Quality of life in the GLARIUS trial randomizing bevacizumab/irinotecan versus temozolomide in newly diagnosed, MGMT-nonmethylated glioblastoma

Abstract

Glioblastoma (GBM) is a devastating disease with a particularly short survival in the large subgroup of patients having a tumor with a nonmethylated O⁶-methylguanine-DNA methyltransferase (MGMT) promoter.¹ Standard temozolomide (TMZ) chemotherapy has virtually no benefit in this subgroup. In search of new first-line therapies for these patients, the randomized phase II GLARIUS trial explored the value of anti-angiogenic therapy using the vascular endothelial growth factor–A antibody bevacizumab (BEV) in combination with the topoisomerase I inhibitor irinotecan (IRI). The GLARIUS trial showed that BEV/IRI is superior to standard TMZ regarding progression-free survival (PFS) at 6 months.² Similar to phase III trials exploring TMZ + BEV versus TMZ in an unselected population of newly diagnosed GBM patients (the AVAglio trial,³ RTOG0825⁴), GLARIUS showed that median PFS was prolonged from 5.99 months in patients treated with TMZ to 9.7 months with primary BEV/IRI therapy. However, overall survival was not prolonged.² In contrast to the similarity of their survival results, data on quality of life (QoL) and neurocognition in AVAglio and RTOG0825 are somehow controversial. RTOG0825 reported a decline in several domains of QoL and neurocognitive functioning in patients treated with BEV, while AVAglio found a prolonged time to deterioration of QoL in some dimensions with primary BEV therapy.^3–5

The GLARIUS trial also included an extensive QoL evaluation program. Regarding QoL analysis, 2 aspects were fundamentally different in GLARIUS as opposed to AVAglio and RTOG0825: (i) In contrast to the AVAglio and RTOG0825 trials, the acquisition of QoL data did not stop at tumor progression in the GLARIUS trial, so that the development of QoL parameters throughout the entire course of the disease was monitored, thus also postprogression QoL could be analyzed; (ii) the protocol of GLARIUS included an optional crossover from the standard arm to BEV therapy as second-line therapy—with 82% of all patients treated with second-line therapy receiving BEV therapy after progression,² GLARIUS is the first trial that allows analysis of the influence of second-line BEV therapy on postprogression QoL. The results of these analyses regarding QoL and other parameters potentially associated with QoL, such as cognitive functioning, performance score, and steroid use, are summarized here.

Patients and Methods

General Features of the Trial

GLARIUS is a randomized, controlled, unblinded phase II trial (EudraCT no. 2009010390-21; clinical trials.gov NCT00967330). The trial protocol was approved by the ethics committees of all participating centers. Written informed consent was obtained from all patients before any study procedures. All trial activities adhered to the principles of the Declaration of Helsinki and the Guidelines of Good Clinical Practice. The trial was sponsored by Roche Pharmaceuticals.

The outline of the trial, inclusion and exclusion criteria, as well as survival endpoints have already been previously reported.² GLARIUS recruited adult patients with a newly diagnosed, MGMT promoter nonmethylated GBM and a Karnofsky performance score (KPS) of 70% or higher. All patients received involved-field radiotherapy (30 × 2 Gy) and were randomized 2:1 into 2 treatment arms. The experimental arm received BEV/IRI combination therapy (bevacizumab 10 mg/kg q14d; irinotecan 125 mg/m² or 340 mg/m² q14d in case of intake of enzyme-inducing anti-epileptic drugs) until progression. IRI could be discontinued in case of grade 3/4 gastrointestinal toxicity. The standard arm received concomitant (75 mg/m² daily) and 6 courses of adjuvant TMZ (150–200 mg/m²/d for 5 days q28d).⁶ The trial included an optional predefined crossover at recurrence: TMZ-treated patients could receive second-line BEV/IRI provided by the sponsor. The choice of second-line therapy for BEV/IRI-treated patients was left to the discretion of the local investigator.

