A practical scoring system to determine whether to proceed with surgical resection in recurrent glioblastoma

Neuro-Oncology, Aug 2013

Background To determine the benefit of surgical management in recurrent glioblastoma, we analyzed a series of patients with recurrent glioblastoma who had undergone surgery, and we devised a new scale to predict their survival.

A PDF file should load here. If you do not see its contents the file may be temporarily unavailable at the journal website or you do not have a PDF plug-in installed and enabled in your browser.

Alternatively, you can download the file locally and open with any standalone PDF reader:

https://neuro-oncology.oxfordjournals.org/content/15/8/1096.full.pdf

A practical scoring system to determine whether to proceed with surgical resection in recurrent glioblastoma

Chul-Kee Park () 0 1 2 Jeong Hoon Kim 0 1 2 Do-Hyun Nam 0 1 2 Chae-Yong Kim 0 1 2 Sang-Bong Chung 0 1 2 Young-Hoon Kim 0 1 2 Ho Jun Seol 0 1 2 Tae Min Kim 0 1 2 Seung Hong Choi 0 1 2 Se-Hoon Lee 0 1 2 Dae Seog Heo 0 1 2 Il Han Kim 0 1 2 Dong Gyu Kim 0 1 2 Hee-Won Jung 0 1 2 0 National University College of Medicine, Seoul National University Hospital , Seoul , South Korea 1 Medicine (J.H.K.); and Department of Neurosurgery, Samsung Medical Center, Sungkyunkwan University School 2 Department of Neurosurgery, Seoul National University College of Medicine, Seoul National University Hospital S.-H.L., D.S.H.), Department of Radiology (S.H.C.), and Department of Radiation Oncology (I.H.K.), Seoul new scale was successfully applied to the validation cohort of patients showing distinct prognosis among the groups (median survivals of 11.0, 9.0, and 4.0 months for the 0-, 1-, and 2-point groups, respectively). Conclusions. We developed a practical scale to facilitate deciding whether to proceed with surgical management in patients with recurrent glioblastoma. This scale was useful for the diagnosis of prognostic groups and can be used to develop guidelines for patient treatment. - Background. To determine the benefit of surgical management in recurrent glioblastoma, we analyzed a series of patients with recurrent glioblastoma who had undergone surgery, and we devised a new scale to predict their survival. Methods. Clinical data from 55 consecutive patients with recurrent glioblastoma were evaluated after surgical management. Kaplan Meier survival analysis and Cox proportional hazards regression modeling were used to identify prognostic variables for the development of a predictive scale. After the multivariate analysis, performance status (P .078) and ependymal involvement (P .025) were selected for inclusion in the new prognostic scale. The devised scale was validated with a separate set of 96 patients from 3 different institutes. Results. A 3-tier scale (scoring range, 0 2 points) composed of additive scores for the Karnofsky performance status (KPS) (0 for KPS 70 and 1 for KPS , 70) and ependymal involvement (0 for no enhancement and 1 for enhancement of the ventricle wall in the magnetic resonance imaging) significantly distinguished groups with good (0 points; median survival, 18.0 months), intermediate (1 point; median survival, 10.0 months), and poor prognoses (2 points; median survival, 4.0 months). The Gcancers, with a median survival of 14.6 months, lioblastoma remains one of the most devastating despite the current best-standard protocol.1 The conventional treatment of glioblastoma includes surgical resection or biopsy, followed by concomitant radiochemotherapy with temozolomide and adjuvant temozolomide.1 Although surgery alone is not sufficient to treat glioblastoma, the maximal safe resection is essential for the efficacy of adjuvant treatment in newly diagnosed cases.2 7 However, the survival benefit of surgical resection in recurrent glioblastoma is reported to be limited in selected conditions.8 11 Recently, a preoperative scale to predict survival after surgery for recurrent glioblastoma (National Institutes of Health [NIH] Recurrent GBM Scale) was devised using factors such as tumor involvement of eloquent/critical brain regions, Karnofsky performance status (KPS), and tumor volume.12 The application of the NIH Recurrent GBM Scale may help clinicians design the appropriate management plan when glioblastoma has recurred or progressed, and this scale may be useful to researchers designing clinical trials. However, the scale has limitations.13,14 To provide more objective and easily applicable guidelines for the management of recurrent glioblastoma, we created a novel, practical scoring system to guide the decision of whether to proceed with surgical resection in recurrent glioblastoma. Materials and Methods Study Cohort Clinical