Clinical Neurology and Neurosurgery 128 (2015) 60–69

Contents lists available at ScienceDirect

Clinical Neurology and Neurosurgery journal homepage:

Treatment results and outcome in elderly patients with glioblastoma multiforme – A retrospective single institution analysis Markus Hoffermann a,∗ , Lukas Bruckmann a , Kariem Mahdy Ali a , Martin Asslaber b , Franz Payer c , Gord von Campe a a

Department of Neurosurgery, Medical University of Graz, Graz, Austria Institute of Pathology, Medical University of Graz, Graz, Austria c Department of Neurology, Medical University of Graz, Graz, Austria b

a r t i c l e

i n f o

Article history: Received 21 July 2014 Received in revised form 14 October 2014 Accepted 9 November 2014 Available online 15 November 2014 Keywords: Elderly Extent of resection Glioblastoma Malignant Primary Brain Tumors Temozolomide

a b s t r a c t Objective: Although glioblastoma multiforme is more common in patients older than 65 years, the elderly population is often excluded from clinical studies. Decision making in this subgroup can be challenging due to the lack of evidence for different neurosurgical and adjuvant treatment strategies. Methods: In this retrospective study, we evaluated clinical, treatment and survival data of 124 consecutive patients over 65 years of age with supratentorial glioblastoma multiforme. Results: Median OS was 6.0 months (std. error 0.783, 95% CI 4.456–7.535). Mean OS was 9.7 months (std. error 0.830, 95% CI 8.073–11.327). In univariate regression analysis, low KPS was of negative prognostic value (p < 0.006 for KPS ≤ 80), while greater advanced age did not have any impact on survival (p = 0.591 for differences between groups). Gross total resection and subtotal resection led to significantly improved overall survival (median 15.0 and 11.0 months; p < 0.02) compared to partial resection or biopsy (both 4.0 months), but complications were more common in subtotal and partial resections. The last observation did not reach statistical significance (p = 0.06). Combinations of irradiation and Temozolomide chemotherapy proved to be more effective than other adjuvant therapies. Extent of resection (gross total resection vs. all others) and form of adjuvant treatment were the only factors of independent prognostic value in multivariate analysis (p = 0.031 and p < 0.001, respectively). Conclusions: It appears that more aggressive treatment regimens can lead to longer overall survival in elderly glioblastoma multiforme patients. Gross total resection should be offered whenever safely possible; otherwise, biopsy may be preferred. Non-surgical treatment should consist of postoperative radiotherapy and concomitant and/or adjuvant chemotherapy. Possibly higher rates of hematological side effects in concomitant chemotherapy need to be further investigated. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Although primary brain tumors are rare [52], increased population aging leads to an increase in tumor frequency among the elderly [21,83], especially regarding glioblastoma multiforme (GBM) [32]. With the introduction of the “Stupp-protocol”, the median survival of glioblastoma multiforme (GBM) patients was prolonged to 14.6 months [72]. While this study excluded patients over 70 years, other studies confirmed that surgery with

∗ Corresponding author at: Department of Neurosurgery, Medical University of Graz, Auenbruggerplatz 29, 8036 Graz, Austria. Tel.: +43 31638583933; fax: +43 316138513368. E-mail address: [email protected] (M. Hoffermann). 0303-8467/© 2014 Elsevier B.V. All rights reserved.

adjuvant radiochemotherapy is also efficient in elderly patient cohorts [7,21,49,55]. Most clinical trials on GBM therapy do not include patients >65 or 70 years, as age itself is considered to be an independent negative prognostic factor [1,3,15,19,24,29,34,39,46,47, 50,60,64,69,70,72,75,89,91,92]. It is not until 2007 that KeimeGuibert et al. have shown that radiotherapy is more effective than supportive care in GBM patients over 70 years [33]. Although recent multicenter randomized controlled trials (RCTs) like NOA08 and NORDIC have helped to improve decision-making in elderly glioma patients, they mainly focused on the impact of different radiochemotherapy regimens [46,85]. The role of surgery, especially extent of resection (EOR), still remains a matter of discussion. In younger trial population groups, there is strong evidence that an EOR around 80–98% is independently prognostic [39]. Some

