Observational real-life study on regorafenib in recurrent glioblastoma: does dose reduction reduce toxicity while maintaining the efficacy?

In the phase 2 REGOMA trial, regorafenib improved overall survival, as compared with lomustine, in glioblastoma (GBM) patients at first progression after chemoradiation. Recently, some real-life trials showed similar impact on survival but a higher rate of adverse events than in REGOMA, thus raising concerns over tolerability. The aim of this study was to assess the efficacy and tolerability of a lower intensity regorafenib regimen. Regorafenib daily dose was gradually increased from 80 to 160 mg across the first 2 cycles. Progression-free survival (PFS) and overall survival (OS) were defined as time from regorafenib initiation and disease progression or death. Sixty-six GBM patients were included. Median age was 60.0 years. Median PFS and OS following regorafenib were 2.7 and 7.1 months, respectively. Best RANO response to regorafenib were partial response (PR) in 10 (15.1%), stable disease in 17 (25.8%), and progressive disease in 39 (59.1%) patients. Forty-six (69.7%) patients presented adverse events of any grade, and 21 (31.8%) grade 3–4 toxicity. In a multivariable analysis, higher age and absence of MGMTp methylation were significantly associated with poorer disease control after regorafenib. Our study is the largest observational real-life study on the use of regorafenib. Our lower intensity regimen proved as effective as the standard 160 mg daily schedule (mPFS and mOS being 2.7 vs 2.0 months and 7.1 vs 7.4 months in our study vs REGOMA, respectively). Moreover, we observed a higher rate of PRs as compared with REGOMA (15.0% vs 3.0%).


3
Background Glioblastoma (GBM) is the most common primary brain tumour of the adult [1]. To date, the standard treatment of newly diagnosed GBM is maximal surgical resection followed by chemoradiation using temozolomide (TMZ) [2,3]. However, disease progression occurs in almost all patients after first line treatment, and median overall survival (OS) is of the order of 20 months. [4] Since, no highly effective treatment has been found for disease recurrence, the choice of the best second-line therapy remains an open issue [5]. To date, the most employed medical treatments at recurrence consist of nitrosoureabased schedules and / or antiangiogenic therapy. Lomustine, a nitrosourea alkylating agent, has been used as control in several phase 2 and 3 trials [6][7][8][9][10][11][12][13], with a low objective response rate (around 10%), almost exclusively limited to patients with O6-methylguanine DNA methyltransferase promoter (MGMTp) methylation [8,10,12]. Bevacizumab is a recombinant humanised antibody that blocks angiogenesis by inhibiting the vascular endothelial growth factor receptor A (VEGF-A). As GBM is a highly vascularised tumour with increased endothelial proliferation and VEGFR expression [14], bevacizumab has been investigated in several trials in both newly diagnosed and recurrent GBM [15]. Overall, it has been shown to increase progression-free survival (PFS), but not OS [10].
To improve drug tolerability and reduce adverse events, a dose escalation regimen was designed: the starting dose of the first cycle was 80 mg for 2 weeks, then 120 mg for 1 week; the starting dose of the second cycle was 120 mg for 2 weeks, then 160 mg. A daily dose of 160 mg was maintained from the third cycle on, if tolerated.
Adverse events were defined according to Common Toxicity Criteria for Adverse Events Version 5 (CTCAE v5.0) [29]. Dose reduction and/or cycle delay and/or interruption applied to patients with poor tolerability (grade 3-4 toxicity).

Magnetic resonance imaging (MRI) monitoring
MRI was performed every 3 months, or earlier in case of clinical deterioration. Response to regorafenib was evaluated on MRI according to RANO criteria [30]. MRI responses were assessed by neuroradiologists with an expertise in Neuro-Oncology in the participating Institutions.

Endpoints
The co-primary endpoints were PFS and OS, which were defined as time from regorafenib initiation to interruption due to recurrence/unacceptable toxicity (PFS) or death (OS). For patients with stable disease at the time of formal analysis, PFS and OS were measured from regorafenib initiation to the last visit (censoring).
The secondary endpoint was to evaluate drug tolerability.

Statistical analysis
Baseline characteristics of patients were summarised using median and interquartile range (IQR), and percentages and frequencies (n, %). Age at surgery was a surrogate of age at diagnosis. The distribution of characteristics between patient subgroups were evaluated by the Mann-Whitney U test for continuous variables and the Chi-square test or Fisher's exact test for categorical variables. Kaplan-Meier curves were drawn for PFS and OS and a Cox proportional hazard model was employed to estimate the crude and the multivariableadjusted hazard ratios (HRs) with 95% confidence intervals (CIs) for predictors of recurrence or survival.
The analysis was performed by IBM SPSS Statistics v.28 software.

