Impact of awake mapping on overall survival and extent of resection in patients with adult diffuse gliomas within or near eloquent areas: a retrospective propensity score-matched analysis of awake craniotomy vs. general anesthesia

Awake craniotomy (AC) with intraoperative mapping is the best approach to preserve neurological function for glioma surgery in eloquent or near eloquent areas, but whether AC improves the extent of resection (EOR) and overall survival (OS) is controversial. This study aimed to compare the long-term clinical outcomes of glioma resection under AC with those under general anesthesia (GA). Data of 335 patients who underwent surgery with intraoperative magnetic resonance imaging for newly diagnosed gliomas of World Health Organization (WHO) grades II-IV between 2000 and 2013 were reviewed. EOR and OS were quantitatively compared between the AC and GA groups after 1:1 propensity score matching. The two groups were matched for age, preoperative Karnofsky performance status (KPS), tumor location, and pathology. After propensity score matching, 91 pairs were obtained. The median EOR was 96.1% (interquartile range [IQR] 7.3) and 97.4% (IQR 14.4) in the AC and GA groups, respectively (p = 0.31). Median KPS score 3 months after surgery was 90 (IQR 20) in both groups (p = 0.384). The median survival times were 163.3 months (95% confidence interval [CI] 77.9–248.7) and 143.5 months (95% CI 94.4–192.7) in the AC and GA groups, respectively (p = 0.585). Even if the glioma was within or close to the eloquent area, AC was comparable with GA in terms of EOR and OS. In case of difficulties in randomizing patients with eloquent or near eloquent glioma, our propensity score-matched analysis provides retrospective evidence that AC can obtain EOR and OS equivalent to removing glioma under GA.

mapping (AC). Does AC improve the EOR and overall survival (OS)? The essential language and motor functions have been at the core of functional preservation efforts [9], and AC is the most suitable language/motor functional preservation method. Although there is little objection that AC is superior to general anesthesia (GA) for functional preservation, its ability to improve the EOR is controversial [3, 5, 8, 11, 13, 18-20, 26, 27, 30, 31, 33, 36, 37, 40]. One reason making such validation difficult is that many observational studies have not evaluated the OS, despite the proven strong correlation between EOR and OS, owing to the difficulty of adjusting the bias induced by the eloquent areas. The lack of an objective definition of eloquent areas on an image is a factor, but the most critical one is the selection bias of patients who underwent surgery performed under GA despite eloquent lesions. Patients who undergo GA may have aphasia and paralysis or are elderly individuals, conditions which also result in poor OS. Only randomized controlled trials (RCTs) can eliminate these unavoidable selection biases in patients with eloquent glioma in the GA group. However, the superiority of AC in functional preservation has been established, and RCT planning has high ethical hurdles. Therefore, we believe that propensity score-matched analysis (or pseudo-randomization) is the best method to examine the effects of AC on EOR and OS while adjusting for treatment selection bias. Hence, this study aimed to compare the long-term clinical outcomes of glioma resection under AC with those under GA. To the best of our knowledge, our propensity score-matched analysis comparing AC and GA is the first and largest study to quantitatively evaluate EOR and OS.

Indication of awake craniotomy
Our indication of AC complies with the Japanese AC guidelines [23]. A precentral gyrus tumor is an absolute indication for AC [35], and other tumors near the pyramidal tract are relative indications. Combined use of AC and motor evoked potential (MEP) monitoring improves the robustness of motor function monitoring and reduces the risk of permanent paralysis [34]. The language-dominant hemisphere is determined by referring to the dominant hand and functional MRI (fMRI) [21]. Right-handedness and left lesion near the language network constitute an absolute indication. However, even in instances where awake surgery would otherwise be indicated, cortical mapping is expected to be difficult if paralysis or aphasia already appears as symptoms, in which case GA is indicated. Furthermore, GA is used in cases where reintubation is difficult, the patient is too young or too old, and for patients judged to be unsuitable for awake surgery due to their mental predisposition.

