Combining intraoperative carmustine wafers and Stupp regimen in multimodal first-line treatment of primary glioblastomas.

Abstract Background. The study investigated if intraoperative use of carmustine wafers, particularly in combination with Stupp regimen, is a viable and safe first-line treatment option of glioblastomas. Methods. Eighty-three consecutive adult patients (50 men; mean age 60 years) with newly diagnosed supratentorial primary glioblastomas that underwent surgical resection with intraoperative carmustine wafers implantation (n = 7.1 ± 1.7) were retrospectively studied. Results. The median overall survival (OS) was 15.8 months with 56 patients dying over the course of the study. There was no significant association between the number of implanted carmustine wafers and complication rates (four surgical site infections, one death). The OS was significantly longer in Stupp regimen patients (19.5 months) as compared with patients with other postoperative treatments (13 months; p = 0.002). In addition patients with eight or more implanted carmustine wafers survived longer (24.5 months) than patients with seven or less implanted wafers (13 months; p = 0.021). Finally, regardless of the number of carmustine wafers, median OS was significantly longer in patients with a subtotal or total resection (21.5 months) than in patients with a partial resection (13 months; p = 0.011). Conclusions. The intraoperative use of carmustine wafers in combination with Stupp regimen is a viable first-line treatment option of glioblastomas. The prognostic value of this treatment association should be evaluated in a multicenter trial, ideally in a randomized and placebo-controlled one.


Introduction
Glioblastoma multiforme [GBM, World Health Organization (WHO) grade IV astrocytoma] is the most common malignant primary brain tumor in adults and among the most aggressive of all tumors. 1,2 Th e current standard post-surgery for patients with newly diagnosed GBM consists of concomitant chemoradiotherapy followed by adjuvant maintenance chemotherapy with temozolomide (TMZ), the so-called Stupp regimen. 3,4 However, as an alternative treatment, biodegradable wafers impregnated with the cytotoxic agent carmustine (1,3-bis(2-chloreoethyl)-1-nitrosourea ϭ BCNU), can be implanted in the surgical bed on the walls of the resection cavity for newly diagnosed and recurrent highgrade gliomas. 5,6 Neither of these unimodal treatments 7 have shown promising results. Multimodal treatment options that combine therapies are needed. Combining carmustine wafers (Gliadel ® , MGI Pharma, Bloomington, MN, USA) with the Stupp regimen is one such multimodal option. Some retrospective 8 -12 and prospective 13,7 studies combining these modalities have had interesting fi ndings. Th ese studies have reported 18-month median overall survival (OS) without a signifi cant increase in complications or adverse events. In addition, a recent prospective study of progression-free survival (PFS) showed that surgery, carmustine wafers, radiotherapy, and 6-month metronomic TMZ chemotherapy produced promising results without a marked increase in toxicities as compared with the Stupp regimen. 14 However, the literature on the use of carmustine wafers in addition to the Stupp regimen is still limited and as a practical consequence, the use of intraoperative carmustine wafers still depends mainly on the neurosurgeon ' s preferences. 15 To add to this limited literature on multimodal therapies for GBMs, here, we present a retrospective observational monocentric study. Th e study reports on an 8-year long single-institution experience of treating adult patients with intraoperative carmustine wafers implantation during fi rst-line treatment of GBM. Specifi cally, we addressed the effi cacy and toxicity of carmustine wafers, particularly when associated with the Stupp regimen.

Data source
We reviewed records of 405 adult patients diagnosed with GBM who were treated at Sainte-Anne University Hospital, Paris, France, between March 2004 and January 2012. To be included in the study patients needed to meet the following criteria: (1) over 18 years of age at diagnosis, (2) histological diagnosis of new-diagnosed primary GBM (WHO grade IV astrocytoma, secondary GBM excluded by an IDH1-positive immunoexpression) based on a central neuropathological review (PV), 16 (3) supratentorial hemispheric location, (4) surgical resection with intraoperative carmustine wafers implantation without fl uorescence-guided resection using 5-aminolevulinic acid, (5) minimal follow-up over the next 12 months from date of surgery, except in cases of progression or death. All enrolled patients gave their written informed consent for storage of the surgical samples for further analyses.

