Dose-dependent efficacy of bevacizumab in recurrent glioblastoma

Bevacizumab (BEV), at a standard dose of 10 mg/kg every 2 weeks is associated with prolonged progression-free survival (PFS) but no improvement in overall survival (OS) in recurrent glioblastoma (rGBM). Few studies have examined the potential dose-dependent efficacy of BEV. In Ontario, reimbursement for the costs of BEV varies, and as a result, our practice began to routinely use lower dose regimens. The main aim of this study was to ensure that there was no harm to patients who received the low dose protocol. A single-center retrospective study of patients given BEV for rGBM between 2015 and 2020 was performed. Clinical and treatment data including BEV dose regimen [SD (10 mg/kg every 2 weeks) vs. LD (5 mg/kg every 2–3 weeks or 10 mg/kg every 3 weeks)] received at the time of rGBM diagnosis were captured. Overall survival (OS) and progression-free survival (PFS) on BEV were compared using the Kaplan–Meier product-limit method. Log-rank test was used to compare potential predictive factors. Cox regression model was performed for multivariable analysis of OS and PFS. A total of 96 patients were included with a median follow-up duration of 6.84 months (range 1.12–50.63 months) from the date of the first infusion. The LD group consisted of 55 of the 96 patients. By virtue of funding mechanisms for BEV, the median age in the LD group was significantly higher (62 vs. 54 years p = 0.009). There was no difference in MGMT status between the two groups (p = 0.60). The LD group had prolonged median PFS (5.89 months versus 3.22 months; p = 0.0112) and OS (10.23 months versus 6.28 months; p = 0.0010). Multivariable analysis including the dose of BEV, the extent of resection, gender, and age revealed that standard dose of BEV, subtotal resection, and female sex were associated with worse overall survival. Nine patients in the SD group vs. 18 patients in the LD group reported an adverse event related to BEV. For patients with recurrent GBM, we found that a low dose regimen of BEV was associated with prolonged OS and PFS compared to the standard dose regimen. Lower dose schedules may be a better and more cost-effective option for patients with rGBM. Lower costs might provide more equitable access to this very important palliative drug.


Introduction
The prognosis for glioblastoma (GBM) remains poor despite advances in our understanding of glioma biology. Despite a large number of clinical trials testing experimental therapies, the standard of care remains surgery, radiation therapy, and temozolomide (TMZ) chemotherapy [1]. There are currently no proven survival-prolonging therapies for recurrent glioblastoma (rGBM) and the median survival after progression is 3-9 months. Bevacizumab is a recombinant humanized monoclonal IgG 1 antibody against VEGF, which is linked to angiogenesis and tumor proliferation in GBM. In 2010, Health Canada approved the use of bevacizumab for rGBM based on promising but preliminary phase II Mary Jane Lim-Fat and James R. Perry contributed equally to this work. data [2]. While bevacizumab can achieve impressive radiographic responses, and provide symptomatic and steroidsparing benefits, improved survival in rGBM has not been demonstrated using bevacizumab alone or in combination with chemotherapy [3,4]. Outcomes in Canadian rGBM patients treated with bevacizumab (alone or with chemotherapy) appear to be consistent with data reported from larger phase II or III clinical trials of bevacizumab in rGBM [5]. Bevacizumab has been increasingly used in Canada, in part because clinical experience suggests that selected patients benefit in terms of symptom palliation and maintenance of the quality of life. However, due to its prohibitive cost, many Canadian patients still do not have access to this agent through provincial healthcare systems, including in Ontario. The approved dose of bevacizumab is 10 mg/ kg (per infusion every 2 weeks), based on the early trials of bevacizumab in GBM, however, there is a lack of doseresponse studies in this population. Our institutional practice has been to administer 10 mg/kg every 2 weeks, but in selfpaying patients who want treatment but are handicapped by the cost, we opted to reduce the regimen to 5 mg/kg every 2-3 weeks. Observational data from retrospective studies using 5-15 mg/kg every 2-3 weeks suggest no difference in progression-free survival (PFS) in rGBM patients [6,7]. In addition, although larger trials are lacking, certain studies seem to suggest that a lower dose of bevacizumab may be related to improved survival [8,9]. In addition to this potentially important finding, the use of lower doses of bevacizumab is compelling as lower costs could make this treatment more widely available for patients who currently lack better alternatives. There is also a potential reduction of side effects and adverse events related to a lower dose of bevacizumab. Finally, the administration of bevacizumab at 10 mg/ kg in conjunction with chemotherapeutic agents has been of limited survival benefit, and salvage chemotherapy after bevacizumab has not been effective [10]. This is thought to be related to the modulation of the blood vessels at this particular dose of bevacizumab, which in turn decreases the tumor penetration of concomitant therapy [11]. It is unclear if lower doses of bevacizumab may allow for more optimized modulation of blood vessels to allow for more effective use of concomitant therapy, which may in turn improve survival.