M. Hoffermann et al. / Clinical Neurology and Neurosurgery 128 (2015) 60–69 Table 1 Major presenting symptoms.

Hemiparesis Mental status change Seizure Speech impairment Cephalea Vertigo Gait disturbance Visual impairment Incidental finding



37 22 17 14 13 8 7 4 2

29.8 17.7 13.7 11.3 10.5 6.5 5.6 3.2 1.6

retrospective analyses – none of these including patients treated later than 2008 – support the role of cytoreductive surgery in elderly patients [3,12,19,29,47,64]. We present a comprehensive data analysis of patients older than 65 years with GBM treated at the University Hospital Graz from 2005 to 2012 and provide an up-to-date overview regarding the influence of surgical and non-surgical treatments on overall survival (OS). 2. Materials and methods The medical charts of 124 patients aged ≥65 years with glioblastoma multiforme treated between 2005 and 2012 were systematically reviewed. Patients were identified using a prospective database. Patients aged ≥65 years with diagnosis of glioma but without histological verification were excluded as well as patients with diagnoses of more favorable prognosis, i.e. Astrocytoma WHO II and III, Oligoastrocytoma or Gliosarcoma. The Follow-up protocol used in clinical routine consists of immediate post-operative MRI (within 48 h), MRI control studies after adjuvant radiochemotherapy or no later than three months postoperatively followed by MRI studies every three months or in any case of clinical deterioration. Clinical data like Karnofsky Performance Status (KPS), age, complications related to surgery and side effects of adjuvant treatment were collected along with treatment data in categorical form. Except for 19 patients, survival data was either directly available or updated from the Austrian Death Records. Tumor locations and major presenting symptoms were categorized as shown in Tables 1 and 2. For cox regression analysis, tumor locations were grouped in frontal vs. all other locations and in central (basal ganglia, corpus callosum, multilocular) vs. all other locations. OS was correlated with patient age as a linear variable and as a categorical variable after stratification into 4 age groups (65–70, 70–75, 75–80, >80 years). EOR was categorized in gross total resection (GTR, removal of complete or at least 95% of the contrast enhancing tumor portion), subtotal resection (STR, removal of 80–95%), partial resection (PR, removal of ≤80%) and stereotactic Biopsy (B). EOR was assessed according to the Macdonald criteria on early postoperative MRI (within 48 h after surgery) [44]. For cox regression analysis, EOR was also grouped in GTR vs. all other locations. Neurosurgical complications were categorized in “permanent neurological deficit” (focal neurological deficits significantly interfering with activities of daily living), “lethal neurological complication” (brain infarction, secondary bleeding or brain edema leading to death during post-op hospitalization, i.e. within a period of 1 week postoperatively), “transient medical complication” (deep vein thrombosis, pulmonary artery embolism, pneumonia or myocardial infarction), “lethal medical complication” (same as transient medical complication, but leading to death during hospitalization) and “infection, impaired wound healing, secondary bleeding or cerebrospinal fluid fistula”. Categories of


