Discriminating uninfected surgical bed cysts from bacterial brain abscesses after Carmustine wafer implantation in newly diagnosed IDH-wildtype glioblastomas

Purpose Carmustine wafers can be implanted in the surgical bed of high-grade gliomas, which can induce surgical bed cyst formation, leading to clinically relevant mass effect. Methods An observational retrospective monocentric study was conducted including 122 consecutive adult patients with a newly diagnosed supratentorial glioblastoma who underwent a surgical resection with Carmustine wafer implantation as rst line treatment (2005–2018). Findings Twenty-two patients (18.0%) developed a postoperative contrast-enhancing cyst within the surgical bed: 16 uninfected cysts and six bacterial abscesses. All patients with an uninfected surgical bed cyst were managed conservatively, all resolved on imaging follow-up, and no patient stopped the radiochemotherapy. Independent risk factors of formation of a postoperative uninfected surgical bed cyst were age ≥ 60 years (p = 0.019), number of Carmustine wafers implanted ≥ 8 (p = 0.040), and partial resection (p = 0.025). Compared to uninfected surgical bed cysts, the occurrence of a postoperative bacterial abscess requiring surgical management was associated more frequently with a shorter time to diagnosis from surgery (p = 0.009), new neurological decit (p < 0.001), fever (p < 0.001), residual air in the cyst (p = 0.018), a cyst diameter greater than that of the initial tumor (p = 0.027), and increased mass effect and brain edema compared to early postoperative MRI (p = 0.024). Contrast enhancement (p = 0.473) and diffusion signal abnormalities (p = 0.471) did not differ between postoperative bacterial abscesses and uninfected surgical bed cysts. wafer


Introduction
Isocitrate dehydrogenase (IDH) wildtype glioblastoma (World Health Organization (WHO) grade IV astrocytoma) is the most common malignant primary brain tumor in adults 1 . Maximal safe resection, which is recommended as the rst-line treatment, improves survival and increases the e cacy of adjuvant therapies [1][2][3] . Treatment guidelines for newly diagnosed IDH-wildtype glioblastomas recommend maximal safe surgical resection, with or without Carmustine (1,3-bis(2-chloreoethyl)-1-nitrosourea, BCNU) biodegradable wafer implantation, followed by the standard radiochemotherapy protocol [4][5][6][7][8] . Carmustine wafer implantation in the surgical bed has been shown to improve survival of newly diagnosed and recurrent high-grade gliomas 7,9,10,10 . Through its direct local application, high concentrations of Carmustine can be delivered to the surgical bed over a 3-week-long period while minimizing systemic adverse effects 11,12 . It provides a therapeutic bridge between surgery and adjuvant therapy onset 11,12 .
Carmustine wafer implantation can result in distinct adverse effects. In addition to an increased risk of brain abscess formation, increased local in ammatory changes may induce surgical bed cyst formation, leading to clinically relevant mass effect 13 . Patients presenting postoperatively with a symptomatic contrast enhancing surgical bed cyst pose diagnostic and management challenges. A revision surgery is regularly indicated if there is clinical suspicion of infection 14,15 . To our knowledge, no dedicated study has evaluated the risk factors for the development of a postoperative surgical bed cyst, and key elements to discriminate between an uninfected surgical bed cyst and a bacterial abscess are lacking.
We assessed: 1) the risk factors for surgical bed cyst formation following surgical resection with Carmustine wafer implantation of a newly diagnosed IDH-wildtype glioblastoma; 2) the clinical and imaging criteria allowing discrimination between an uninfected surgical bed cyst and a brain abscess.