data from 55 consecutive patients with histologically confirmed recurrent glioblastoma after a craniotomy in Seoul National University Hospital from 2000 through 2010 were retrospectively collected and used for the study. Initially, all of the patients underwent surgical resection, followed by radiotherapy and nitrosourea-based chemotherapy, including temozolomide or nimustine. All of the patients underwent surgical resection at the time of radiological progression and had cases histologically confirmed as a recurrence. Tumor recurrence, or disease progression, was defined using the Macdonald criteria15 by T1-weighted magnetic resonance (MR) imaging. Surgical resection at recurrence was attempted when there was either a good chance to achieve total resection or the need to control an increase in intracranial pressure. Adjuvant chemotherapy after reoperation involved the initiation of any type of chemotherapy regimen within 1 month after surgery without any evidence of disease progression. Adjuvant chemotherapy was indicated when incomplete resection of the tumor was suspected in the postoperative MR images. The survival time was measured from the date of reoperation to the date of the patients death. All of the patients were tracked until their time of death. The surviving patients were submitted to censored observations at the last follow-up. Institutional review board approval was obtained for the retrospective analysis. Validation Cohort The validation cohort of 96 patients was collected from 3 separate institutes (Asan Medical Center [n 50], Samsung Medical Center [n 30], and Seoul National University Bundang Hospital [n 16]). All of the patients in the validation cohort were from a consecutive series at each institution and collected from 1995 through 2011. Their initial treatment, diagnosis of tumor recurrence, and maximal safe surgical resection of their recurrent tumors were similar to the study cohort of patients. The assessment of the clinical and radiographical data was performed in a blinded manner by collaborators at each institute, and the results were combined for the statistical analysis. Prognostic Variables Clinical data, such as age, KPS, extent of resection, recurrence interval, tumor volume, ependymal involvement of the tumor in the MR images at recurrence, and adjuvant chemotherapy, were collected and used in the univariate analysis of overall survival. The variables selected for the prognosis analysis were those variables determined to be significant based on previous reports on recurrent glioblastoma surgery.8 12 Tumor volumes were determined using the Osiris software (version 4.8; Service of Medical Informatics, Geneva University Hospital, Geneva, Switzerland) by summing a series of regions of interest in a set of multiplanar MR images. The tumor was defined as having ependymal involvement when there was an enhanced lesion at the ventricular wall. We also validated the NIH Recurrent GBM Scale with use of the same method proposed previously.12 Statistical Analysis The KaplanMeier method was used to estimate the overall survival distributions, and a Cox proportional hazards model was used to adjust for covariates. The Breslow test was used to identify differences in the overall survival distributions with respect to the prognostic variables and scale groups. To select the candidate variables for the scale components, a significance level of a 0.10 was used. For the validation of the NIH scale and the newly developed scale, a significance level of a 0.05 was used. These analyses were performed using IBM SPSS Statistics software (version 16.0; SPSS Inc., Chicago, IL). Patient Demographic Characteristics The baseline clinical data of the study and the validation cohorts are summarized in Table 1. Among the 55 patients in the study cohort, the median age was 50 years (range, 24 73 years). Thirty-three patients were male (60.0%), and 22 patients were female (40.0%). The median recurrence interval from the initial operation to the second operation was 10 months (range, 3 101 months). The median overall survival among all the patients after reoperation was 13.0 months (95% confidence interval [CI], 10.6 15.4). Validation of the NIH Recurrent GBM Scale We applied the NIH Recurrent GBM Scale to the current patient group. As in previous reports,12 the NIH Recurrent GBM Scale was meaningful in distinguishing prognostic groups (Table 2). However, the difference in the median survival between the intermediate prognosis group (NIH Recurrent GBM Scale 1 or 2) at 14.0 months (95% CI, 10.4 17.6) and good prognosis group (NIH Recurrent GBM Scale 0) at 14.0 months (95% CI, 10.7 17.3) was not statistically significant (P .113) (Fig. 1). Univariate analysis