High-grade glioma (World Health Organization Grades III and IV) Salvage Therapy in Patients with Glioblastoma. Is There any Benefit?

Transcription

1 2678 Salvage Therapy in Patients with Glioblastoma Is There any Benefit? Peter Hau, M.D. 1 Ulrike Baumgart, M.D. 1 Katharina Pfeifer, M.D. 1 Anne Bock, R.N. 1 Tanya Jauch, R.N. 1 Jörg Dietrich, M.D. 1,2 Klaus Fabel, M.D. 1,3 Oliver Grauer, M.D. 1 Caecilia Wismeth, M.D. 1 Monika Klinkhammer-Schalke, M.D. 4 Michael Allgäuer, M.D. 5 Gerhard Schuierer, M.D. 6 Horst Koch, M.D. 1 Juergen Schlaier, M.D. 7 Winfried Ulrich, M.D. 7 Alexander Brawanski, M.D. 7 Ulrich Bogdahn, M.D. 1 Andreas Steinbrecher, M.D. 1 1 Klinik und Poliklinik für Neurologie der Universität Regensburg im Bezirksklinikum, Regensburg, Germany. 2 Department of Biomedical Genetics, University of Rochester, Rochester, New York. 3 Department of Neurosurgery, Stanford University Medical Center, Stanford, California. 4 Tumorzentrum der Universität Regensburg, Regensburg, Germany. 5 Abteilung für Strahlentherapie, Krankenhaus der Barmherzigen Brüder Regensburg, Regensburg, Germany. 6 Neuroradiologische Abteilung im Bezirksklinikum Regensburg, Regensburg, Germany. 7 Klinik und Poliklinik für Neurochirurgie der Universität Regensburg, Regensburg, Germany. Address for reprints: Peter Hau, M.D., Klinik und Poliklinik für Neurologie der Universität Regensburg im Bezirksklinikum Universitätstrasse 84, D Regensburg, Germany; Fax: (011) ; peter.hau@bkr-regensburg. de Received May 27, 2003; revision received July 25, 2003; accepted September 11, BACKGROUND. Survival after first-line therapy is poor for patients with glioblastoma. The role of second-line treatment for recurrent disease is controversial. The authors studied the outcome in a subset of patients with glioblastoma who were selected for an aggressive reintervention strategy at the time of progression. Their objectives were to improve patients overall survival with sustained quality of life and to make comparisons with overall survival in unselected patients. METHODS. Overall, 168 patients were eligible for retrospective analysis. Ninety patients received specific therapy for disease recurrence (reintervention group) by specific criteria. RESULTS. In the reintervention group, promising median overall survival (mos) results after diagnosis (61.5 weeks) and progression (33 weeks) were obtained. The progression-free survival (PFS) rate at 12 months and the overall survival rate were superior in the reintervention group (71% at 12 months and 32% at 24 months) compared with the total cohort (45% and 20%, respectively) and the standard group (15% and 5%, respectively). A matched-pair analysis (n 46 in each group), with an mos period of 65.5 versus 28.5 weeks, confirmed these data. Quality of life was stable or slightly improved during reinterventions in a subset of patients treated within clinical studies. CONCLUSIONS. The majority of patients in the current series were treated with a reintervention strategy, which had an impact on PFS and mos. A second resection, focal radiotherapy (in selected cases), and additional chemotherapeutic regimens should be considered for patients with recurrent glioblastoma. Cancer 2003;98: American Cancer Society. KEYWORDS: glioblastoma, disease recurrence, salvage therapy, reintervention. High-grade glioma (World Health Organization Grades III and IV) has an incidence of approximately 5 in 100,000 per year and represents approximately 40% of primary brain tumors in adults, with 50% of them belonging to the most malignant phenotype, glioblastoma. 1 Median overall survival (mos) after first-line therapy is poor, i.e., months after treatment with surgery, radiotherapy, and adjuvant chemotherapy. 2 Tumor grading, 1 age, 3 extent of resection, 4 and Karnofsky performance score (KPS) at diagnosis 5 are the most important prognostic factors for a favorable outcome. Adjuvant chemotherapy is beneficial only for a subset of patients. 6 In one analysis, 7 chemotherapy was found to be associated with an increase in long-term survivors (survival period 5 years). The overall survival rate at 18 months was 11.1% with radiotherapy alone and 22.9% with radiotherapy and N,N -bis(2-chloroethyl)-nnitrosourea (BCNU) in patients with glioblastoma. This is consistent with a recently published metaanalysis that included data from 12 randomized controlled trials comprising 3004 patients. Overall, re American Cancer Society DOI /cncr.11845

