An exploratory prospective phase II study of preoperative neoadjuvant bevacizumab and temozolomide for newly diagnosed glioblastoma

Enrollment was conducted at three university hospitals between January 2017 and November 2021, with the goal of enrolling 15 patients. To avoid possible incorrect diagnosis due to lack of histological verification, patients with lesions with ring-like enhancement on contrast administration with extensive perifocal edema were selected. Selection and exclusion criteria are shown in Table 1.

All patients provided written informed consent. The protocol was approved by the ethics committees, and the study was conducted in accordance with the Declaration of Helsinki.

Patients were treated preoperatively with Bev at a dose of 10 mg/kg on day 0 and TMZ at a dose of 150 mg/m² on days 1–5. Surgical resection was performed between days 21 and 30 with careful evaluation of the extent of tumor bulk based on preoperative MRI including T1-weighted MRI with contrast-enhancement (T1CE) and diffusion-weighted imaging (DWI). Following resection, radiotherapy and concurrent and adjuvant TMZ were prescribed according to the published protocol no later than 4 weeks after resection [8]. Radiotherapy was planned as per daily practice of each institution based on baseline MRI before preoperative Bev and TMZ. Adjuvant TMZ was intended to be administered for more than five courses. The extent of resection was evaluated by T1-weighted MRI with contrast-enhancement (T1CE) within 72 h after resection, and classified as gross total resection (GTR; with no residual tumors), subtotal resection (STR; removal of ≥ 90%), or partial resection (PR; removal of < 90%). After a case of surgical wound infection (Case 2), the protocol was amended with the addition of the following instructions regarding wound handling; (1) subdermal sutures should be firmly placed; (2) suture removal is to be performed ≥ 14 days after surgery; and (3) use of carmustine wafer is to be withheld during the resection surgery.

Brain MRI was performed at the prespecified timing (Table 1 and Supplementary Fig. 1). MRI sequence included T1CE, T2-weighted imaging, fluid-attenuated inversion recovery (FLAIR), and DWI. In addition, apparent diffusion coefficients (ADCs) were evaluated in all cases by verification of region of interests encompassing the entire tumor volume, whereas perfusion MRI was performed in several cases. Karnofsky Performance Scale (KPS) and the Mini-Mental State Examination (MMSE) were evaluated before and after Bev administration (Table 1).

The primary endpoint was progression-free survival (PFS) as assessed by Response Assessment in Neuro-Oncology criteria [9]. Secondary endpoints were safety, response to preoperative Bev and TMZ, and 2-year overall survival (OS). Subsidiary endpoints included changes in KPS and MMSE by preoperative Bev and TMZ. Survival was calculated from day 0 of preoperative Bev and TMZ. All enrolled patients were followed until death or 2 years after treatment initiation. Response was assessed by changes in tumor volume as the sum of products of perpendicular diameters of all slices of all measurable lesions on T1CE and FLAIR images. Adverse events (AEs) were evaluated using CTCAE version 4.0. The initial safety evaluation was made for the initial 7 cases. For these 7 cases, AEs within 3 months of resection surgery were evaluated, with the study to be stopped and not proceed beyond Case 7 in the event of more than two non-hematological severe AEs (SAEs) possibly related to preoperative Bev and TMZ, or multiple cases that did not undergo resection surgery during the prespecified period due to reasons possibly related to the preoperative Bev and TMZ. Efficacy was evaluated in total of 15 cases by intention to treat. Both of safety and efficacy was to be evaluated by the investigators and a data and safety monitoring committee (DSMC). PFS was compared with representative historical data for newly diagnosed GB including EORTC26981/22981/NCIC CE.3, and JCOG0911 [8‐10]. Associations between response to preoperative Bev and TMZ and patient survival were also investigated on an exploratory basis according to T1CE, FLAIR and ADCs for each.

Continuous variables are provided as the mean ± standard deviation or median, and categorical variables as numbers and percentages. The paired t test was used for comparisons of continuous variables between groups. The log-rank test was used to compare survival differences following the Kaplan–Meier method. All p-values were two-sided with the significance level set to < 0.05. Statistical analyses were performed using STATA 14 software (Stata Corp LP, College Station, TX).

Patient characteristics are summarized in Table 2 and Supplementary Table 1. Mean age was 62.2 years (range, 36–74 years). All tumors showed strong, ring-like enhancement on T1CE with extensive perifocal edema in accordance with the eligibility criteria (Supplementary Fig. 1). Pathological diagnoses of tumors resected following the single dose of Bev and 5 days of TMZ were GB, IDH-wild type in 14 cases and GB, IDH-mutant in 1 case, according to the WHO 2016 criteria [11]. Mean MIB-1 index was 31.6% (range, 10–80%) (Table 2).

Safety evaluations for the initial 7 cases demonstrated 1 non-hematological SAE within 3 months of resection surgery possibly related to the preoperative therapy, as grade 3 wound infection in Case 2 in which a carmustine wafer had been placed. Neither cancelation nor postponement of resection surgery was required for any of the 7 cases and resection surgery was performed as scheduled between days 21 and 30. The DSMC therefore judged the study protocol as tolerable and approved enrollment of the additional 8 cases with a protocol amendment regarding wound handling (see the Methods section).