(radio-)chemotherapy side effects were: “no severe side effects”, “fatigue”, “thrombopenia” and “infection”. The side effects were graded according to CTCAE v3.0. Patients were stratified into 6 groups of adjuvant treatment: (1) “none”: Patients who received no further treatment other than surgery. (2) “radiochemotherapy only”: At least 4 weeks of combined radiochemotherapy with 75 mg Temozolomide per square meter of body-surface area (mg/m2 ) per day, 7 days per week from the first to the last day of radiotherapy (fractionated, 40–60 Gy). (3) “Stupp-protocol”: fractionated radiotherapy (60 Gy) plus continuous daily 75 mg/m2 Temozolomide, followed by six (at least 3 actually administered) cycles of adjuvant Temozolomide (150 mg/m2 for 5 days during each 28-day cycle) [72]. (4) “radiotherapy only”: Postoperative irradiation (fractionated focal irradiation, 40–60 Gy). (5) “chemotherapy only”: at least 3 cycles of adjuvant Temozolomide chemotherapy (150 mg/m2 for 5 days during each 28-day cycle) without irradiation. (6) “radiotherapy followed by chemotherapy”: postoperative irradiation followed by at least 3 cycles of adjuvant Temozolomide chemotherapy (150 mg/m2 for 5 days during each 28-day cycle). Information about date of death was not available in 19 cases. Therefore, we corrected their survival data by presuming that patients with severely reduced overall performance at the last follow-up visit (KPS ≤40) would most likely die within one month. This correction did not influence the statistically significant results. Progression-free survival is not presented, as inconsistencies regarding follow-up intervals seemed prevalent in this patient cohort. That data would have been prone to misinterpretation. Statistical analysis was performed using SPSS® (IBM Corporation) Version 20.0. Categorical data for KPS and EOR were correlated with categorical data for surgical complications using Pearson Chi Square cross tabulations. We used Kaplan–Meier plots and tested for equality of OS distribution for Age, tumor location, KPS, EOR and form of adjuvant treatment categories using Log Rank (Mantel–Cox), Breslow (Generalized Wilcoxon) and Tarone-Ware tests. Age as a linear parameter was correlated with OS using the Pearson product-moment correlation coefficient while using one way ANOVA and cox regression when defining age as a categorical parameter. We performed univariate cox regression analyses and pairwise comparisons providing differences in OS and Hazard Ratios (HR) for all categorical factors. For multivariate regression analysis, we used a cox regression model and the forward: conditional method. In all applied tests, a p-value of less than 0.05 was used as level of significance. The study protocol was approved by the local ethics committee of the Medical University of Graz (ID: 25-2812 ex 12/13).

3. Results Data from 124 patients were analyzed. 51 women and 73 men with a mean age of 71.0 years (min. 65, max. 84) were included. The major presenting symptoms, i.e. the reasons why imaging was performed, are outlined in Table 1. At the time of analysis, 93 patients had died, 14 were in progressive, 9 in stable disease status and 8 patients were lost to follow-up. Median OS was 6.0 months (std. error 0.783, 95% CI 4.456–7.535). Mean OS was 9.7 months (std. error 0.830, 95% CI 8.073–11.327).

3.1. Location Tumor locations were distributed as shown in Table 2. Survival analysis, one-way ANOVA and univariate cox-regression did not depict significant differences in OS between the location groups.


M. Hoffermann et al. / Clinical Neurology and Neurosurgery 128 (2015) 60–69

Table 2 Tumor locations and OS in months.

Basal ganglia Corpus callosum Multilocular Left frontal Left rolandic Left parieto-occipital Left temporal Right frontal Right rolandic Right parieto-occipital Right temporal