2 Salvage Therapy in Glioblastoma/Hau et al sults showed that significant prolongation of survival was associated with chemotherapy. 8 The role of reinterventions at the time of tumor recurrence is unclear. Recurrent glioma appears to be relatively resistant to currently used second-line chemotherapies. 7,9 Most authors report overall response rates 10%. In contrast to most solid tumors, for which response is defined as decreasing tumor mass, treatment of high-grade glioma can be considered successful if stable disease is achieved. 10 In a recent metaanalysis of 8 Phase II studies involving 225 patients with recurrent glioblastoma, mos after disease recurrence was 25 weeks and the progression-free survival (PFS) rate at 6 months was 15%. 11 KPS, but not age, was defined as a prognostic factor. A less favorable outcome was observed in patients receiving more than two surgical interventions and/or more than two different chemotherapies overall. All chemotherapeutic agents used in the current study have shown efficacy in treating glioblastoma. 1-(4-amino-2-methyl-5-pyrimidinyl)methyl-3-(2- chloroethyl)-3-nitrosourea (ACNU) plus teniposide 12 appears promising, with an mos of 74 weeks and an overall survival rate of 25% two years after diagnosis. This regimen, traditionally used in Germany and Austria, is comparable to regimens combining the other commonly used nitrosoureas, BCNU and 1-(2-chloroethyl)-3-cyclohexyl-1-nitrosourea (CCNU), with topoisomerase inhibitors. The combination of procarbazine, CCNU, vincristine, and temozolomide exhibited efficacy in several trials, 6,13 16 as did pegylated liposomal doxorubicin 17 in a published single-institution study. 13-cis retinoic acid was used as a maintenance strategy in same patients. 18 Considering the efficacy of the regimens listed above, we hypothesized that reinterventions are feasible for individual patients and might improve survival in selected patients with recurrent glioblastoma. To improve our understanding regarding which subgroup of patients is likely to benefit from a reintervention strategy, we analyzed our total cohort of patients with glioblastoma and compared the outcome of patients treated with reinterventions with that of the total unselected cohort, a standard group receiving no reinterventions, and with data from the literature. The purpose of the current study was to verify our initial assumption that it is possible to select and treat patients with a reintervention strategy and improve their overall survival. Although an improved overall survival might be expected because of selection bias, a comparable outcome would instead argue against a reintervention strategy. We believed it was important to determine the existence and magnitude of this effect in our population, the number of reinterventions that is feasible and effective, and the impact of the reintervention strategy on patients quality of life. MATERIALS AND METHODS Patient Selection and Prognostic Factors All consecutive patients referred to the Neurology Department, University of Regensburg (Regensburg, Germany) with a diagnosis of glioblastoma were analyzed retrospectively in a single-institution, nonrandomized observational trial. Tumor histology was evaluated at diagnosis and after any additional surgery. Patients treated within clinical trials were subject to confirmatory histology at the German Reference Center (Department of Neuropathology, University of Bonn, Bonn, Germany) for brain tumors. The last available histology data were used to decide on inclusion in the analysis. All patients who had a change in histology during follow-up (e.g., radionecrosis) were excluded. No other neuroimaging tools (e.g., positron emission tomography) were used in the current study. Treatment groups (reintervention group, total cohort, and standard group) were analyzed for relevant demographic data (gender and age), KPS at diagnosis and at first disease recurrence, extent of resection at diagnosis (macroscopically complete or incomplete), and treatment (surgery, radiotherapy, chemotherapy) performed at first disease occurrence and at recurrence. Treatment Criteria Some patients were treated in other centers and were referred to our tertiary care center at first disease recurrence. First-line treatment was performed according to our institutional standards or the guidelines of the referring institutions. Evidence of recurrent or progressive disease after first-line therapy was documented by a magnetic resonance imaging (MRI). MRI analysis was performed in three planes (i.e., T1 without and with gadolinium, T2, and FLAIR imaging). Disease recurrence was evaluated according to the criteria introduced by Macdonald et al. 19 Reinterventions were initiated if patients fulfilled specific treatment criteria. For example, surgery was performed first if a substantial benefit was expected. 20 The major criteria guiding this decision were location and tumor size (both influencing the chance of a macroscopic total resection), mass effect, and the impact of surgery on additional strategies. Reirradiation was performed only for patients eligible for stereotactic radiotherapy or if the maximum dose had not been reached at the previous treatment location. Chemotherapy with different regimens or therapy with experimental drugs was administered to patients with a KPS of 70 and an expected survival period of 3 months. Written informed consent was obtained from

