All patients were recruited at our neurosurgical department, which covers a population of approximately 1.7 million inhabitants in a system with referrals based on area of residence. The department manages all patients requiring a neurosurgical procedure due to an intracranial lesion in the region of Västra Götaland, Sweden.

Data were derived from two cohort studies, one retrospective and one prospectively recruited ( Figure 1 ). The retrospective cohort was obtained using the electronical health records (EHR), pathology database, and operation logs. Patients were ≥18 years old with histopathological verified supratentorial dLGG classified according to the 2007 World Health Organization (WHO) classification of tumors of the central nervous system (8) and graded as grade 2 or 3. Patients who underwent biopsy or tumor resection during the period January 2010 through December 2016 were identified.

The prospectively recruited cohort consists of patients ≥18 years old with radiologically suspected dLGG based on MRI scan(s). These patients underwent biopsy or tumor resection during the period January 2017 through December 2020. As part of pre-operative work-up and approximately 3 months postoperatively, these patients were invited to provide quality-of-life measurement in terms of EQ-5D 3L. Subsequently, only patients with dLGG classified according to the WHO 2016 classification (9) as grade 2 or grade 3 were selected for the study.

Immunohistochemistry staining for isocitrate dehydrogenase 1 R132H (IDH) mutant protein was performed as the initial step in assessment of IDH mutation. Negative immunohistochemistry analyses were subsequently tested with next-generation sequencing to detect rarer IDH mutations (10). Codeletion of chromosomal arms 1p and 19q was evaluated with fluorescence in situ hybridization, multiplex ligation-dependent probe amplification or evaluation from methylation arrays as reported previously (11). A minority of the retrospective material was not evaluated according to WHO 2016 due to lack of tissue and therefore excluded from the study. The 2021 fifth edition of the WHO Classification of the Tumors of the Central Nervous System (WHO 2021) was not available during the design of the study. As a result, all remaining material, including IDH wild-type dLGG where the majority show clinical features of glioblastoma and are currently assessed as glioblastoma according to WHO 2021, was classified according to WHO 2016 for this study (12).

KPS is an ordinal scale designed to measure levels of patient activity and medical requirements. Patients are classified into 11 categories from 100 (no evidence of disease) to 0 (dead) (5, 13). KPS at admission was retrospectively scored based on data extracted from EHR. The mRS was originally designed for stroke patients; it focuses on patient disabilities, and patients are classified on 7 categories from 0 (no symptoms) to 6 (dead) (5, 14). For all patients, the mRS was retrospectively estimated from the EHR at follow-up visit (1–3 months postoperative) by clinicians (AJ and DB). To assess whether mRS was affected by surgical complications in patients registering adverse events, mRS at follow-up was qualitatively and retrospectively estimated from EHR.

Adverse events related to post-operative complications were evaluated using the LIC. LIC focuses on general postoperative morbidity using a four-grade severity scale based on the therapy administered to treat a postoperative adverse event within 30 days of surgery; it also considers whether the complication is medical or surgical (2). Complications were recorded based on EHR.

Neurological deficit and any information concerning seizures and seizure control was routinely assessed at admission, discharge, and follow-up at the neurosurgical department. Also based on EHR, neurological deficits recorded included motor, language, cognitive, and visual domains. Any post-operative new or worsened neurological deficit, including transient or suspected ones like the supplementary motor area syndrome, was registered. A deficit was considered permanent if deterioration compared to baseline was still present at 3 months, even if significant recovery had occurred.

Anatomical magnetic resonance imaging (MRI) T2-weighted image (T2) or fluid-attenuated inversion-recovery (FLAIR) sequences were used to assess tumor volume using software 3D Slicer (15) according to our previously reported method (16). Multifocal lesions were classified according to the largest tumor. Main tumor location and presumed eloquent brain areas were routinely identified as part of the preoperative work, and both were recorded based on EHR. Location taxonomy followed the anatomical lobe mainly involved by the lesion. Presumed eloquent brain areas were identified following the areas listed in the University of California San Francisco classification system (17).

