Introduction

Brain metastases (BM) are the most common form of malignant central nervous system tumors, and their incidence is increasing due to improved systemic therapies and therefore survival rates of patients with solitary tumors. In recent years new treatment modalities have arisen mostly incorporating targeted and immunotherapeutic agents [1] and yet microneurosurgical resection with adjuvant radiotherapy has remained the mainstay for local control, especially for patients with solitary BM [2, 3]. Traditionally, microneurosurgical resection of BM has been considered primarily for solitary or symptomatic lesions, or in patients with unknown primary tumor for histopathological diagnosis. Multiple and/or asymptomatic metastasis, or those in deep or so-called “eloquent” regions are preferentially referred to radiotherapy, either in single or in multiple sessions [2]. However, the introduction of new therapeutic modalities and concepts [4,5,6] together with continuously improving microneurosurgical techniques warrants the need to reassess the value of surgical tumor burden reduction within multimodal treatment concepts [7], and the safety of microneurosurgical BM resection not only in solitary and superficial lesions but also in multiple, highly eloquent and deep-seated tumors.

Adverse events (AE), defined as any deviation from the expected postoperative course, can significantly affect patients’ outcome due to the delay of adjuvant or subsequent treatment with a possible negative effect on overall survival [8, 9]. Historically, the rate of AEs after radiotherapy, particularly radiosurgery, is assumed to be lower [10, 11] than after micro-neurosurgery [12,13,14], but modern and large series with a clear definition and systematic reliably reproducible grading of AEs are scarce, particularly with respect to different anatomical BM locations.

With the introduction of new treatment modalities for BM, it is crucial to reassess the safety of modern microneurosurgical BM resection especially considering multiple, central region and deep-seated lesions. Here we report a large series of 666 patients of which 537 underwent microneurosurgical resection in the last decade for histologically confirmed BM. Particular emphasis is put on surgical outcome and standardized assessment of AEs characteristics.

Methods

Study cohort, data acquisition and ethical considerations

All patients who underwent microneurosurgical resection or biopsy of histologically confirmed BM between January 2012 and June 2022 at our institution were included. Patient records were extracted from our prospectively recorded institutional registry [15]. The registry was approved upfront by the local ethics review board (PB-2017-00093) and internationally registered at clinicaltrials.gov (NCT01628406). Parameters extracted from the registry contained age, sex, American Society of Anesthesiologists (ASA) risk classification, length of stay (LOS), modified Rankin scale (mRS), Karnofsky Performance Scale (KPS), surgical characteristics and AEs at discharge. mRS and KPS were used as general clinical performance scales. Complications are defined as any deviation of the usual, expected postoperative course and are graded according to the Clavien–Dindo Grading system (CDG, see Online resource 1) [16,17,18]. For ease of handling, for each case only the complication with the highest CDG entered further analysis, if not stated otherwise.

All data were collected by staff neurosurgeons at the time of hospital admission, surgery, hospital discharge, and at each outpatient follow-up visit. Discharge reports are validated by the attending neurosurgeon responsible for the patient. In addition, all AEs are validated at the monthly department meeting and at the monthly morbidity and mortality meeting to ensure an accurate data collection. Data entry is performed by neurosurgeons only, and each new team member is provided with introductory training and written instructions in the form of a standardized operating procedure and is required to obtain a certificate on the correct use of the clinical and complication scores.

For each case, the number of craniotomies, surgical modalities such as intraoperative ultrasound, neuromonitoring and intraoperative magnetic resonance imaging (MRI), as well as number of metastases and anatomic gyral location of resected tumors were additionally extracted from the hospital’s electronic medical record system. Pre- and postoperative T1-weighted MRI images with gadolinium contrast were used to determine anatomical location of the BM and were confirmed by neuroradiological reports and an experienced consultant neurosurgeon. Tumor location was divided into infra- and supratentorial and further stratified into superficial and deep-seated. Deep lesions were defined as tumor location in the cuneus, precuneus, corona radiata, basal ganglia, thalamus, cingulate gyrus, ventricles, operculum, medial and lateral occipitotemporal gyrus, orbital gyrus, insula, clivus, parahippocampal gyrus, corpus callosum, pineal region, gyrus rectus, medulla oblongata, pons as depicted in Online resource 2.

Statistical analysis

All data processing and analysis steps were performed with R Studio (Version 1.4, R Studio Inc.) [19] using open-source libraries. Plotting of anatomical locations on a reference brain atlas was done using the coldcuts R package [20]. Missing values were considered missing at random and therefore omitted from all analysis. Continuous variables are given as means and standard deviation (SD) whereas categorical variables are reported as numbers and percentages of total. Uni- and multivariate logistic regression analysis was performed to find predictive features for AEs. The statistical tests used are additionally indicated in the figure captions or the main text. P-values < 0.05 were considered statistically significant. Further study cohort stratification was based on the occurrence of major AEs at discharge which are defined as the occurrence of any AE with CDG ≥ 2 or a new neurological deficit at discharge.

