Resection and survival data from a clinical trial of glioblastoma multiforme‐specific IRDye800‐BBN fluorescence‐guided surgery

Abstract Supra‐maximum surgical tumor resection without neurological damage is highly valuable for treatment and prognosis of patients with glioblastoma multiforme (GBM). We developed a GBM‐specific fluorescence probe using IRDye800CW (peak absorption/emission, 778/795 nm) and bombesin (BBN), which (IRDye800‐BBN) targets the gastrin‐releasing peptide receptor, and evaluated the image‐guided resection efficiency, sensitivity, specificity, and survivability. Twenty‐nine patients with newly diagnosed GBM were enrolled. Sixteen hours preoperatively, IRDye800‐BBN (1 mg in 20 ml sterile water) was intravenously administered. A customized fluorescence surgical navigation system was used intraoperatively. Postoperatively, enhanced magnetic resonance images were used to assess the residual tumor volume, calculate the resection extent, and confirm whether complete resection was achieved. Tumor tissues and nonfluorescent brain tissue in adjacent noneloquent boundary areas were harvested and assessed for diagnostic accuracy. Complete resection was achieved in 82.76% of patients. The median extent of resection was 100% (range, 90.6–100%). Eighty‐nine samples were harvested, including 70 fluorescence‐positive and 19 fluorescence‐negative samples. The sensitivity and specificity of IRDye800‐BBN were 94.44% (95% CI, 85.65–98.21%) and 88.24% (95% CI, 62.25–97.94%), respectively. Twenty‐five patients were followed up (median, 13.5 [3.1–36.0] months), and 14 had died. The mean preoperative and immediate and 6‐month postoperative Karnofsky performance scores were 77.9 ± 11.8, 71.3 ± 19.2, and 82.6 ± 14.7, respectively. The median overall and progression‐free survival were 23.1 and 14.1 months, respectively. In conclusion, GBM‐specific fluorescent IRDye800‐BBN can help neurosurgeons identify the tumor boundary with sensitivity and specificity, and may improve survival outcomes.


| INTRODUCTION
Glioblastoma multiforme (GBM) is considered a grade IV malignant glioma according to the World Health Organization (WHO) and is among the malignancies with the worst prognoses. No cure is presently available, and affected patients have a median overall survival (OS) of 14.6 months after treatment with a standard combination of surgery, radiation therapy, and chemotherapy. [1][2][3] A maximum safe tumor resection is the current goal in the treatment of GBM. [4][5][6] However, this goal is extremely difficult to achieve because characteristically, GBM is diffusely infiltrating and highly proliferative and generally does not display a location preference. 7 Additionally, more than 50% of GBM tumors are located near or within the eloquent areas of the brain. 8 Damage to a functionally eloquent area can cause inevitable postoperative neurological deficits, including motor weakness, sensory deficits, language difficulties, and visual deficits. 9 During the past two decades, advanced neurosurgical imaging technologies, such as neuronavigation, 10,11 intraoperative ultrasonography (iUS), 12,13 and intraoperative magnetic resonance imaging (iMRI), 14,15 have been developed to achieve the maximum degree of tumor resection without incurring neurological deficits. Although these technologies have improved the potential to achieve a complete resection of GBM, all are associated with limitations and technical issues. 16 For example, iMRI may cause image distortion and inaccurate target registration, 17 and imaging artifacts that arise during iUS guidance could reduce the tumor detection sensitivity and specificity. 18 Moreover, a residual disease measuring <1 cm in diameter might be missed by iUS. 19 Optical fluorescence imaging is a cost-effective and time-efficient alternative technique that can be used during GBM resection. 5-Aminolevulinic acid (5-ALA), a natural biochemical precursor of hemoglobin, can elicit the synthesis and accumulation of fluorescent porphyrins in malignant glioma tissue. A Phase III trial verified that 5-ALA enables a more complete resection of tumors and has led to improvements in 6-month progression-free survival (PFS) outcomes. 20 Fluorescein sodium (FS) is a fluorophore that, when intravenously injected, accumulates selectively in high-grade glioma (HGG) cells via an altered blood-brain barrier (BBB). A Phase II study reported that FS injection is both safe and feasible and enables a high rate of complete resection. 21 However, both 5-ALA and FS lack active targeting capabilities and are characterized by limited penetration depths, strong background fluorescence, and autofluorescence. Near-infrared fluorescent (NIRF) dyes, which have excitation and emission wavelengths between 700 and 900 nm, could be used to overcome these negative effects. 22,23 Gastrin-releasing peptide receptor (GRPR), also known as bombesin (BBN) receptor subtype II, is overexpressed in multiple tumor types, including glioma. 24 We previously verified the feasibility and safety of BBN conjugated to the NIRF dye IRDye800CW (IRDye800-BBN) during GBM surgery. 25 In this study, we aimed to determine the resection efficiency, sensitivity, specificity, and survivability of IRDye800-BBN-assisted neurosurgery for patients with GBM.