Data collection
Clinical and imaging characteristics collected at the time of histopathological diagnosis included: gender, age, neurological defi cit, Karnofsky performance status (cutoff 70), main T1-weighted sequence contrast-enhanced tumor anatomical location (enhanced contrast obtained through injection of gadopentetate dimeglumine), contrast-enhanced tumor ventricular contact area and contrast-enhanced tumor volume (cutoff 40 cm 3 ). Th e revised RTOG-RPA classifi cation system for GBM was used retrospectively to stratify comprehensive clinical risk. 17 Th is was done by classifying each patient (while blinded to patient survival) according to the RTOG-RPA classifi cation system using the data collected at the time of histopathological diagnosis.
Neuropathological characteristics obtained from patient records included the histopathological GBM subtype: pleomorphic cell GBM, gliosarcoma, giant cells GBM, GBM with oligodendroglial component and small cell GBM. Th ese characteristics were assigned according to the 2007 WHO classification of tumors of the central nervous system (CNS). 16 Th e research of IDH1 R132H mutation was retrospectively performed by immunohistochemistry. Th e O-6-methylguanine-DNA methyltransferase (MGMT) methylation status and the MGMT expression by immunohistochemistry were not available for the patients under study.
Treatment-related characteristics obtained from patient records included: number of carmustine wafers implanted (cutoff of maximum of eight wafers), quantifi cation of the extent of surgical resection based on early postoperative MRIs (within 48 h) on T1-weighted sequence following injection of gadopentetate dimeglumine, 18,19 postoperative non-surgical and surgical complications (new neurological defi cit or worsened neurological condition, seizures, postoperative hematoma requiring surgical evacuation, wound infection, meningitis, abscess), postoperative oncological treatments [none, radiotherapy alone, chemotherapy with TMZ alone, other chemotherapy alone, concomitant chemoradiotherapy with TMZ without adjuvant maintenance chemotherapy, Stupp regimen (cutoff of maximum of six courses)], complications of oncological treatments, and adverse events using WHO toxicity scale to grade the severity, 20 oncological treatments at progression stages (such as: none, surgical resection with or without carmustine wafers implantation, chemotherapy (TMZ, other), radiotherapy, antiangiogenic therapy with Bevacizumab).

Study endpoints
Th e primary endpoint was OS, measured from the date of histopathological diagnosis after fi rst surgical resection to the date of death. Th is interval was censored from the date of last follow-up for survivors. Th e secondary endpoint was PFS that was measured from the date of histopathological diagnosis to the date of fi rst radiological evidence of progression according to MacDonald ' s criteria 21 or to the date of death. Th is interval was censored at the date of last follow-up for survivors.

Statistical analysis
Statistical analyses were performed using JMP software, version 11.0.0 (SAS Institute, Cary, NC, USA). Results are reported as mean Ϯ standard deviation (SD) and range for continuous data. Categorical data are reported as percentages. A significance level of p Ͻ 0.05 was used. Comparisons among groups were performed 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 comparing continuous variables, as appropriate. Associations between variables were determined using non-parametric Spearman ' s rank order correlation. Kaplan -Meier analysis was performed for unadjusted survival curves, using log-rank tests to assess signifi cance. Cox proportional hazards models were also constructed, adjusted for predictors associated with mortality or tumor progression in univariate analyses. Variables associated at the p Ͻ 0.2 level in unadjusted analysis were then entered into models, with the fi nal model retaining only the variables signifi cant at the p Ͻ 0.05 level.

Patients and tumors characteristics
Eighty-three patients (50 men, 33 women) met the study inclusion criteria. Clinical, imaging, and surgical fi ndings are summarized in Table I. Th e mean age at surgery was 59.9 Ϯ 10.0 years (range, 21 -78). Presenting symptom comprised neurological defi cit in 62 cases (74.7%), intracranial hypertension in 31 cases (37.4%) and seizures in 15 cases (18.1%). Pleiomorphic cell GBM was the most frequent histopathological subtype (n ϭ 79, 95.2%), giant cell glioblastoma, and gliosarcoma were diagnosed in two and two cases, respectively.
Th e mean contrast-enhanced tumor volume at fi rst surgery was 45.4 Ϯ 30.2 cm 3 (range, 0.16 -170.4). Postoperative contrast-enhanced imaging control included MRI in 64 cases (77.11%) and CT-scan in 15 cases (18.07%). We restricted the quantitative analysis of the extent of resection for cases with available preoperative and postoperative digitized contrastenhanced MRI (n ϭ 57): a total surgical removal was obtained in 28 cases (49.1%), a subtotal resection (78 -99% of initial tumor volume) was obtained in 20 cases (35.1%), and a partial resection ( Ͻ 78% of initial tumor volume) was obtained in 9 cases (15.8%). A mean 7.1 Ϯ 1.7 (range, 3 -13) carmustine wafers were implanted per surgery. Th ere was a positive correlation between the preoperative contrast-enhanced tumor volume and the number of implanted carmustine wafers (r ϭ 0.388, p Ͻ 0.001), the number of implanted carmustine wafers increasing with the tumor volume. Th ere was no signifi cant correlation between the extent of surgical resection or the percentage of surgical resection and the number of implanted carmustine wafers (r ϭ 0.228, p ϭ 0.131 and r ϭ 0.006, p ϭ 0.967, respectively).