Mean OS in months

Std. error

8 8 6 5 10 16 14 13 12 16 16

4.63 4.22 5.67 11.80 9.30 12.88 12.13 7.85 12.55 9.11 11.80

0.96 1.25 3.93 5.99 2.90 3.69 1.76 1.82 2.84 2.04 2.33

All frontal tumors as one group vs. all other locations did not show significantly longer survival. In comparison of all centrally located (basal ganglia, corpus callosum) or multifocal tumors (n = 22) with lobar located tumors, patients with centrally located tumors had significantly shorter OS (p = 0.003, Log Rank). Lobar location was predictive of longer OS in univariate cox regression (HR 0.489, p = 0.005). 3.2. KPS Minimum KPS at time of presentation was 40, maximum 100, mean 70–80. A higher number of permanent neurological deficits in low KPS groups (≤70) was statistically significant in Pearson Chi-Square (p = 0.017). The influence of KPS at time of presentation on survival is shown in Fig. 1. The OS difference between groups reached statistical significance (p < 0.001). In univariate cox regression with KPS 100 as reference, HR were 1.3 for KPS90, 2.4 for KPS80, 1.8 for KPS70, 2.4 for KPS60, 2.5 for KPS50 and 7.9 for KPS40 (p ≤ 0.006 for all below KPS80). 3.3. Age Age at time of diagnosis did not show significant linear correlation to OS (Pearson Correlation 0.033, p = 0.733). Patients were stratified into 4 age groups: Median OS was 6.0 months in patients aged 65–70 (std. error 1.7, n = 48), 7.0 for 70–75 years (std. error 2.1, n = 44), 5.0 for 75–80 (std. error 1.5, n = 25) and 9.0 for over 80 years (std. error 2.6, n = 7) (see Fig. 2). In univariate cox regression, no significant impact on HR through the age groups was evident. 3.4. Surgery The achieved extent of resection was as follows: GTR: n = 35, STR: n = 28, PR: n = 34, B: n = 27). Surgical mortality was 4.0% (n = 5); all of these patients had undergone either STR (n = 3) or PR (n = 2). Morbidity was 19.3% (12.1% permanent neurological deficits, 4.0% infections and wound healing disturbances, 3.2% transient medical complications). Complications were most frequent in STR (42.9%), followed by GTR (28.6%), PR (14.7%) and B (7.4%). The differences between groups did not reach statistical significance (p = 0.06 for crosstabulation with complication categories, Pearson Chi square). The GTR group showed a mean OS of 15.0 months (std. error 1.9, 95% CI 11.4–18.7), followed by STR with 11.0 (std. error 1.6, 95% CI 7.9–14.2), PR with 6.4 (std. error 1.2, 95% CI 4.1–8.8), B with 5.6 (std. error 1.1, 95% CI 3.4–7.8). Median OS was 15.0 months in GTR (std. error 3.6, 95% CI 7.9–22.1), 9.0 in STR (std. error 3.3, 95% CI 2.5–15.5), 4.0 in PR (std. error 1.1, 95% CI 1.9–6.1), 4.0 in B (std. error 0.8, 95% CI 2.7–5.3) (see Fig. 3).

95% CI 2.74–6.51 1.76–6.68 0–13.37 0.59–23.54 3.61–14.99 5.64–20.11 8.68–15.58 4.28–11.41 6.98–18.12 5.12–13.10 7.23–16.37

Median OS in months

Std. error

4.0 3.0 1.0 3.0 5.0 5.0 13.0 6.0 7.0 6.0 9.0

2.9 0.57 0.39 1.10 4.74 0.99 1.73 1.35 6.30 0.99 2.76

95% CI 0–9.54 1.88–4.12 0.25–1.75 0.85–5.15 0–14.3 3.06–6.95 9.61–16.39 3.36–8.64 0–19.35 4.06–7.95 3.58–14.42

Pairwise comparisons revealed significantly longer OS in GTR or STR compared to PR and B (all p < 0.05) and no significant difference between GTR and STR (p = 0.11 in Log-Rank). In univariate cox regression with B as reference, GTR and STR proved to significantly decrease the risk for early death (HR 0.344 and 0.514, respectively; p < 0.02), while PR procedures did not (HR 0.878, p = 0.63). Patients undergoing one (n = 6) and two (n = 2) additional resections in case of tumor recurrence showed a mean survival of 17.3 and 14.0 months, respectively. 3.5. Adjuvant therapies Therapy according to the “Stupp-protocol” was administered in 39, only radiochemotherapy in 21, only radiotherapy in 7, only chemotherapy in 6, radiotherapy with adjuvant chemotherapy in 6, and no adjuvant therapy in 45 patients. Temozolomide was the only administered alkylating agent. The total dose per patient was heterogeneous. The maximum number of completed cycles was 17, minimum 3 and mean 8.57 (std. deviation 3.4). Severe side effects of chemotherapy leading to discontinuation of Temozolomide were present in 19 cases. Severe fatigue was present in 8 patients (CTCAE Grade 3 in 2, Grade 4 in 6 patients), 7 patients developed significant thrombopenia (CTCAE Grade 3 in 4 patients, Grade 4 in 3 patients) and 4 patients suffered from infection (all CTCAE Grade 3). All patients with thrombopenia had received concomitant radiochemotherapy, i.e. either “Stupp-protocol” or radiochemotherapy only. No chemotherapy related mortality was observed. Mean OS was 3.0 months (std. error 0.4, 95% CI 2.2–3.8) without therapy, 3.4 for radiotherapy (std. error 0.4, 95% CI 2.7–4.2), 6.5 for radiochemotherapy (std. error 1.1, 95% CI 4.4–8.6), 10.7 for chemotherapy (std. error 2.3, 95% CI 6.2–15.1), 15.8 for radiotherapy followed by adjuvant chemotherapy (std. error 1.6, 95% CI 12.7–19.0) and 19.4 for “Stupp-protocol” (std. error 1.5, 95% CI 16.5–22.4). Median OS was 2.0 months without adjuvant treatment (std. error 0.4, 95% CI 1.3–2.7), 4.0 for radiotherapy (std. error 0.0), 5.0 for radiochemotherapy (std. error 0.7, 95% CI 3.7–6.3), 8.0 for chemotherapy (std. error 3.7, 95% CI 0.8–15.2), 15.0 for radiotherapy with adjuvant chemotherapy (std. error 1.8, 95% CI 11.4–18.6) and 18.0 for “Stupp-protocol” (std. error 1.6, 95% CI 14.8–21.2) (see Fig. 4). For significant differences in pairwise comparison analysis, see Table 3. 3.6. Multivariate analysis A multivariate cox regression model including age, KPS, EOR (GTR vs. all others), form of adjuvant therapy, complications of surgery and/or adjuvant therapy and tumor location (central vs.