Study design and participants
This is a single-center retrospective study performed at the Odette cancer center and encompassed a cohort of patients with a diagnosis of rGBM. Clinical charts from all patients with rGBM who received bevacizumab between 2015 and 2020 were reviewed using our institutional electronic medical records (SunnyCare, Connecting Ontario, PRO, and Sovera) and chart abstraction was done by trained clinicians/ abstractors in accordance with institutional guidelines and ethics review board policies. Detailed patient demographics, IDH and MGMT promoter methylation status, treatment modalities (type and extent of resection, type, dose, and cycles of chemo-radiotherapy), adverse event, and outcome data were collected and stored using encrypted computers onto REDCAP, a widely used and secure data platform, and real-time data verification was performed periodically. Specifically, the dose of bevacizumab, number of infusions, and frequency of infusions were captured. A regimen of less than 10 mg/kg every 2 weeks was annotated as being a "low dose (LD)", and a dose of 10 mg/kg every 2 weeks as the "standard dose (SD)".

Definitions and outcomes
An integrated diagnosis adhering to the fifth edition of the WHO Classification of Tumors of the Central Nervous System (WHO CNS5) [12] was assigned in our study. In rare instances where the initial histopathological diagnosis showed an Astrocytoma, IDH-wildtype, WHO Grade 2 or 3 they were grouped with Glioblastoma, IDH-wildtype at the time of recurrence based on the radiological features rather than molecular features. Recurrence or progression definition was based on clinical and/or radiological criteria after the traditional surgical and/or chemo-radiation therapy period. Both focal and non-focal neurological symptoms/ signs that can't be attributed to the traditional therapy side effects or be better explained by an alternative diagnosis were used for the clinical recurrence. Radiologically, the Response Assessment in Neuro-Oncology (RANO) criteria was used to define recurrence and was assessed by an experienced neuro-radiologist [13].
Overall survival (OS) was measured in months from the diagnostic surgery date to the date of death and OS on bevacizumab was measured in months from the first infusion date to the date of death. Progression-free survival (PFS) on bevacizumab was measured in months from the first infusion date to the date of progression on bevacizumab or date of death whichever occurred first. Patients lost to follow-up were censored at their last appointment date.

Statistical analysis
Patient and radiological characteristics were compared across the two treatment groups (low dose vs. standard-dose bevacizumab) using Wilcoxon't sum rank test for continuous measures and the chi-squared and/ or Fisher's exact test, as appropriate, for categorical variables. OS, as well as PFS rates, were calculated using the Kaplan-Meier product-limit method. Log-rank test was used as a univariate analysis to compare levels of patient characteristics and other potential predictive factors. Cox regression model was performed for multivariable analysis for OS and PFS to assess the joint effect of those factors that were found potential in the univariate analysis. All P-values were 2-sided and for the statistical analyses, P < 0.05 was considered to indicate a statistically significant result. Statistical analysis was performed using version 9.4 of the SAS system for Windows, 2002-2012 SAS Institute, Inc., Cary, NC, USA.

Results
Between 2015 and 2020, 96 patients with a diagnosis of recurrent GBM were captured within this single-center cohort. Median follow-up was 20.3 months (range: 2.5-77.1 months) from the time of surgery. Of the 96 patients, 55 (57%) received the LD schedule and 41 (43%) were given the SD schedule. From the first day of BEV administration, the median follow-up was 6.84 months (range: 1.12-50.63 months). The median age of the entire cohort was 57 (range: 19-80 years) with a significantly older age seen in the LD group (median 62 years for LD, versus 54 years for SD, p = 0.009). Extent of resection was not significantly different between the two treatment groups (p = 0.97). There was no statistically significant difference in the number of cycles of TMZ administered in the adjuvant phase (p = 0.11). One patient in the low-dose cohort did not have an IDH status report but was presumed to be IDH-wildtype GBM based on age and radiographic appearance. MGMT status was not known for 58 patients, due to inconclusive results or lack of testing. Of the remaining patients, only 12 (31.58%) patients had a methylated MGMT promoter and 26 (68.42%) were unmethylated and MGMT status was not significantly different between the tow treatment groups (with the proportion of methylated being 34.78% versus 26.67%, p = 0.60) ( Table 1).

