Early repeat resection for residual glioblastoma: decision-making among an international cohort of neurosurgeons

Abstract

Objective

The importance of extent of resection (EOR) in glioblastoma (GBM) has been thoroughly demonstrated. However, few studies have explored the practices and benefits of early repeat resection (ERR) when residual tumor deemed resectable is unintentionally left after an initial resection, and the survival benefit of ERR is still unknown. Herein, the authors aimed to internationally survey current practices regarding ERR and to analyze differences based on geographic location and practice setting.

Methods

The authors distributed a survey to the American Association of Neurological Surgeons and Congress of Neurological Surgeons Tumor Section, Society of British Neurological Surgeons, European Association of Neurosurgical Society, and Latin American Federation of Neurosurgical Societies. Neurosurgeons responded to questions about their training, practice setting, and current ERR practices. They also reported the EOR threshold below which they would pursue ERR and their likelihood of performing ERR using a Likert scale of 1–5 (5 being the most likely) in two sets of 5 cases, the first set for a patient’s initial hospitalization and the second for a referred patient who had undergone resection elsewhere. The resection likelihood index for each respondent was calculated as the mean Likert score across all cases.

Results

Overall, 180 neurosurgeons from 25 countries responded to the survey. Neurosurgeons performed ERRs very rarely in their practices (< 1% of all GBM cases), with an EOR threshold of 80.2% (75%–95%). When presented with 10 cases, the case context (initial hospitalization vs referred patient) did not significantly change the surgeon ERR likelihood, although ERR likelihood did vary significantly on the basis of tumor location (p < 0.0001). Latin American neurosurgeons were more likely to pursue ERR in the provided cases. Neurosurgeons were more likely to pursue ERR when the tumor was MGMT methylated versus unmethylated, with a resection likelihood index of 3.78 and 3.21, respectively (p = 0.004); however, there was no significant difference between IDH mutant and IDH wild-type tumors.

Conclusions

Results of this survey reveal current practices regarding ERR, but they also demonstrate the variability in how neurosurgeons approach ERR. Standardized guidelines based on future studies incorporating tumor molecular characteristics are needed to guide neurosurgeons in their decision-making on this complicated issue.

Despite maximal safe resection and adjuvant therapies including temozolomide and radiotherapy, the survival of patients with glioblastoma (GBM) remains poor.¹ This low survival is driven by extensive tumor infiltration of adjacent parenchyma, which inevitably leads to recurrence even after resection of evident tumor. Given the diffuse nature of this disease, aggressive cytoreductive strategies are often considered. Landmark studies have demonstrated that a 70%–98% extent of resection (EOR) threshold is associated with a survival benefit.^2–4 Moreover, recent studies have suggested that supratotal resection (i.e., resection of signal change on FLAIR MRI sequences) can provide overall survival advantages to GBM patients.^5–7

With this emphasis on achieving a high EOR, what does one do when unintended residual disease is seen on immediate imaging following initial surgery or in a referred patient who has undergone incomplete initial resection? Clinical studies regarding decision-making for early repeat resection (ERR) or early reoperation to resect residual tumor before applying adjuvant therapies are scarce, and the efficacy of ERR has not been studied through randomized controlled trials. Moreover, surgical neuro-oncology best practices are largely unknown when residual enhancing tumor amenable to gross-total resection (GTR) is unintentionally left unresected.

In a recent survey by the American Association of Neurological Surgeons (AANS) and Congress of Neurological Surgeons (CNS) Tumor Section, although 59% of respondents reported never performing re-resections within 30 days of initial surgery, 31% indicated performing between 1 and 4 early re-resections per year. The most common reason for early re-resection was incomplete initial resection (26.03%) or other (38.36%).⁸ This survey highlighted how rare early re-resection is in most surgeons’ practices and gave some insight into why ERR is performed. We wished to build on this analysis with a case-based survey that assessed how tumor location, tumor mutation status, case context, and patient variables change a surgeon’s likelihood of pursuing ERR. Also, the AANS/CNS Tumor Section survey represented the opinions of surgeons from the US and Canada, but we wished to survey an international cohort of surgeons.