3 2680 CANCER December 15, 2003 / Volume 98 / Number 12 patients before the start of chemotherapy. The local ethics review board approved all therapies, which were performed according to good clinical practice guidelines. Patients treated with individual regimens were monitored at least every 12 weeks by clinical examination and by MRI. Phase II protocols performed at our institution include regular follow-up visits and an MRI scan at least every 8 weeks. If exact follow-up data were not available, available data were averaged to the 15th day of the relevant month for statistical evaluation. If a patient was lost to follow-up, data were used as if the patient had reached his or her endpoint at this time. In addition to the total cohort, two subgroups were defined for analysis. The reintervention group consisted of all patients who received at least one additional therapy after first disease recurrence. The standard group consisted of all patients who received no specific therapy after first disease recurrence, except for symptomatic and palliative care. Progression and Survival All patients were included in an analysis of event data. The median time to first progression and mos after diagnosis (mttp-1 and mos-1, respectively) and median time to second progression and mos from first progression (mttp-2 and mos-2, respectively) were analyzed. In addition, PFS at 6, 12, and 24 months and overall survival at 12 and 24 months were evaluated. All demographic and survival data were evaluated descriptively, with medians and corresponding 95% confidence intervals (CI). The Kaplan Meier product-limit method was used to yield survival curves. 21 Nonparametric two-sample (Gehan generalized Wilcoxon test) and multisample tests (Peto and Peto generalized Wilcoxon test) were applied. 22,23 All calculations were performed with commercially available software packages (Statistica; Statsoft, Tulsa, OK; Number Cruncher Statistical Software, Kaysville, UT). Results were compared with those from a recent metaanalysis of eight Phase II studies. 11 PFS rates for recurrent glioblastoma were 15%, 8%, and 4% at 6, 12, and 24 months, respectively. Overall survival at 12 months was 21% in this metaanalysis. To assess results with respect to the bias introduced by the selection of appropriate patients for the reintervention group, a matched-pair analysis was performed for 46 patients from each group. Patients were matched for age, KPS, extent of resection at diagnosis, and treatment with first-line chemotherapy. Subgroup analysis (mttp, mos, and overall survival at 12 months) was performed for patients with more than 2 resections and/or chemotherapy regimens, based on the data of Wong et al., 11 which predicted an inferior outcome for patients who received more than 2 resections and/or chemotherapy regimens. Furthermore, we performed subgroup analysis for the following prognostic factors: gender, age, extent of resection at diagnosis, and KPS at diagnosis and at disease recurrence. In addition, a subgroup analysis was performed for patients treated within clinical studies. Overall survival was determined for patients with late progression (later than mttp) to evaluate its impact on the survival of the subgroup and the total cohort, respectively. Finally, mos for patients with more than the median number of chemotherapeutic reinterventions was compared with the mos for the entire reintervention subgroup to exclude the negative effect of serial therapies. Known significant prognostic factors (age, KPS, and extent of resection) were evaluated for correlations to the major survival data (number of disease recurrences, mttp, and mos) using the exponential regression method. 24 Quality of Life and Karnofsky Performance Score Quality of life was monitored and evaluated for a subset of patients treated within clinical trials by the certified Version 3.0 of the quality of life questionnaire (QLQ-C 30), provided by the European Organization for Research and Treatment of Cancer (EORTC). 25 The QLQ-C 30 is divided into three broad categories (global health status, functional scales, and symptom scales). To determine treatment-related deterioration of quality of life as opposed to deterioration induced by tumor progression, first measurements were performed at initiation of therapy and the next measurements early in treatment for recurrent disease (normally, the fourth course of chemotherapy). In case of missing items, scores were calculated based on items that were completed, by skipping missing items. 25,26 If complete questionnaires were missing, the next available questionnaire replaced the missing one. Statistics was performed according to the EORTC guidelines with SPSS (Chicago, IL) software. 26 Because scores are not linear, we used the Wilcoxon test as an estimate of significance. Performance was monitored using the KPS, 27 which rates general daily-life skills on a scale from 0 to 100. We determined patients performance before the start of each reintervention and at each follow-up visit during the course of therapy. The scale was used not only to monitor the overall performance of patients, but also to compare it with the quality of life data; deterioration of quality of life may not only be due to toxicity of chemotherapy but may also reflect a general tumor-related decline. We assumed that a decline in quality of life without a decline in KPS most prob-

