The primary objective of neurosurgery for a neuro-oncological condition is to decrease local recurrence rates while avoiding neurological harm. Despite advancements in adjuvant therapies, achieving complete resection continues to influence prognoses. The aim of this review was to examine the concept of complete resection in neuro-oncological surgical procedures, exploring its definition and its implications for existing and upcoming intraoperative techniques.

The definition of complete resection in neurosurgical literature varies based on factors such as tumor type, postoperative investigation methods, delay after surgery, and surgical goals. In earlier studies, it might have relied on the surgeon's perioperative impression, a practice that still occurs in clinical settings. However, contemporary studies increasingly rely on more objective assessments. Neurosurgery provides notable examples of diverse complete resection definitions for different tumors such as meningiomas, glioblastomas, metastases, and low-grade gliomas. An essential criterion for defining complete resection is the tumor resection margin, whether microscopically positive or negative. Generally, the goal of tumor resection is to attain a microscopically negative margin while minimizing complications. The optimal margin varies, contingent upon tumor characteristics, patient circumstances, and techniques used to detect tumor cells at the resection boundaries. Additionally, numerous tumor types infiltrate brain tissue, complicating matters.

For meningiomas, postoperative complete resection usually means complete removal of the contrast-enhanced mass on postoperative magnetic resonance imaging (MRI), which corresponds to Simpson grades 1–3—macroscopically complete removal of the tumor with excision of its dural attachment (grade 1), or with coagulation of its dural attachment (grade 2), or without it (grade 3),1 i.e., based on the resection margin. Decades of literature show that complete or supra-complete resection should be performed whenever possible, but that the aim of surgery depends on the tumor location.2 Comparing surgeon-assessed resection with 68Ga-DOTATATE positron emission tomography (PET)/computed tomography results revealed a 40% rate of false-negative evaluations, especially for convexity and falcine meningiomas. This suggests that many purported complete resections are incomplete, casting doubt on prior literature.3

For malignant tumors, complete resection means complete removal of the contrast-enhanced mass on MRI performed <72 hours postoperatively. Although it is proven that glioblastoma cells invade the whole brain,4 authors often use the term complete resection to refer to a contrast-enhanced lesion complete resection, which can be misleading, especially for patients or clinicians not familiar with the clinical course of the disease. In contrast, complete resection of a metastasis might involve removing all metastatic cells, sometimes with a healthy margin. Nevertheless, even after a so-called complete resection of the metastasis, local radiotherapy has shown a benefit in terms of local recurrence, showing it is indeed often subtotal.5 For low-grade gliomas, complete resection is based on fluid attenuated inversion recovery signal on MRI. Therefore, the extent of removal is based either on anatomic landmarks such as the sulci or on functional limits, e.g., in awake surgery. Because tumor cells invade the brain further than MRI can show, some neurosurgeons argue that supratotal removal, based on functional landmarks, is the best treatment making the concept of complete resection obsolete.6

These examples underscore that complete resection is a more intricate concept than it appears. Furthermore, neurosurgery faces challenges in achieving complete resection in certain locations due to factors such as interruption of functional white matter tracts, cortex preservation, cranial nerve protection, or vascular structures. The advancement of adjuvant treatments has shifted the balance between neurologically acceptable surgery and oncological objectives. Survival enhancements can be achieved without subjecting patients to unreasonable surgical risks, thanks to radiotherapy or targeted chemotherapy.

To assess whether complete resection remains a significant predictor of outcomes in neuro-oncological surgery given the advancements in radiosurgery and targeted therapy, we conducted a comprehensive literature review. Our aim was to analyze the prevalence of local recurrence following cranial surgery and identify associated risk factors across various tumor types. Employing a systematic Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) approach, we reviewed studies related to local tumor recurrence in neurosurgery on PubMed from the years 2012–2022. Our inclusion criteria encompassed cases involving adult patients, with a focus on series containing at least 10 cases. We excluded pediatric cases, literature reviews, articles lacking details on surgical extent, and non-English publications. Our data collection process involved extracting demographic, surgical, recurrence, and histological information from each selected article. Statistical analyses were conducted using Microsoft Excel 2010 (Microsoft Corp., Redmond, Washington, USA). Further comprehensive details of the review can be obtained by reaching out to the authors.

For meningiomas, the extent of resection was complete in 70 ± 10% of cases.7, 8, 9, 10, 11 Recurrence was observed in a median 15 ± 3% of patients (ranging from 4% in a mixed cohort of meningiomas,12 to up to 100% for a cohort of malignant meningiomas13 after 50 ± 7 months. Among all patients undergoing meningioma surgery, 7.9% were underwent reoperation for a local recurrence.14 As expected from long-term data, complete resection was identified as an independent predictive factor for recurrence, but was associated with higher neurological impairment.7 Two cohorts included only completely resected tumors, with a long-term recurrence rate of 15% for histological grade 1 meningiomas15 and 19.6% for mixed histological grades.16

