Second Surgery for Recurrent Glioblastoma
The first important reports on this issue appeared in 1987, when Ammirati et al. studied 55 cases of recurrent anaplastic astrocytoma and glioblastoma (20 and 35, respectively); in all cases, second surgery was performed; 41 of the 55 patients (75%) underwent postoperative chemotherapy and/or radiotherapy. The overall median survival was 92 weeks, whereas median survival time after surgery was 36 weeks. The outcome was better in patients with a low-grade recurrence at histology and a good Karnofsky performance status (KPS; >70) who underwent radical resection. At multivariate analysis, the most powerful variables were KPS and extent of resection. However, the reliability of the data reported in this study, one of the first to suggest a correlation between reoperation and survival, was undermined by the fact that it was conducted on a small series of patients with different histologies. In their retrospective review in 1994, Landy et al. analyzed 33 recurrent glioma patients, all of whom had second surgery; 12 for glioblastoma, eight for anaplastic astrocytoma, nine for astrocytoma, three for oligodendroglioma and two for mixed glioma. After surgery, nine patients had reirradiation and 24 had chemotherapy. The median survival after reoperation was 8, 13, 22 and 47 months for glioblastoma, anaplastic astrocytoma, astrocytoma and oligodendroglioma/mixed group, respectively. Despite a trend toward improvement in survival depending on tumor grade, the difference did not attain statistical significance; a significant correlation was found between age/KPS and prognosis. Younger patients with a better KPS have a longer survival. According to the authors, 11 patients benefited from second surgery and any consideration of aggressive treatment for recurrent gliomas should be balanced against the predicted outcome since symptoms of pressure, unlike those of infiltration, are usually ameliorated by surgery. The small series precluded any sound conclusion: patients with a better histological diagnosis would probably have enjoyed a better outcome, irrespective of the surgical approach.
In 1999, Keles et al. reported on a series of 92 patients with hemispheric glioblastoma who underwent a total of 107 operations; 52 were undertaken on patients who had undergone previous surgery or biopsy; the percentage of resection and volume of residual disease were calculated using volumetric image technique analysis. At unsorted analysis, it was found that preoperative KPS, chemotherapy, percentage of resection and volume of residual disease have a statistically significant effect on time to tumor progression: improvement in outcome was proportionate to residual disease reduction.
In 2001, Pinsker and Lumenta reported on 38 patients who underwent repeat surgery for recurrent glioblastoma, and those given total resection lived longer than those who had subtotal resection (STR; 21 and 18 weeks, respectively), although the difference was not of statistical significance. Moreover, on evaluating the effect of conventional variables (age: <50 years; time between first and second resection >26 weeks; KPS: >90 at second resection, total primary resection) were correlated with a longer survival; for postoperative survival, extent of resection, unlike age and recurrence-free interval, KPS was predictive of outcome improvement. These data demonstrate that second surgery enhances efficacy of treatment in patients with a good KPS. Although this study is valuable because it was performed on a homogeneous, well defined and stratified cohort, the size of the series compromises the value of the evidence provided.
In Mandl et al.'s retrospective analysis of 32 patients with recurrent glioblastoma, the inclusion criteria were a good clinical condition (KPS at least 60), local recurrence without multifocality and the feasibility of debulking. The cohort was split into three subgroups: nine patients received only surgical resection as salvage therapy (four had more than one operation), 11 had surgery plus chemotherapy or stereotactic radiosurgery (SRS) and 12 were given chemotherapy alone or SRS. The median overall survival was 34 weeks in patients who had chemotherapy or SRS plus surgery, 28 weeks for those given chemotherapy alone or SRS and 13 weeks for those resected without further treatment. The authors, therefore, advised that patients with severe mass effect symptoms should be considered candidates for resection followed by further salvage therapy but that patients with mild symptoms could be spared surgery since a similar survival advantage can be gained by other approaches. Importantly, in 40% of re-operated patients, performance status deteriorated; this suggests that second surgery may be detrimental. This study, however, was retrospective and the series is very small; a reliable comparison between different treatment strategies should therefore be accomplished through a prospective analysis.
In 2009, McGirt et al. made a retrospective analysis on a large series of patients with malignant astrocytoma; altogether, 949 patients were considered: 700 were WHO grade IV and 249 WHO grade III. Of these patients, 549 had primary resection and 400 had second surgery; 294 of the latter had glioblastoma. Resection was considered gross-total resection (GTR) in the absence of contrast enhancement at postoperative MRI, near-total resection (NTR) if a rim enhancement of the resection cavity was evidenced and STR if nodular enhancement was found. In patients given second surgery, the median survival after GTR, NTR and STR was 11.9 and 5 months, respectively. At adjustment for variables independent at multivariate analysis (i.e., age, KPS and temozolomide adjuvant therapy), GTR and NTR were associated with improved survival; GTR provided a 10% greater reduction of the risk of death than NTR, which provided a further 37% greater decrease than STR. The time interval between two subsequent surgeries and the tumor site was not a significant factor. The author of the paper concluded that extensive resection, even at the time of recurrence, may provide a better outcome and patients receiving optimal surgery are more likely to benefit from subsequent therapy. This study is significant because the series was large and any bias obviated by precautions such as a blind review of the neuroradiological images.