When making decisions about ERR, one must consider many variables. Aside from subjecting patients to the inherent risks associated with any operation, neurosurgeons must also consider the eloquence of the residual tumor location, patient characteristics (age, Karnofsky Performance Status [KPS], presence of neurological deficits), EOR and residual volume (RV) of the initial surgery, and, if available, tumor molecular characteristics. Additionally, the potential benefit of resecting a greater tumor volume must be weighed against the possibility of causing a delay in adjuvant therapy, discharge to a rehabilitation facility, or an increase in the financial burden on patients.

In this study, we aimed to survey how different factors impacted the likelihood that neurosurgeons would pursue ERR. Additionally, by contacting established national and international neurosurgical organizations, we were able to reach neurosurgeons practicing in a variety of settings accounting for geographic, cultural, and socioeconomic factors. Our goal was not only to represent the heterogeneity of neurosurgical practice, but also to identify ERR decision-making variations across different regions and practice settings.

Methods

Survey Design and Data Collection

Qualtrics software version VM (Qualtrics) was used to design, distribute, and collect responses from survey participants. An English version of the survey (Supplementary Document 1) was written by our team and revised and approved by the Survey Committee of the Executive Committee of the AANS and CNS Tumor Section prior to its distribution. The survey was then translated to Spanish (Supplementary Document 2) by two native Spanish speakers (A.D., Y.E.).

The survey first collected demographic data from participants, including general views and practices regarding ERR. The patient characteristics evaluated to play a role in the decision-making about ERR were those factors known to affect survival (age, KPS, presence of deficits, initial EOR, and RV). EOR thresholds were assessed as percentages, and RV thresholds were measured in cm³. Respondents were asked to rate their likelihood of performing ERR on a Likert scale⁹ ranging from 1 (extremely unlikely) to 5 (extremely likely). Then, surgeons were asked to judge individual cases with pre- and postoperative T1-weighted MRI studies with and without contrast so that we could investigate the impact that tumor location had on decision-making. Finally, neurosurgeons were invited to share more nuanced comments through free-text boxes.

Cases were split into two groups of 5 cases. The first set of cases was described as patients with an unintentionally unresected residual tumor within an initial hospitalization, whereas the second set was described as patients referred by other physicians after having undergone an operation at an outside institution. In terms of this second set of 5 cases, respondents were also asked how tumor MGMT methylation and IDH mutation status would affect their decisions given that, in these referred cases, such results would already be available and therefore taken into consideration during the decision-making process. Such information was not available for the first set of cases, as the residual tumor was identified on immediate (< 24 hours) postoperative MRI. These two clinical scenarios reflected the two most common reasons for ERR within 30 days according to a recent survey study by the AANS/CNS Tumor Section, that is, incomplete initial resection (26%) and referral (22%).⁸

Image Selection

De-identified images were selected from 10 different GBM patients (ages > 18 years) who had undergone surgery by both senior authors at The University of Texas Health Science Center at Houston and Johns Hopkins University. Preoperative pre- and postcontrast T1-weighted MRI studies were included in the survey, as well as one postoperative postcontrast T1-weighted MR image. EOR and RV were calculated using semi-automated lesion segmentation from standard-of-care MRI studies on routine postcontrast 2D or 3D T1-weighted images before and after resection. FLAIR signal abnormality volume was measured on standard 2D or 3D T2-FLAIR images. Volumes were calculated within either Carestream Vue PACS version 12.10 or TeraRecon Aquarius iNtuition viewer.

Survey Recruitment

A link to our survey was provided to members of the AANS/CNS Tumor Section, Society of British Neurological Surgeons, European Association of Neurosurgical Society, and multiple Latin American neurosurgical societies. The Latin American neurosurgical societies were contacted through the Latin American Federation of Neurosurgical Societies (FLANC) and included the following: Argentine Association of Neurosurgery, Chilean Society of Neurosurgery, Dominican Society of Neurology and Neurosurgery, Colombian Association of Neurosurgery, Peruvian Society of Neurosurgery, Ecuadorian Neurosurgical Society, Uruguayan Society of Neurosurgery, Brazilian Society of Neurosurgery, Brazilian Academy of Neurosurgery, Panamanian Society of Neurosurgery and Neurology, Neurosurgical Association of El Salvador, and Mexican Society of Neurological Surgery. We accepted survey responses from August 28 to November 7, 2020. Full and partial responses were accepted, and the number of opened survey links was also recorded.