Survival outcomes from the start of BEV infusions
From the first BEV infusion, the median PFS and OS were 4.14 months (95% CI 3.25-5.82 months) and 7.96 months (95% CI 6.58-9.9 months), respectively (Fig. 1). The 12-month PFS and OS survival rates were 18.7% and 26.3%, respectively. Stratified by the diagnostic surgery extent of resection, biopsy versus subtotal versus gross total, there was a statistically significant difference in the PFS (p = 0.0037) but not the OS (p = 0.1248). The median PFS was 2.88 months (95% CI 2.04-3.81 months) for a subtotal resection, 5.29 months (95% CI 3.26-8.25 months) for a gross total resection, and 7.4 months (95% CI 3.06-14.17 months) for a biopsy. There was no statistically significant difference in the overall survival when stratified by sex (p = 0.1247).

Survival outcomes for low versus standard dose from the start of BEV infusions
In the SD group, all patients received BEV 10 mg/kg every 2 weeks. In the LD group, 63.6% received 5 mg/kg every 3 weeks, 20% were 10 mg/kg every 3 weeks and 16.4% were 5 mg/kg every 2 weeks. No patients had a cross-over between the standard-and low-dose regimens ( Table 2). The median number of BEV cycles was numerically higher in the SD (median of 4 cycles; range: 0-12) but not statistically different (p = 0.32) from the LD regimen (median of 2 cycles; range 0-12). At the time of the first infusion, the median dexamethasone dose (median of 4 mg per day; range: 0-16) and the proportion of patients with functional independence (65.30% in the LD versus 60% in the SD) were similar between both groups. Bevacizumab salvage therapy was started at an earlier point in the disease course in those receiving the SD (48.78% in SD versus 35.54% in LD at 1st recurrence) ( Table 1). Compared to the SD group those in the LD group had a statistically significant improvement in PFS (p = 0.0112) and OS (p = 0.0010) (Fig. 2).
In the low dose regimen, the median PFS was 5.89 months (95% CI 3.72-7.5 months; 6-month PFS 49.8%) compared to a median of 3.22 months (95% CI 2.27-4.70 months; 6-month PFS 23%) in the standard dose regimen from the time of the first infusion (Fig. 2a). Gross total resection at the time of the diagnostic surgery and low dose of BEV were factors found to have a positive impact on PFS with those receiving a lower dose found to be 2.67-fold less likely to progress during our follow up time (HR = 0.375, 95% CI 0.23-0.61; p < 0.001).
In contrast to the median OS of 6.28 months (95% CI 3.98-7.79 months; 6-month OS 51.7%) seen with the SD regimen, the LD group had a median OS of 10.23 months (95% CI 7.79-11.87 months; 6-month OS 77.3%) resulting in an approximate survival gain of 4 months (Fig. 2b). Multivariable analysis revealed that low dose of BEV,  (Fig. 3). Table 3 summarizes the safety profile of BEV in our patients. Overall, the infusions were well tolerated in both  the LD and SD groups. The most common adverse effect was fatigue followed by arthralgia and hypertension, more commonly encountered with the lower dose of bevacizumab. Serious adverse events necessitating discontinuation of treatment were seen in three patients receiving the LD. These included venous thromboembolism and severe fatigue/ arthralgia while on the 5 mg/kg every 3 weeks regimen and nephrotic syndrome while on the 10 mg/kg every 3 weeks regimen. One patient on SD had asymptomatic thrombocytopenia (nadir at 51,000), however, the reason for treatment discontinuation was lack of bevacizumab efficacy. No Grade 4 or 5 toxicities were encountered.