Surgical outcomes
Th irteen patients (15.7%) developed a total of 16 adverse postoperative events (Supplementary Table I to be found online at http://informahealthcare.com/doi/abs/10.3109/ 02688697.2015.1012051). Four patients (4.8%) had bacterial infectious surgical site complications (two abscesses, three meningitis), one of whom (1.2%) died 2.5 months after surgery. One patient (1.2%) had postoperative hematoma requiring surgical evacuation. In addition, a worsening of the neurological condition was observed in eight patients (9.6%) with four of these patients (4.8%) also developing nonsurgical adverse postoperative events (pneumonia, pyelonephritis, pulmonary embolism, acute pulmonary edema). Th ere was no signifi cant association between the number of implanted carmustine wafers and the overall likelihood to adverse postoperative events (p ϭ 0.599), overall surgical complications (p ϭ 0.468), infectious surgical site complications (p ϭ 0.896), or worsening of a patient ' s neurological condition (p ϭ 0.908).

First-line oncological treatments
Postoperative fi rst-line oncological treatments were started in 80 cases (96.4%) (Table II). Radiation therapy was started in 79 cases (95.2%) at a mean 6.3 Ϯ 1.7 weeks (range, 3 -16) after surgery. Th e radiation therapy had to be discontinued in one case (1.3%). Th e fi rst-line chemotherapy was TMZ in all cases. A Stupp regimen was administered in 61 cases (73.5%). Twenty-two patients did not receive a Stupp regimen: eight patients were treated before the Stupp era, in twelve cases based on a decision of a corresponding oncological center (1 case for altered postoperative performance status, 6 cases for age over 70, 5 cases for other causes), and two cases for mean, 8.8 Ϯ 0.9) or in patients who had none of these therapeutic modalities (4.7; mean, 4.8 Ϯ 0.7; p Ͻ 0.001).

Overall survival
During the follow-up period, 56 patients (67.5%) died of tumor progression at a mean of 20.6 Ϯ 1.7 months (median, 15.8; range, 2 -53) since histopathological diagnosis. At second and third year follow-ups, 36% and 9.9% of patients were alive, respectively (Fig. 1A). Whatever the number of implanted carmustine wafers, the median OS was signifi cantly longer in patients who had a subtotal or total surgical resection (21.5 months; mean, 21.7 Ϯ 1.7) than in patients who had only a partial surgical resection (13 months; mean, 12.9 Ϯ 2.5; p ϭ 0.011). Th e median OS was signifi cantly longer in patients who had eight or more implanted carmustine wafers (24.5 months; mean, 23.3 Ϯ 2.3) than in patients who had seven or less implanted carmustine wafers (13 months; mean, 15.9 Ϯ 1.8; p ϭ 0.021) (Fig. 1B). Th e median OS was signifi cantly longer in patients who had a Stupp regimen (19.5 months; mean, 23.5 Ϯ 2.2) than in patients who had any other type of postoperative treatments (13 months; mean, 13.6 Ϯ 1.8; p ϭ 0.002) (Fig. 1C). In the subgroup of patients who had a Stupp regimen (n ϭ 61), the median OS was signifi cantly longer in patients who had six or more courses of adjuvant TMZ (n ϭ 28; 27 months; mean, 30.5 Ϯ 3.5) than in patients who had less than six courses of adjuvant TMZ (n ϭ 33; 16 months; mean, 15.5 Ϯ 1.8; p Ͻ 0.001) (Fig. 1D). Th e median OS was signifi cantly longer in patients who had a maximal treatment (including subtotal and total resection plus eight or more implanted carmustine wafers plus a Stupp regimen with six or more courses of adjuvant TMZ) (27 months; mean, 26.8 Ϯ 1.9) than in patients who had one or two of these therapeutic modalities (18.5 months; mean, 19.5 Ϯ 1.9) or in patients who had none of these therapeutic modalities (11.5 months; mean, 10.9 Ϯ 1.6; p Ͻ 0.001).