Assessments and Statistical Evaluation of QoL

Quality of life was determined using the European Organisation for Research and Treatment of Cancer (EORTC) 30-item core QoL questionnaire (QLQ-C30) and 20-item Brain Neoplasm questionnaire (QLQ-BN20).^7–9 These are validated questionnaires provided on paper that are completed by patients without assistance prior to meeting the physician and discussing examination results, treatments, and perspectives. Each patient had to complete the questionnaires at baseline and at least every 3 months throughout the course of the disease. The results were transformed to a 0- to 100-point scale according to algorithms detailed in a scoring manual supplied by the EORTC,¹⁰ and summarized in 26 dimensions (6 functional, 20 symptom scores). Higher functional scores are favorable, while higher symptom scores are unfavorable.

The longitudinal development of each dimension compared with baseline was analyzed using a model with generalized estimating equations (GEE) with independence working correlation with weights to account for missing data.¹¹ Additionally, the time to first deterioration of each patient in each domain was determined and Kaplan–Meier plots (log-rank test for significance) were calculated for each arm. No correction for multiple testing was applied due to the explorative nature of the analyses. Similar to AVAglio,⁵ a 10-point deterioration (as seen from the patient’s perspective, ie, higher values on symptom scores, lower values on functional scores) on the 100-point scale was regarded as relevant. Deterioration was only recorded if the QoL change was not reverted with the next measurement to a value of overall less than 10 points in comparison with the baseline value. Death from any cause but not tumor progression was regarded as an event. Similar analyses were performed for the time to postprogression deterioration. Values of QoL measurement obtained up to day 30 after progression¹² were taken as baseline. Only patients having a second measurement after progression entered analysis. Effect sizes were calculated by using the Cox regression analysis. Finally, the timepoint of first deterioration of QoL was set in relationship to the timepoint of progression¹² for each patient. Differences between the arms in the rate of patients experiencing QoL deterioration prior to tumor progression versus after tumor progression were analyzed using contingency tables and a chi-square test.

For the figure arrangements, we selected a total of 6 domains for preferred presentation. The domains of physical functioning, social functioning, global health status, motor dysfunction, and communication deficit were selected to offer an easy comparison with the results of AVAglio, where the preferred presentation of these items was already prespecified in the protocol. The sixth item (cognitive functioning) was also selected for preferred presentation, since the question of BEV-induced cognitive changes is one of the main discussion points between AVAglio and RTOG0825.

Evaluation of KPS, MMSE, and Steroid Use

At the same timepoints that the EORTC QLQ questionnaires had to be completed, further parameters which may be associated with general physical and cognitive functioning were recorded: KPS and Mini Mental State Examination (MMSE). The KPS, which grades performance of activities of daily life on a 100-point scale, was recorded by the treating physician or the study nurse. The MMSE is a composite test for different aspects of neurocognitive functioning¹³; the results of the MMSE are graded on a 30-point scale. As for the QoL dimensions described above, KPS and MMSE were analyzed using a GEE model for changes from baseline, a time to first deterioration, and a time to postprogression deterioration analysis. Persisting drops of 20 points or more on the KPS scale and 3 points on the MMSE scale were regarded as significant deterioration and therefore as an event for Kaplan–Meier analysis. For both analyses, progressive disease according to Macdonald and colleagues¹² was not regarded as an event.

Additionally, the steroid use which may be associated with QoL was recorded. The group of patients not having a significant amount of steroids (<2 mg dexamethasone equivalent within 5 days of baseline examination) at baseline was analyzed for the parameter “time to (re)initiation of steroid medication” using Kaplan–Meier statistics. Differences between the treatment arms were analyzed using the log-rank test. The analysis was performed irrespective of the detection of progressive disease.

Results

Patients

The GLARIUS trial included 170 MGMT-nonmethylated patients in the modified intention-to-treat population (116 in the experimental arm, 54 in the standard arm) recruited in 22 German centers. The CONSORT diagram and patients’ characteristics have already been published.² Briefly, the median age was 56 years (range, 25–78) and about 50% of patients in both treatment arms had a macroscopic complete resection according to early postoperative MRI.