For glioblastomas, complete and near-total resection (>90%) are often deemed similar. With this in mind, the extent of resection was complete in 50 ± 10%. Although long-term recurrence is 100%, the time to recurrence varied depending on adjuvant therapies from 6 to 7.2 months,17, 18, 19 and the decision for a second surgery from 22% to 100%.17,18,20,21 The use of perioperative techniques to achieve gross total resection was highly variable, even in single series. One study retrospectively comparing 11C-methionine amino acid PET with MRI to evaluate local recurrence, as defined by histological result following a redo surgery, found that amino acid PET is more reliable, with up to 22% discordance with MRI sequences.22

For metastases, data collected from series focusing on local recurrence included patients with all types of primary cancer (non–small cell lung carcinoma, small cell lung carcinoma, melanoma, breast, colorectal, renal cell, ovarian, cervical, gastric, testicular cancers, sarcoma). Complete resection, as defined on postoperative MRI, was achieved in 66 ± 9% of cases, and the local recurrence rate varied from 12% to 57%, depending on the duration of follow-up and the type of adjuvant treatment.5,23, 24, 25, 26 Incomplete resection on brain MRI is illustrated in Figure 1. Even in patients with complete resection,5 local recurrence occurred in up to 57% of patients in the first year in the absence of local adjuvant radiotherapy. Interestingly, after supratotal resection, the analysis of the peritumoral region showed that neoangiogenesis, in the absence of tumor cells, was predictive of local recurrence.24 In all series, the overall survival was influenced by the presence of multiple metastases, the type and efficacy of adjuvant treatments, and the histological characteristics of primary cancer, making cerebral metastasis extent of resection a secondary factor.

Numerous intraoperative techniques are available to help the neurosurgeon identify tumor cells, to avoid leaving any tumor remnant. Recent systematic reviews of intraoperative techniques allowing maximal safe resection in glioblastomas and gliomas found that all techniques (sodium fluorescein, 5-aminolevulinic acid [5-ALA], neuronavigation, intraoperative MRI, intraoperative ultrasound, intraoperative cellular and molecular profiling of tumors, improved microscopic imaging, intraoperative mapping, and augmented and virtual reality) increase the extent of resection and that combining those techniques leads to the best results.27, 28, 29 In addition to anatomical tools, such as neuronavigation, intraoperative imaging may offer insight into the tumoral nature of the tissue. All techniques described hereafter can in theory be coupled with imaging and mapped on neuronavigation or preoperative MRI. Figure 1 shows techniques used in clinical practice and in clinical research to maximize neurosurgical outcomes. Although each technique has advantages and pitfalls (Table 1), conservative methods will be presented separately from invasive methods, as their use in neurosurgical settings is fundamentally different.

Morphological imaging remains the standard information needed to perform resection surgery. Because the tumor is identified on MRI, intraoperative MRI is the gold standard to achieve an MRI-based complete resection. Notwithstanding its practical and financial burden, it seems efficient to improve the extent of resection and survival.30 Intraoperative ultrasound, especially B-mode ultrasound and elastosonography, which reflects the tissue mechanical properties, has shown good results to define tumor margins and infiltration in both high- and low-grade gliomas.31 Optical coherence tomography is a promising tool in terms of microanatomy, detects back-scattered near-infrared light, and is harmless to biological tissue, offering a spatial resolution of 1–15 μm.32 It has been shown to differentiate healthy tissue from brain melanoma metastases and from high-grade glioma,33,34 but its intraoperative relevance is mostly in identifying vascular structures, e.g., in extra-axial tumors. As such, it is a relevant intraoperative tool for noninvasive tumors such as meningiomas, where anatomical dissection of functionally essential vessels is the major obstacle to complete resection.32

To compensate for the limits of frozen sections, discussed under Intraoperative Invasive Techniques, intraoperative neurosurgical confocal laser microscopy has shown feasibility of cellular level imaging during neurosurgery after fluorescein sodium injection.35 The microscope produces grayscale images of tissue histoarchitecture, providing valuable diagnostic clues that need to be validated by an experienced neuropathologist. The clinical benefit of this system, which is available commercially, is under investigation. Its reliability in vivo and ex vivo has been proved retrospectively in some cases of meningiomas, gliomas, and brain metastases.35

More specific techniques allow direct identification of tumor cells based on fluorescence. Endogenous autofluorescence has proved useful to separate glioma from healthy tissue due to time-resolved fluorescence spectroscopy in preliminary clinical data.36 However, exogenous fluorescence-guided surgery, provoked by fluorophore intake preoperatively, is the only fluorescence-based technique available in clinical practice, using 5-ALA, the metabolism of which specifically leads to the accumulation of protoporphyrin IX in cancerous cells. Other fluorophores have been tested over the decades, but 5-ALA is the only one used in clinical practice due to its sensitivity, specificity, and low phototoxicity.37 5-ALA fluorescence–guided surgery, using a specific microscope filter, is mostly used in glioblastoma surgery, where its systematic use improves gross total resection from 36% to 65%,38 but its use is also proposed for other tumors such as metastases.24 As a shift of paradigm, in addition to allowing intraoperative photodiagnosis, 5-ALA is also a photosensitizer and is used in clinical trials to perform local postresection phototherapy, combining the surgeon's removal with a cellular cleaning of the walls of the cavity.39 Near-infrared visualization of tumors after indocyanine green injection is used in solid tumor surgery, but is rarely considered in neurosurgery, with only 1 pilot series reporting meningioma and dural tail visualization with a high sensitivity.40 Research to overcome current limitations of fluorescence-guided surgery offers quantitative imaging techniques as a solution to subjectivity measurements, interobserver dependence, and decreased sensitivity for residual disease.41