Grouping of Respondents

Respondents were grouped into three separate regions based on geographic location. A list of the countries in each region appears in Supplementary Table 1. Additionally, we chose 15 operative cases per year as the cutoff to evaluate differences between respondents by glioma caseload, as it represented both the median in our population and a previously described number of cases to consider high-volume facilities in a glioma resection study.¹⁰

Statistical Analyses

Differences in demographic variables by region and experience were calculated using Fisher’s exact test and a t-test or Mann-Whitney U-test for categorical and continuous variables, respectively. Normality was determined using the Shapiro-Wilk and Kolmogorov-Smirnov tests. The resection likelihood index for each respondent was calculated as the mean Likert score across all cases, with 5 being extremely likely to perform ERR and 1 being extremely unlikely to do so. The correlation between mean EOR thresholds and the resection likelihood index was determined through simple linear regression analysis. Differences in the resection likelihood index by geographic region and among cases were calculated using one-way ANOVA and Tukey’s multiple comparison test. Spearman’s rank correlation was performed between variables. Stepwise multiple linear regression analysis based on the p value was performed to adjust for multiple variables. Cochran’s Q was used to analyze differences in IDH and MGMT patient status between cases, and Mann-Whitney U-tests were used to calculate differences in the overall resection likelihood index between IDH wild-type and mutant tumors and between MGMT methylated and unmethylated tumors. Statistical analyses were performed using R (version 4.0.3, R Foundation for Statistical Computing) with EZR software¹¹ (1.54, Yoshinobu Kanda) and Prism version 9.0.1 (GraphPad Software Inc.).

Results

Respondent Demographics by Region

Overall, this survey was emailed to 9608 people and had a 1.9% response rate (180 responses). The response rate of surgeons who opened the link to our survey (230 links opened) was 78.2%. The demographic variables collected through the survey are shown in Table 1. When these variables were broken down by region, we found that Latin American neurosurgeons had spent more years in practice (p = 0.048) and that the US/Canada had the highest rate of respondents who had completed intraoperative mapping training (p < 0.001). Latin American neurosurgeons had the lowest number of GBM resections (p < 0.001) but also had the highest percentage of ERR (p < 0.001). Respondents from the US/Canada and Europe/Asia more frequently practiced in academic settings, whereas Latin American neurosurgeons were more evenly distributed among academic, community, and private practices (p < 0.001). Latin American neurosurgeons were more frequently identified as general neurosurgeons (p = 0.002), whereas US/Canadian neurosurgeons more frequently focused on intraaxial tumors (p = 0.001). Intraoperative tool use varied regionally, with both the US/Canada and Europe/Asia more frequently using most of the tools listed in our survey (5-aminolevulinic acid [5-ALA], awake craniotomy, and mapping) compared to Latin America (p < 0.001 for all 3 tools), while there was no statistically significant difference in ultrasound use across regions. However intraoperative MRI was heavily used in the US/Canada, with much less use among respondents in Latin America and Europe/Asia.

TABLE 1.