In a publication by Park et al. that appeared in 2010, a series of 34 consecutive patients with recurrent supratentorial hemispheric glioblastoma was investigated with a view to identifying a prognostic model predictive of outcome. Patients had radical resection for recurrence in a single institution, the procedures being performed by the same surgeon. At survival analysis, significant variables for survival were KPS <80, tumor mass >50 cm and a motor-speech-middle cerebral artery (a scale to assess the involvement of three eloquent/critical brain areas) score of >2. Based on findings with these parameters, used to determine the overall score for each case, the patients were divided into three prognostic groups; median survival was 10.8 months for patients with null score and 1 month for those with a score of 3. In their comment concerning this study, Xu et al. argued that the since the number of patients believed to have the worst prognosis was only 10%, there might have been a selection bias that precluded a reliable conclusion; moreover, they stated that performance status alone cannot be considered an independent prognostic factor because it is frequently influenced by tumor location; finally they pointed out that although all the patients underwent surgery, some may have benefited from upfront chemotherapy without any adverse effect on the final outcome.
In 2011, Clarke et al. analyzed 758 patients with recurrence from glioblastoma; of the cohort, enrolled in the North American Brain Tumor Consortium (NABTC) Phase II clinical trials, 208 underwent second surgery at the time of disease progression/relapse. Patients who underwent surgery were compared with those who did not for progression-free survival (PFS) at 6 months and overall survival. No difference was found between the surgical and nonsurgical groups, either for PFS and overall survival, which ranged from 8 to 19 weeks and from 24 to 34 weeks, respectively. Chamberlain and Silbergeld questioned the methodological approach used in this study, arguing that it did not report data on tumor volume or extent of resection; nor did the authors evaluate whether there was any correlation between surgery and improvement of symptoms – this would have strengthened their assumption that second surgery does not affect outcome.
In 2012, De Bonis et al. made a retrospective evaluation of 76 recurrent glioma patients, 17 of whom had second surgery alone, 24 had chemotherapy alone, 16 had surgery and chemotherapy and 19 had no treatment. It was found that patients undergoing surgery and chemotherapy lived longer than patients who had alternative treatment. Moreover, unlike age and extent of resection, performance status was a significant independent prognostic factor.
In 2007, Carson et al. reported the results of a recursive partitioning analysis (RPA) conducted on 333 recurrent glioma patients in Phase I and Phase II trials in order to investigate systemic or local chemotherapy and brachytherapy in the New Approaches to Brain Tumor Therapy (NABTT) CNS Consortium. Glioblastoma patients made up 67.4% of the study population; 44.5% (n = 146) underwent second surgery. The RPA selected different prognostic factors for the glioblastoma and nonglioblastoma group. The former was split on the basis of KPS (60–70 vs 80–100), age (50 vs <50) and tumor location (outside frontal lobe vs confined to frontal lobe) while the latter was split on the basis of KPS (60–80 vs 90–100), age (50 vs <50) and corticosteroid use. At analysis, seven classes were identified. Survival ranged from 3.8 months for patients with histology other than glioblastoma and a low KPS, to 10.4 months for glioblastoma patients younger than 50 years and a high KPS, to 25.7 months for nonglioblastoma patients with a frontal tumor and a good performance status. The number of surgical procedures performed was not significant at multivariate analysis being ruled out by RPA.
In order to investigate prognostic factors in recurrent glioma patients, in 2012, Gorlia et al. considered 300 patients recruited in Phase I or II trials conducted by the EORTC Brain Tumor Group: among them, 138 had received temozolomide concomitant with and adjuvant to radiotherapy, 158 received radiotherapy alone or combined with another chemotherapy regimen and four received chemotherapy without previous radiotherapy. From progression, the median overall survival and the PFS were 6.2 and 1.8 months, respectively. The prognostic variables that proved significant were performance status, neurological deficit, time since primary diagnosis, baseline administration of steroids, number of target lesions, tumor size, frontal tumor location and prior chemotherapy with temozolomide; at multivariate analysis, variables impacting on overall survival were performance status, baseline steroids, number of target lesions, frontal location and tumor diameter. Despite this, the study did not directly address the issue of reoperation, 8% of patients underwent second surgery, but this approach did not affect survival.