In all 15 cases, tumor resection was performed between days 21 and 30 (mean, day 23.7; range, day 21–29). Tumors appeared light brownish in coloration, with less vascularity and improved brain swelling, rather than the greyish to brownish with prominent hypervascularity seen in usual glioblastoma surgery. Extent of resection was GTR in 11 cases, STR in 3, and PR in 1. Although no difficulty with hemostasis was encountered during resection, postoperative hemorrhage in the resection cavity was identified in Case 8.

Intraoperative fluorescence diagnosis with 5-ALA was used in all patients and was effective, suggesting that protoporphyrin IX fluorescence was undiminished by the preoperative therapy (Fig. 3I).

One patient underwent trephination surgery under local anesthesia, since the intensive care unit was unavailable due to the COVID-19 pandemic (Fig. 3).

Radiotherapy and concurrent TMZ was commenced after suture removal, and no later than 4 weeks after resection.

In comparison between MRI within 14 days before registration (baseline) and within 7 days before resection, tumor volume was decreased by preoperative therapy in all but one case (Case 7) on T1CE, and in all cases on FLAIR (Fig. 1A, Supplementary Fig. 2). Mean tumor volume decrease rates on T1CE and FLAIR were − 36.2% and − 54.0%, respectively. Importantly, the decrease rate on T1CE did not correlate with that on FLAIR. Mean ADCs before and after preoperative therapy were similar, and no obvious trends toward changes in ADCs were noted with therapy (Supplementary Table 1).

KPS significantly improved with preoperative therapy, with a mean pre-therapy (within 5 days before registration) score of 83.3 and a mean post-therapy (within 7 days before resection) score of 94.7 (Fig. 1B; p < 0.0024; paired-t test). MMSE score also significantly improved after preoperative therapy, with a pre-therapy mean of 21.3 and a post-therapy mean of 27.5 (Fig. 1C; p < 0.0026; paired-t test).

Median PFS and OS in the present cohort were 9.5 months and 16.5 months, respectively. The 2-year OS rate was 20%. To determine whether any correlation existed between radiological response and patient survival, PFS and OS were compared between good responders (G-Res) and poor responders (P-Res) on each of T1CE and FLAIR images. G-Res and P-Res on T1CE were defined as showing radiological response rates of ≥ 40% and < 40% on T1CE based on the median response rate. In the same way, G-Res and P-Res on FLAIR were defined as radiological response rate of ≥ 55% and < 55% on FLAIR, respectively. Median PFS of G-Res and P-Res on T1CE were 11.0 months and 7.7 months, respectively (p = 0.196) (Fig. 2A). Median OS of G-Res and P-Res on T1CE were 19.8 months and 12.9 months, respectively (p = 0.0667) (Fig. 2B). In contrast, median PFS of G-Res and P-Res on FLAIR were 9.4 months and 9.6 months, respectively (p = 0.950) (Fig. 2C), and median OS were 16.0 months and 17.0 months, respectively (p = 0.963) (Fig. 2D).

Three SAEs possibly related to preoperative Bev and TMZ were observed in 2 patients, comprising wound infection in one, postoperative hematoma in the resection cavity and thrombocytopenia in the other. In a patient with postoperative wound infection (Case 2), a carmustine wafer had been implanted in the resection cavity. Actually, the protocol was amended after this SAE with the addition of the instructions regarding wound handling (see Treatment in Methods). In the patient with a multifocal deep-seated tumor, postoperative hematoma in the resection cavity occurred immediately after partial resection performed on day 27 (Case 8). The hematoma was removed on the day of resection. The platelet count in this patient was 228,000/μL before administration of Bev and TMZ, 132,000/μL before resection surgery, and 30,000/μL immediately before surgery for hematoma removal. Fortunately, the patient fully recovered to her preoperative status immediately. The postoperative hemorrhage was likely attributable to residual tumor and insufficient hemostasis, and thrombocytopenia was considered mostly attributable to consumption of platelets during the resection surgery. However, the DSMC pointed out that the decrease in platelet count before resection could have been associated with preoperative TMZ.

The Food and Drug Administration in the United States provides a warning that use of Bev should be withheld within 28 days before and after major surgery to avoid wound-healing complications (WHC). Indeed, perioperative use of Bev could be associated with an increased risk of WHC, by the nature of inhibiting the angiogenesis essential to wound healing [12‐15]. However, we have not previously experienced any WHC or other complications in 8 cases where tumor resection was performed within 28 days of Bev administration [7, 16]. We therefore considered preoperative use of Bev in the neoadjuvant setting as feasible. Given the half-life of Bev and postoperative peak level of VEGF [17], we proposed that resection surgery might be best performed 3–4 weeks after Bev administration to maximize the benefits and minimize the risks. TMZ was included to minimize the risk of tumor extension due to insufficient efficacy of Bev during the wait for surgery.