Characteristics of survey respondents, divided by region

Variable	Total Cohort	Latin America	US/Canada	Europe/Asia	p Value*
No. of respondents	180	75	71	34
Yrs in practice	14 (6, 21.25)	15 (10, 23.5)	11 (5, 21)	11.5 (7, 17.5)	0.048
Intraop mapping training	95 (52.8)	23 (30.7)	51 (71.8)	21 (61.8)	<0.001
Completed a neuro-oncology fellowship	72 (40.0)	23 (30.7])	32 (45.1)	17 (50.0)	0.096
GBM resection experience
Resections for GBM	15 (7, 25)	10 (5, 13.5)	20 (10, 30)	25 (18, 40)	<0.001
ERR percentage	0 (0,1)	5 (0, 20)	0 (0, 0)	0 (0, 3)	<0.001
Practice setting
Academic hospital/practice	118 (65.6)	32 (42.7)	55 (77.5)	31 (91.2)	<0.001
Community hospital	49 (27.2)	35 (46.7)	13 (18.3)	1 (2.9)
Private practice, solo/partners	48 (26.7)	40 (53.3)	5 (7.0)	3 (8.8)
Type of practice†
General neurosurgery	94 (56.6)	49 (73.1)	32 (45.7)	13 (44.8)	0.002
Intra-axial tumors	92 (55.4)	26 (38.8)	45 (64.3)	21 (72.4)	0.001
Skull base tumors	47 (28.3)	17 (25.4)	24 (34.3)	6 (20.7)	0.312
Spine tumors	25 (15.1)	8 (11.9)	12 (17.1)	5 (17.2)	0.653
Intraop tools‡
5-ALA/fluorescence	64 (43.8)	10 (16.7)	31 (56.4)	23 (74.2)	<0.001
Awake craniotomy	85 (58.2)	22 (36.7)	41 (74.5)	22 (71.0)	<0.001
Mapping	94 (64.4)	27 (45.0)	46 (83.6)	21 (67.7)	<0.001
Ultrasound	78 (53.4)	30 (50.0)	29 (52.7)	19 (61.3)	0.607
MRI	23 (15.8)	6 (10.2)	16 (29.1)	1 (3.2)	0.002
Postop MRI for EOR assessment§	117 (81.3)	37 (62.7)	53 (96.4)	27 (90.0)	<0.001
EOR % threshold for reoperation	80 (50, 90)	70 (50, 90)	80 (50, 90)	50 (0, 80)	0.043
RV threshold for reoperation in cm3	5 (1, 10)	5 (1, 10)	5 (0, 10)	5 (0, 10)	0.210
Factors affecting reoperation likelihood
KPS ‖	125 (89.3)	50 (89.3)	50 (92.6)	25 (83.3)	0.421
Age ¶	109 (76.2)	40 (67.8)	49 (89.1)	20 (69.0)	0.017
Neurological deficit #	111 (79.9)	39 (68.4)	48 (90.6)	24 (82.8)	0.014

Values are expressed as median (interquartile range) or number (%), unless indicated otherwise.

^*Fisher exact test and Mann-Whitney U-test were performed for categorical and continuous variables, respectively. Boldface type indicates statistical significance.

^†Data were missing for 14 respondents.

^‡Data were missing for 34 respondents.

^§Data were missing for 36 respondents.

^‖Data were missing for 40 respondents.

^¶Data were missing for 37 respondents.

^#Data were missing for 41 respondents.

Respondent Demographics by Number of GBM Cases

Our study revealed that neurosurgeons performing > 15 GBM resections per year were more likely to have been trained on intraoperative mapping (p < 0.0001; Table 2) and to have completed a neuro-oncology fellowship (p = 0.001). Moreover, they were more frequently from the US/Canada or Europe/Asia (p < 0.0001) than from Latin America. Neurosurgeons performing more GBM resections more commonly practiced at academic institutions (p = 0.003) and had an intraaxial tumor practice (p < 0.0001). Additionally, neurosurgeons with a higher number of GBM resections per year more frequently utilized intraoperative tools than the neurosurgeons with a lower number of resections (5-ALA/fluorescence, p = 0.0007; awake craniotomy, p < 0.0001; mapping, p = 0.0004; intraoperative MRI, p = 0.022). Finally, neurosurgeons with a higher number of GBM resections per year assessed EOR through postoperative MRI (< 48 hours) more frequently (p = 0.019).

TABLE 2.

Characteristics of survey respondents, divided by number of GBM cases operated on per year