Discussion
Bevacizumab has been incorporated into the care of IDH-wt GBM patients, particularly in the recurrent setting where salvage therapies are limited. Bevacizumab has never been compared to best supportive care alone but evidence from large phase 2/3 studies [2,4,14,15] has shown an improvement in the median PFS ranging from 3 to 5.6 months with BEV-based regimens. In contrast to this improvement in PFS [16,17], no overall survival benefit has been shown [18]. Our SD cohort had a median PFS of 3.22 months (95% CI 2.27-4.70 months; 6-month PFS 23%) and a median OS of 6.28 months (95% CI 3.98-7.79 months; 6-month OS 51.7%) which is consistent with historical outcomes. Two phase II trials led to an accelerated approval by the U.S. Food and Drug Administration on May 5, 2009, for the use of BEV as a monotherapy in rGBM. Shortly afterward, in 2010, approval was granted by Health Canada [2,15,19]. Both trials assessed the efficacy of BEV standard dose of 10 mg/kg in patients with rGBM who were previously treated with traditional radiotherapy and TMZ chemotherapy. While none of the later studies in rGBM showed an improvement in OS, the 10 mg/kg regimen has since become the standard regimen.
In this retrospective study, we demonstrated that the LD schedule is associated with a statistically significant improvement in both progression-free and overall survival in the SD group. The median overall survival (b) was 10.23 months (95% CI 7.79-11.87 months) in the LD group compared to 6.28 months (95% CI 3.98-7.79 months) in the SD group as compared to the SD schedule of 10 mg/kg given every 2 weeks. Our study demonstrates that there appears to be dose and schedule-dependent efficacy of BEV for recurrent glioblastoma. This finding is compelling for several reasons. While many studies have looked into the role of bevacizumab in rGBM (Table 3), dose-finding studies have generally not been pursued in the rGBM population. In colorectal cancer, alternative schedules of 5 mg/kg every 2 weeks or 7.5 mg/kg every 3 weeks are widely used. Initial reports of responses using dosing schedules as low as 5 mg/kg every 2 weeks encouraged the development of BEV for this indication [20,21]. A single-arm phase 2 study confirmed the antitumor activity of 7.5 mg/kg every 3 weeks rGBM [6]. However, a meta-analysis of 548 patients with rGBM found no difference in the BEV dose-response benefit between 5 mg/kg and 10-15 mg/kg. Intriguingly, a single-center retrospective study of 181 patients compared the use of 10 mg/ kg every 2 weeks and 7.5 mg/kg every 3 weeks and found a trend towards better survival in the lower dose group [9].
Indeed, it was this study by Levin et al. that helped to inform our practice of using lower-dose schedules for self-paying patients with financial hardship. Our main purpose of this study was to see if there was any evidence of worse survival outcomes with the lower dose schedule. Not only did we find no evidence of harm, but our findings also suggest a survival benefit in the lower dose group.
There may be other advantages to an LD schedule beyond survival alone. Although the adverse event rates in our study were low, we wondered if there would be less toxicity with the LD schedule. Our study doesn't have the power to detect differences in these subgroups. On the surface, there appear to be more AEs reported in the LD group; however, there was a very large difference in median age between our study cohorts. Older patients tend not to have third-party prescription insurance and financial toxicity is a concern with expensive anti-cancer agents. In our study, the age bias may have led to the appearance of more adverse effects in the LD group. Age of course is a very powerful predictor of worse survival in GBM, making it all the more remarkable that we were able to detect better survival in the much older low-dose group. Drug costs are a barrier to equity and access in all areas of oncology. Lower dose, less frequently administered drug regimens that are more cost-effective may improve access to these expensive therapeutics.
Our study has several limitations including its retrospective nature, potential bias in patient selection, and relatively small sample size. MGMT methylation status was unavailable for most patients. As 60% of patients are unmethylated and progress sooner, an imbalance in the distribution of methylation status could bias results, although BEV does not have an MGMT-dependent mechanism of action. The majority of our patients in the LD group received 5 mg/kg q 3 weekly, a cost reduction of 66% compared to standard dosing. We were unable to compare the different LD schedules due to the small number of patients in each group. A prospective clinical trial comparing different lower dose regimens with a well-captured clinician and patient-reported outcomes would be needed to better evaluate the potential survival and quality of life benefits of a specific lower dose regimen in rGBM.
Author contributions JP and ML contributed to the study's conception and design. Data collection was performed by IG, EC, and AT. Data analysis was performed by EA. The first draft of the manuscript was written by JM and ML and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Conflict of interest
The authors have no relevant financial or non-financial interests to disclose.
Ethical approval This study was approved by our local REB (SUN-3574) and a waiver for patient consent was obtained due to the retrospective nature of this study .