As already reported in the previous, survival-focused report of GLARIUS,² the rate of surviving patients providing QoL data was higher than 70% and similar in both arms during the first 9 months after randomization (availability rates experimental vs standard arm at baseline 96%/94%, 3 mo 92%/94%, 6 mo 87%/94%, and 9 mo 76%/72%). Starting at 12 months after randomization, the availability rate of QoL data was lower in the experimental arm (availability rates experimental vs standard arm at 12 mo 66%/79%, at 15 mo 59%/83%, at 18 mo 46%/68%, and at 21 mo 47%/84%. For details, see Supplementary Table S7 of Herrlinger and colleagues²).

Development of QoL over Time

The development of all 26 domains of QoL over the whole course of the disease is shown in Fig. 1 and Supplementary Figure S1. Overt differences between the 2 treatment arms could not be identified.

Fig. 1

Development over time in 6 selected domains of QoL according to the EORTC QLQ C30 and BN20 questionnaires. For functional scales a high number means high functionality, for symptom scales high numbers mean a high symptom burden. Median PFS was 5.99 months in the standard arm and 9.7 months in the experimental arm.

Open in new tabDownload slide

QoL was additionally analyzed for differences in the development over time concentrating solely on values obtained prior to the individual timepoint of progression (“preprogression QoL”; see Fig. 2A for selected domains and Supplementary Figure S2 for the other 24 domains of QoL). No overt differences between the treatment arms could be observed. The same parameters were also analyzed purely with individual values obtained after the individual timepoint of progression (“postprogression QoL”; see Fig. 2B for selected domains and Supplementary Figure S3 for the other 24 domains of QoL). In the postprogression analysis, although there are no marked differences between the treatment arms, some of the domains summarized in Fig. 2B and Supplementary Figure S3 tended to show inferior scores in the experimental arm at month 3 (global health status) or at month 6 (physical functioning, KPS) after progression. Of note, the patients in the tentatively inferior experimental arm mostly had second-line TMZ therapy, while the patients in the standard arm mostly had crossover BEV/(IRI). Forty-four of 54 patients in the standard arm received additional therapies after study treatment. Of them, 36 patients received BEV and 22 received IRI.

Fig. 2

Development over time in KPS, MMSE, and 2 selected domains of QoL according to the EORTC QLQ C30 and BN20 questionnaires. For functional scales a high number means high functionality, for symptom scales high numbers mean a high symptom burden. Median PFS was 5.99 months in the standard arm and 9.7 months in the experimental arm. (A) QoL, KPS, and MMSE prior to tumor progression. (B) Postprogression QoL, KPS, and MMSE.

Open in new tabDownload slide

Table 1 summarizes the results of the GEE analysis. In the experimental arm, 10/20 symptom scores and 2/6 functional scores (role and social functioning) significantly worsened (ie, increase in symptom scores, decrease in functional scores) over time. In the standard arm, 7/20 symptom scores and 2/6 functional scores (physical and emotional functioning) significantly worsened. When the development of scores over time are compared between both treatment arms, there are no statistically significant differences except for the domain “diarrhea” and at borderline for “nausea/vomiting,” which, as typical side effects of IRI, developed significantly worse with BEV/IRI.

Time to First Deterioration of QoL

GLARIUS also evaluated the time to first deterioration of QoL for each of the QoL domains. In all QoL dimensions of the EORTC C30 and BN20 questionnaires, there were no significant differences regarding time to first deterioration (ie, 10-point increase in symptom scores or decrease in functional scores) between the 2 treatment arms (Fig. 3 and Supplementary Figure S4). However, borderline significance (P = 0.05) was found for the time to deterioration of social functioning and nausea/vomiting in favor of the standard arm (Fig. 3).

Fig. 3

Kaplan–Meier plot of time to first deterioration irrespective of disease progression. Deterioration was assumed if symptom scores increased by 10 points or more or functional scores decreased by 10 points or more. Median PFS was 5.99 months in the standard arm and 9.7 months in the experimental arm. (A) Six selected domains of QoL. (B) Irinotecan-induced gastrointestinal toxicity affecting QoL.