of the NIH Recurrent GBM Scale variables revealed that the Motor-Speech-MCA (MSM) score may be responsible for the difference in the survivals between the intermediate prognostic group Validation cohort (n 5 96) Number of patients Study cohort (n 5 55) Number of patients Characteristics Age, years 50 Male 33 Female 22 Extent of resection at recurrence 95% resection 37 ,95% resection 18 Performance status, KPS 100 12 90 12 80 4 70 10 60 13 50 4 Tumor volume 50 cm3 33 ,50 cm3 22 MSM score 0 13 1 21 2 19 3 2 Ependymal involvement Yes 25 No 30 Adjuvant chemotherapy Yes 22 No 33 Abbreviations: MSM, Motor-Speech-MCA; N.A., not available. Table 2. Prognostic groups as defined by the NIH Recurrent GBM Scale (n 55) Prognostic group Good Intermediate Poor NIH scale Number of patients 16 Kaplan Meier & Breslow tests. Median survival (months) 10.7 17.3 10.4 17.6 0.0 12.7 and good prognostic group being indistinguishable (Table 3). New Scale for Recurrent Glioblastoma Surgery Among the prognostic variables that significantly affected overall survival according to the univariate analysis, a Fig. 1. Kaplan Meier survival plots of patients in study cohort stratified by prognostic groups according to the NIH Recurrent GBM Scale. Table 3. Survival differences according to the NIH Recurrent GBM Scale variables (n 55) Variables Median survival (months) Yes Kaplan Meier & Breslow tests. KPS of , 70 (hazard ratio, 0.395; 90% CI, 0.166 0.940; P .078) and ependymal involvement (hazard ratio, 0.411; 90% CI, 0.214 0.789; P .025) remained significant prognostic factors in the multivariate analysis (Table 4). Using these 2 variables, we devised a new scoring system to help decide whether to proceed with surgical resection in recurrent glioblastoma. A 3-tier scale (range, 0 2 points) composed of additive scores of the KPS (0 for KPS 70 and 1 for KPS , 70) and ependymal involvement (0 for no enhancement and 1 for enhancement of the ependymal lining in MRI) was established. On the basis of the new scale, we could distinguish good (0 points), intermediate (1 point), and poor (2 points) prognostic groups (Table 5). A survival analysis confirmed the significant differences between the groups (Fig. 2A). Patients with a good prognosis (score, 0; median survival, 18.0 months; 95% CI, 10.3 25.7) differed significantly from those patients with intermediate (score, 1; median survival, 10.0 months; 95% CI, 5.2 14.8; P .006) or poor (score, 2; median survival, 4.0 months; 95% CI, 1.1 6.9; P .000) prognoses. Moreover, there were significant differences in survival between the intermediate group and poor group (P .010). Variables Median survival (months) Table 5. A new scale for recurrent glioblastoma surgery and its prognostic significance Prognostic group New scale Median survival (months) Study cohort (n 55) Good 0 Intermediate 1 Poor 2 Validation cohort (n 96) Good 0 Intermediate 1 Poor 2 KaplanMeier and Breslow tests. The significance of the newly devised scale was confirmed using a separate validation cohort (Table 1). The median age for the 96 patients from 3 different institutes in the validation cohort was 49 years (range, 13 74 years). Sixty-five patients were male (67.7%), and 31 patients were female (32.3%). The median recurrence interval from the initial operation to the second operation was 12 months (range, 1 98 months). The median overall survival among all the patients after reoperation was 10.0 months (95% CI, 8.5 11.5). Patients with a good prognosis (score, 0; n 65; median survival, 11.0 months; 95% CI, 9.4 12.6) lived significantly longer compared with patients with intermediate (score, 1; n 27; median survival, 9.0 months; 95% CI, 7.7 10.3; P .023) or poor (score, 2; n 4; median survival, 4.0 months; 95% CI, 1.1 6.9; P .000) prognoses (Table 5, Fig. 2B). There were significant differences in survival between the intermediate group and poor group (P .013). Decision to Proceed with Repeated Surgery for Recurrent Glioblastoma The subgroup analysis of each prognostic group in the study cohort revealed that there were no significant differences in survival after reoperation, regardless of whether adjuvant treatment was administered, in either the good prognosis group (median survival, 20.0 months [n 12] Univariate analysis, P value 90% confidence interval Multivariate analysis, P value 9.412.6 7.710.3 1.16.9 0.1660.940 0.5041.847 0. 