| Patients
Patients with preoperative enhanced MRI imaging and/or pathological evidence of a newly diagnosed GBM considered suitable for surgical removal were recruited at Peking Union Medical College Hospital and Beijing Tiantan Hospital. The exclusion criteria were as follows: preoperative Karnofsky Performance Status (KPS) score < 70; mental disease; severe liver or kidney illness with a serum creatinine concentration >3.0 mg/dl; any liver enzyme level ≥5× above the normal upper limit; a severe allergy to intravenous radiographic contrast; claustrophobia or an inability to accept positron emission tomography (PET)/computed tomography (CT) or PET/MRI scanning; pregnancy or breast feeding; and an inability to voluntarily provide informed consent.
The tumor locations relevant to the eloquent brain areas were categorized as Grades I, II, and III, indicating noneloquent, neareloquent areas, and eloquent areas, respectively. The ethics committee of the Peking Union Medical College Hospital and Beijing Tiantan Hospital approved this study. All included patients provided signed informed consent preoperatively and underwent surgery at Beijing Tiantan Hospital. The clinical trial number of the study is NCT 02910804.

| IRDye800-BBN preparation
Clinical-grade IRDye800-BBN (Figure 1a) was produced using current good manufacturing practices (cGMP). Briefly, 10.5 mg of BBN and 8.0 mg of IRDye800CW (LI-COR Biosciences Inc.) were dissolved in 4 ml of dimethyl formamide (DMF) in a 20 ml glass vial, to which 0.05 ml of N,N-diisopropylethylamine was added. The mixture was stirred at room temperature for 2 hr and monitored via analytical high-performance liquid chromatography (HPLC). Once IRDye800CW was completely consumed, the mixture was diluted with 4 ml of water purified on a C-18 prep-HPLC in two separate injections with a linear gradient starting from 6% A (0.1% TFA in acetonitrile) and 94% B (0.1% TFA in water) for 5 min; this was increased to 65% A in 35 min at a flow rate of 12 ml/min. The fractions containing the desired product were collected, combined, and lyophilized to give 11.4 mg of final product with an 81.4% yield. The purity of the product was >97% by

| Fluorescence imaging system
We developed a customized imaging system (DPM-III-01, Zhuhai Dipu Medical Technology Co., Ltd.) based on the fluorescence properties of IRDye800-BBN. [26][27][28] We designed a new 778 nm laser and re-optimized the optical path and optical elements to capture the emission fluorescence signal at 795 nm with maximum efficiency. The DPM-III-01 system could simultaneously acquire white light images while performing NIR imaging with an overlaid imaging capability. In addition, it can convert NIR images into heat maps.