Awake craniotomy
All patients received sleep-awake-sleep protocol anesthesia. First, craniotomy was performed after inducing sleep using a supraglottic airway device. Next, the first iMRI was performed. The patient was then awakened, and cortical mapping was used to identify language and motor areas. Tumors were removed by performing appropriate cortical/ subcortical mapping with a positive mapping strategy. After evaluating the brain tumor's resection on a second MRI, the scalp was closed under mild sedation without intubation. An Ojemann cortical bipolar stimulator (OCS-1; Integra Radionics, Inc., Burlington, MA) was used for cortical stimulation (repetitive square-wave biphasic currents of alternating polarity; intensity, 0-6 mA [biphasic currents 0-12 mA]; frequency, 50 Hz; duration, 2 s; interpolar distance, 5 mm). The distance between the stimulation sites was 5-10 mm, and a surface electroencephalogram (bandpass filter of 10 Hz to 1.5 kHz) was recorded to detect epileptic seizures.

Intraoperative MRI-guided surgery
MR images were obtained using iMRI (0.3-T AIRIS II™ Hitachi Medical, Chiba, Japan). The first intraoperative MR image was obtained after a dural incision to minimize the effects of brain shift. Sequential iMRI was performed before duraplasty to allow repeated resection of the residual tumor. The final iMR image was obtained after achieving the best possible resection. The tumor site, which showed positive findings in the cortex/white matter mapping during AC, was not removed and was preserved even if it remained on intraoperative MRI. A three-dimensional volumetric measurement of the first and final iMRI studies was retrospectively conducted, as previously described [14,15,24]. Manual segmentation was performed with the region-of-interest analysis to measure tumor volume based on T2-weighted images of WHO grades II and III and contrast-enhanced T1-weighted MR images of grade IV tumors. If the tumor did not show gadolinium enhancement, the T2 hyperintensity area was measured. EOR was defined as follows: (initial MR tumor volume − final MR tumor volume) / (initial MR tumor volume).

Statistical analysis
Statistical analyses were performed using SPSS Statistics, Version 25.0 (IBM Corp, Armonk, NY, USA). Categorical variables were compared using the chi-square test. The Mann-Whitney U test and t test were used for continuous nonparametric and parametric variables, respectively. OS analysis was performed using Kaplan-Meier curves and log-rank tests. p values < 0.05 were considered significant.

Propensity score-matched analysis
To overcome the bias arising from the lack of randomization and heterogeneity of glioma patients, one-to-one matching without replacement was performed using the nearest-neighbor match on the logit of the propensity score with caliper width set to 0.2 times the standard deviation of the logit of the propensity score. The distribution of each characteristic mentioned in Table 1 in the two groups was assessed with the standardized mean difference, calculated as the difference in the means or proportions of a variable divided by the pooled estimate of its standard deviation [7]. Standard mean difference < 0.1 between the two groups was considered an adequate balance of matching [2]. Propensity score matching was conducted using age, preoperative Karnofsky   [38].
Most patients with eloquent gliomas underwent surgery under AC. The patients who underwent GA also had a selection bias that affects prognosis because of their relatively low KPS and older age. Few studies have considered eloquent lesions as a prognostic factor. Therefore, we used the tumor's main region as an alternative variable to investigate the effect of AC on EOR and OS.

Patient characteristics
A total of 335 patients who underwent surgery with iMRI for newly diagnosed gliomas of WHO grade II-IV between 2000 and 2013 were analyzed. Patient characteristics are shown in Table 1. Before propensity score matching, patients were significantly older in the AC group than in the GA group (p < 0.001). The male/female ratio was not significantly different (p = 0.499). The left hemisphere, with a high proportion of eloquent lesions, was significantly more affected in the AC group (p < 0.001). The main tumor location in the AC group was the fronto-temporo-parietal cortex, including the insula near the language network and motor pathway (p < 0.001). Grade II-III tumors were predominant over glioblastoma (p < 0.001). The preoperative KPS score was higher in the AC group than in the GA group (p < 0.001), while the preoperative tumor volume was not significantly different between the two groups (p = 0.738).

Analysis after propensity score matching
After matching, no differences were found in the mean age (p = 0.997), tumor location (p = 0.959), pathology (p = 0.966), or mean preoperative KPS (p = 0.722). Moreover, the standardized mean difference results demonstrated negligible or small differences in all characteristics between the two groups.
The adjusted covariates were well balanced in the propensity score-matched cohort with a standardized mean difference of < 0.1 (Table 1), although the proportion of left hemisphere tumors, a factor that was not adjusted, remained higher in the AC group.