Analysis of the Literature
The first important reports on this issue appeared in 1987, when Ammirati et al. studied 55 cases of recurrent anaplastic astrocytoma and glioblastoma (20 and 35, respectively); in all cases, second surgery was performed; 41 of the 55 patients (75%) underwent postoperative chemotherapy and/or radiotherapy. The overall median survival was 92 weeks, whereas median survival time after surgery was 36 weeks. The outcome was better in patients with a low-grade recurrence at histology and a good Karnofsky performance status (KPS; >70) who underwent radical resection. At multivariate analysis, the most powerful variables were KPS and extent of resection. However, the reliability of the data reported in this study, one of the first to suggest a correlation between reoperation and survival, was undermined by the fact that it was conducted on a small series of patients with different histologies. In their retrospective review in 1994, Landy et al. analyzed 33 recurrent glioma patients, all of whom had second surgery; 12 for glioblastoma, eight for anaplastic astrocytoma, nine for astrocytoma, three for oligodendroglioma and two for mixed glioma. After surgery, nine patients had reirradiation and 24 had chemotherapy. The median survival after reoperation was 8, 13, 22 and 47 months for glioblastoma, anaplastic astrocytoma, astrocytoma and oligodendroglioma/mixed group, respectively. Despite a trend toward improvement in survival depending on tumor grade, the difference did not attain statistical significance; a significant correlation was found between age/KPS and prognosis. Younger patients with a better KPS have a longer survival. According to the authors, 11 patients benefited from second surgery and any consideration of aggressive treatment for recurrent gliomas should be balanced against the predicted outcome since symptoms of pressure, unlike those of infiltration, are usually ameliorated by surgery. The small series precluded any sound conclusion: patients with a better histological diagnosis would probably have enjoyed a better outcome, irrespective of the surgical approach.
In 1999, Keles et al. reported on a series of 92 patients with hemispheric glioblastoma who underwent a total of 107 operations; 52 were undertaken on patients who had undergone previous surgery or biopsy; the percentage of resection and volume of residual disease were calculated using volumetric image technique analysis. At unsorted analysis, it was found that preoperative KPS, chemotherapy, percentage of resection and volume of residual disease have a statistically significant effect on time to tumor progression: improvement in outcome was proportionate to residual disease reduction.
In 2001, Pinsker and Lumenta reported on 38 patients who underwent repeat surgery for recurrent glioblastoma, and those given total resection lived longer than those who had subtotal resection (STR; 21 and 18 weeks, respectively), although the difference was not of statistical significance. Moreover, on evaluating the effect of conventional variables (age: <50 years; time between first and second resection >26 weeks; KPS: >90 at second resection, total primary resection) were correlated with a longer survival; for postoperative survival, extent of resection, unlike age and recurrence-free interval, KPS was predictive of outcome improvement. These data demonstrate that second surgery enhances efficacy of treatment in patients with a good KPS. Although this study is valuable because it was performed on a homogeneous, well defined and stratified cohort, the size of the series compromises the value of the evidence provided.
In Mandl et al.'s retrospective analysis of 32 patients with recurrent glioblastoma, the inclusion criteria were a good clinical condition (KPS at least 60), local recurrence without multifocality and the feasibility of debulking. The cohort was split into three subgroups: nine patients received only surgical resection as salvage therapy (four had more than one operation), 11 had surgery plus chemotherapy or stereotactic radiosurgery (SRS) and 12 were given chemotherapy alone or SRS. The median overall survival was 34 weeks in patients who had chemotherapy or SRS plus surgery, 28 weeks for those given chemotherapy alone or SRS and 13 weeks for those resected without further treatment. The authors, therefore, advised that patients with severe mass effect symptoms should be considered candidates for resection followed by further salvage therapy but that patients with mild symptoms could be spared surgery since a similar survival advantage can be gained by other approaches. Importantly, in 40% of re-operated patients, performance status deteriorated; this suggests that second surgery may be detrimental. This study, however, was retrospective and the series is very small; a reliable comparison between different treatment strategies should therefore be accomplished through a prospective analysis.
In 2009, McGirt et al. made a retrospective analysis on a large series of patients with malignant astrocytoma; altogether, 949 patients were considered: 700 were WHO grade IV and 249 WHO grade III. Of these patients, 549 had primary resection and 400 had second surgery; 294 of the latter had glioblastoma. Resection was considered gross-total resection (GTR) in the absence of contrast enhancement at postoperative MRI, near-total resection (NTR) if a rim enhancement of the resection cavity was evidenced and STR if nodular enhancement was found. In patients given second surgery, the median survival after GTR, NTR and STR was 11.9 and 5 months, respectively. At adjustment for variables independent at multivariate analysis (i.e., age, KPS and temozolomide adjuvant therapy), GTR and NTR were associated with improved survival; GTR provided a 10% greater reduction of the risk of death than NTR, which provided a further 37% greater decrease than STR. The time interval between two subsequent surgeries and the tumor site was not a significant factor. The author of the paper concluded that extensive resection, even at the time of recurrence, may provide a better outcome and patients receiving optimal surgery are more likely to benefit from subsequent therapy. This study is significant because the series was large and any bias obviated by precautions such as a blind review of the neuroradiological images.
