Clinical and prognostic implications of rim restriction following glioma surgery

Rim restriction surrounding the resection cavity of glioma is often seen on immediate post-op diffusion-weighted imaging (DWI). The etiology and clinical impact of rim restriction are unknown. We evaluated the incidence, risk factors and clinical consequences of this finding. We evaluated patients that underwent surgery for low-grade glioma (LGG) and glioblastoma (GBM) without stroke on post-operative imaging. Analyses encompassed pre- and postoperative clinical, radiological, intraoperative monitoring, survival, functional and neurocognitive outcomes. Between 2013 and 2017, 63 LGG and 209 GBM patients (272 in total) underwent surgical resection and were included in our cohort. Post-op rim restriction was demonstrated in 68 patients, 32% (n = 20) of LGG and 23% (n = 48) of GBM patients. Risk factors for restriction included temporal tumors in GBM (p = 0.025) and insular tumors in LGG (p = 0.09), including longer surgery duration in LGG (p = 0.008). After a 1-year follow-up, LGG patients operated on their dominant with post-op restriction had a higher rate of speech deficits (46 vs 9%, p = 0.004). Rim restriction on postoperative imaging is associated with longer duration of glioma surgery and potentially linked to brain retraction. It apparently has no direct clinical consequences, but is linked to higher rates of speech deficits in LGG dominant-side surgeries.

Incidence and risk factors. Rim restriction was demonstrated on the postoperative DWI studies of 68 patients (25%), including 20 LGG and 48 GBM patients.
We did not find significant differences in the rate of post-op rim restriction between LGG and GBM patients (32% vs 23%, p = 0.184).
We found no significant association between postoperative rim restriction and the evidence of blood in the surgical cavity that was demonstrated on the postoperative CT of 30% of the restriction cases and 28% of the non-restricted ones in LGG (p = 0.864), as well as 51% of GBM patients with restriction vs 49% in those without restriction (p = 0.177).
A univariate analysis that compared patients with and without rim restriction with respect to demographic, clinical, and pathological parameters showed no significant differences between those with and without rimrestriction in the combined GBM + LGG cohort (Table 1). Among GBM patients, rim restriction was more common in tumors located in the temporal lobes (32% Vs 18% in non-temporal areas, p = 0.025) and non-significantly more common in LGG of insular location (54% vs. 24%, p = 0.09). The calculated risk (odds ratio, OR) for post-op rim restriction among temporal GBM cases was 1.74 (95% CI 1.09-4.10, p = 0.027). www.nature.com/scientificreports/ IOM. Full IOM of MEPs, anesthesiology and awake monitoring reports were available in 152, 267 and 98 of the combined study population (n = 272), respectively. Neither MEPs changes nor awake intraoperative monitoring parameters showed any significant associations with rim restriction. Duration of anesthesia in the combined GBM and LGG cohort was significantly longer among patients with rim restriction, when compared to those without it (306 ± 87 min and 276 ± 81 min, respectively, P = 0.009). The difference was even more pronounced in the LGG group analysis (375 ± 17 min vs 316 ± 13 min; p = 0.008), Table 2.
Motor deficits. New or worsening immediate post-op motor deficits among LGG occurred in 4/20 (20%) of those with post-op restriction, similar to the rate in the non-restriction group (10/43 ,23%, p = 0.772). The rates were 9/48 (19%) in patients with post-op rim restriction and 19/161 (12%) of those without restriction (p = 0.215) in the GBM sub-population. In either group, no significant changes were noted in the rate of motor deficits over time, Tables 3 and 4.
Speech deficits. Speech deficits were analyzed in a subgroup of patients that underwent surgeries for tumors located in the dominant hemispheres (33 LGG and 103 GBM patients). We found a significantly higher rate of LGG patients with speech deficits in the post-op follow-up period in the restriction group, as compared to the non-restriction group (p = 0.004). All rim-restriction cases with speech deficits (n = 7) involved eloquent areas: either the left frontal, temporal or insular lobes. The percentage of patients with speech deficits in the post-op period decreased from 63 to 46% in the restriction group (p = 0.873) and from 18 to 9% (p = 0.989) in the nonrestriction group after 1 year of follow-up, Table 3. Among GBM patients, no significant differences were noted in the rate of speech deficits in the post-op follow-up period between the restriction and non-restriction groups (p = 0.226), Table 4. Table 2. Correlations between post-op rim restriction on diffusion-weighted imaging and intraoperative parameters in the entire study population, high-grade glioma subgroup and low-grade glioma subgroup: intraoperative monitoring, awake surgery monitoring and anesthesiology parameters (univariate analysis). IOM intra-operative monitoring, MEP motor evoked potentials, ASA American Society of Anesthesiology score, TIVA total intra-venous anesthesia, SBP systolic blood pressure, MAP mean arterial pressure, SEM standard error of mean.