4 Salvage Therapy in Glioblastoma/Hau et al TABLE 1 Demographic Data at Start of First-Line Therapy, at Start of Salvage Therapy, and Overall for All Patients, Patients with Reinterventions, and Standard Patients a Characteristic All patients (n 168; 100%) Reintervention group (n 90; 54%) Standard group (n 78; 46%) Demographic data No. of males (%) 103 (61) 57 (63) 46 (59) Median age in yrs (range) 55 (24 82) 53 (26 81) 55 (24 82) Median KPS (range) 80 (50 100) 80 (60 100) 80 (50 100) No. of complete resections at diagnosis (%) 65 (43) 43 (48) 22 (28) First-line therapy Resection (%) 168 (100) 90 (100) 78 (100) Radiotherapy (%) 161 (96) 90 (100) 71 (91) Chemotherapy (%) 90 (53) 48 (53) 42 (54) Therapy at disease recurrence No. of recurrent events (%) 164 (98) 90 (100) 74 (95) Median no. of recurrent events (range) 1 (0 4) 1 (0 4) 2 (0 4) Median KPS at disease recurrence (range) 70 (40 100) 70 (60 100) 60 (40 100) No. of resections at disease recurrence n.a. 47 (52) n.a. Median no. of resections (range) n.a. 1 (0 2) n.a. No. of radiotherapies at disease recurrence n.a. 13 (14) n.a. Median no. of radiotherapies (range) n.a. 0 (0 2) n.a. No. of chemotherapies at disease recurrence n.a. 142 (99) n.a. Median no. of chemotherapies (range) n.a. 2 (0 5) n.a. Therapy overall Median no. of therapies (range) 3 (1 9) 5 (3 9) 2.5 (1 4) Median no. of reinterventions (range) 1 (0 6) 2 (1 6) 0 (0 0) Median no. of resections (range) 1 (0 3) 2 (1 3) 1 (0 1) Median no. of radiotherapies (range) 1 (0 3) 1 (1 3) 1 (0 1) Median no. of chemotherapies (range) 1 (0 5) 2 (0 5) 0.5 (0 2) n.a.: not applicable; KPS: Karnofsky performance score. a Standard patients received up to four different therapies and up to two chemotherapy regimens, because some received maintenance therapy with 13-cis-retinoic acid without tumor progression. ably would have to be attributed to the toxicity of reinterventions if no other evidence for clinical tumor progression was present. RESULTS Data Accrual One hundred sixty-eight consecutive patients with glioblastoma were referred to our Neuro-Oncology Unit between January 1997 and September Fourty-one patients were lost at some time during follow-up, but their event data, quality of life, and KPS were analyzed until dropout when applicable. Ninety patients in the total cohort (54%) received reinterventions, and 78 patients (46%) received standard therapy only. Demographic data were obtained from all 168 patients. With respect to event data, mttp-1 was determined for 165, mttp-2 for 86, overall survival for 105, and mos-2 for 108 patients, with censored data for some patients who had not reached the endpoints at the time of the current evaluation. PFS and proportional overall survival were evaluated in corresponding subsets of patients. Although KPS data were sufficient for 107 patients, quality of life was evaluable only for patients treated within clinical studies and available for a subgroup of 43 patients. Incomplete data on quality of life and KPS measurements were due to physical or mental inability or the refusal of patients to comply with self-assessment (quality of life questionnaires). Patient Characteristics and Comparability of Groups All 168 patients had a histologically proven diagnosis of glioblastoma. This diagnosis was confirmed by reference pathology in 69 (41%) patients treated within clinical trials. Number of patients, gender, median age at diagnosis, and median KPS at diagnosis and at disease recurrence were well matched (Table 1), except for the extent of resections in both subgroups. To control for a possible bias induced by the difference in surgical radicality, a subset analysis with matched pairs was performed and included 46 patients in each group (see Progression and Survival in Materials and Methods). First-line therapy in the standard and reintervention groups was comparable and is outlined in Table 1. All patients receiving second or further resec-

5 2682 CANCER December 15, 2003 / Volume 98 / Number 12 tions at disease recurrence underwent debulking surgery. No diagnostic biopsies were performed for patients with recurrent disease. Types of Reinterventions Reinterventions were performed only when the criteria defined above were fulfilled. A total of 202 reinterventions were performed. Different therapeutic strategies were used, as summarized in Table 1. With 142 regimens applied, chemotherapy was the most common type of reintervention and was used in 89 patients (99%). The following chemotherapeutic regimens were initiated: 90 mg/m 2 ACNU on Day 1/60 mg/m 2 teniposide on Days 1 3 for 42 days in 18, procarbacine, CCNU and vincristine (PCV) in 6, temozolomide in 57, pegylated liposomal doxorubicin in 25, and other experimental regimens in 36 patients with recurrent disease. Eighty-nine patients received one, 43 patients two, 9 patients three, and 1 patient four chemotherapeutic regimens after disease recurrence. 