The primary aim of neoadjuvant therapy is to reduce the size or extent of the tumor before surgery, to decrease the difficulty and morbidity of procedures and increase the probability of success. For example, the utility of alkylating agent chemotherapy for oligodendrogliomas in the neoadjuvant setting has been reported [18, 19]. Given the highest expression of VEGF in GB, preoperative use of Bev is expected to reduce tumor volume, tumor vascularity, and peritumoral edema, possibly leading to safer surgery with a reduced risk. Indeed, most tumors in the present study appeared less vascular and brain swelling was improved, rather than the prominent hypervascularity usually seen for GB.

Moreover, the reductions of tumor volume and perifocal edema by Bev could improve the preoperative PS of patients, as the most important prognostic factor in GB patients [20]. Indeed, both KPS and MMSE scores in the present study were significantly improved by a single injection of preoperative Bev, which was shown to contribute to a reduced surgical burden and improved preoperative PS in patients. As a result, preoperative Bev enabled awake surgery in 2 cases due to improvements in language function or limb weakness in a short period.

At the planning of the study, we considered possible increases in WHC, perioperative hemorrhagic events, and thromboembolic events. Among the 15 cases, we encountered 1 case with CTCAE grade 3 wound dehiscence and infection, and 1 case with postoperative hematoma. In both cases, the patient fortunately recovered to their full preoperative status by salvage surgery. The former adverse event occurred in the case for which carmustine wafers were placed during tumor resection, and was likely associated with both Bev and carmustine wafer. Indeed, WHC was not seen in any of the 8 cases in our previous experience in which carmustine wafers were not placed, and in any of the other 14 cases in this study with the protocol amendment prohibiting the use of the wafer[7]. This latter event might have been associated with decreased platelets due to preoperative TMZ. Importantly, we have not encountered any preoperative hemorrhagic events and have not realized any problems in hemostasis during resection [7]. Moreover, the protoporphyrin IX fluorescence by 5-aminolevlinic acid was equivalent to that in usual practice. We therefore consider that preoperative Bev and subsequent tumor resection can be safely performed with careful wound handling.

In the present study, tumor enlargement during the waiting period was noted on T1CE in one case (Case 7). Given the decrease in tumor volume on FLAIR and weakened contrast enhancement on T1CE, some benefits were obtained with reduced vascularity and brain swelling. However, the possibility of ineffective therapy should be kept in mind.

A lack of histological verification can lead to incorrect diagnosis and subsequent ineffective therapy. In the present study, preoperative MRI findings with ring-like enhancement and extensive perifocal edema were among the eligibility criteria. In addition to this criterion, careful exclusion of other pathological conditions would minimize the risk of incorrect diagnosis. In fact, in the present study, all 15 tumors enrolled based on imaging findings were GB according to the WHO 2016 criteria.

Anti-VEGF therapies are known to often weaken contrast enhancement accompanied by decreases in hyperintensity volume on T2WI/FLAIR due to reduced vascular permeability. In the present study, the volume reduction rates on T1CE and FLAIR by preoperative Bev were not correlated, and the larger reduction rate on T1CE rather than FLAIR showed a tendency toward more prolonged OS. Because the volume reduction of FLAIR abnormality is likely to be susceptible to pseudoresponse, our results might provide evidence that tumor volume on T1CE is more reliable indicator of response to Bev that could be associated with patient OS, than that on FLAIR. Indeed, previous reports have supported our results, demonstrating that early progression on T1CE but not FLAIR after Bev-containing chemoradiotherapy offers a prognostic indicator for OS in recurrent and newly diagnosed GB [21, 22].

Previous reports have suggested decreases in ADC, relative cerebral blood volume, and uptake on positron emission tomography with [¹⁸F]-fluoromisonidazole after Bev administration as predictors of good response to Bev [23]. In the present study, consistent tendencies in ADCs were not seen before and after neoadjuvant Bev. Some cases showed an apparent decrease in TBV after preoperative Bev.

As mentioned earlier, preoperative Bev would help decrease the difficulty and morbidity of tumor resection, not only by reducing tumor vascularity and edema, but also by improving KPS. We expect that preoperative Bev might prove particularly useful in the following situations: a bulky hypervascular tumor for which the feeding arteries cannot be intercepted during the early phase of surgery, such as tumor encasing sylvian middle cerebral arteries, and a hypervascular tumor with extensive edema in the eloquent area, for which preoperative Bev might provide conditions allowing more accurate functional monitoring. Moreover, preoperative Bev might be beneficial for improving the PS of patients suffering from early clinical deterioration before initial resection.

On the other hand, the initial response to one-shot Bev thus did not translate into patient survival in this exploratory phase II study. In this regard, the concerns might be that the tumor invading front does not withdraw despite the initial volume reduction on T1CE. Nonetheless, preoperative Bev might provide a greater likelihood of supratotal resection or FLAIRrectomy of GBs, and thus could contribute to better local control in those tumors [24, 25]. Further studies including preoperative Bev and subsequent FLAIRectomy might be warranted.