In a publication by Park et al. that appeared in 2010, a series of 34 consecutive patients with recurrent supratentorial hemispheric glioblastoma was investigated with a view to identifying a prognostic model predictive of outcome. Patients had radical resection for recurrence in a single institution, the procedures being performed by the same surgeon. At survival analysis, significant variables for survival were KPS <80, tumor mass >50 cm and a motor-speech-middle cerebral artery (a scale to assess the involvement of three eloquent/critical brain areas) score of >2. Based on findings with these parameters, used to determine the overall score for each case, the patients were divided into three prognostic groups; median survival was 10.8 months for patients with null score and 1 month for those with a score of 3. In their comment concerning this study, Xu et al. argued that the since the number of patients believed to have the worst prognosis was only 10%, there might have been a selection bias that precluded a reliable conclusion; moreover, they stated that performance status alone cannot be considered an independent prognostic factor because it is frequently influenced by tumor location; finally they pointed out that although all the patients underwent surgery, some may have benefited from upfront chemotherapy without any adverse effect on the final outcome.
In 2011, Clarke et al. analyzed 758 patients with recurrence from glioblastoma; of the cohort, enrolled in the North American Brain Tumor Consortium (NABTC) Phase II clinical trials, 208 underwent second surgery at the time of disease progression/relapse. Patients who underwent surgery were compared with those who did not for progression-free survival (PFS) at 6 months and overall survival. No difference was found between the surgical and nonsurgical groups, either for PFS and overall survival, which ranged from 8 to 19 weeks and from 24 to 34 weeks, respectively. Chamberlain and Silbergeld questioned the methodological approach used in this study, arguing that it did not report data on tumor volume or extent of resection; nor did the authors evaluate whether there was any correlation between surgery and improvement of symptoms – this would have strengthened their assumption that second surgery does not affect outcome.
In 2012, De Bonis et al. made a retrospective evaluation of 76 recurrent glioma patients, 17 of whom had second surgery alone, 24 had chemotherapy alone, 16 had surgery and chemotherapy and 19 had no treatment. It was found that patients undergoing surgery and chemotherapy lived longer than patients who had alternative treatment. Moreover, unlike age and extent of resection, performance status was a significant independent prognostic factor.
In 2007, Carson et al. reported the results of a recursive partitioning analysis (RPA) conducted on 333 recurrent glioma patients in Phase I and Phase II trials in order to investigate systemic or local chemotherapy and brachytherapy in the New Approaches to Brain Tumor Therapy (NABTT) CNS Consortium. Glioblastoma patients made up 67.4% of the study population; 44.5% (n = 146) underwent second surgery. The RPA selected different prognostic factors for the glioblastoma and nonglioblastoma group. The former was split on the basis of KPS (60–70 vs 80–100), age (50 vs <50) and tumor location (outside frontal lobe vs confined to frontal lobe) while the latter was split on the basis of KPS (60–80 vs 90–100), age (50 vs <50) and corticosteroid use. At analysis, seven classes were identified. Survival ranged from 3.8 months for patients with histology other than glioblastoma and a low KPS, to 10.4 months for glioblastoma patients younger than 50 years and a high KPS, to 25.7 months for nonglioblastoma patients with a frontal tumor and a good performance status. The number of surgical procedures performed was not significant at multivariate analysis being ruled out by RPA.
In order to investigate prognostic factors in recurrent glioma patients, in 2012, Gorlia et al. considered 300 patients recruited in Phase I or II trials conducted by the EORTC Brain Tumor Group: among them, 138 had received temozolomide concomitant with and adjuvant to radiotherapy, 158 received radiotherapy alone or combined with another chemotherapy regimen and four received chemotherapy without previous radiotherapy. From progression, the median overall survival and the PFS were 6.2 and 1.8 months, respectively. The prognostic variables that proved significant were performance status, neurological deficit, time since primary diagnosis, baseline administration of steroids, number of target lesions, tumor size, frontal tumor location and prior chemotherapy with temozolomide; at multivariate analysis, variables impacting on overall survival were performance status, baseline steroids, number of target lesions, frontal location and tumor diameter. Despite this, the study did not directly address the issue of reoperation, 8% of patients underwent second surgery, but this approach did not affect survival.