Open in new tabDownload slide

Since the timepoints of first QoL deterioration and MRI-based progression are recorded independently, GLARIUS also analyzed whether the sequence of these 2 timepoints was different in the 2 treatment arms. The results for 6 selected domains are shown in Table 2: While in the experimental arm the majority of patients showed deterioration of QoL prior to tumor progression, the deterioration of QoL in 4 of 6 domains in the standard arm occurred predominantly after the diagnosis of progression. In these 4 domains (physical functioning, social functioning, global health status, and motor dysfunction) the differences between the treatment arms were statistically significant (chi-square test, P < 0.05).

Postprogression Deterioration of QoL

Upon tumor progression, 75 of 87 patients (86%) in the experimental arm receiving any second-line therapy after BEV/IRI were switched to TMZ and 36 of 44 patients (82%) in the standard arm receiving any second-line therapy were switched to BEV/(IRI). In 101 patients, data on the development of QoL after progression (under these second-line therapies) are available for analysis of time to postprogression deterioration of QoL (ie, 10-point increase in symptom scores or decrease in functional scores). The time to postprogression deterioration in motor dysfunction (shown in Fig. 4) and headaches (Supplementary Table S1) was prolonged in the standard arm predominantly receiving BEV at this time. The hazard ratios (HRs) for deterioration of headaches (HR 0.41, 95% CI: 0.25–0.79) and motor dysfunction (HR 0.38, 95% CI: 0.18–0.81) were greatly reduced in the standard arm mainly receiving postprogression BEV therapy. The remaining domains are summarized in Fig. 4 (selected domains) and Supplementary Table S1 (other domains) and did not show significant differences.

Fig. 4

Time to postprogression deterioration of QoL in 6 selected domains. Deterioration was assumed if symptom scores increased by 10 points or more or functional scores decreased by 10 points or more.

Open in new tabDownload slide

Other Parameters with Potential Influence on Quality of Life: KPS, MMSE, and Steroid Use

The development of KPS and MMSE over time has already been reported.² Briefly, no differences between the 2 treatment arms were found. Similar to the analyses of QoL, the time to first deterioration and time to postprogression deterioration of KPS and MMSE were analyzed. As shown in Supplementary Figure S5a for KPS and in Supplementary Figure S5b for the MMSE, none of the analyses detected significant differences between the experimental arm and the standard arm.

Steroid use, which may interfere with QoL, was also analyzed. In the group of patients not receiving steroids at baseline, the time to (re)initiation of steroids was significantly longer in the experimental arm than in the standard arm. The median time was prolonged from 12 to 15.8 months (P = 0.027, log-rank test; Supplementary Figure S5c).

Discussion

The GLARIUS QoL data show that primary BEV therapy neither impairs nor improves QoL compared with standard TMZ therapy. Beyond that, postprogression time to QoL deterioration, however, was prolonged in the standard arm for the motor dysfunction and headache domains, an effect which could be related to a beneficial effect of second-line BEV therapy that most patients received in this arm.

The GLARIUS QoL data have to be compared with data of the 2 larger phase III trials, AVAglio^3,5 and RTOG0825.⁴ Comparability is, however, limited, since GLARIUS exclusively recruited patients with MGMT-nonmethylated tumors, while AVAglio and RTOG0825 recruited irrespective of MGMT status and did not perform QoL subgroup analyses according to MGMT status. GLARIUS has in common with both larger trials that the examination rates in the first progression-free interval of treatment (ie, in the first 9 mo) were high, beyond 70%. Later timepoints cannot be compared, since, in contrast to AVAglio and RTOG0825, GLARIUS also collected QoL data after tumor progression. With a median PFS of about 6 months in the standard arm and 9.7 months in the experimental arm,² the data from timepoints beyond 9 months after randomization are in the vast majority provided by patients having already experienced tumor progression and receiving second-line therapy. The GLARIUS data confirm the AVAglio results that BEV does not have a negative impact on health-related QoL.^3,5 However, the observation of the AVAglio trial that the time to deterioration of global health status, communication deficits, and social functioning was prolonged by BEV therapy could not be confirmed by GLARIUS. To explain this difference, one has to take into account that AVAglio counted tumor progression itself as an event, while GLARIUS analyzed QoL deterioration irrespective of tumor progression and, most importantly, beyond tumor progression. Of note, the majority of the patients in the standard arm experienced deterioration of QoL only after the diagnosis of tumor progression (Table 2). Therefore, the application of second-line therapy (crossover BEV for most patients of the standard arm) may have had a strong prolonging influence on the time to deterioration in the standard arm.