Post hoc analysis for prognostic factors
Isocitrate dehydrogenase (IDH) mutation and 1p19q codeletion status were examined in the cohorts after propensity score matching. In 26 cases, the 1p19q status could not be identified. However, the percentage of 1p19q codeletion was not significantly different between the groups (Table 2; p = 0.397). The IDH status of 16 individuals was unknown, but no significant difference was found in the prevalence of IDH mutations (Table 2; p = 0.131). We also examined the proportion of patients who received radiation therapy and chemotherapy in the two groups, finding no significant differences in nimustine hydrochloride or temozolomide usage (p = 0.741) nor in the administration of radiation therapy ( Table 2; p = 0.532).

EOR, postoperative KPS, and OS
The EOR calculated by volumetry and the KPS scores are shown in Table 3. The median removal rates were 96.1% and 97.4% in the AC group and GA group, respectively, showing no significant difference. Three months after surgery, the median KPS values in both groups decreased by 10 points compared with the preoperative values. Interestingly, KPS scores did not decrease in grade II tumors but decreased the most in the grade IV GA group (Table 3; p = 0.137). The median survival time was not significantly different between the AC (163.3 months) and GA (143.5 months) groups (Fig. 1a). Although the median survival time could not be calculated for grade II patients, the mean survival time was 143.4 months (95% CI 125.2-161.7) in the AC group and 135.8 months (95% CI 121.0-150.6) in the GA group, without significant difference (Fig. 1b). No  Fig. 1d).

Discussion
In a comparative study of GA and AC methods, the largest analysis was performed with the prognostic factors adjusted by propensity score. This study showed that AC for eloquent glioma was comparable with GA in terms of EOR and OS.

Superiority in propensity score-matched analysis for bias adjustment
We compared our study to previous articles comparing AC versus GA-adjusted background factors (Table 4). When considering the effect of AC on OS, it is challenging to consider a control group for AC other than GA. However, in retrospective or prospective designs, there will always be significant differences between the characteristics of patients undergoing AC and GA. The AC group had better KPS scores and younger age before propensity score matching ( Table 1). The same table also shows that lowergrade cases are biased toward AC, inevitably implying that cases with good prognosis are biased toward the AC group. The most direct way to establish effectiveness of AC is to randomize patients with glioma near eloquent areas. Only one RCT concluded that removal rate and functional preservation were significantly worse in the AC group [20]. Since superiority of AC in terms of functional preservation has become commonly accepted, it is becoming challenging to implement RCTs ethically. To our knowledge, only one RCT is planned for glioblastoma resection [17]. The selection criteria include the condition that the neurosurgeon can remove the tumor in both surgical procedures, cleverly avoiding ethical issues. We considered propensity score-matched analysis as the most suitable method for examining the effects of AC on OS in a retrospective analysis.

EOR and OS
There is much debate on whether AC improves the resection rate. Some reports have stated that the removal rate is improved [5,8,11,18,27,33,36], while others reported a decrease or no significant difference [3,13,19,20,30,37,40]. The best way to address this issue is to confirm that the resection rate correlates with prognosis. Most studies are limited to the analysis of the functional outcome. Gerritsen et al. used propensity score-matched analysis limited to grade IV to show that AC improved EOR, but not OS [18], and attributed the lack of significance to a sample size issue. Pichierri et al. showed the same resection rate for AC and GA, but better OS with AC [30]; postoperative neural function might have affected prognosis. Although Sacko et al. evaluated the OS of glioma patients, data from tumors other than glioma were included in the analysis of resection rates [33]. OS data of Duffau et al. and Gravesteijn et al. were not quantitative evaluations using Kaplan-Meier curves [11,19]. Our data are consistent with the absence of significant differences in EOR and OS between the AC and GA groups, proving the validity of the results.