TDN grades were calculated following the TDN criteria (4), here referred to as TDN scores. Adverse events were ordered in relation to the therapy, disability, and neurological deficits they involved. Therapy was evaluated using the LIC; Disability was assessed with the mRS at follow-up (mRS was not considered for TDN classification if it was affected by documented tumor progression); Neurological deficit was assessed using a binary definition for any new or worsened neurological deficit following surgery. According to TDN criteria, the dimensions of Disability and Neurology were only considered for TDN scoring when their deterioration resulted from the adverse event ( Figure 2 ).

HRQoL was measured with EuroQoL-5-dimension, three levels of response (EQ-5D 3L) questionnaire (18). Patients completed the EQ-5D 3L at the time of first visit to the outpatient clinic and at follow-up approximately 3 months postoperatively. EQ5D 3L was only available in parts of the prospective cohort. The results of EQ-5D 3L questionnaire were transferred to a utility index ranging from −0.594 to 1 (19), where higher scores indicated better quality of life. A minimal clinical important difference (MCID) of ±0.14, previously reported in patients undergoing glioma surgery, was used for this study (20). Change in EQ-5D3L index value was calculated subtracting the follow-up value from the preoperative value. Negative values (postoperative better than preoperative) indicated improvement; positive values indicated a decline.

All analyses were carried out in IBM SPSS version 28 (IBM Corp., Armonk, NY, US). Central tendencies for descriptive statistics are presented with either percentages, means with standard deviation (SD) or 95% confidence intervals (CI), or medians with first and third quartile (Q1, Q3). Statistical significance level was set to p < 0.05. All tests were 2 sided. Comparisons between groups were conducted with unpaired t-test, Mann–Whitney U test, Pearson χ ², or Fisher’s exact test as appropriate. Collinearity was assessed with Pearson correlation, ANOVA, and Cramer test between covariables, covariables and factors, and between factors, respectively. Collinearity between variables was set to values higher than 0.80 for Pearson’s eta, 0.64 for ANOVA’s eta squared, and 0.80 for Cramer’s V. Multivariable logistic regression was used with complications related to surgery as response. Age, sex, preoperative KPS, epilepsy, neurological deficits at admission (motor, cognitive, visual, and language), type of neurological intervention, tumor classification, main tumor location, tumor volume, and preoperative eloquence were used as dependent variables. Additional Sankey diagram, bar plots, and box plots were generated using Python programming language version 3.8.3 (Python Software Foundation, Delaware, US).

A total of 260 patients were screened for inclusion. Two patients in the retrospective cohort were excluded due to lack of tissue. From the prospectively recruited cohort, 27 patients were excluded because final histopathological diagnosis was other than dLGG grade 2 or grade 3, or because patients declined to participate in the study ( Figure 1 ).

A total of 231 patients were included in the study. The mean age at surgery was 48.3 years (SD 14.5), and 134 patients (58.0%) were males. There were 69 patients (29.9%) with oligodendrogliomas, 75 (32.5%) with astrocytoma IDH-mutant, and 87 (37.6%) with astrocytoma IDH wild type. The distribution of grade showed that 119 patients (51.5%) had tumors of WHO grade 2 and 112 patients (48.5%) of WHO grade 3. Within 3 months of surgery, a total of 46/231 patients (19.9%) started chemotherapy only (either temozolomide or procarbazine-lomustine-vincristine), and 52/231 patients (22.5%) had started radiotherapy only, and 61/231 patients (26.4%) had started both chemotherapy and radiotherapy.

At admission, 157 patients (68%) reported a history of seizures. From surgery to the 3 months follow-up, five of these patients (2.2% of total cohort) had worsening of seizures. During the same period, three patients (1.3%) had first-onset seizures. All worsened and first-onset seizures occurring withing 30 days of surgery were scored according to LIC grading. The proportion of any neurological deterioration post-operatively was 89/231 (38.5%). Deficit in more than one function occurred in 36/231 patients (15.6%). Also, out of 89 patients with deficits, 40 patients (44.9%) had complete recovery at 3 months postoperatively. A detailed list of patient characteristics and clinical variables is provided in Table 1 .