Data source
We screened consecutive adult patients harboring a supratentorial newly diagnosed IDH-wildtype glioblastoma treated with surgical resection plus Carmustine wafer implantation as rst treatment between December 2005 and December 2018 in a tertiary neurosurgical oncology center. Inclusion criteria were: 1) patients ≥ 18 years; 2) newly diagnosed supratentorial IDH-wildtype glioblastoma; 3) surgical resection with Carmustine wafer implantation as rst-line treatment; 4) available postoperative MRI (including T1-weighted sequence with and without intravenous administration of contrast agents, uidattenuation inversion recovery (FLAIR) sequence, and diffusion weighted sequence) to quantify the extent of resection; and 5) available clinical and MR imaging follow-up during the rst six postoperative months.
The decision to implant Carmustine wafers was not randomized but was decided by the treating neurosurgeon on clinical bases according to: 1) the guidelines from the French Neurosurgical Society 8 ; 2) the presence of a contrast-enhanced and necrotic mass on preoperative MRI; 3) the preoperative neurosurgical expectation of a gross total removal of the contrast enhanced tumor mass; 4) the preoperative obtained inform consent of the patient; 5) the intraoperative extemporaneous histopathological diagnosis of a malignant glioma; and 6) the gross total resection suspected intraoperatively.
According to the inclusion criteria, 122 out of 160 screened patients were included in the nal analysis. We excluded 27 patients without available follow-up postoperative MRIs, ve patients without available early postoperative MRI, three patients who were lost to follow-up, and three patients without an IDHwildtype glioblastoma (two IDH-mutant glioblastomas, one G43R diffuse glioma).

Data collection
Data at surgery included: sex, age, Karnofsky performance status (KPS) score, revised Radiation Therapy Oncology Group -Recursive Partitioning Analysis (RTOG-RPA) classi cation system for glioblastoma 16 , tumor location, quanti ed volume of the contrast-enhanced tumor, quanti ed extent of surgical resection based on early postoperative MRI (within 48 h) on three-dimensional contrast-enhanced T1-weighted sequence (subtotal resection de ned by the removal of ≥ 90% of enhancing tumor) 2 , search for restricted diffusion coe cient and gadolinium enhancement in parenchyma adjacent to Carmustine wafers, number of Carmustine wafers implanted, adverse postoperative events, histomolecular diagnosis, and O6-methylguanine-DNA methyltransferase (MGMT) promoter methylation status, if available. Follow-up data during the rst six-postoperative months included: systematic clinical examination and MRI (six weeks, three months, and six months postoperatively), surgical bed cystic lesion occurrence, cystic lesion volume, contrast enhancement of the cystic lesion walls, signal on diffusion-weighted imaging (DWI), restricted diffusion on apparent diffusion coe cient (ADC) on the wall and inside the cystic lesion, residual air in the surgical bed, edema-related mass effect, related clinical symptoms, and practical management of the cyst (conservative, surgery, steroid therapy).

Statistical analyses
To determine factors associated with the development of a postoperative surgical bed cyst or a brain abscess, univariate analyses were performed, computing unadjusted Odds Ratios (OR) and using the Chi square or Fisher's exact tests for comparing categorical variables, and the unpaired t-test or Mann-Whitney rank sum test for continuous variables, as appropriate. Variables associated with a p < 0.200 level in unadjusted analysis were entered into logistic backward stepwise regression models. The nal model retained only the variables signi cant at the p < 0.05 level. Statistical analyses were performed using JMP software (version 14.1.0, SAS Institute Inc Cary, USA).

Data availability
Data not provided in the article because of space limitations may be shared (anonymized) at the request of any quali ed investigator for purposes of replicating procedures and results.

Patient population
One hundred and twenty-two patients were included (63.9% men, mean age 60.1 ± 11.0 years), 87 of them (71.3%) were previously reported in another study 17 . Patient and tumor characteristics are detailed in Table 1. The mean number of Carmustine wafers implanted was 7.5 ± 1.7 per patient (median 8, range 4-16). follow-up after these rst six postoperative months. Sixteen presented uninfected surgical bed cysts and six presented bacterial abscesses according to postoperative follow-up. Three patients (2.5%) presented with recurrent disease con rmed by further imaging follow-up and requiring oncological treatments.

Incidence of a postoperative bacterial abscess
Six patients (4.9%) had a surgical site bacterial abscess con rmed by the identi cation of an organism on cultures, which was diagnosed at a mean of 26.2 ± 10.6 postoperative days (median 30, range 12-40) ( Fig. 1). All cases bene ted from a surgical procedure, which consisted in the evacuation of the bacterial abscess, an abundant washing of the operating site and bacteriological samples. The bone ap was systematically left in place. A bacterium was found in all patients ( Supplementary Fig. 1). New postoperative related symptoms (focal neurological de cit, increased intracranial pressure) were present in the six patients (100%). An associated healing defect was present in 2 out of 6 patients (33.3%). Fever was present in 4 out of 6 cases (66.7%). The volume of the postoperative bacterial abscess (mean of 44.9 ± 23.0 cc, median 35.5, range 23.0-80.8) was superior to the volume of the initial glioblastoma in 5 out of 6 cases (83.3%). All postoperative bacterial abscesses showed linear contrast enhancement of their walls, high signal and restricted diffusion on ADC inside the resection cavity on DWI, and residual air in the surgical bed. Increased mass effect compared to the early postoperative MRI was present in 5 out of 6 cases (83.3%).