Variable	≤ 15 GBM Resections/Yr	> 15 GBM Resections/Yr	p Value*
No. of respondents	76	67
Yrs in practice	14 (6, 24)	14 (6, 20)	0.486
Intraop mapping training	28 (37)	50 (75)	<0.0001
Completed a neuro-oncology fellowship	24 (32)	40 (60)	0.001
Early reoperation percentage	0 (0, 0)	0 (0, 0)	0.983
Region
Latin America	48 (63)	11 (16)	<0.0001
US/Canada	21 (28)	34 (51)
Europe/Asia	7 (9)	22 (33)
Practice setting
Academic hospital/practice	45 (59)	55 (82)	0.003
Community hospital	29 (38)	8 (12)	0.0004
Private practice, solo/partners	30 (39)	8 (12)	0.0003
Type of practice†
General neurosurgery	50 (74)	21 (33)	<0.0001
Intra-axial tumors	24 (35)	56 (89)	<0.0001
Skull base tumors	22 (32)	16 (25)	0.433
Intraop tools‡
5-ALA/fluorescence	23 (30)	40 (60)	0.0007
Awake craniotomy	32 (42)	52 (78)	<0.0001
Mapping	39 (51)	54 (81)	0.0004
MRI	7 (9)	16 (24)	0.022
Ultrasound	40 (53)	40 (60)	0.867
Postop MRI for EOR assessment	56 (74)	60 (90)	0.019

Values are expressed as median (interquartile range) or number (%), unless indicated otherwise.

^*Fisher exact test and Mann-Whitney U-test were performed for categorical and continuous variables, respectively. Boldface type indicates statistical significance.

^†Data were missing for 17 respondents.

^‡Data were missing for 50 respondents.

Stepwise multiple linear regression analysis demonstrated that the number of glioma cases performed by neurosurgeons was influenced by geographic region. The Latin American region was correlated with a reduced number of GBM cases per year (Spearman’s rank correlation −0.422, p < 0.0001; multiple regression p = 0.046). An intraaxial tumor practice (p < 0.0001) and the performance of awake craniotomy (p = 0.016) were positively correlated with a greater number of GBM cases in the multiple regression analysis.

Characteristics That Influence ERR Among Neurosurgeons

With regard to the demographic characteristics evaluated to influence ERR (KPS, age, neurological deficit), we revealed that neurosurgeons from the US/Canada and Europe/Asia were less likely than Latin American neurosurgeons to perform ERR in patients who had a KPS < 70 (p = 0.023; Fig. 1) or a new neurological deficit (p < 0.001). Additionally, Latin American neurosurgeons were more likely to perform ERR in patients 18–65 years of age (p = 0.004) and in patients with no new neurological deficit (p = 0.05). Other characteristics, such as older patient age groups (65–80 years and > 80 years) and higher KPS (> 70), were not significantly different among the neurosurgeons based on geographic location.

FIG. 1.

There were significant differences among regions in the likelihood of pursuing ERR: in patients with a KPS < 70 (A), in patients between the ages of 18 and 65 years (B), in patients with new deficits that began after the initial surgery (C), and in patients with no new deficits after the initial surgery (D). Figure is available in color online only.

We also calculated a mean resection likelihood index for each surgeon on the basis of how high they scored their likelihood of performing ERR in each case (1, extremely unlikely; 5, extremely likely). To validate this resection likelihood index, we compared it to the respondent’s mean EOR threshold. Linear regression analysis revealed a weak but significant positive correlation between the two (R² = 0.1811, p < 0.001), indicating that respondents who ranked themselves as more likely to perform ERR also had higher EOR thresholds for choosing to perform ERR, as expected. Neurosurgeons from Latin America had a significantly higher mean resection likelihood index (p < 0.001) than surgeons from the US/Canada and Europe/Asia (Fig. 2A).

FIG. 2.

A: Neurosurgeons from Latin America were more likely to pursue ERR across all 10 cases. B: Neurosurgeons were more likely to pursue ERR in patients with MGMT methylated tumors. C: There was no statistically significant difference in neurosurgeons’ ERR likelihood between IDH-mutant and wildtype tumors. The resection likelihood index represented the average of their Likert scores across the survey, with 1 being extremely unlikely to pursue ERR and 5 being extremely likely. Figure is available in color online only.

Neurosurgeon Responses Across Cases

When comparing neurosurgeon responses across cases (Fig. 3A–J), the location of the tumor did not significantly alter the EOR threshold below which respondents would perform ERR (p = 0.260). The average EOR threshold for all cases was 80.2% (75%–95%). There was no significant difference in EOR thresholds between immediate repeat resections (first 5 cases) and referrals from an outside institution (last 5 cases; p = 0.928).

FIG. 3.