Entire study population
Low-grade glioma Glioblastoma

No restriction (n = 90) Restriction (n = 28) Sig
MEPs decline (%) 13 Tables 3  and 4). Cognitive function tests, as previously described 11 , were performed in 33 LGG patients before and 3 months after surgery. We compared the results of 6 patients with post-op rim restriction and 27 with no restriction and could not detect any significant differences between the 2 groups in any of the functions assessed. Pre-and post-operative global cognitive scores were 98 ± 11and 95 ± 9 in the restriction group, and 95 ± 9 and 95 ± 9 in the non-restriction group, respectively (p = 0.194).

Discussion
To the best of our knowledge, this is the first study to specifically describe the incidence as well as the short-and long-term clinical correlations of a phenomenon, defined as rim restriction on postoperative imaging following resection of low-grade gliomas and GBM, while excluding cases that got complicated by ischemic stroke. The relevance of this data lies in the question whether this quite prevalent abnormal post-operative imaging finding is linked to certain intra-operative parameters and can predict surgical outcomes.
Rim restriction was seen on the postoperative DWI studies in 32% and 23% of LGG and GBM patients, respectively. The incidence rates of postoperative rim restriction reportedly ranged from 16 to 39%, yet previous studies encompassed various types of tumors and also included patients that sustained intra-operative ischemic strokes, which as well present with restrictive changes on post-operative DWI. 12,13,23 We found that the duration of surgery was significantly longer among patients that developed rim restriction on postoperative imaging as opposed to those without ( Table 2) and these differences were particularly noted in the LGG group. Proposed mechanisms for the development of postoperative DWI restrictive changes included local intraoperative brain ischemia by direct vascular damage, coagulation, or vasospasm, as well as kinking of small arteries or mechanical tissue pressure by brain retraction 24,25 . Prolonged brain retraction is a potential cause for rim restriction in surgeries of longer duration and minimizing retraction might reduce its occurrence 9,24-26 .
We found that rim restriction is more common after surgeries for GBM involving the temporal lobe (p = 0.025) and tends to be higher in LGG surgeries involving the insula (p = 0.09). Previous reports demonstrated a higher    www.nature.com/scientificreports/ risk for restrictive changes on DWI as well as for ischemic complications in insular region surgeries 13 . The insula is mainly supplied by frequent perforating arteries with no collateral flow, and surgeries in this location often require prolonged brain retraction 13,24-28 . As we often use retractors during trans-cortical resections of either temporal or insular gliomas, we raise the possibility that even mild surgical retraction in a diseased temporal lobe might be enough to induce the micro-ischemic changes which are often seen post-op. We did not detect any additional peri-operative parameters to be associated with the occurrence of rim restriction, including previously reported ones, such as recurrent operations and decreased MAP or IOM abnormalities (Tables 1  and 2). 15,25,26,29 Findings of rim restriction have been associated with evidence of remnant blood products in the surgical cavity on postoperative imaging, yet no significant association were seen in our study 9,30 . We found a higher rate of speech deficits among LGG patients with rim restriction after they had undergone dominant-side surgeries in either the frontal, temporal or insular lobes, even after adjusting for the duration of surgery. (Table 3). It was previously reported that rim-pattern DWI abnormalities were non-significantly more common among patients with new postoperative neurological deficits (35%), as opposed to those without (25%), p = 0.177 10 . However, most studies viewed this imaging pattern as displaying either normal postoperative changes or as part of a larger group of ischemic complications, which did not enable subgroup analyses 9,13,15,23 . It is possible that rim restriction in dominant-side surgeries, particularly insular cases, is an indicator for subtle parenchymal injuries during surgery that manifest in speech deficits. Such injuries may be related to retraction forces applied on the tissue, as discussed above 24,27,28 , which seem to lead to a long-lasting damage, as there was no significant decrease in the rate of speech deficits over time. The GBM population showed a significant decrease in the rate of seizures following surgical resection, and these results are in line with previous reports regarding the beneficial effects of surgery in alleviating seizures among glioma patients, yet no differences were noted in relation to post-op restriction status. 31,32 .