Incidence of a postoperative uninfected surgical bed cyst
The 16 non-infected surgical bed cysts were diagnosed at a mean of 54.0 ± 16.9 postoperative days (median 60.0, range 20.0-90.0), three before radiotherapy onset and 13 at a mean 20.2 ± 13.8 days after radiotherapy onset (median 18.0, range 2.0-55.0) (Fig. 1). The volume of these postoperative surgical bed cysts (mean 24.6 ± 15.9 cc, median 25.1, range 4.9-68.4) was inferior to the volume of the initial glioblastoma in 12 out of 16 cases (75.0%). All postoperative uninfected surgical bed cysts showed thin and linear membrane-like contrast enhancement of their walls. High signal on DWI was observed in 9 of the 11 available cases (81.8%). Restricted diffusion on ADC maps was observed in 9 of the 12 available cases (75.0%), which was limited to the rim of the resection cavity. Residual air was observed within the cyst in 9 out of 16 cases (56.2%). Concomitant brain edema and mass effect were present in 5 out of 16 cases (31.3%). Increased mass effect compared to the early postoperative MRI was present in 5 out of 16 cases (31.3%). New postoperative symptoms were present in 8 out of 16 patients (50.0%): increased intracranial pressure in 7 (43.7%), epileptic seizures in 4 (25.0%), and neurological de cits in 1 (6.3%) of the 16 patients. The remaining 8 patients (50.0%) were asymptomatic.
All above mentioned patients were managed conservatively. Symptomatic patients were treated with high dose steroid therapy and asymptomatic patients did not receive additional steroid therapy.