Regarding surgical outcomes, 121/231 patients (52.4%) did not present with any postoperative complications and consequently received a TDN score of 0. The remaining 110 patients with complications related to surgery were scored according to TDN criteria. In 10 out of 110 patients (9.1%) with complications, mRS measurements could not be considered for TDN scoring. Five of these ten patients had notable deterioration in mRS at follow-up due to tumor progression, which is therefore unrelated to complications following glioma surgery. For the remaining five patients, mRS was missing due to loss of follow-up. A comparison between TDN score and LIC grade results is provided in Table 2 and Figure 3 .

Patient characteristics and clinical variables in cases with and without complications related to surgery are shown in Table 3 . Of the included variables, only preoperative cognitive deficit and type of surgery were found to be significantly differently distributed among groups of patients with and without complications. Patients with complications more often had cognitive impairment prior to surgery than patients without complications (23.6% versus 10.7%, p = 0.01) and more often underwent tumor resection than tumor biopsy as primary surgery strategy (90% versus 70.2%, p < 0.001 see Table 3 ).

A multivariable logistic regression was performed to ascertain the effects of age, sex, preoperative KPS, epilepsy, neurological deficits at admission, type of neurological intervention, tumor classification, main tumor location, tumor volume, and preoperative eloquence on the likelihood that complications arise following glioma surgery. Of these, cognitive impairment at admission, surgical resection, and tumors located in eloquent regions were associated with an increased likelihood of complications following glioma surgery (respectively p = 0.01, p ≤ 0.001, and p = 0.01, see Supplementary Table 1 ).

In 110 patients experiencing adverse events, a total of 133 new or worsened neurological deficits were found. Of the 110 patients experiencing adverse events, 21 patients (19.1%) did not present with new or worsened neurological deficit. On average, patients scoring TDN 2 experienced more neurological deficits (112 deficits in 89 patients) and more often postoperative neurological deficit only (51/89 patients, 57%) than patients with TDN higher than 2 (21 deficits in 20 patients; and 7/20 patients, 35% respectively). A detailed list of neurological deficits in groups of patients by TDN score is shown in Table 4 .

EQ-5D 3L was reported for 45 patients. The patients experiencing complications related to surgery had similar pre- to postoperative change in EQ-5D 3L index values (n = 27; mean 0.03; 95% CI −0.06 to 0.11) compared to patients without complications (n = 18; mean −0.06; 95% CI −0.21 to 0.08). Although subgroups were small, there was no apparent difference in change in EQ-5D 3L index values for TDN scores 3–5 (n = 3; mean 0.04; 95% CI −0.12 to 0.21) compared to TDN scores 1–2 (n = 24; mean 0.02; 95% CI −0.07 to 0.12). Changes in EQ-5D 3L index value in subgroups of patients based on TDN scores are shown in Table 5 and Figure 4 .

In patients grouped by the presence of neurological deficit at follow-up, patients without new or worsened neurological deficit at follow-up had better HRQoL change with mean −0.08 change in pre- to post-operative EQ-5D 3L index value (n = 21; 95% CI −0.20 to 0.05) compared to patients experiencing any new or worsened neurological deficit (n = 24; mean 0.05; 95% CI −0.04 to 0.13). Changes in EQ-5D 3L index value in patients grouped by presence of neurological deficit at follow-up are shown in Table 6 and Figure 5 .

Patients with new-onset neurological deficit(s) at follow-up had worse HRQoL change, with mean 0.11 change in pre- to postoperative EQ-5D 3L index value (n = 13; 95% CI 0.0 to 0.22) compared to other patients (n = 32; mean −0.06; 95% CI −0.15 to 0.03). Changes in EQ-5D 3L index value in patients grouped by change in neurological deficit from admission to follow-up are shown in Table 7 and Figure 6 .

Patients undergoing any adjuvant treatment within 3 months of surgery showed a similar change in EQ-5D 3L index values mean −0.06 (n = 28, CI −0.17 to 0.06). Despite small subgroups, no major trend was seen with the different adjuvant treatment (n = 11, mean −0.05, CI −0.23 to 0.14 for patients undergoing chemotherapy only; n = 8, mean −0.19, CI −0.45 to 0.07 for patients undergoing radiotherapy only; and n = 9, mean 0.05, CI −0.09 to 0.19 for patients undergoing chemotherapy and radiotherapy). The patients without any adjuvant treatment within 3 months of surgery had comparable results (n = 17, mean 0.07, CI 0.00 to 0.13).