24 months was significantly better in the reintervention group (Table 3). Overall survival after diagnosis was significantly better for patients receiving treatment in a clinical trial Time to Progression, Overall Survival, and Proportional Survival The mttp after diagnosis was 26.5 weeks in the reintervention group, compared with 26 weeks in all patients and 21 weeks in the standard group; these results were not significantly different. The mttp after first disease recurrence was 13 weeks in the reintervention group; however, mos after diagnosis (Fig. 1A) was significantly greater (P ) in the reintervention group (61.5 weeks) than in the total cohort (46 weeks) or the standard group (26 weeks). The mos after disease recurrence was 33 weeks in the reintervention group, compared with 30 weeks in the total cohort and 9 weeks in the standard group. These results were further substantiated by a matched-pair analysis of mos (P 0.05; Fig. 1C; Table 2). The PFS rate at 6 months was 56% in the reintervention group and 50% and 48% in the total and standard groups, respectively; these results were comparable. However, the PFS rate at 12 months was superior in the reintervention group (21%) compared with the standard group (8%). In addition, the overall survival at 12 and FIGURE 1. (A) Cumulative proportional survival (overall survival) for patients with reinterventions versus standard patients and all patients. (B) Cumulative proportional survival (overall survival) for patients with one to seven reinterventions compared with standard patients. The first reintervention adds significantly to overall survival. There is only a trend toward prolonged survival for further reinterventions. (C) Cumulative proportional survival (overall survival) for patients with reinterventions versus standard patients from the matched-pair analysis.

6 Salvage Therapy in Glioblastoma/Hau et al TABLE 2 Efficacy of Treatment for All Patients and Patients with or without Reinterventions (i.e., Standard Therapy) mttp-1 mttp-2 mos-1 mos-2 Characteristic No. of patients Kaplan Meier (in weeks) 95% CI Kaplan Meier (in weeks) 95% CI Kaplan Meier (in weeks) 95% CI Kaplan Meier (in weeks) 95% CI All patients Reintervention group Standard group n.a. n.a All patients, matched-pairs a n.a. n.a Matched-pair reintervention Matched-pair standard n.a. n.a n.a.: not applicable; CI: confidence interval; mttp-1: median time to first progression; mttp-2: median time from first to second progression; mos-1: median overall survival; mos-2: median overall survival from first progression. a Results from the matched-pair analysis for patients with reintervention, standard patients, and all patients. TABLE 3 Proportional Survival from Diagnosis for All Patients, Patients with Reinterventions, and Standard Patients Characteristic PFS-6 (%) PFS-12 (%) PFS-24 (%) OS-12 (%) OS-24 (%) All patients Reintervention group Standard group PFS-6, -12, -24: progression-free survival 6, 12, or 24 months after diagnosis or start of first reintervention; OS-12, -24: overall survival 12 or 24 months after diagnosis or start of first reintervention. either at first diagnosis or at disease recurrence (n 48; overall survival, 66.5 weeks; 95% CI, weeks) compared with the total patient population, including patients with reinterventions and standard patients (n 168; overall survival, 46 weeks; 95% CI, weeks; P 0.05). Patients with late progression (using mttp as the cutoff) were analyzed separately to determine the impact of above-average mttp on overall survival. Fifty-one patients with reinterventions and 36 patients with standard therapy had late progression. A late first progression predicted a good overall outcome, and mos was superior for the entire group of these patients (56 weeks vs. 46 weeks), as well as for the subgroups of patients with reinterventions (73 weeks vs weeks) and those receiving standard therapy (39 weeks vs. 26 weeks). More than one chemotherapeutic reintervention did not further improve the outcome in the reintervention subgroup. Patients who received more than 1 chemotherapeutic reintervention after their first disease recurrence had an overall survival of 31 weeks, compared with an mos of 33 weeks for all patients with reinterventions (Fig. 1B). TABLE 4 Effect of Prognostic Factors on Overall Survival Characteristic a mos-1 P value (Cox) Karnofsky at first incidence Karnofsky at first incidence Karnofsky at first incidence ( 70 vs. 70) Karnofsky at disease recurrence ( 70 vs. 70) Complete resection at first-line therapy 53 Incomplete resection at first-line therapy 40 Resection at first-line therapy (complete vs. incomplete) Age 55 yrs 48.5 Age 55 yrs 47 Age ( 55 yrs vs. 55 yrs) mos-1: median overall survival. a A Karnofsky performance score 70 is not favorable at incidence of glioblastoma but is highly favorable at first disease recurrence. A macroscopic complete resection at first surgery and age 55 years are not favorable compared with a partial resection or biopsy and age 55 years. Prognostic Factors and Correlations All standard prognostic factors were analyzed for their impact on survival. Age, macroscopic complete resection, and KPS 70 at diagnosis were not associated with a significantly better outcome, although there was a near-significant trend for KPS. At disease recurrence, KPS 70 significantly predicted better outcome (P ; Table 4). A longer time to first and second progression was correlated significantly with better overall survival (Table 5), as supported by the