Case images presented in the survey. The first 5 cases (A–E) were presented as an unintentionally unresected residual tumor within a patient’s initial hospitalization, and the second set of 5 cases (F–J) was described as patients referred by other physicians after having undergone an operation at an outside institution. The three T1-weighted MRI studies presented for each case are, from right to left, presurgery postcontrast, postsurgery precontrast, and postsurgery postcontrast, with arrows marking the location of residual tumor in the final image. Mean ERR likelihood scores (K) differed significantly among cases. Figure is available in color online only.

However, the mean resection likelihood index did vary among cases (p < 0.0001), and respondents ranked themselves as more likely to perform ERR in the right temporal case (Fig. 3A; mean ERR likelihood rating of 3.70), right parietal case (Fig. 3F; 3.88), and right occipital case (Fig. 3J; 3.81) than in all other cases (1.78–3.01, p < 0.01, Tukey’s multiple comparisons test). There was no significant difference among cases A, F, and J (p = 0.99).

IDH (p < 0.001) and MGMT (p = 0.049) status also influenced the surgeons’ decision-making among cases. In asking about the role of IDH and MGMT status in decision-making, we found that mutation status in the right parietal case mattered the most to respondents (Fig. 3F), with 43.4% and 26.4% stating that IDH and MGMT status, respectively, would play a role in their decision to perform ERR. This location was followed by the right temporoparietal case (Fig. 3G; 37.2% and 21.0%, respectively), occipital case (Fig. 3J; 35.6% and 22.0%), left temporal case (Fig. 3H; 29.5% and 17%), and right frontal case (Fig. 3I; 23.8% and 17.8%).

The neurosurgeons who responded that IDH and MGMT statuses would play a role in decision-making were then asked how likely they would be to perform ERR given this additional information on tumor mutation status. In looking at their pooled resection likelihood across cases, we found that neurosurgeons were more likely to pursue ERR when the tumor was MGMT methylated than when it was unmethylated, with a resection likelihood index of 3.78 and 3.21, respectively (p = 0.004; Fig. 2B). Although surgeons tended to be more aggressive with IDH mutant tumors than with IDH wild-type tumors, the difference between the two did not reach statistical significance (3.58 and 3.28, respectively, p = 0.083; Fig. 2C).

Discussion

Our study aimed to evaluate decision-making and practices regarding ERR for residual GBM among a broad cohort of neurosurgeons, in terms of both geographic location and practice setting. Unexpectedly, the location of the tumor did not significantly impact respondents’ EOR threshold for performing ERR; however, a participant’s ERR likelihood index did vary significantly among tumor locations. Surveyed Latin American neurosurgeons were less likely to have a dedicated neuro-oncology practice and had a lower median GBM caseload, as well as lower usage of intraoperative adjunctive therapies. Even though the percentage of GBM cases in which ERR was performed was consistently low for all surgeons surveyed, the Latin American neurosurgeons more frequently performed ERR for unresected tumors deemed amenable to GTR. In addition, neurosurgeons with a higher volume of GBM cases, defined as ≥ 15 GBM surgeries per year,¹⁰ were more likely to be from the US/Canada or Europe/Asia, to work in an academic setting, to have a dedicated intraaxial tumor practice, to have completed a neuro-oncology fellowship and intraoperative mapping training, and to utilize intraoperative adjunctive therapies more frequently. Additionally, molecular factors such as IDH and MGMT methylation statuses appear to be important factors in ERR decision-making. This reflects a better understanding of the importance of tumor molecular makeup on patient survival.