In contrast to GLARIUS, RTOG0825 observed a BEV-dependent decrease of QoL in some domains (cognitive functioning, motor dysfunction, and communication deficit), mainly at timepoint 34 weeks after randomization.⁴ One could argue that GLARIUS failed to detect the negative impact of BEV on QoL domains, since GLARIUS, in contrast to RTOG0825, recorded QoL data irrespective of tumor progression, so that early tumor progression with associated clinical deterioration as a confounder would be partially responsible for early deterioration of QoL in the standard arm, thus underestimating a potentially negative effect of BEV in the experimental arm in arm-to-arm comparisons. However, there is a strong argument speaking against this notion and in favor of the interpretation that there was truly no detrimental effect of BEV on these QoL domains: Focusing exclusively on patients without tumor progression (similar to RTOG0825), GLARIUS did not see any BEV-associated unfavorable decline in these QoL domains at week 34 (Supplementary Table S2) and no indications for a reduced rate of favorable scores for these domains in the BEV arm compared with the standard arm (Supplementary Table S3).

The results of neurocognitive testing are not comparable between GLARIUS, AVAglio, and RTOG0825. The MMSE results of AVAglio and GLARIUS are similar, with no differences between the treatment arms, but the MMSE is clearly inferior to detect small differences, especially in the most important domains of central executive functions.¹⁴ The test battery applied by RTOG0825 is therefore more sensitive and valid. With the RTOG0825 test battery, BEV-treated patients are underperforming in an oral word association test and the Trail Making test.⁴ However, the constraints mentioned in the previous paragraph regarding the exclusion of patients with tumor progression may also apply to the results of neurocognitive testing.

Regarding postprogression QoL, GLARIUS found a clinically relevant increased time to postprogression deterioration of motor dysfunction and headaches in the standard arm. Since 82% of patients receiving second-line therapy in the standard arm received BEV/(IRI) therapy, the increase in postprogression time to QoL deterioration may be mainly due to an effective second-line BEV therapy (as opposed to presumably ineffective TMZ second-line therapy in the experimental arm). The prolonged time to deterioration of headaches in the (secondarily BEV-treated) standard arm would very well fit in with an antitumoral and/or anti-edematous effect of second-line BEV therapy.

It is not fully clear why BEV has a positive impact on QoL only after progression. Although it is very difficult to dissect the effects of IRI from the effects of BEV, we believe that IRI is not responsible for the differences seen in first-line versus second-line BEV treatment. The main reason for this statement is found in the fact that even in first-line therapy many patients stopped additional IRI therapy early, so that IRI may not have a strong impact on QoL. Also, a differential effect of BEV on newly diagnosed versus recurrent GBM is not very likely a main reason for the observed differences. The main reason for the more apparent beneficial effect may be found in the fact that the standard arm treated with BEV at the time of rapid progression was treated with an active treatment, while for the patients in the experimental arm no active treatment at all was applicable, despite the rapid progression. The lack of an effective treatment in the experimental arm in times of rapid tumor progression may be the driving force for the observed differences. In the standard arm, this gives BEV the opportunity to show its impact on tumor progression–associated QoL. This impact may be particularly prominent with rapidly progressive tumorous lesions usually seen in the relapsed situation compared with the first-line situation characterized by a high percentage of completely resected patients and radiotherapy as an at least partially effective therapy slowing tumor growth and deterioration of QoL in both arms.

Of course, the postprogression results have to be taken with caution, since there is an increasing rate of missing data at timepoints later than 9 months after randomization and the examination rates were higher in the BEV-treated standard arm than in the TMZ-treated experimental arm.² One may speculate whether the reduced rate of examinations in the predominantly TMZ-treated experimental arm could be attributed to the fact that patients with low performance scores are not able to complete the questionnaires. If that is true, the differences in QoL between the 2 treatment arms might be even larger than suggested in our analysis.