3981.819 0.2140.789 1.0103.654 vs 14.0 months [n 10]; P .492) or poor prognosis group (median survival, 4.0 months [n 2] vs 3.0 months [n 7]; P .486). However, the intermediate prognosis group exhibited a survival gain after reoperation when adjuvant treatment was administered compared with when it was omitted (median survival, 18.0 months [n 8] vs 8.0 months [n 16]; P .066). The decision to administer adjuvant chemotherapy was based on the surgeons impression of the operative findings. Discussion Repeated surgery has been a relatively infrequent option for the treatment of recurrent glioblastoma because the disease is frequently beyond the limits of local therapies when it is diagnosed. However, there is a group of patients with recurrent glioblastoma that will benefit from maximal surgical resection. The only problem is the lack of reliable guidelines that identify favorable surgical candidates among patients with recurrent glioblastoma. The NIH Recurrent GBM Scale was the first attempt to address this issue and has been shown to be a valuable and objective standard.12 However, there are obstacles with the application of the NIH Recurrent GBM Scale in daily clinical practice. The estimation of the eloquent/critical area is subjective, and the measurement of tumor volume is not an intuitional process. The scale suggested in the present study provides a simple and practical method to classify prognostic groups, which can be helpful in designing the management plans for patients with recurrent glioblastoma who are candidates for surgery. Ependymal involvement in gadolinium-contrast MR images and the performance status obtained using the KPS 70 criteria are intuitive and simple, which renders discordance between clinicians to be rare. Moreover, the suggestion of appropriate management guidelines based on the new scale can be helpful in daily clinical practice. Considering the results of the survival analysis for the various prognostic groups, the new scale can be used to guide the treatment of recurrent glioblastoma. When the preoperative estimation of a new scale for recurrent glioblastoma achieves a score of 0, surgical resection and observation until progression represent a reasonable approach to treatment. When the score is 1, surgical resection and adjuvant chemotherapy may be beneficial in extending the survival period. For patients with a score of 2, surgery is not recommended and conservative management may be better. However, this proposal needs to be validated by a prospective study because the retrospective context involves a study population subject to selection bias. Performance status based on the KPS 70 criteria is a well-documented predictive factor of recurrent glioblastoma surgery and glioblastoma in general.8,9,11,16 Ependymal or subventricular zone involvement in glioblastoma is a prodromic sign of multifocal progression, which is a contraindication for surgical treatment.17 Lim et al. reported that glioblastomas that have no contact with the subventricular zone recur at the border of the primary lesion, and local treatment, such as surgery, is recommended as a reasonable treatment option.17 Bohman et al. found that glioblastomas abutting the ventricle are significantly larger in size than those that do not.18 Young et al. showed that the involvement of the subventricular zone and the tumor growth rate during radiotherapy were significantly worse prognostic factors for progression-free and overall survival among patients with glioblastoma.19 However, it remains unproven whether the mechanism of poor prognosis of ependymal involvement of tumor is the result of specific ontogeny or a reflection of the stage progression of the glioblastoma.18,19 Serial inspection of the patients and periodic MR images obtained during follow-up are basic and routine practices in the treatment of patients with glioblastoma. Despite the modern technological innovations facilitated by molecular diagnoses, recurrent glioblastoma is likely to be treated with surgery. One of the limitations of this new scale is that it is useful only for measurable lesions that can be excised, not for extensive tumors or tumors in inoperable areas. The result of the present study may serve as a starting point for the prospective studies on evaluating the value of surgical resection for recurrent glioblastoma. This work was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF), which was funded by the Ministry of Education, Science and Technology (2012R1A1A2003779), and by a grant from the Seoul National University Hospital Research Fund (04-2011-0190).


This is a preview of a remote PDF: https://neuro-oncology.oxfordjournals.org/content/15/8/1096.full.pdf

Chul-Kee Park, Jeong Hoon Kim, Do-Hyun Nam, Chae-Yong Kim, Sang-Bong Chung, Young-Hoon Kim, Ho Jun Seol, Tae Min Kim, Seung Hong Choi, Se-Hoon Lee, Dae Seog Heo, Il Han Kim, Dong Gyu Kim, Hee-Won Jung. A practical scoring system to determine whether to proceed with surgical resection in recurrent glioblastoma, Neuro-Oncology, 2013, 1096-1101, DOI: 10.1093/neuonc/not069