7 2684 CANCER December 15, 2003 / Volume 98 / Number 12 TABLE 5 Correlations of Important Prognostic Factors with Median Time to First Progression, Median Time to Second Progression, and Median Overall Survival a Characteristic Karnofsky at first incidence > 70 (P value) Complete resection at first-line therapy (P value) Age < 55 yrs at diagnosis (P value) mos-1 (P value) Karnofsky 70 n.r ( 0.05) 0.11 ( 0.05) Complete resection n.r. n.r ( 0.05) Age 55 yrs 0.27 ( 0.05) n.r ( 0.05) mttp ( 0.05) 0.04 ( 0.05) 0.08 ( 0.05) 0.80 ( 0.05) mttp-2 n.r. n.r. n.r ( 0.05) n.r.: not reported; mttp-1: median time to first progression; mttp-2: median time to second progression; mos-1: median overall survival. a Only the correlations of higher KPS, younger age, and longer interval to TTP-1 and TTP-2 with longer mos-1 are statistically significant. Negative values indicate negative correlations. analysis of the subgroup of patients with late progression. Quality of Life and Performance during Therapy Overall, the observed changes were not statistically significant according to the EORTC guidelines. For the patients who underwent quality of life evaluation, KPS did not significantly deteriorate during reinterventions, according to a t test (P 0.182). Therefore, no deterioration in either quality of life or KPS occurred during the observation period, generally intended to be the time from first treatment to the fourth cycle of the first reintervention chemotherapy. It is to be emphasized that these data can only be applied to a subset of patients treated within clinical trials, because patients outside clinical trials were not subject to systematic quality of life assessments. DISCUSSION The impact of salvage therapy on patients with recurrent glioblastoma is a matter of debate. We hypothesized that it might be possible to select patients who are candidates for an aggressive reintervention strategy at first disease recurrence or progression and that the overall survival of these patients may be improved with sustained quality of life. We performed a singleinstitution, nonrandomized observational trial that stratified patients according to the reintervention strategy received. Our reintervention strategy included neurosurgical resections, as second resection reduces tumor mass, whereas improvements in survival are not universally accepted. 20 Only a limited number of patients underwent a second course of irradiation. The majority of patients, however, received one or more chemotherapeutic regimens after first disease recurrence. Patients were selected according to standard criteria as follows: clear indication of disease recurrence on an MRI scan with gadolinium, 19 expected substantial benefit of neurosurgery (macroscopic total resection, symptomatic relief, or impact on additional therapy), KPS 70, biologic age 55 years, and expected survival 3 months if chemotherapy was planned, as well as informed consent for all interventions. Altogether, 168 consecutive patients were analyzed. We found that the subset of patients (n 90) selected for reinterventions after first disease recurrence lived significantly longer overall compared with the total cohort and the standard group. They attained an mos of 61.5 weeks. The PFS (21%) and proportional overall survival rates (71%) at 12 months also were superior in the reintervention group. Because the current study was not prospective or randomized, a selection bias in favor of the reintervention strategy has to be expected. The time to first progression was slightly longer in the reintervention group, possibly lacking significance due to relatively small patient numbers. However, median TTP after disease recurrence and overall survival differed significantly, with a clear advantage for the patients receiving reinterventions. We conclude that the more favorable overall outcome in the reintervention group reflects the impact of the reintervention strategy rather than the slight advantage of the reintervention group with regard to primary progression. The matched-pair analysis, which corrected for major inequalities in demographic and previous treatment data, showed an almost identical time to first disease recurrence/progression and confirmed the better overall survival of patients in the reintervention group. The most likely explanation for this advantage is the efficacy of reintervention therapy beginning at first disease recurrence. In the analysis of all patients and matched pairs, this effect appears to be based on an advantage generated within the first 6 months after disease re-