Also based on the injection of a specific marker, nuclear methods led to the development of intraoperative PET probes, both for gamma and beta detection. Any radiotracer validated for PET imaging can be used, making these techniques highly versatile and preoperatively individualized42,43: 99mTc for gliomas44 or 111In-DTPA-octreotide or 111In-pentetreotide, radio-labeled somatostatin receptor analogues, for meningiomas. Data about safety for the operating room personnel are reassuring with low exposure, which allows daily use of the technique, with appropriate radioprotection controls.43 Radioguided resection led to a significantly higher rate of radical resection in a series of patients with gliomas, without prolonging surgery duration.44 In meningiomas, radioguidance improved resection in cranial base meningiomas.45,46 Preoperative and postoperative nuclear imaging, either scintigraphy or PET/computed tomography/MRI, allows determination of the tumor uptake and resection control. The advantage of beta detection, although more technically challenging because of nuclear background elimination, is the short distance detection that allows a lower radiotoxicity and a more precise spatial resolution, approximately 1 mm.47, 48, 49 Clinical trials are ongoing to evaluate short-term and long-term clinical benefits in neurosurgical patients as well as most types of solid cancer surgery.50

Several promising spectroscopy-based methods are currently under development and are poised to become integral tools in clinical practice. Among these, intraoperative Raman spectroscopy emerges as a potent technique enabling noninvasive molecular analysis of brain tissue.51, 52, 53 Whereas magnetic resonance spectroscopy effectively quantifies specific metabolite concentrations on preoperative imaging, Raman spectroscopy identifies molecular structures and compositions, enabling the differentiation between healthy and tumor brain tissue. Grounded in the principle of Raman scattering, this method entails subjecting a sample in the surgical cavity to monochromatic light, inducing shifts in frequency attributed to molecular vibrations. These frequency shifts are indicative of the molecular makeup of the sample, thereby enabling discrimination between healthy and tumoral brain tissue. However, it is essential to acknowledge the presence of interindividual and intraindividual variations, stemming from the complex molecular heterogeneity inherent in glioblastomas. To expedite clinical integration, intraoperative systems have already been devised and are presently undergoing validation procedures, mostly refining data analysis methodologies and ensuring the reproducibility of results. Still in a research setting, hyperspectral analysis, based on hyperspectral images of in vivo brain tissue captured and processed during neurosurgical operations, is also able to discriminate between normal and tumor tissue.54

Although techniques based on the photoacoustic effect are used in solid cancer imaging, e.g., breast,55 it has been scarcely reported in neurosurgery. When a biological tissue is exposed to a short pulse of laser light, the absorbed energy causes local thermal expansion, generating ultrasound waves due to the sudden increase in pressure. The ultrasound spectrum is processed to create images. It has been shown in vitro and in animals that this effect can be used to differentiate healthy brain from glioma, the specificity and contrast enhancement being improved by injection of nanoparticles.56

Mass spectrometry is a destructive technique, where the spectrum of molecules present in the tissue is evaluated according to their molecular mass. It has been improved as almost noninvasive due to the invention of desorption electrospray ionization mass spectrometry, which allows analysis of microscopic samples from the margins.57 In a cohort of 10 patients, desorption electrospray ionization mass spectrometry was shown to be feasible and reliable in glioma surgery, with a high correlation with frozen section results, based on lipid or N-acetylaspartate58 or 2-hydroxyglutarate content.59 Similar to all spectrum analysis methods, mass spectrometry gives an exhaustive description of the tissue and can distinguish molecular subtypes of tumors according to quantitative analysis of tissue composition, such as IDH mutation status in glioma.58 Although this opens infinite diagnostic opportunities, it makes the process of validating healthy versus tumoral brain difficult—rather high-infiltrate versus low-infiltrate versus healthy—due to the heterogeneity and molecular complexity of both tissues and requires a thorough medical assessment of the final results obtained with refined data analysis software.

Intraoperative invasive techniques include techniques that are performed during the surgery on excised samples, the results of which are communicated to the neurosurgeon so that the surgical strategy can be adapted. Frozen section histological analysis remains the gold standard for determining the tumor boundaries, but it is time-consuming, requires an available expert pathologist, and samples are often small. Many techniques are developed to reduce the preparation time of samples, avoid staining, and make the microscopic images available digitally, allowing remote and multiple analyses of the images—fundamentally variations of ex vivo confocal microscopy.35 Once the tissue has been resected, virtually any analysis can be proposed on the sample, which includes all discussed techniques, depending on their relevance and timely acceptability. Artificial intelligence is also often used to increase efficacy in all the procedures, which raises the question of the need for an expert pathologist to confirm the diagnosis, creating a legal responsibility void regarding the surgeon's responsibility in using artificial intelligence–generated results in his or her intraoperative decision.