Data and script availability

Raw data and analysis scripts are available from the corresponding author upon reasonable request.

Results

Study cohort characteristics

A total of 664 patients which underwent surgery of BM were included in the study cohort. Sex was roughly equally distributed with 51% male (n = 337) and 49% female (n = 327). Mean patient age was 61 years and the BM most frequently originated from the lung (44%, n = 294) followed by melanoma (25%, n = 163) and breast cancer (11%, n = 74) as depicted in Table 1. In most patients a singular BM was present (50%, n = 336), whereas 2–4 lesions were present in 33% and > 10 lesions in 8.6% of the patients (see Table 1).

Table 1 Patient cohort and surgical characteristics
Full size table

BM and surgical characteristics

The most frequently used intraoperative tools were intraoperative neuronavigation (89%) and intraoperative ultrasound (78%). Not surprisingly, neuromonitoring (19%) and intraoperative MRI (4%) were used less often as shown in Table 1. Most of the surgeries were resections (96%) via one craniotomy (89%), while 6% (n = 40) of the patients underwent two and four patients underwent three craniotomies during the same surgery. Biopsy only cases accounted for 4% (n = 27) and 19% of cases were reoperations with previous BM resections (see Table 2, Overall).

Table 2 Lesion and surgery characteristics
Full size table

78% of cases were located supratentorial, whereas 22% were located infratentorial (n = 145, see Table 2, Overall). Of the supratentorial lesions, 71% had a convexity and 29% a subcortical/deep location. 11% (n = 76) BM were located in the central region, which is comprised of the pre-, post- para- and subcentral gyrus [21]. Online resource 2 and Fig. 1 illustrate the gyral localization of the resected BM in the study cohort. Roughly half of the lesions were either located on the right or left side (50% vs. 44% respectively) while the minority were in the midline (2.6%, n = 17).

Fig. 1
figure 1

Anatomical BM location. BM brain metastases

Full size image

Frequency of adverse events and clinical outcomes

The occurrence of any AE at discharge was recorded in 8% (n = 53, see Online resource 3) of the cases, most of which were recorded as CDG 1 and 2 (see Online resource 3), meaning that these AEs did not require any invasive interventions, but only pharmacological or no treatment. Severe AE requiring invasive interventions (CDG 3) or ICU stay (CDG 4) were present in only 1.9% (n = 12) and the mortality rate was 0.6% (n = 4, CDG 5). New neurological deficits occurred in 2% of the entire study cohort (n = 13, see Online resource 3), followed by urinary tract infections, pulmonary artery embolism (each 0.8%, n = 5), postoperative hemorrhage and pneumonia (each 0.6%, n = 4).

Stratifying the study cohort according to the occurrence of major AEs (defined as CDG ≥ 2 or new neurological deficits at discharge), Table 2 shows the occurrence of major AEs for the different anatomic and surgical parameters. Cases with major AEs were more frequently located infratentorial (40% vs. 20%, p ≤ 001, Pearson’s Chi-squared test) or in the midline (8% vs. 2%, p = 0.013, Fisher’s exact test; see Table 2). Major AEs were not associated with deep-seated lesions, the number of craniotomies or emergency operations (see Table 2). Figure 2 shows the postoperative change in clinical outcome scales for the two groups: Patients who experienced an AE had higher mRS scales at discharge compared to admission (see Fig. 2B) compared to patients who did not experience any AE (see Fig. 2A). In addition, considering the relative changes in mRS (see Fig. 2C) and KPS (see Fig. 2D) between discharge and admission, more patients with an AE suffered from a worsening of mRS (44%) and KPS (49%) compared to patients without AEs (11% each, p < 0.001 for mRS and KPS, Pearson’s chi-squared test).

Fig. 2
figure 2

Clinical outcome scale changes at discharge. Percentages of mRS scores at admission (upper row) and discharge (lower row) for patients without AE (A) and for patients with AE (B). Changes of mRS (C) and KPS (D) at discharge relative to admission stratified for the occurrence of AE. AE at discharge were considered as CDG ≥ 2 or new neurological deficits. AE adverse event, mRS modified Rankin Scale, KPS Karnofsky Performance Status

Full size image

Logistic regression and survival analysis

In univariate logistic regression analysis, ASA status (OR 1.6, 95% CI 1.01–2.54, p = 0.044), tumor location in the midline (OR 3.94, 95% CI 1.04–12.3, p = 0.037) and cerebellar tumor location (OR 3.44, 95% CI 1.73–6.95, p = 0.005) were significant predictors of postsurgical major AEs (see Online resource 4). Furthermore, in a subsequent multivariate logistic regression analysis, cerebellar BM location (OR 5.46, 95% 2.31–13.8, p = 0.001), reoperation (OR 2.31, 95% 1.07–4.81, p = 0.033, see Table 3) and central region tumors (OR 3.03, 95% 1.03–8.60) were all associated with the occurrence of postoperative AEs. Looking at overall survival (OS), Kaplan–Meier curve analysis showed significant differences in OS between patients with major AEs and the ones without AEs (p = 0.044, log-rank test, see Online resource 5). Additionally, patients with higher BM load (n ≥ 5 BM) showed significantly impaired OS rates compared to patients with fewer BM (n < 5; p < 0.001 log-rank test, see Online resource 6).