Predictors
In univariate analysis (Table III), neurological defi cit (p ϭ 0.022), number of carmustine wafers (p ϭ 0.037), and Stupp regimen (p ϭ 0.020) were predictors of PFS. Tumor volume (p ϭ 0.043), extent of surgical resection (p ϭ 0.011), number of carmustine wafers (p ϭ 0.021), and Stupp regimen (p ϭ 0.002) were predictors of OS. In multivariate analysis (Table IV), independent prognostic factor for PFS was the presence of a neurological defi cit (p ϭ 0.022). Independent prognostic factors for OS were tumor volume (p ϭ 0.014) and extent of surgical resection (p ϭ 0.018). Of note, the parameters " number of implanted carmustine wafers " and " Stupp regimen " were statistically linked: the number of implanted carmustine wafers was signifi cantly higher in the subgroup of patients who had a Stupp regimen (mean, 7.3 Ϯ 1.5; range, 4 -13; 67.2% with eight or more carmustine wafers) than in the subgroup of patients who had any other type of postoperative treatments (mean, 6.5 Ϯ 2.0; range, 3 -12; 40.9% with eight or more carmustine wafers) (p ϭ 0.029). To avoid the potential bias of Stupp regimen in the assessment of the prognostic signifi cance of the number of implanted carmustine, we conducted a multivariate analysis in the subgroup of patients who had a Stupp regimen (n ϭ 61). Independent postoperative complications precluding further oncological treatment (cases 2 and 3). Th e mean number of courses of adjuvant TMZ was 5.7 Ϯ 3.6 (range, 1 -27) with 28 patients (45.9%) having six or more courses of TMZ and seven of them (11.5%) having nine or more courses of adjuvant TMZ.

Treatments at progression
Sixty-nine fi rst progressions (83.1%) were observed and new oncological treatments were started in 47 (68.1%) of those cases (Table II). A second surgery was performed in six cases (8.7%) with carmustine wafers implantation in four of these cases. Th ree patients had re-irradiation (4.3%). A systemic chemotherapy was administered in 44 patients (63.8%). Th irty-six secondary progressions (43.4%) were observed and new oncological treatments were started in 25 (69.4%) cases (Table II). A third surgical removal was performed in one case (2.8%) without carmustine wafers implantation and followed by systemic chemotherapy. A systemic chemotherapy was performed in the remaining 24 patients (96%).

Adverse events during oncological treatments
During fi rst-line oncological treatment, 21 patients (26.6% of 79 patients that started radiotherapy) experienced 28 adverse events during and after radiotherapy (Supplementary Table II
Whatever the number of implanted carmustine wafers, the median PFS tended to be longer in the subgroup of patients with a subtotal and total surgical resection (9.5 months; mean, 10.0 Ϯ 0.9) than in patients with a partial surgical resection (5.5 months; mean, 7.9 Ϯ 1.2), without reaching statistical signifi cance (p ϭ 0.278). Th e median PFS was signifi cantly longer in patients who had eight or more implanted carmustine wafers (8.5 months; mean, 10.7 Ϯ 1.2) than in patients who had seven or less implanted carmustine wafers (5.0 months; mean 7.4 Ϯ 1.0; p ϭ 0.037) (Fig. 1B). Th e median PFS was signifi cantly longer in patients treated with the Stupp regimen (8.5 months; mean, 10.6 Ϯ 1.1) than in patients treated with any other type of postoperative treatments (4.7 months; mean, 6.8 Ϯ 1.0; p ϭ 0.020) (Fig. 1C). Th e median PFS was signifi cantly longer in patients who had a maximal treatment (including subtotal and total resection plus eight or more implanted carmustine wafers plus a Stupp regimen with six or more courses of adjuvant TMZ) (12 months; mean, 13.8 Ϯ 1.3) than in patients who had one or two of these therapeutic modalities (7.0 months; prognostic factors for OS were tumor volume (p ϭ 0.003), extent of surgical resection (p ϭ 0.005), and number of carmustine wafers (p ϭ 0.008) (Table IV).