M. Hoffermann et al. / Clinical Neurology and Neurosurgery 128 (2015) 60–69


Fig. 1. Kaplan–Meier plot of KPS and cumulative overall survival in months.

lobar) revealed form of adjuvant therapy and EOR (GTR vs. all other groups) as independent risk factors. HR for the “Stupp-protocol” group was lowest with 0.054 (95% CI 0.03–0.11, p < 0.001), followed by radiotherapy with adjuvant chemotherapy with 0.103 (95% CI 0.04–0.27, p < 0.001). GTR vs. all other surgical procedures was independently prognostic with a HR of 0.576 (95% CI 0.35–0.95, p = 0.031).

4. Discussion The impact of patient characteristics and different treatment forms on OS and adverse events in 124 elderly patients suffering from GBM was assessed. The primary objective of this retrospective analysis was to reflect the influence of extent of resection within current adjuvant treatment standards.

Fig. 2. Kaplan–Meier plot of age groups and cumulative overall survival in months. Differences between age groups did not reach statistical significance in one-way ANOVA (p = 0.654).


M. Hoffermann et al. / Clinical Neurology and Neurosurgery 128 (2015) 60–69

Fig. 3. Kaplan–Meier plot of EOR categories and cumulative overall survival in months.

Fig. 4. Kaplan–Meier plot of adjuvant treatment forms and cumulative overall survival in months.

Mean OS was 9.7 months and median OS 6.0 months. These results are consistent with previous studies in elderly glioma patients and confirm the general assumption of poorer prognosis when compared to younger study populations, e.g. in the EORTC/NCIC study, or data from the glioma outcomes project [11,13,54,72]. To find possible correlations between age and survival time, we performed regression analysis and also stratified into four age categories. Neither method depicted significant correlations between

age and survival, although there was a trend toward poorer survival in the 75–80 years subgroup. These interesting results are consistent with a comparable analysis from Chaichana et al. and support the hypothesis that age itself is not a reliable predictor for outcome and treatment response in elderly glioma patients [11]. Oszvald et al. were able to show in a retrospective study of 361 GBM patients that age is only prognostic in patients undergoing biopsy but not in those undergoing resection [54]. On the other hand, our results are in contradiction with retrospective studies from Barker et al.

M. Hoffermann et al. / Clinical Neurology and Neurosurgery 128 (2015) 60–69


Table 3 Pairwise comparison analysis depicting the influence of different adjuvant treatment forms on overall survival in months. None

None Radiotherapy only Chemotherapy Radiotherapy followed by Chemotherapy Radiochemotherapy only “STUPP” regimen



0.05 12.4 21.4 9.3 82.7


Treatment results and outcome in elderly patients with glioblastoma multiforme--a retrospective single institution analysis.

Although glioblastoma multiforme is more common in patients older than 65 years, the elderly population is often excluded from clinical studies. Decis...
1MB Sizes 0 Downloads 5 Views