8 Salvage Therapy in Glioblastoma/Hau et al currence (Fig. 1A, C), indicating the importance of early reinterventions after disease recurrence. Selecting patients for a reintervention strategy clearly introduces a bias that is likely to affect survival. Therefore, it is not surprising that the data for our standard group are worse than the outcome data for unselected cohorts from the literature. 2 The mos of our total population of patients, however, was slightly greater than what could be expected from the literature. 2 Therefore, our total cohort can be claimed to be unselected. The outcome of our reintervention group is very similar to the outcome observed in a recently published clinical trial using PCV in patients with recurrent glioblastoma, with mttp and overall survival after disease recurrence being 13 and 33 weeks, respectively. 28 In that study, age 40 years and KPS 90 were associated with longer TTP and survival. The matched-pair analysis indicated that age, initial KPS, and extent of resection were not important prognostic factors for subgroups of patients with or without reinterventions. The reintervention and standard groups differed slightly, albeit not significantly, with respect to KPS at the time of disease recurrence. However, the correlation between a KPS 70 at disease recurrence and an improved overall survival was highly significant. These results indicate that patients with a KPS 80 at disease recurrence are most likely to receive a benefit from reinterventions. Therefore, KPS at disease recurrence should be a major factor in the decision on whether to propose a further reintervention. These results are confirmed by the analysis of Wong et al., 11 who identified KPS at disease recurrence, but not age, as a prognostic factor. Clearly, a definitive answer as to whether patients with a KPS 70 would benefit from reinterventions can be obtained only after a prospective randomized, controlled trial that stratifies patients at the time of disease recurrence by KPS. In further subset analyses, longer times to first and second progression were determined to be prognostic factors associated with a favorable outcome. We also found that patients included in clinical studies had a significantly better outcome compared with other patients, which is consistent with published literature and recent results for temozolomide. 29,30 In a metaanalysis of patients in Phase II trials, 11 patients receiving salvage treatment consisting of more than 2 surgical procedures and/or 2 chemotherapy regimens before study entry fared worse than did patients receiving fewer interventions (PFS rate at 6 months after disease recurrence, 10% vs. 23%). Patients receiving 0, 1, or 2 chemotherapy regimens before their respective Phase II trials had a similar PFS rate at 6 months (15%), regardless of whether they had 1 or 2 previous surgical procedures. In the current study, patients who received one reintervention fared better than did patients who did not receive reintervention, but any further reinterventions did not significantly prolong overall survival, although there was a trend toward doing so. Because many patients with recurrent disease did not undergo debulking surgery before chemotherapy, an independent effect of reresection on survival in patients with high-grade glioma is not evident, but it cannot be excluded. Therefore, re-resection and chemotherapy in patients with recurrent disease were analyzed separately. Adverse treatment side effects, increased hospitalization time, and psychosocial stress are factors that may outweigh survival data. Therefore, quality of life measurements were a major issue in the current analysis for showing that reinterventions can be performed without compromising quality of life. More than 60% of patients treated within clinical studies responded to questionnaires. Quality of life analyses including 60% of patients commonly are considered to be sufficient. 25 Our data demonstrate that reinterventions are associated with sustained performance and quality of life in the subset of patients treated within clinical studies. This finding has a major impact, because all patients with glioblastoma receive palliative treatment, and a treatment-related decline in quality of life would not be acceptable without major treatment effects on overall survival. We conclude that reinterventions are feasible and advisable in selected patients with recurrent glioblastoma. In the current series of 168 consecutive patients, more than 50% were selected for a reintervention, which had a major impact on overall survival. Assuming that patients considered for second-line treatment are referred to tertiary care centers, the true proportion of patients eligible for reintervention may be slightly smaller. A long time to first progression is a favorable prognostic factor, as is a KPS 70 at disease recurrence. Based on our experience and that of others, patients should be included in clinical studies if it is feasible to do so. A second resection, radiotherapy in selected patients, and additional chemotherapeutic regimens should be considered for reintervention. This approach appears to be associated with a better overall survival without a negative impact on quality of life. Further chemotherapeutic reinterventions after the first attempt, i.e., at the time of the second or later disease recurrence, did not add any obvious benefit to overall survival. Although the retrospective nature of the current study and the limited number of patients preclude definite conclusions, reinterventions have