Regorafenib regimen and response to treatment
Regorafenib was prescribed with a mean dose of 93.3 mg/ day during the first cycle, 133.3 mg/day during the second cycle, and 160 mg/day from the third cycle. However, as 20/66 (30.3%) patients needed a dose reduction due to poor tolerability, the mean regorafenib dose for the whole cohort from the third cycle was 147.8 mg daily. Overall, best RANO response on gadolinium-enhanced T1 sequences was partial response (PR) in 10 (15.2%), stable disease (SD) in 17 (25.8%), and progression of disease (PD) in 39 (59.1%) patients ( Table 2). All 10 PRs were seen on the first MRI after three months of therapy: at the 6-months MRI that followed, 5/10 patients with initial PRs showed PD and 3/10 patients were stable. For the remaining 2 patients with initial PR it was not possible to assess the duration of the radiological response, as one patient interrupted regorafenib due to the occurrence of adverse events shortly after the 3-months MRI, and one had not yet undergone the 6-months evaluation at the time of formal analysis.
All PRs on gadolinium-enhanced T1 sequences in the enhancing areas displayed a SD in the T2/FLAIR sequences.

Progression-free survival and overall survival following regorafenib
Median time of follow-up was 6.8 months (range 1.0-24.8 months). At the end of the study, 63 (95.5%) patients were out-of-treatment: regorafenib was discontinued due to disease progression in 62 patients (93.9%), and drug-related toxicity in 1 patient (1.5%). The remaining 3 (4.5%) patients were still on treatment at the time of formal analysis. All progressions were local: no patient had multicentric progression or leptomeningeal spread.

Prognostic factors
In a univariate analysis of PFS following regorafenib, MGMTp methylation was the sole factor associated with a reduced risk of progression. In a univariate analysis of OS, MGMTp methylation was significantly associated with a prolonged OS, whereas use of steroids was associated with a shorter OS. However, in a multivariable analysis, MGMTp methylation only retained a prognostic importance on PFS and OS. Noteworthy, in a univariate analysis of survival, reresection before regorafenib initiation did not significantly impact either mPFS (2.3 versus 2.7 months, p = 0.823) or mOS (8.9 versus 6.9 months, p = 0.924), nor did it influence the multivariable analysis (Table 4).
To explain why patients who received a reduced dose of regorafenib had a longer mOS, we explored the characteristics of this subgroup as compared with the other patients (Supplementary Table 1). We found that these patients had a higher prevalence of MGMTp methylation (13/20, 65.0%, versus 17/46, 37.0%, p = 0.035), underwent a longer period of TMZ chemotherapy before TMZ-failure (median TMZ cycles 8.0 versus 4.5, p = 0.046; time-to-TMZ failure 11.9 months vs 8.0 months, p < 0.001), and displayed a higher rate of objective RANO response to regorafenib (PRs being 6/20, 30.0%, vs 4/46, 8.7%, p = 0.021). Thus, some of these factors (especially the prevalence of MGMTp methylation) might have influenced the longer mPFS of this subgroup of patients. Finally, dose reduction did not retain a prognostic importance in a multivariable analysis on mPFS or mOS (Supplementary Table 2).