General ERR Practices in GBM

While a greater EOR provides survival benefits,^12–15 it must be weighed against the risks of complications, as studies have shown that when new deficits occur the survival benefit from resection is lost.^16,17 Additional considerations when performing ERR include socioeconomic factors and the burden of an additional procedure. The survival benefit of ERR for tumors deemed amenable to GTR is largely unknown. A study evaluating ERR for GBM within 1 week of the initial resection demonstrated that ERR was feasible and safe to achieve GTR. However, ERR was performed in only 6% of patients, and its survival benefit was not assessed.¹⁸ A recent study evaluating the impact of ERR during the initial hospitalization showed that patients who had undergone ERR had improved progression-free and overall survival. However, ERR patients had a higher KPS, fewer initial resection complications, and less eloquent tumors.¹⁹ Therefore, these variables could bias the observed survival improvement. Surprisingly, in this study, 19% of the study population underwent ERR (32% of patients with residual disease), a high proportion of patients compared to that in a prior study evaluating neurosurgical ERR practice.²⁰ Moreover, current neurosurgical recommendations for such cases are relatively unknown, with few studies evaluating neurosurgeon practice. In the current study, we observed that most neurosurgeons do not regularly perform ERR for tumors that are deemed amenable to GTR. This finding agrees with that in a study evaluating the practice in the United Kingdom, in which even though 16.3% of patients had residual enhancing disease that was amenable to resection, no ERRs were performed. The lack of ERR in that study was largely attributable to the perceived absence of clinical benefit; however, the molecular factors that influence the prognosis of patients were not evaluated.²⁰ Interestingly, our study showed that Latin American neurosurgeons were more likely to perform ERR in their practice. Although the resection likelihood index was higher among our Latin American respondents, the resection threshold was similar between Latin America and the US/Canada (70% vs 80% respectively).

Neurosurgeon responses to hypothetical patient characteristics also shifted among regions. Latin American neurosurgeons were significantly more likely to consider ERR when a patient’s KPS was < 70 or their age was 18–65 or when they had a new deficit or no new deficit after the initial surgery. Although, overall, the neurosurgeons from Latin America were consistently more likely to pursue ERR, the spread of their responses relative to those of other countries differed. In the younger age group and the no new deficit group, they almost universally selected extremely or somewhat likely to pursue ERR, whereas other groups were less consistent in their responses. Meanwhile, in the variables associated with poorer outcomes after surgery (KPS < 70 and new deficit), the US/Canada and Europe/Asia more heavily stated that they were unlikely to perform ERR. The responses from Latin American neurosurgeons were more evenly distributed, and the most popular response was still “somewhat likely” to reoperate.

It is possible that the difference between neurosurgical approaches among the various regions is attributable to complex socioeconomic factors that push neurosurgeons toward ERR, as other therapeutics are not easily accessible. Radiotherapy, a well-known adjuvant therapy that improves survival,²¹ is not as accessible in low- and middle-income countries in Latin America.^22,23 Moreover, on average, Latin American countries invest only approximately $7–$8 per patient with cancer, a much lower amount than in countries like the US ($460) or Japan ($244).²⁴ Therefore, it is understandable that the surveyed Latin American neurosurgeons would be more likely to treat residual disease with ERR given the adjuvant treatment constraints that some patients face, in which resection becomes the sole option.

Influence of MGMT and IDH Status on ERR Decision-Making

Over the last decade, studies have revealed that the genetic makeup of gliomas has important prognostic implications.^25–29 The 2016 WHO classification incorporates IDH status into tumor diagnosis, providing for a more accurate prognosis.³⁰ Recent studies have demonstrated that the benefit of increased resection of contrast-enhancing tumor is consistent regardless of IDH and MGMT statuses.⁵ In the current study, we evaluated if the IDH and MGMT statuses would change the ERR decision-making of neurosurgeons in a referred patient. A considerable percentage of neurosurgeons deemed IDH (23.8%–43.4%) and MGMT (17%–26.4%) statuses as important factors when considering ERR.

Fewer neurosurgeons reported that MGMT was an important decision-making factor compared to IDH. This finding is interesting given that MGMT status plays an important role in treatment. MGMT unmethylated tumors respond poorly to temozolomide,^31,32 which has led some to argue against temozolomide use in these patients.³³ Given the lack of effective adjuvant chemotherapies in MGMT unmethylated GBM patients, resection may play a critical role in survival.

