline assessment characteristics with other clinical studies. The median MoCA score was 25 (IQR 19-28), which is the cutoff for an abnormal result. The median time to complete the TMTB was 124 sec, an improved score compared with other GBM populations^20; however, the completion rate of the TMTB did not meet target and therefore may be subject to selection bias.

Survival Analysis

All parameters were assessed for association with OS using univariate analysis (Table 3). None of the baseline demographics (age group, sex, MGMT methylation status, tumor laterality, focality, and mass effect) had a statistically significant relationship with OS.

Of the neurologically focused GA scores, impairment as measured by MoCA score was associated with an increased risk of death (HR = 2.7, 95% CI 1.128-6.530), with a similar association noted for IADL impairment (HR = 2.9, 95% CI 0.983-8.541) and those patients with impaired mobility needing a walker (HR = 10.7, 95% CI 2.2-51.6).

Multivariable analysis was performed using Cox regression models (see Table 4). Three factors suggested an association with survival in the multivariable model: receiving radical treatment (HR = 0.01, 95% CI 0.0-0.1), abnormal MoCA score (HR = 20.34, 95% CI 2.4-172.1), and mobility impairment (walking with a walker; HR = 23.28, 95% CI 1.07-508.75). These results must be interpreted with caution as the original study was not designed to be powered to detect statistical significance and there is a risk of overfitting. However, the observations are consistent with clinical experience and emphasize the importance of functional assessments in this patient group.

Toxicity

Toxicity data were collected on the 41 patients who received active oncological management for up to 3 months after the end of treatment. Toxicity was scored using the CTCAE v4.0 classification²¹ and is summarized in Table 5. Severe toxicities were relatively uncommon but grade 3 or 4 fatigue was observed in 17% of patients and confusion in 5%. Steroid dosing data also were collected during this period: Twenty-three (56%) participants had no change in steroid dose, 8 (20%) decreased their steroid dose, and 10 (24%) increased. This is in line with the published data.⁴ No attempt was made to correlate assessment results or baseline characteristics with toxicities or steroid requirements in this feasibility study.

Discussion

The GOLDEN study is the first prospective study to examine the use of GAs in elderly patients with GBM, and has shown the use of these assessments to be feasible and acceptable both to staff and patients. The long-held belief that GAs are too time- and resource-consuming to be performed regularly has been disproved.^22–25 The study has also identified a number of clinical factors and assessment tools that provide useful prognostic information and may have potential value in informing treatment decisions in this previously underresearched patient group.

Some parts of the assessment were more time-consuming than others, and completion rates differed slightly from site to site. Subjective feedback obtained during the study period indicated that the nurses’ personal opinions on how useful they found the questionnaires or how taxing they felt they were to participants influenced the enthusiasm with which they encouraged participants to complete them. This was particularly evident in the TMTB, which did not meet its target completion rate. In one center the nurse felt it was too onerous a task for some participants to undertake, so the threshold at which it was attempted was higher. This emphasizes the need for GA tools to be acceptable to participants as well as assessors if they are to become accepted in routine clinical practice.

Owing to the heterogeneity of our patient cohort, a number of different treatment regimens were employed, and the study provides some useful insights into their tolerability. Of the 14 patients who were prescribed radical CRT (60 Gy in 30#), 86% completed concurrent CRT but only 43% completed the 6 months of adjuvant TMZ. This is a similar proportion to that observed in the Stupp trial of 2005, which included adults younger than 70 years and supports the idea that age alone should not be a prerequisite for not giving radical CRT regimens.²

As was expected within this cohort, no cases of IDH1-mutated tumors were found. Of the tumors in which MGMT testing was successfully performed, only 31% were found to show promoter methylation. This is slightly lower than some published data, which suggests that MGMT methylation is not age dependent and occurs in 45% to 50% of GBMs,²⁶but is consistent with the Neuro-Oncology Working Group (NOA)-08 study.⁷ The low rate of methylation within our cohort may explain why MGMT methylation status was not a significant factor in univariable analysis for survival.

The overall prevalence of grade 3-4 toxicity levels in this cohort was 20% in those who received radiotherapy or CRT. The majority of these were grade 3 fatigue (17%). This is slightly higher than the 6% to 10% reported in the radiotherapy arms of the NOA-08 and Nordic trials,^6,7 perhaps reflecting the poorer PS of our cohort.

The analysis of the data using univariable and multivariable Cox regression techniques must be interpreted with caution as the study was not powered for these as primary endpoints. However, the effect of impaired mobility and abnormal MoCA are likely to be clinically relevant and are worthy of further investigation on a larger scale.

The study was successful in its intention to evaluate a heterogeneous population, including patients with poorer PS and individuals both with a radiological and a histological diagnosis, so as to provide a realistic representation of the older GBM cohort without the selection bias of suitability for interventional clinical trials. As this is the first evaluation of a panel of GA domains in GBM patients, we are unable to compare the outcomes and scores with those of other studies.

There are now a number of validated treatment options available to older GBM patients; however, the process of deciding on treatments for each individual participant is still unstandardized and subject to clinician bias. There is also a lack of information available from patients and carers about their experience of the toxicities and benefits of treatment, having received it, and whether the impact on their quality of life was adequately discussed prior to making these treatment decisions. We have shown that a neurologically focused GA is feasible within NHS clinics, and there is a suggestion that the results of a quick and simple cognitive screening tool can predict survival, independent of treatment received.

It was beyond the scope of this feasibility study to assess whether GA outcomes influenced treatment decisions. Future work should focus on using a neurologically focused GA routinely within outpatient clinics and incorporating a GA within interventional clinical trials involving older neuro-oncology patients. This would enable exploration of the predictive value of GA and cognitive assessment tools in larger patient numbers. Using a GA in a greater number of outpatient centers would enable researchers to explore the effects of GA results on treatment-decision pathways as well as providing information on perceived or actual barriers to their use.

Geriatric oncology research in other tumor sites has evolved into implementing interventions aimed at reversing the impact of biological aging and assessing whether these can affect treatment tolerance and survival. Neuro-oncology is lagging behind and currently lacks evidence regarding patient assessment. Our study has provided preliminary evidence to support the use of a neurologically focused GA in older GBM patients, and this may facilitate future attempts to implement interventions aimed at improving outcomes.

Further research will be aimed at developing models to help predict which patients are likely to benefit from active treatment and which are not, enabling clinicians to have more informed and individualized treatment discussions with patients. In this vulnerable patient cohort, discussions should emphasize quality of life as well as OS.

Funding

This work was supported by BrainsTrust [grant number 212524]. The study required only administrative costs (£150 in total), which were funded by contributions from BrainsTrust, 4 Yvery Ct, 36 Castle Rd, Cowes PO31 7QG, UK.

Conflict of interest statement. None declared.

References