Difference between AC and GA resection rates
According to Chang et al., a false eloquent area is presumed to be eloquent based on preoperative fluid-attenuated inversion recovery (FLAIR)/T2 imaging but is not eloquent based on the AC mapping [6]. A true eloquent area is presumed to be eloquent based on anatomical preoperative FLAIR/ T2 imaging and is confirmed as such by AC mapping. They concluded that AC mapping improved the removal rate and prognosis when the T2/FLAIR tumor area was a false eloquent. By contrast, if it was a true eloquent, the prognosis did not change, and the EOR was the same as in the GA group. Considering the lack of significant difference in KPS scores between the AC and GA groups before and 3 months after surgery of diffuse gliomas including grades II-IV, the lack of difference in AC and GA removal rates suggested that a false eloquent contained more of the iMRI removal target than the true eloquent. In the sub-analysis of grade II gliomas, no difference was found in the removal rate, OS, or postoperative KPS between the two groups ( Table 3, Fig. 1b). Our data suggest that there was little true eloquence in the grade II T2 regions. Grade IV sub-analysis showed no difference in EOR, but KPS scores tended to be superior in the GA group than in the AC group, contrary to what was observed in other grades. The EOR result can confirm the long-standing belief that there are no functional fields in the gadolinium-enhanced lesions of glioblastoma. Although there was no significance for OS, a divergence could be observed at the beginning of the Kaplan-Meier curves. Surprisingly, in glioblastoma, AC may have reduced early mortality by preventing an early postoperative decline in KPS (Fig. 1d, Table 3). The recent meta-analysis by Gerritsen et al. of AC for high-grade glioma also suggests the efficacy of AC and is in the process of being validated by RCTs in grade IV tumors [16][17][18].

Effect of genetics and postoperative treatment
This study was not based on the WHO 2016 classification, as we could not analyze the 1p19q codeletion status in the early 2000s. In principle, we could have eliminated these cases, but since OS analysis was among the main purposes of the study, these cases and those with grades II-III require long follow-up. Recently, however, IDH status and 1p19q coding have been considered strict prognostic factors [12,39]. The IDH status of 16 patients was unknown, but no significant difference was found between the AC and GA groups (p = 0.131). Most of the cases whose IDH status could not be identified were glioblastoma. At our facility, gene mutation search was not actively performed in the past for cases pathologically classified as glioblastoma. Approximately 90% of glioblastoma cases have wild type IDH, in agreement with previous estimates [39]. In 26 cases, the 1p19q status could not be identified. However, since the percentage of 1p19q codeletion was not significantly different between the groups (p = 0.397), the effects of prognostic genetic factors are considered small, although we could not analyze O 6 -methylguanine (O 6 -MeG)-DNA methyltransferase promoter methylation status.
The inclusion of all gliomas in our analysis has an effect on post-treatment and may affect the analysis of prognosis. Our facility does not actively perform AC when we suspect glioblastoma. Indeed, glioblastoma is typically already symptomatic at the time of discovery and is often not indicated for AC, as shown in Table 1. Furthermore, it is not possible to accurately assess the WHO grade in the preoperative image. The reason for simultaneously analyzing tumors of different WHO grades was to reduce the selection bias. Furthermore, in the post hoc analysis, no difference was noted after treatment between the AC group and the GA group, and the effect on prognosis was considered small.

Limitations
Our study is inherently limited by its retrospective design, leading to selection bias, lack of randomization, limited control of confounding factors, and difficulty in establishing causes and effects. Moreover, it did not adjust for eloquent area lesions when comparing the AC and GA groups.
The selection of glioma patients should be limited to near eloquent lesions to investigate the effects of AC on EOR and OS, and previous studies have attempted to do. Paradoxically, adjusting for eloquent lesions made it difficult to adjust for KPS and age, which affected the OS; moreover, the small number of eloquent cases in the GA group led to small sample sizes. There were 12/91 (13%) patients with near eloquent lesions in the GA group, although the analysis performed when excluding them did not change the results (data not shown). However, excluding the eloquent group from the GA group created a bias for OS analysis. Furthermore, the cohort adjusted for eloquence was fundamentally enriched in perirolandic glioma cases, because eloquent glioma, which is an indication for GA, is abundant in the motor area and is considered less prevalent in the languagerelated area. Given that the most prominent feature of AC is preserving language function, this method was also limited. Nevertheless, we believe that propensity score-matched analysis is the best adjustment for assessing EOR and OS in non-RCT studies.
Most previous studies did not use iMRI, resulting in more significant variability of the EOR in the GA group. AC and GA under iMRI were compared in only two studies [30,37], but even in these, quantitative evaluations of resection rate were not performed. Interestingly, the removal rate under iMRI was consistent, as also in our study, in the AC and GA groups. Nevertheless, it is possible that image-guided awake surgery with and without iMRI may have different effects on OS and EOR.

Conclusions
Our extensive study quantitatively evaluated AC and GA removal rates and adjusted for bias with propensity score matching. As it is difficult to plan an RCT for patients with eloquent glioma, our propensity score-matched study provided retrospective evidence that AC can obtain EOR and OS equivalent to removing glioma under GA.