In summary, GLARIUS did not find indications for a detrimental effect of BEV in first-line therapy of MGMT-nonmethylated GBM patients. Moreover, GLARIUS provided some indirect corroborative data supporting the notion that BEV may have beneficial effects upon QoL in relapsed GBM favoring a place of BEV in treatment of relapsed GBM.

Supplementary Material

Supplementary material is available at Neuro-Oncology online.

Funding

The trial was funded by Roche Pharmaceuticals.

Acknowledgments

We thank patients and relatives for their participation and support. We are grateful to all local investigators and study nurses in the participating centers as well as the data managers, statisticians, and monitoring personnel of RPS Research. We thank the independent Data Monitoring and Safety Board (Manfred Westphal, Carsten Bokemeyer, Monika Warmuth-Metz, Heinz Schmidberger).

This work was presented in part at the 50th Annual Meeting of the American Society of Clinical Oncology, Chicago, IL, May 30–June 3, 2014.

Conflict of interest statement

Niklas Schäfer received honoraria and travel fees from Roche.

Martin Proescholdt has a consulting role for Roche and a speakers’ bureau for Novocure.

Joachim P. Steinbach received honoraria and travel fees from Roche, Medac, Boehringer, and BMS; consulting roles for Roche, Mundipharma, and BMS; and speakers’ bureau for Medac.

Peter Hau received honoraria from Novocure and travel fees from Roche, has consulting roles for Roche, BMS, and Novocure and speakers’ bureau for Novocure.

Oliver Grauer received honoraria from Roche and Medac and travel fees from Magforce and has a consulting role for BMS.

Roland Goldbrunner received travel fees from Roche, Magforce, BMS and has consulting roles for Magforce and BMS.

Franziska Friedrich received honoraria from Novartis.

Veit Rohde received honoraria and travel fees from BBraun, Storz, and Ulrich and has a consulting role for BBraun and speakers’ bureau for Storz, Ulrich, and Johnson & Johnson.

Florian Ringel has a consulting role for BrainLab; speakers’ bureau for Integra, Signus, Ulrich, BrainLab, Medtronic; and research funding from Medtronic.

Uwe Schlegel received honoraria from Medac and GSK and has consulting roles for Roche and Novocure.

Michael Sabel received honoraria and travel fees from Integra and Depuy, has a speakers’ bureau for Integra, and consulting roles for Integra and Depuy.

Martin Uhl received travel fees from Medac, Novocure, and BMS.

Stefan Grau received honoraria from Roche.

Mathias Hänel received honoraria from Roche and Novartis.

Oliver Schnell received honoraria from Roche and has a consulting role for Novocure.

Dietmar Krex received honoraria from Northwest Biotherapeutics, travel fees from Medac, and has a speakers’ bureau for Novocure and a consulting role for Baxter.

Frederic Mack received honoraria from Roche.

Michael Nießen received travel fees from Roche.

Barbara Leutgeb is employed by F. Hoffmann-La Roche

Horst Urbach received honoraria from Bayer, Bracco, UCB Pharma, and Stryker and has a consulting role for Springer.

Claus Belka received honoraria and travel fees from Merck KGaA and Amgen, received research funding from Merck KGaA, and has speakers’ bureau for Merck KGaA and Amgen and a consulting role for Merck KGaA.

Walter Stummer receivend honoraria and travel fees from Medac, Magforce, Zeiss, and Leica.

Martin Glas received honoraria from Roche, travel fees from Medac, research funding from Novartis, and has consulting roles for Roche, Novartis, and Mundipharma.

Ulrich Herrlinger received honoraria from Roche, Medac, Novocure, Noxxon, Riemser, and BMS; travel fees from Roche and BMS; research funding from Roche; speakers’ bureau for Roche, Medac, and Riemser; and consulting roles for Roche, Novocure, Noxxon, and BMS.

No other authors have disclosures.

References

Author notes