Discussion
Together with TMZ and Bevacizumab, carmustine wafers is one of the only three drugs approved by FDA for treatment of high-grade gliomas during the past decades. 9,27 It was approved in 1996 and in 2003 for use in recurrent GBM and in newly diagnosed high-grade gliomas, respectively. 22 Th is approval arose from two phase III trials, 23,6 where the postoperative treatment consisted of external radiotherapy. Only few retrospective 8 -12 and prospective 13,7 studies have previously analyzed the impact of the combination of carmustine wafers implantation together with the Stupp regimen for the treatment of newly diagnosed primary GBM. Th ey have reported interesting results. in the subgroup of 61 patients that received a Stupp regimen, and a 4.8% rate of postoperative infectious complications. Several teams reported similar outcomes regarding this multimodal treatment. 13,11,7,24 Interestingly, even if the study by Th e present study reports a single institution ' s experience of intraoperative carmustine wafers either alone or in combination with the Stupp regimen for newly diagnosed primary GBM in 83 patients, paying close attention to patient survival and complication rates. We report a median OS at  both MGMT-methylated and unmethylated patients, suggest an additional benefi t from carmustine wafers implantation plus conventional chemoradiotherapy with TMZ. Finally, the assessment of a multimodal oncological treatment combining intraoperative carmustine wafers together with the Stupp regimen for the fi rst-line treatment of newly diagnosed primary GBM would benefi t of a multicenter trial. Th e need of such a study, which should comprise molecular analyses, including the MGMT methylation status and the IDH mutation status, 29,30 would help to identify the subgroups of patients that could benefi t from such combination treatments and better characterize the modalities of use.

Conclusion
Th e intraoperative use of carmustine wafers in combination with Stupp regimen is a viable fi rst-line treatment option of glioblastomas. Th e prognostic value of this treatment association should be evaluated in a multicenter trial, ideally in a randomized and placebo-controlled one.
Noel et al. that combined grades III and IV tumors concluded that there was no signifi cant infl uence of carmustine wafers on OS, one can observe that, in the GBM subgroup, the median OS was higher in cases with (20.8 months; n ϭ 20) than without (13.8 months; n ϭ 16) carmustine wafers implantation and that the diff erence was almost signifi cant (p ϭ 0.067). 25 In line with previously published studies 26,19 we confi rm that the extent of surgical resection and the Stupp regimen are predictors of longer OS. We report for the fi rst time the association between the number of implanted carmustine wafers and OS. In the literature, the reported number of implanted carmustine wafers has varied between one and nine and to date no association between patient survival and number of implanted wafers has been reported. 9,7 In our study the number of implanted wafers varied between 3 and 13. Previous studies did not fi nd a signifi cant infl uence of the number of implanted carmustine wafers on OS but did report a significant increase in the risk of adverse events with the number of wafers implanted. 28,11 Here, the OS and PFS were signifi cantly longer in patients that had eight or more implanted carmustine wafers. We did not observe an increased risk of adverse events with a higher number of wafers implanted: the toxicity analyses confi rm that the association of carmustine wafers implantation with the adjuvant therapies was well tolerated during fi rst-line treatment of GBM. Particularly, the adverse events during adjuvant radiation therapy or conventional chemoradiotherapy were similar to those observed without carmustine wafers implantation. Th e positive correlation we observed between the number of implanted carmustine wafers and the tumor volume is explained by the large cavity performed by the removal of such tumor, whatever the extent of resection. Th e lack of correlation between the number of implanted carmustine wafers and the extent of resection is explained by the diffi culty, in several cases, to ascertain intraoperatively the subtotal or total removal of the tumor. Indeed, some patients under study have had the carmustine wafers implantation in case of partial surgical resection. It was attributed to diffi culties in intraoperative assessment of the extent of resection, when intraoperatively the neurosurgeon concluded to an " apparently gross total surgical removal " , while the postoperative MRI demonstrated a partial removal. As the NICE technology 15 indicates the use of carmustine wafers only in case of surgical resection Ͼ 90%, the intraoperative implantation of carmustine wafers should be guided by surgical techniques such as fl uorescenceguided resection with 5-aminolevulinic acid and intraoperative functional mapping that allow a maximal resection and a control of its extent.
We confi rm that the tumor volume, the number of carmustine wafers, and the extent of resection were all independent prognostic factors for OS in the subgroup of patients who had a Stupp regimen. Th us, our fi ndings of better outcomes in patients implanted with eight or more carmustine wafers and treated with the Stupp regimen regardless the extent of resection and initial tumor volume, strongly suggests that some patients could really benefi t from this multimodal oncological treatment. Th e lack of MGMT methylation status precludes postoperative subgroup analyses, but the observed survivals in this unselected mixed population, including