No restriction Restriction No restriction Restriction No restriction Restriction No restriction Restriction No restriction Restriction
Interestingly, the GBM population showed a unique yet non-significant trend towards increased overall and progression-free survival among patients with post-op rim restriction (p = 0.076 and p = 0.065, respectively). We could not clearly explain this trend and additional research is needed in order to validate and explore these findings.
Glioma patients often sustain various preoperative cognitive abnormalities that may further deteriorate after surgery, depending upon the eloquence of tumor location. These functions often improve when analyzed up to 1 year after surgery 33,34 . With mild cognitive abnormalities having been related to small strokes in proximity to the resection cavity 35 , we now searched for any association between rim restriction and cognitive changes following surgery, yet our cohort was too small to detect significant differences between the groups.
Interestingly, a recent study on perilesional rim restriction after surgical removal of convexity meningiomas found that restriction thicker than 1 cm was associated with post-op neurological deficits that lasted more than 3 months, including motor and speech deficit and seizures, with the majority improving over time. Main risk factors for post-op restriction were increasing age, intra-operative blood loss, tumor location over the motor strip and pre-op peri-tumor edema 36 .
The main limitation of this study is its historical cohort design, which forced us to estimate and define the degree of several study parameters, such as MRS and KPS, based on the study patients' medical records. In addition, in the case of LGG, the study follow-up period was too short (5 years) to reveal differences in long-term survival between groups of patients. Furthermore, molecular IDH status was not available for the entire study population, as it included patients treated between the years 2013-2017. Finally, this study was not large enough to perform certain subgroup analyses, such as cognitive changes following dominant-side surgeries.

Conclusions
Rim restriction on postoperative imaging is associated with longer durations of glioma surgeries, particularly of LGG, and more often occurs in cases of temporal and insular tumors. These imaging findings generally showed no apparent direct clinical consequences in either LGG or GBM, but they might be linked to a higher rate of speech deficits in dominant-side LGG surgeries.