Table 3 Multivariate logistic regression model of major AE at discharge
Full size table

Discussion

The treatment of patients with metastatic cancer is rapidly changing due to the introduction of novel therapeutic modalities, most notably targeted and immunotherapeutic systemic agents [1]. This resulted in an improved overall survival in these patients and an increasing incidence of the development of BM [4,5,6]. However, due to the changing multimodal treatment regiments in BM patients, the role and safety of microneurosurgical BM resection needs to be reassessed in the light of other treatment modalities. Technological development has advanced modern micro-neurosurgery forward with numerous technical developments to maximize intraoperative patient safety and improve the extent of resection [22,23,24,25,26]. This is reflected in the low overall rate of AE in our patient cohort with only few AEs considered as severe based on the CDG grading and without any difference between cortical, central or deep-seated lesions. Our data shows that modern microneurosurgical resection can be considered as safe also in the case of deep located or central region lesions. Infratentorial BM or non-primary surgical resections both were independent predictors of postsurgical major AEs, which has also been described in the literature before [12, 27], whereas central region location did not reach statistical significance in our cohort. Extra care in perioperative management must be applied in these situations to avoid AEs.

Patients with AEs showed decreased overall survival in our Kaplan–Meier analysis, which might be due to the subsequent delay of postoperative therapies (which is already known from glioma surgery [8]) with a corresponding effect on overall survival. Furthermore, patients with AEs did show worse functional outcome as measured by the mRS and KPS scales. Since both overall survival and functional outcome decreased after AEs, the importance of perioperative safety must be emphasized.

The reported rates of AEs might appear at first sight similar to previous reports [12, 14]. However, on the one hand modern case series are scarce [14, 28, 29], on the other hand AEs are often reported in a non-standardized or non-reproducible manner, or focused primarily on “neurosurgical relevant” AEs [13, 30] leading to a possible underestimation of total number of AEs. We adopt a broad definition of AEs as any deviation from the normal postoperative course. As suggested in the literature [4], this methodological approach is more rigorous in assuring that no AE is missed.

In general, there seems to be a continuing trend of decreasing incidence of AEs, presumably due to newly introduced technical nuances that increase operative safety. If in the ‘80s reported AE could be as high as 27% [28, 30, 31], more recent series report a AE rate of 10–12% [12], Our results, although they originate from a prospective registry (which notoriously leads to much higher reported AE rate [32]) and rely on a broader definition of AE, seem to suggest an even lower AE rate. As such, based on our data, modern microneurosurgical resection can be considered as safe in terms of the overall AE rate and severity. Recent studies of radiosurgical AEs reported an overall AE rate similar to our findings with 2.9% [11], 7% [10] and 6.6% [33] respectively. Interestingly, the definition of AEs was rather narrow and included mostly the occurrence of new neurological deficits only, without considering other forms of AEs as it is the case of the CDG grading. Still, our data suggest that in the current era, thanks to continuous improvement of microneurosurgical technique, the overall rate of AEs after microneurosurgical BM resection might be comparable to that after radiotherapy both in incidence and in severity, although our study was not designed to specifically address this issue.

In our study, anatomical infratentorial tumor location as well as reoperation were shown to be associated with postoperative AEs, which has been described previously [12, 27]. Particularly interesting appears to be the fact that the rate of AEs for supratentorial lesions did not seem to significantly differ between superficial, central and deep-seated BM. This confirms the role of microneurosurgical resection techniques [21, 34] and supports that microneurosurgical resection can be safely performed also in the case of central as well as deep-seated lesions.

We are completely aware of possible limitations of this study, which consists of its retrospective as well as single-center design and the inter-surgeon variability regarding the anatomical localization of the lesions. However, regarding the last point, each anatomical localization was confirmed by the authors of this study and thus bias should be minimized. Taken together, the AE rates presented in this study should be compared to other centers to confirm the overall low incidence of severe AE following microsurgical resection of deep or superficially seated BM.

Conclusions

Microneurosurgical resection is an excellent option in modern treatment of BM in terms of safety. Incidence of any AE, particularly severe ones, is low even in multiple, central and deep-seated lesions. Infratentorial tumor location and reoperation are associated with postoperative AEs and should warrant a closer perioperative patient monitoring.