Interpretation
To the best of our knowledge, no previous report has evaluated the risk factors for surgical bed cyst formation following Carmustine wafers implantation during the rst resection of a newly diagnosed IDH-wildtype glioblastoma. We report contrast-enhancing cysts within the surgical bed during the rst six postoperative months of imaging follow-up in 18.0% of patients under study, corresponding to either uninfected surgical bed cysts or bacterial abscesses. The prevalence of surgical bed cysts after Carmustine wafer implantation varies in the literature from 3-58% 15,18−20 . This may be explained by a lack of systematic imaging follow-up data, since 50% of surgical bed cysts are asymptomatic according to our cohort, and by the varying level of inclusion of postoperative bacterial abscesses between reports.
Interestingly, surgical bed cysts occur around the 8th postoperative week 15 while postoperative bacterial abscesses usually occur earlier in the rst postoperative month 21 . This could be linked to the bacterial proliferation which would grow faster than the cystic formation which, also explaining the faster clinical deterioration. Interestingly, Hasegawa et al. reported a retrospective study involving 19 patients who bene ted from a surgical resection plus Carmustine wafer implantation for newly-diagnosed and recurrent high-grade gliomas 22 . They found one case of surgical bed cyst, which required a new surgical procedure without bacteria identi ed on intraoperative samples. In the present study, the 16 uninfected surgical bed cysts showed a benign course, with presenting symptoms in 50% of cases and clinical and radiographic resolution in all cases after conservative management without surgery. High dose steroids were administered for symptomatic patients, as recommended 19 . These observations contrast with previous studies, which emphasize and justify the need for surgical treatment to treat elevated intracranial pressure despite corticosteroid treatment 14,15,20 . Of note, most of the previously published surgical bed cysts requiring surgical management occurred after resection and Carmustine wafer implantation for a recurrent glioblastoma. This may be related to in ammatory changes we did not encounter in the present series, focused on newly diagnosed glioblastomas.
We identi ed independent risk factors of developing an uninfected surgical bed cyst following surgical resection plus Carmustine wafer implantation as rst treatment of IDH-wildtype glioblastomas: age ≥ 60 years, number of Carmustine wafers implanted ≥ 8, and partial surgical resection. This will help in clinical practice to better identify patients at risk of developing a surgical bed cyst and at tailoring the postoperative management (clinical and imaging follow-up, duration and dose of steroid therapy). One study, including 36 cases of newly and recurrent glioblastomas, analyzed the relationship between cyst occurrence and some clinical and surgical data 19 , and identi ed partial resection as a risk factor, in accordance with our results.
The analysis of MRI ndings in the setting of IDH-wildtype glioblastomas treated by surgical resection plus Carmustine wafer implantation followed by standard radiochemotherapy protocol is challenging, particularly in the initial postoperative months 23 . Discriminating between early recurrent disease and surgical bed cyst is a concern. Here, all surgical bed cysts showed a thin circular rim of contrast enhancement caused by the Carmustine wafers, which contrasts with the nodular appearance of recurrent disease. MR perfusion and MR spectroscopy might be helpful in identifying early recurrences 15,20,24,25 . Discriminating between postoperative surgical site bacterial abscess and an uninfected surgical bed cyst is another major concern. In our series, contrast enhancement pattern and diffusion signal abnormalities were not discriminating imaging parameters because, in the context of Carmustine wafer implantation and of post-radiation in ammatory changes, the resection cavity and adjacent parenchyma showed a peak of restricted diffusion signal, from three weeks up to 6 months postoperatively, in addition to contrast enhancement 12,24,26 . We identi ed clinical and imaging parameters that may help discriminate between a bacterial abscess and an uninfected surgical bed cyst: cyst-related new postoperative focal neurological de cits, fever, residual air in the surgical cavity, volume of the cyst superior to that of the initial tumor, and increased mass effect compared to early postoperative MR imaging related to brain edema. In the six cases with a postoperative bacterial abscess, patients had worsening of their general (fever and asthenia) and/or neurological (focal neurological de cit and signs of increased intracranial pressure) condition, and their MRI showed signs consistent with a brain abscess (contrast enhancement, restricted diffusion, and residual air in the surgical bed).
When both clinical and MRI ndings were highly suspicious for a bacterial abscess, surgical intervention was performed, con rming the diagnosis. If the clinical and radiological criteria are not present, then clinic-radiological monitoring should be started. These observations, together with the timing of cyst occurrence, will orient the treatment strategy -conservative management versus surgical intervention -in each individual patient harboring a contrast-enhancing cyst within the surgical bed following surgical resection plus Carmustine wafer implantation.

Generalizability
Strengths of this study include the data collection of a large case series in a tertiary neurosurgical oncology center and the homogeneous postoperative imaging follow-up. We controlled for patient-related and methodological biases by selecting a homogeneous and consecutive population of newly diagnosed supratentorial IDH-wildtype glioblastomas in adults who all underwent the same surgical procedure with the histopathological re-assessment of all cases according to the 2016 update WHO classi cation. The present study could help: 1) identify patients at risk of developing a postoperative surgical bed cyst after resection of a newly diagnosed IDH-wildtype glioblastoma with Carmustine wafer implantation; 2) discriminate between postoperative uninfected surgical bed cysts and bacterial abscesses; 3) manage these patients accordingly with early surgical intervention for postoperative bacterial abscesses and conservative management for postoperative uninfected surgical bed cysts. We cannot extend the results to patients harboring another subtype of high-grade glioma or a recurrent glioblastoma with previous oncological treatments.

Limitations
These ndings should be interpreted with caution, given the retrospective and monocentric design of the study, the exploratory design of the statistical analyses, the lack of a control group, and the lack of an external validation set, all limiting the generalizability of the results. Given the number of observed events, the application of multivariable models was limited, particularly regarding isolation of parameters discriminating between postoperative bacterial abscesses and uninfected surgical bed cysts. In addition, we lack bacteriological evidence excluding the possibility of a bacterial abscess in surgical bed cysts that were all managed conservatively with positive outcomes. Further con rmatory analyses are required to reproduce the present results.

Conclusion
Postoperative surgical bed cysts occurred in about 20% of cases following Carmustine wafer implantation during the initial resection of newly diagnosed IDH-wildtype glioblastomas. The identi cation of risk factors for formation of a postoperative surgical bed cyst in this particular patient population may help tailor postoperative steroid therapy and imaging follow-up schedule. In addition, we identi ed clinical and imaging characteristics that may help discriminate between an uninfected surgical bed cyst and a bacterial abscess, the latter requiring early surgical management.