Among the neurosurgeons who reported IDH and MGMT as important factors in ERR decision-making, they were significantly more likely to choose ERR for MGMT methylated tumors than for unmethylated tumors. Recent studies have shown that the EOR cutoff for improved survival in MGMT methylated tumors can be as high as 98%, whereas unmethylated tumors only needed to reach a 70% EOR to observe a survival benefit.³⁴ This finding supports the respondents’ increased aggressiveness in MGMT methylated tumors. There was also a higher but not statistically significant resection likelihood index associated with IDH mutant cases compared to wild type. This lack of significance could be due to variability in surgeons’ responses to IDH cases depending on their anatomical locations. It is also important to note that IDH mutant tumors are almost always MGMT methylated.³⁵ In summary, neurosurgeons in this survey were more likely to pursue ERR when patients had tumor molecular characteristics associated with improved survival.³⁶ This increased likelihood may be related to the thinking that a greater EOR for these tumors further improves outcomes, although one could argue that a more aggressive surgical approach for IDH wild-type or MGMT unmethylated tumors should be considered, as adjuvant therapies are less effective. How molecular biomarkers shape the practices of neurosurgeons is a fascinating topic that will evolve as more biomarkers of prognosis and response to therapy are discovered and translated into clinical practice. Preoperative diagnosis through minimally invasive techniques such as liquid biopsies,^37,38 as well as intraoperative diagnostic technologies such as stimulated Raman histology and deep neural networks,³⁹ may be incorporated into surgical decision-making on an individual basis in the future.

Some other parameters variably impacted surgeon decision-making. Surprisingly, the setting of the case (early unintentional incomplete resection vs referral from an outside institution) did not change the mean ERR likelihood. Moreover, the location of the tumor did not significantly affect the EOR thresholds below which surgeons would pursue an ERR. However, we did note that the mean ERR likelihood indexes varied among tumor locations, with respondents most heavily favoring ERR in the right temporal, parietal, and occipital cases. Although, oddly, this increased aggressiveness was not noted when comparing EOR thresholds, it is understandable that GBM location would influence a surgeon’s aggressiveness with ERR, particularly in tumors located in the right (presumed nondominant) hemisphere. Prior studies have shown that postoperative deficits are linked to the functional grade of a tumor’s location.⁴⁰ Additionally, when a group of neurosurgical oncologists was surveyed about GBM resection preferences, right parietal, anterior temporal, and occipital lobe tumors were in the top six locations at which surgeons chose more aggressive resection options (supratotal resection),⁴¹ similar to our findings in Fig. 3.

Study Limitations

Although we were able to obtain a sizeable response from North and South America and European countries, our survey was not able to reach all continents equally. We had a small number of respondents from Africa and Asia, mostly because of difficulties in accessing their neurosurgeon networks. Also, we reached neurosurgeons only through preexisting networks; therefore, neurosurgeons who had chosen not to join these professional organizations were not surveyed. Finally, we were not able to collect the demographic variables of nonrespondents, and because we were unable to compare respondents to nonrespondents, we cannot exclude the possibility that our results were influenced by sampling bias.

The estimated survey completion time was 10–15 minutes. Survey length deterred some neurosurgeons from completing it. This decrease in responses as the survey progressed could have skewed our analysis.

Conclusions

Through a web-based survey, we investigated the heterogeneity in neurosurgeons’ ERR decision-making. Overall, neurosurgeons rarely performed ERRs in their practice (< 1% of all GBM cases), and case context (initial resection or referred patient) did not significantly influence a neurosurgeon’s ERR decision-making in our cases. However, neurosurgeons were more likely to pursue ERR in tumors with molecular markers that are associated with increased survival and in right parietal, anterior temporal, and occipital lobe tumors. Latin American neurosurgeons were more likely to pursue ERR at a higher EOR threshold and in more of our sample cases. The regional variation in responses could be due to a decreased caseload among the Latin American neurosurgeons we studied, as well as important socioeconomic barriers that some GBM patients face in accessing adjuvant therapy. Although this survey sheds some light on current views of ERR, it also highlights the variability in neurosurgeons’ approaches to ERR. Further studies evaluating the survival benefit of ERR in GBM patients may improve best practices regarding ERR.

Abbreviations :

5-ALA =

5 aminolevulinic acid

AANS =

American Association of Neurological Surgeons

CNS =

Congress of Neurological Surgeons

EOR =

extent of resection

ERR =

early repeat resection

GBM =

glioblastoma

GTR =

gross-total resection

KPS =

Karnofsky Performance Status

RV =

residual volume