Discussion
Patients with GBM at recurrence show a dismal prognosis and scarce response to treatments. REGOMA was the first phase II trial to suggest the superiority of regorafenib over lomustine in terms of mPFS (2.0 vs 1.9 months, with 16.9% vs 8.3% progression-free patients at 6 months, p = 0.022),  and mOS (7.4 vs 5.6 months, with 38.9% vs 15.0% surviving patients at 12 months, p = 0.0009) [12]. Of note, the OS of patients treated with lomustine from REGOMA trial was remarkably short (5.6 months) as compared with that of patients included in the lomustine single-arms of the other randomised controlled trials (mOS ranging from 7.1 months to 10.4 months [6, 10, 31]), which could imply an overestimation of regorafenib efficacy. Aside from REGOMA, few other studies have investigated the efficacy and tolerability of regorafenib at recurrence: to date, an observational real-life study on 54 patients and few smaller retrospective series have been published (Table 5) [12,[23][24][25][26][27][28]. Therefore, we are presenting the largest real-life observational study that has been performed so far.
The baseline characteristics of patients from our cohort were similar to those of patients included in other large studies [12,23]. However, some differences should be mentioned: first, patients from our cohort had a higher median age (60 years) than those from REGOMA (54.8 years) and from the larger observational real-life study (56 years) [23]. Also, patients with MGMTp methylation included in our study (30/66, 45.5%) were slightly less represented than in REGOMA trial (29/59, 49.0%) and the observational reallife study (28/54, 52.8%). [23] As in REGOMA trial, we included glioblastoma patients treated with regorafenib as second-line treatment. We also included a small subgroup (4, 6.1%) of patients with IDH-mutant high-grade astrocytomas, formerly defined as glioblastomas IDH-mutant by the 2016 WHO Classification, which was our reference when the study was started. Likewise, in the REGOMA trial few patients with GBM IDHmutant were included (2/59, 3.4%): however, in this trial the IDH status was unknown in 15/59 (25.4%) patients, which could disclose a significant underestimation of IDH-mutant tumours with more indolent clinical course [12]. Patients with GBM IDH-mutant were 5/54 (9.3%) in the observational real-life trial from Lombardi et al., [23] whereas other retrospective studies included more heterogeneous cohorts of patients with both GBM IDH-wildtype and lower-grade gliomas IDH-mutant. [24][25][26][27][28] In our study, we employed for the first time a dose escalation protocol to improve patient tolerability and compliance. We reported a lower rate of adverse events of any grade (69.7%) according to CTCAE v5.0 as compared with other real-life studies (90.7%-100.0%) [23,24,27]. Similarly, we observed a lower rate of adverse events grade 3-4 (31.8%) compared with other series (53.0-83.3%) [24,27,28]. Furthermore, we reported a reduced incidence of some adverse events that were described as frequent complications in other studies: for example, the incidence of liver enzymes elevation was lower in our series than in REGOMA trial (15.2% vs 48.0%), as well as that of hypertension (12% vs 22%), low platelet count (10.6% vs 20.0%), hypothyroidism (12.1% vs 19.0%), or increased pancreatic enzymes (4.5% vs 20.0%); on the other hand, we reported a higher incidence of fatigue (33.3% vs 24.0%) and a similar incidence of hand-foot syndrome (27.3% vs 22.0%) as compared with REGOMA. Interestingly, we found a significant association between the development of grade 3-4 adverse events and a previous history of temozolomide-related adverse events during the first line of treatment, which may help to identify which patients are at higher risk of toxicity with regorafenib.
What is of critical importance is that dose reduction did not negatively affect treatment efficacy in our cohort. In fact, the mPFS from our study (2.7 months) was similar to that reported in REGOMA trial (2.0 months) and other studies (2.1-3.5 months) [23][24][25][26][27][28]. Noteworthy, mPFS spans in a quite narrow gap (2-3.5 months) across different studies, also depending on the different timing of the MRI monitoring (either 2 or 3 months).
Conversely, OS after regorafenib shows a slight variation across different studies. In our study, the mOS was similar to REGOMA trial (7.1 months vs 7.4 months), and longer than that reported in the retrospective series by Kebir [28]. The shorter survival of patients included in the smaller series was probably a consequence of the use of regorafenib in patients with advanced disease and/or after two or more relapses. Conversely, patients included in the observational real-life study from Lombardi et al. had a longer mOS (10.2 months) as compared with patients included in the other series (including ours) [23]. In this case, the longer survival was partly due to a prevalence of patients with higher performance status who underwent second surgery before starting regorafenib (16/54, 29.6%).
We observed a higher proportion of objective RANO responses as compared with other studies. We reported 10 PRs out of 66 patients (15.1%), whereas CRs and PRs were 1/59 and 2/59 respectively (overall, 5.0%) in REGOMA trial, and PRs were 4/54 (7.4%) in the observational study by Lombardi et al. In our study all PRs were observed within the first 3 cycles of treatment, and were rarely maintained over time. In the next future, the use of advanced MRI (with integration of perfusion and diffusion-weighted imaging with ADC) and/or amino acidic PET will help to identify cases of pseudoresponse, as some initial data have suggested. [25,32] Whether some biological factors are associated with a higher benefit from regorafenib is under investigation. To date, some molecular markers have been proposed as predictors of response to regorafenib in two post-hoc analyses on patients from the REGOMA trial. First, a mini-signature of 2 gene transcripts (HIF1A, CDKN1A) and 3 miRNAs (miR-3607-3p, miR-301a-3p, miR-93-5p) was associated with a prolonged survival after regorafenib [33]. Second, the phosphorylation of the acetyl-CoA carboxylase (pACC), a surrogate of the activation of the AMPK pathway, has been suggested to prolong the mOS of patients on regorafenib, but not lomustine (mOS 9.3 months vs 5.5 months, p = 0.0013) [34]. These preliminary data are of interest, as the identification of molecular predictors of response may help to select patients who can benefit from regorafenib.
Finally, the AGILE trial, which is currently investigating the role of regorafenib in patients with newly diagnosed GBM without MGMTp methylation, will help to clarify some open issues by collecting prospective data from a controlled population of patients. [35]

Conclusion
Our study is the largest observational real-life study on the use of regorafenib in a cohort of GBM patients at first recurrence. We employed an escalation dose protocol that was useful to reduce regorafenib-related toxicity and improve compliance, while proving as effective as the standard schedule.
Long-lasting responses to regorafenib are rare, and probably limited to patients with peculiar clinical and molecular features. Also, to define criteria for pseudoresponse or pseudoprogression following regorafenib is needed to better evaluate the MRI response.
The ongoing AGILE trial will address many open issues in a larger prospective cohort of patients.