| Surgical protocol
IRDye800-BBN was infused intravenously at a dose of 1 mg in 20 ml of sterile water and was administered 16 hr before the induction of anesthesia ( Figure 1b). 25 Neuronavigation was allowed only for the incision, bone flap, and cortex incision region, but not while resecting the tumor or dissecting the residual tumor around the tumor cavity.
During GBM resection, the DPM-III-01 system was used to determine the extent of resection. Other techniques, such as iUS and iMRI, were not utilized.
After opening the dura, the DPM-III-01 was used to determine the area and boundary of the tumor (Figure 1c). If IRDye800-BBN fluorescence was present in a safe area, the tumor was resected using a white-light microscope (M205FA, Leica, Germany). The two processes were alternated until wound bed was reached. Finally the tumor cavity was re-examined using the DPM-III-01. If additional fluorescence was detected in a safe area, the tissue was further resected and the specimens from different locations were submitted for neuropathological analysis. When safe and feasible, some biopsies were randomly harvested from nonfluorescent brain tissue in noneloquent areas around the tumor cavity to assess the diagnostic accuracy. To avoid inter-surgeon variability, all the operations were performed by the same neurosurgeon, who had 25 years of experience in brain tumor practice. All biopsies were analyzed according to standard pathological procedures (Figure 1d).

| MRI analysis
All the patients were intravenously injected with 0.1 mmol/kg body weight of Magnevist (Omniscan, GE healthcare) and underwent MRI scans on a 3.0T scanner. The slice thickness was set to 5 mm. Preoperative enhanced MRI imaging was performed within 1 week before surgery. Early postoperative MRI scans were performed within 72 hr postoperatively. All MRI data were analyzed in the Department of Neuroradiology at Beijing Tiantan Hospital.
The tumor volume was defined as a high-signal area after T1-weighted enhancement and was calculated as follows. The tumor contour of each slice was segmented by two experienced neuropathologists using Picture Archiving and Communication Systems (PACS) and was then cross-verified. If discrepancies arose, a consensus opinion was obtained from another higher-level pathologist. The areas on each slice were added, and this sum was multiplied by the thickness.

| Statistical analysis
The signal-to-background ratio (SBR) of the IRDye800-BBN fluorescence region to the peripheral brain parenchyma (PBP) was calculated as follows. ImageJ software (Version: 1.52v, National Institutes of  Table 1.

| Sensitivity and specificity
Eighty-nine samples were harvested at the tumor margin, including 70 fluorescence-positive and 19 fluorescence-negative samples (  Figure 3b). Regarding GBM surgery, increasing evidence indicates that maximal resection could improve the life expectancy of the patient.

| Progression-free survival and OS
However, it is crucial to minimize the risk of perioperative morbidity, especially neurological damage. 16 To circumvent the brain shift and provide the neurosurgeon with surgical information updated provided. Therefore, these conclusions should be interpreted with scientific accuracy and caution.
We acknowledge that the complete resection of a GBM tumor is not always feasible, even when using targeted IRDye800-BBN. Excision should be avoided when a tumor has invaded the functional areas of the brain. However, the FGS method, including 5-ALA and FS, is not sensitive for low-grade gliomas because the BBB remains intact, and it is therefore difficult for contrast agents to bind to tumor tissues. 35 In our present study, all four false negative specimens were pathologically confirmed to be low-grade components of a secondary GBM pathological analysis confirmed that the two false positive specimens were from the reactive gliosis zone where the density of cancer cells was very low. However, the BBB might have been partially damaged. Therefore, IRDye800-BBN was found in tissues where other fluorophores, such as 5-ALA, were also found. Moreover, we found that patients in the IDH1 mutation subgroup had a median OS of 26.9 months, whereas those in the IDH1 wild-type subgroup had a median OS of 13.5 months. Patients in the methylated MGMT subgroup had a median OS of 26.9 months, compared to 12.8 months in the unmethylated MGMT subgroup. However, our analysis revealed no statistical effect of the MGMT promoter (p = .077) or IDH1 gene status (p = .289) on OS, although this may be due to our small sample size. Therefore, the main limitations of this study were the small sample size and lack of a control group with randomization. Given our current research progress, we will perform a randomized controlled trial of this method versus routine clinical operations using white-light microscopy or other fluorescent dyes (e.g., 5-ALA).

| CONCLUSIONS
This study indicated that GBM-specific NIRF IRDye800-bombesin can help neurosurgeons sensitively and specifically identify the tumor boundary for complete resection, which may improve survival outcomes. FGS with targeted contrast agents can also be applied in different diseases to benefit patients.