9 2686 CANCER December 15, 2003 / Volume 98 / Number 12 been valuable in selected patients in our clinical experience and in that of others. Decisions on secondary or tertiary reinterventions in patients with glioblastoma are made on an individual basis, taking into account the patients KPS, age, medical history, motivation for therapy, and support by family and friends. The success of current chemotherapeutic regimens is limited, but new therapies may soon be available that more specifically target the biology of malignant glioma. If these emerging options are rigorously studied in patients with malignant glioma, patients may look forward to longer survival and better quality of life. REFERENCES 1. Kleihues P, Sobin LH. World Health Organization classification of tumors. Cancer. 2000;88: Davis FG, Freels S, Grutsch J, et al. Survival rates in patients with primary malignant brain tumors stratified by patient age and tumor histological type: An analysis based on Surveillance, Epidemiology, and End Results (SEER) data, J Neurosurg. 1998;88: Burger PC, Vogel FS, Green SB, et al. Glioblastoma multiforme and anaplastic astrocytoma. Pathologic criteria and prognostic implications. Cancer. 1985;56: Wood JR, Green SB, Shapiro WR. The prognostic importance of tumor size in malignant gliomas: a computed tomographic scan study by the Brain Tumor Cooperative Group. J Clin Oncol. 1988;6: Should SC, Burger PC, Mendeksohn DB, et al. Primary tumors of the brain and spinal cord. Boston: Butterworth- Heinemann, 1997: Burton EC, Prados MD. Malignant gliomas. Curr Treat Options Oncol. 2000;1: DeAngelis LM, Burger PC, Green SB, et al. Malignant glioma: who benefits from adjuvant chemotherapy? Ann Neurol. 1998;44: Stewart LA, for the Glioma Trial Metaanalysis Group. Chemotherapy in adult high-grade glioma: a systematic review and meta-analysis of individual patient data from 12 randomised trials. Lancet. 2002;359: Prados MD, Levin V. Biology and treatment of malignant glioma. Semin Oncol. 2000;27(Suppl 6): Brandes AA, Fiorentino MV. The role of chemotherapy in recurrent malignant gliomas: an overview. Cancer Invest. 1996;14: Wong ET, Hess KR, Gleason MJ, et al. Outcomes and prognostic factors in recurrent glioma patients enrolled onto Phase II clinical trials. J Clin Oncol. 1999;17: Weller M, Muller B, Koch R, Bamberg M, Krauseneck P. Neuro-Oncology Working Group of the German Cancer Society. Neuro-Oncology Working Group 01 trial of nimustine plus teniposide versus nimustine plus cytarabine chemotherapy in addition to involved-field radiotherapy in the first-line treatment of malignant glioma. J Clin Oncol. 2003; 21: Medical Research Council. Randomized trial of procarbazine, lomustine, and vincristine in the adjuvant treatment of high-grade astrocytoma: a Medical Research Council trial. J Clin Oncol. 2001;19: Boiardi A. PCV chemotherapy for recurrent glioblastoma multiforme. Neurology. 2001;56: Brada M, Hoang-Xuan K, Rampling R, et al. Multicenter Phase II trial of temozolomide in patients with glioblastoma multiforme at first relapse. Ann Oncol. 2001;12: Yung WK, Albright RE, Olson J, et al. A Phase II study of temozolomide vs. procarbazine in patients with glioblastoma multiforme at first relapse. Br J Cancer. 2000;83: Fabel K, Dietrich J, Hau P, et al. Long-term stabilization in patients with malignant glioma after treatment with liposomal doxorubicin. Cancer. 2001;92: Fabel-Schulte K, Hau P, Dietrich J, et al. 13-cis-retinoic acid as a potential treatment in glioma patients who show stable disease after combined modality treatment [abstract]. Proc Am Soc Clin Oncol. 2000;36:A Macdonald DR, Cascino TL, Schold SC Jr. et al. Response criteria for Phase II studies of supratentorial malignant glioma. J Clin Oncol. 1990;8: Sawaya R. Extent of resection in malignant gliomas: a critical summary. J Neurooncol. 1999;42: Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53: Lee ET. Statistical methods for survival data analysis. Belmont, CA: Lifetime Learning, Mantel N. Ranking procedures for arbitrarily restricted observations. Biometrics. 1967;23: Zar JH. Biostatistical analysis. Englewood Cliffs, NJ: Prentice Hall, Aaronson NK, Ahmedzai S, Bergman B, et al. The European Organization for Research and Treatment of Cancer QLQ- C30: a quality-of-life instrument for use in international clinical trials in oncology. J Natl Cancer Inst. 1993;85: Young T, de Haes JC, Curran D, et al. Guidelines for assessing quality of life in EORTC clinical trials. Brussels: EORTC, Karnofsky DA, Abelmann WH, Craver LF, et al. The use of nitrogen mustard in the palliative treatment of carcinoma: with particular reference to bronchial carcinoma. Cancer. 1948;1: Kappelle AC, Postma TJ, Taphoorn MJ, et al. PCV chemotherapy for recurrent glioblastoma multiforme. Neurology. 2001;56: Winger MJ, Macdonald DR, Schold SC Jr, Cairncross JG. Selection bias in clinical trials of anaplastic glioma. Ann Neurol. 1989;26: Stupp R, Dietrich PY, Ostermann Kraljevic S, et al. Promising survival for patients with newly diagnosed glioblastoma multiforme treated with concomitant radiation plus temozolomide followed by adjuvant temozolomide. J Clin Oncol. 2002;20: