Introduction

Glioblastoma (GBM) is the most common malignant brain tumor in adults, comprising 15% of all intracranial tumors, 45–50% of all primary brain tumors, and 55–57% of all gliomas [1, 2]. This World Health Organization (WHO) grade IV astrocytic neoplasm is characterized by its aggressive and diffusely infiltrative nature, which portends its poor prognosis [3]. Despite genomic, technologic, therapeutic, and operative advancements in the treatment of GBM, it remains invariably fatal, with an average overall survival (OS) of 15–18 months and just 6.8% of patients surviving 5 years after diagnosis [1, 2, 4, 5]. The role of surgery in the treatment of GBM and other similar-behaving IDH wild-type high-grade gliomas (HGGs) has been extensively studied and refined in recent decades, with particular attention towards so-called “cytoreductive surgery” or improved extent of resection (EOR) [6,7,8,9]. Numerous studies have demonstrated that greater EOR is independently associated with longer progression-free survival (PFS) and OS, symptom palliation, and improved quality of life in patients with both low- and high-grade gliomas [7, 10,11,12,13,14,15,16]. Although much attention has been paid to EOR as an independent prognosticator, several recent studies have argued that contrast-enhancing residual tumor volumes (CE-RTVs) are a more important predictor of survival after HGG resection—because EOR is significantly influenced by preoperative tumor volume [17,18,19,20,21]. Accordingly, several assistive technologies—including intraoperative magnetic resonance imaging (iMRI) and fluorescence-guided resection using 5-aminolevulinic acid—have been developed to safely improve EOR and decrease CE-RTV [22,23,24,25,26,27,28].

Since its introduction in the mid-1990s, iMRI has been demonstrated to maximize EOR and survival while reducing CE-RTV in patients with HGGs when compared with conventional stereotactic navigation alone [24, 26, 27, 29, 30]. Many of these studies have focused primarily on the relationship of survival parameters (i.e., PFS and OS) with EOR or immediate CE-RTV measurements [29, 31,32,33]. We present a single-institution analysis of patients with HGGs who were treated surgically over a 10-year period using iMRI. In this study, we compared pre-, intra-, immediate post-, and three-month postoperative T1-weighted (T1W) and T2-weighted (T2W) imaging volumes and their relationship with PFS and OS. Further, we examine those patients who underwent additional resection after initial iMRI scans, the likelihood of positive pathologic margins, and the relationship of additional resection with PFS and OS.

Methods

Patient population and study design

The study was a retrospective review of a prospectively maintained database of patients with HGG treated surgically using a 3.0T iMRI at the University of Utah and the 1.5T iMRI at the Huntsman Cancer Institute over a 10-year period from December 2011 to October 2021. The initial cohort consisted of 125 total patients with pathology-proven GBM (isocitrate dehydrogenase [IDH] wild-type or IDH mutated) or WHO grade III anaplastic astrocytoma (IDH wild-type only) who underwent resection using iMRI. Exclusion criteria included patients who did not have progression or survival data, IDH1/2 or O6-methylguanine-DNA methyltransferase (MGMT) biomarker data, or those patients whose tumors were determined to be IDH1/2 mutant. Patients were also excluded if they lacked imaging data available at the time points of interest: preoperative MRI with and without gadolinium contrast, iMRI with and without gadolinium contrast, immediate postoperative MRI with and without gadolinium contrast, and three-month postoperative MRI with and without gadolinium contrast. For patients who underwent more than one iMRI scan, the last scan performed was used to obtain the volumetric data for analysis. A total of 24 patients were excluded from the analyses based on these exclusion criteria. The final patient cohort used in the analyses included 101 patients with new or recurrent IDH wild-type HGGs. Patient demographics, progression and survival data, pathological reports with tumor-specific markers, and imaging records were obtained. Although many other patients with HGG were treated surgically during the study’s 10-year duration, a large number were resected using conventional stereotactic navigation and/or 5-aminolevulinic acid fluorescence-guided resection, thus precluding them from the present study cohort.

Data collection

Minimum MRI sequences obtained included axial T1W non-contrast, axial T1W post-contrast, axial T2W, axial T2 fluid-attenuated inversion recovery (FLAIR), and axial diffusion-weighted imaging (DWI) with apparent diffusion coefficient sequences. Preoperative, intraoperative, immediate postoperative (within 24 h of surgery), and three-month postoperative scans were downloaded from the hospital picture archiving and communication system to an OsiriX DICOM viewer (Pixmeo SARL, Bernex, Switzerland). Radiological volume data from each time point of interest included T1W non-contrast hyperintense volumes, T1W contrast-enhancing (CE) tumor volumes, T2W hyperintense volumes, and FLAIR hyperintense volumes.

Clinical variables were collected in prospective fashion, and survival data were updated retrospectively. Clinical variables included age, sex, primary presenting sign/symptom, tumor location (lobe and laterality), date of surgery, date of progression, recurrence status, date of death, preoperative and 1-month postoperative Karnofsky Performance Scale (KPS) scores, whether patients underwent subsequent resection after the initial iMRI scan and whether these residual margins were positive on pathological analysis, and tumor-specific markers (i.e., MGMT, TP53, PTEN, EGFR, and MIB-1).

Imaging and volumetric analysis

Using the OsiriX DICOM viewer, three authors (N.T.G., J.C.H., and R.L.J.) measured the quantitative volumes by using contours of axial T1W or T2W imaging. CE-RTVs were calculated by subtracting T1W non-contrast hyperintense volumes from T1W-CE volumes to subtract out postoperative blood products. Volumetric data for each patient was measured after surgery. The EOR was calculated using the following formula: [preoperative CE tumor volume – CE residual tumor volume / preoperative CE tumor volume] × 100%. The initial iMRI scans were performed only after a gross total resection was achieved intraoperatively using both stereotactic neuronavigation and visual inspection with the microscope and with the attending surgeon’s independent approval. To obtain a more granular understanding of tumor resection for each patient at different timepoints of their care, ratios were calculated for individual patients using postoperative/intraoperative and three-month postoperative/intraoperative T1W-CE, T2W, and FLAIR imaging volumes. The ratios were calculated to better understand relative disease burden for each patient at their various timepoints of treatment and to determine whether there was any relationship of these variables with PFS or OS.

Statistical analysis

All data analyses were performed using Jupyter Notebook software v4.4.0 in Python v3.6.8 programming language (Project Jupyter, open source; Python Software Foundation, Fredericksburg, VA) or using GraphPad Prism v9.4.1 (GraphPad Software, San Diego, CA). Descriptive statistics were obtained to define the patient cohort, including analyses of PFS and OS using T-test or Chi-square test. Patients were separated into groups that were above or equal to and below the median imaging volumes for all survival analyses. Using a log-rank test, imaging volume distributions were analyzed for their relationships with PFS and OS. A log-rank test was performed to examine EOR and its relationship with PFS and OS, as well as the relationship between further resection (after initial iMRI scan) and PFS and OS. Multivariable Cox proportional hazard regression analyses were also carried out for the imaging volumes from T1W-CE, T2W, and FLAIR sequences at the timepoints of interest to assess their relationship with PFS and OS. Lastly, we examined the relationships between PFS and OS and the following imaging volume ratios using log-rank tests: immediate postoperative/intraoperative volumes and three-month postoperative/intraoperative volumes for T1W-CE, T2W, and FLAIR sequences. A two-tailed p < 0.05 was deemed statistically significant in all analyses.

Results

Patient and tumor characteristics

Patient characteristics and survival data for the 101 patients with GBM (IDH wild-type only) or anaplastic astrocytoma (IDH wild-type only) surgically resected using iMRI are shown in Table 1. The median age was 57 years old (range 22.9–85.0 years); younger patients had significantly better PFS and OS (p = 0.015 and p = 0.002, respectively). Although preoperative KPS was not significantly associated with PFS, it was associated with OS (p = 0.04). Postoperative KPS was significantly associated with both PFS and OS (p = 0.0015 and p = 0.0002, respectively). The median PFS for the patient cohort was 7.8 months (range 0.13–101.2 months), and the median OS was 14.0 months (range 0.13–101.2 months) (Fig. 1a). The most common location of tumors was the temporal lobe (31/101; 30.7%), followed by the parietal lobe (28/101; 27.7%) and frontal lobe (27/101; 26.7%). The tumor location was significantly associated with PFS (p < 0.001). Tumors were roughly equal with respect to laterality (right: 52/101; left: 49/101).

Fig. 1
figure 1

(a) Scatter plot showing median values and interquartile range for PFS and OS in the 101-patient cohort with new or recurrent IDH wild-type HGGs. Median values are represented by a dashed line. Error bars represent the interquartile range. Temporal trends of median T1W-CE (b), T2W (c), and FLAIR (d) imaging volumes (cm3). T1W-CE, T2W, and FLAIR tumor volumes from preoperative (red), intraoperative (green), immediate postoperative (blue), and three-month postoperative (purple) MRI scans. The median value is represented by a colored circle. Error bars represent the interquartile range for each of the volumes

Full size image
Table 1 Baseline patient, tumor, and surgical characteristics in 101 patients with new or recurrent HGG who underwent surgical resection using iMRI
Full size table

Tumor and surgery characteristics are shown in Table 1. Of the 101-patient cohort, 42 patients (41.6%) were newly diagnosed, and 59 patients (58.4%) had recurrent disease and had undergone prior maximal safe resections. Recurrence was significantly associated with decreased PFS and OS (p = 0.019 and p = 0.04, respectively). Separate log-rank analyses were performed for newly diagnosed HGGs and recurrent HGGs and are included in Online Resource Table 1 and Online Resource Figs. 1 and 2. All patients were IDH1/2-wild type (101/101; 100%), and the majority of patients were MGMT unmethylated (67/101; 67.7%). MGMT promoter methylation was associated with longer PFS but not OS (p = 0.03 and p = 0.08, respectively). Of patients with available data, 49/81 (60.5%) were EGFR non-amplified, 73/90 (81.1%) had a TP53 mutation, and 48/71 (67.6%) carried a PTEN mutation. The mean MIB-1% was 40% (range 5.0–95.0%). EGFR amplification status, TP53 mutation, PTEN mutation, and MIB-1% were not associated with PFS or OS. The median EOR was 93.6% (range 11.4–100%), with 52 patients (51.5%) undergoing further resection after the initial iMRI scan. Of those 52 patients, 46 (88.5%) had pathology-proven positive margins on subsequent resection. EOR was significantly associated with both PFS and OS, but further resection after initial iMRI was associated with neither.

Fig. 2
figure 2

Relationship of T1W-CE (a, b), T2W (c, d), and FLAIR (e, f), imaging volumes on PFS and OS. Log-rank analyses of preoperative, intraoperative, immediate postoperative, and three-month postoperative imaging volumes—separated above and below the median—and their relationship with PFS and OS

Full size image

T1-weighted contrast-enhancing tumor volumes and survival

The median volumes and results of the log-rank analyses for the different imaging modalities at specific timepoints are summarized in Table 2. The temporal trend of T1W-CE tumor volumes for the patient cohort is depicted in Fig. 1b. It should be noted that a consistent trend of increase was seen in T1W-CE tumor volumes between intraoperative and immediate postoperative scans despite subtraction of blood products. Using a log-rank analysis of patients above and below the median volumes, there were no statistically significant associations observed for T1W-CE and PFS (Fig. 2a) or OS (Fig. 2b), but a consistent trend towards longer PFS and OS is seen with below-median T1W-CE tumor volumes at all timepoints.

Table 2 Imaging volumes and their relationships with PFS and OS at the timepoints of interest
Full size table

T2-weighted imaging volumes and survival

The temporal trend of T2W hyperintense volumes is depicted in Fig. 1c. In the log-rank analysis of patients above and below the median volumes at various timepoints, there was no association between intraoperative T2W volumes and PFS or OS (Fig. 2c,d), as well as no association found between immediate postoperative T2W volumes and PFS (Fig. 2c). However, there was a statistically significant association between smaller preoperative T2W volumes and longer PFS and OS (p = 0.045 and p = 0.026, respectively; Fig. 2c,d). A statistically significant association between lower immediate postoperative T2W volumes and better OS (p = 0.002, Fig. 2d), as well as lower three-month postoperative T2W imaging volumes and longer PFS and OS (p = 0.031 and p = 0.006, respectively; Fig. 2c,d).

T2-FLAIR weighted imaging volumes and survival

The temporal trend of FLAIR hyperintense volumes is shown in Fig. 1d. On log-rank analysis of patients above and below the median volumes, there was no association between preoperative or intraoperative FLAIR volumes and PFS or OS (Fig. 2e,f) nor between immediate postoperative FLAIR volumes and PFS (Fig. 2e). A statistically significant association between smaller immediate postoperative FLAIR volumes and longer OS (p = 0.02, Fig. 2f), as well as smaller three-month postoperative FLAIR volumes and better PFS and OS (p = 0.018 and p = 0.004, respectively; Fig. 2e,f).

Volumetric-based multivariable Cox proportional hazard regression analyses

A summary of the median volumes and results of the multivariable Cox proportional hazard regression analyses for the different imaging modalities at specific timepoints is summarized in Table 2. There was a statistically significant association between smaller intraoperative and three-month postoperative T1W-CE tumor volumes and PFS (p = 0.04 and p < 0.005, respectively). Only smaller three-month postoperative T1W-CE tumor volumes were significantly associated with better OS (p < 0.005). There was no association seen for preoperative, intraoperative, or immediate postoperative T2W imaging volumes with PFS or OS. However, there was a statistically significant association between lower three-month postoperative T2W imaging volumes and PFS and OS (p < 0.005 and p = 0.01, respectively). Finally, no association was observed for preoperative, intraoperative, or immediate postoperative FLAIR imaging volumes with PFS or OS; however, a statistically significant association was seen between three-month postoperative FLAIR imaging volumes and PFS and OS (p < 0.005 and p = 0.04, respectively).

Extent of resection and survival

The median calculated EOR was 93.6% (range 11–100%). By log-rank analysis of EOR, patients were dichotomized above and below the median value. Statistically significant associations between greater EOR and both PFS and OS were seen (Fig. 3a,b).

Fig. 3
figure 3

Relationship of EOR (a, b) and additional resection (c, d) on PFS and OS. Log-rank analysis of EOR separated above and below the median value and its relationship with PFS (a) and OS (b). Log-rank analysis of no additional resection versus additional resection and its relationship with PFS (c) and OS (d)

Full size image

Further resection, margin positivity, and survival

Fifty-two of 101 patients (51.5%) underwent additional resection after the initial iMRI scan. Tumor margins sent for final pathological analysis were positive for HGG in 46 patients (88.5%). In a log-rank analysis of patients who did or did not have additional resection, there was no observed improvement in either PFS (p = 0.387) or OS (p = 0.592) for patients undergoing additional resection after the initial iMRI scan (Fig. 3c,d).

Imaging sequence ratios and survival

Volumetric imaging ratios were calculated using the T1W-CE, T2W, and FLAIR sequence imaging volumes for each patient. There was no observed association with PFS or OS for any of the imaging volume ratios at any timepoint.

Discussion

In our single-center cohort of 101 patients with HGG, we evaluated the relationship of imaging volumes on different pre- and postoperative sequences with PFS and OS, along with the effect of additional resection after initial iMRI on survival. Our median EOR and OS were 93.6% and 14.0 months in this patient group, respectively. We identified that EOR and 3-month imaging volumes (i.e., T1W-CE, T2W, FLAIR) had the greatest association with OS. We also showed that additional surgical resection in 51.5% of cases after iMRI resulted in positive specimen in 88.5% of cases but did not improve survival in this patient group.

T1W imaging volumes

Numerous prior studies have shown that greater EOR and decreased CE-RTV (i.e., decreased intraoperative and postoperative T1W-CE tumor volumes) are both independently associated with longer PFS and OS [7, 10, 14,15,16,17,18,19, 21]. In our log-rank analysis of T1W-CE tumor volumes, there did not appear to be any association with PFS or OS; however, there was a consistent trend at all timepoints for longer PFS and OS with below-the-median T1W-CE tumor volumes. Although this was not entirely an expected result, we attribute the lack of statistical significance primarily to the median EOR of 93.6% for this cohort. Our T1W-CE-based EOR calculations and log-rank analysis did show a statistically significant impact on both PFS (p = 0.012) and OS (p = 0.006).

T2W and FLAIR imaging volumes

Although much research has focused on T1W-CE tumor volumes and survival in patients with HGGs, a number of more recent studies have also demonstrated the prognostic importance of T2W and FLAIR imaging volumes [34, 35]. In the 1980s, Hochberg and Pruitt [36] first demonstrated via postmortem examinations that 80% of HGGs recurred within 2 cm of the resection margin—where infiltrating tumor cells are highest in concentration. Accordingly, resection of hyperintense T2W/FLAIR tumor margins has been an area of significant interest. In 2016, Li et al. [12] examined 876 patients with GBM who underwent complete resection and found through multivariate analysis that those patients with > 50% non-enhancing hyperintense FLAIR volume resection had significant improvement in OS. Grossman et al. [37] examined 103 patients with HGGs undergoing resection and found no significant association between immediate postoperative FLAIR volumes and OS. However, their study did show a statistically significant OS benefit when three-month postoperative FLAIR imaging volumes were below 19.3 cm3 [37].

In our log-rank analyses, lower immediate postoperative T2W and FLAIR imaging volumes were significantly associated with longer OS (p = 0.002, p = 0.02, respectively) but not PFS (p = 0.054, p = 0.189, respectively). However, on multivariate analyses there was no statistically significant association between PFS or OS and immediate postoperative T2W (p = 0.06, p = 0.81, respectively) or immediate postoperative FLAIR volumes (p = 0.05, p = 0.48, respectively). Similar to our T1W-CE volumetric analyses, we did see a consistent survival benefit for smaller T2W and FLAIR imaging volumes, likely suggesting our study was inadequately powered to detect a small but appreciable survival effect.

Our data showed a significant association between decreased volumes on three-month postoperative T2W and FLAIR and longer PFS (p < 0.005, p < 0.005) and OS (p = 0.01, p = 0.04). Collectively, we believe these data strongly support the prognostic value of T2W and FLAIR imaging volumes at three months postoperatively—with patients having significantly longer PFS and OS when three-month T2W volume was < 23.0 cm3 and three-month FLAIR volume was < 28.7 cm3.

iMRI utility

Lastly, we examined the effect of additional resection after initial iMRI scan on survival. Half of our patients underwent additional resection after iMRI use, with 88.5% of specimens containing tumor; however, there was no association of PFS or OS with undergoing additional surgical resection. This suggests that additional surgical resection after initial iMRI scan does not significantly improve nor worsen PFS or OS in patients with HGGs. This finding has been reported by other studies and meta-analyses evaluating iMRI use in HGGs [26, 32]. Likely, these results are not able to adjust for all the other heterogenous patient variables (e.g., age, tumor mutation, tumor location, EOR, complications, or postoperative treatment).

Limitations

This study represents a single-center series of patients with pathologically confirmed HGGs who underwent resection using iMRI. The limitations of this study include the likely skewed EOR distribution of patients in our cohort, with a median EOR of 93.6%, which makes our 101-patient cohort likely underpowered with respect to assessing T1W imaging-based survival associations. Additionally, most patients included in our cohort had recurrent disease (59/101; 58.4%), which likely confounded the overall PFS and OS of the cohort, as demonstrated in Online Resource Table 1 and Supplemental Figs. 1 and 2.

Conclusions

Our study of 101 patients with new or recurrent HGGs shows that three-month postoperative T2W and FLAIR imaging volumes may serve as significant prognosticators with respect to PFS and OS. In addition, our study did not demonstrate any improvement of PFS or OS for patients undergoing further resection after the initial iMRI scan.