Study population. Our historical cohort included patients who underwent resection of LGG and GBM
between January 2013 to December 2017. Preoperative clinical and intraoperative monitoring data were retrieved and documented, as were short-and long-term clinical outcomes. We selected patients that underwent surgical resection of LGG (World Health Organization 1 or 2) or GBM (WHO 4), and who had a full radiological and clinical dataset. The 2016 WHO classification became available in the middle of our study follow-up period (2013-2017). Due to the high ambiguity of WHO grade 3 that occasionally overlaps with grade II, but often tends to be more aggressive and overlap with grade IV, we decided not to include patients with the diagnosis of anaplastic astrocytoma in order to keep both study groups as uniform as possible. Patients who only underwent biopsies and those without full radiological and clinical data were excluded. In addition, in order to specifically focus on peri-resection rim restriction on post-operative imaging, we excluded patients who sustained intraoperative ischemic stroke (30 GBM and 19 LGG patients). The most recent surgery was considered as the index one in cases of patients who underwent more than one operation during the study period. This study was approved by the Tel-Aviv Sourasky Medical Center institutional ethics committee, reference number: 0768-17-TLV.
Clinical and demographic data. We reviewed admission, surgical, and discharge reports for each case.
Clinical data on follow-up in the ambulatory or hospital setting up to one year after surgery were also collected. The following clinical and demographic pre-and post-operative parameters were analyzed: age, sex, body mass index, hand dominance, other malignancies, brain radiotherapy, cerebrovascular, cardiac and metabolic Mean overall and progression free survival were also measured. Neurological manifestations were documented, including motor and speech deficits as well as report of seizures before and immediately (within hours) after surgery, and at 3, 6, and 12 months of follow-up, when available. Due to the inherent prognostic differences between each pathology groups, these outcomes were measured separately, up to 6 months in GBM patients and up to 12 months among LGG patients.
Radiological data. Data on tumor volume and location, tumor enhancement, and extent of resection (EOR) were derived from pre-and immediate postoperative MRI studies, which are routinely performed within 48 h after surgery. 3,16 . The EOR for GBM and LGG was calculated using the following formula: (preoperative − postoperative tumor volume)/preoperative tumor volume × 100). In case of GBM, the volume of blood products rather than the volume of the residual tumor was confirmed by comparing T1-weighted gadoliniumenhanced and non-enhanced MRIs. FLAIR sequence was used for measuring the non-enhancing component of the tumors 17 . Rim-pattern restriction surrounding the resection cavity was detected by a neuroradiologist who was blinded to clinical outcomes, and it was based on DWI studies and apparent diffusion coefficient techniques (Fig. 1). Infarcts, which were excluded from this study, were distinguished by their typical wedge-shaped arterial territory and relatively rapid appearance on immediate post-op DWI. Areas of T1-weighted hyper-intensities, accompanied by areas of hyperdensity on post-operative computerized tomographs (CT) were considered as blood products 11,14 .  LGG patients were evaluated by a battery of computerized cognitive tests before surgery and at the 3-month follow-up ( Table 5). The NeuroTrax testing platform was used for the assessment of various cognitive functions, including visual and verbal memory, executive function, attention, naming, and visual spatial processing as was previously described 11,[18][19][20] . Test scores were calculated by the software normalized for age and education level relative to a large normative database of healthy individuals and fit to an IQ-style scale, with higher scores reflecting better performance (mean:100, SD: 15).

Intraoperative neurophysiologic and anesthetics data. Changes in transcranial and direct cortical
and subcortical motor evoked potentials (MEPs) of intraoperative monitoring (IOM) were evaluated as reported by our group 3,16,21,22 . Monitoring during awake craniotomies was performed by a trained neuropsychologist and it was based on continuous physical examination for motor functions, as well as language assessments for the detection of production and comprehension decline. Data on anesthetics included anesthesia duration and blood pressure measurements before and during surgery.
Statistical analysis. Characteristics of categorical data were compared using the Pearson's χ 2 test. Multivariate logistic regression analysis was used to evaluate risk factors for the occurrence of rim restriction. Overall and progression free survival (PFS) were analyzed based on the Kaplan-Meier method. Repetitively measured variables were analyzed using generalized linear models. Statistics were performed using SPSS 21.0 software (SPSS Inc, Chicago, IL).
Ethics approval. This study was approved by the Tel-Aviv Sourasky Medical Center institutional ethics committee, reference number: 0768-17-TLV. The study was performed in accordance with the relevant guidelines and regulations.
Consent to participate. The ethics committee of the Tel-Aviv Sourasky Medical Center waived the requirement of informed consent due to the retrospective nature of the study.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. In the future we may consider asking our patients for their permission to share clinical data for research puropses.