Is age an additional factor in the treatment of elderly patients with glioblastoma? A new stratification model: an Italian Multicenter Study

View More View Less
  • 1 Neurosurgery Unit, Department of Neurosciences, Santa Maria della Misericordia University Hospital, Udine;
  • 2 Division of Neurosurgery, Department of Neurosciences, Reproductive and Odontostomatological Sciences, Università degli Studi di Napoli Federico II, Naples;
  • 3 Department of Medical and Surgical Sciences and Advanced Technologies (G.F. Ingrassia); Neurological Surgery, Policlinico “G. Rodolico - San Marco” University Hospital, University of Catania;
  • 4 Interdisciplinary Research Center on Brain Tumors Diagnosis and Treatment, University of Catania;
  • 5 Division of Neurosurgery, BIOMORF Department, University of Messina;
  • 6 Department of Neuroscience “Rita Levi Montalcini,” Neurosurgery Unit, University of Turin;
  • 7 Department of Neurosurgery “Giampaolo Cantore”–IRCSS Neuromed, Pozzilli;
  • 8 Institute of Neurosurgery, Fondazione Policlinico Gemelli, Catholic University, Rome;
  • 9 Department of Human Neurosciences–“Sapienza” University of Rome;
  • 10 Division of Neurosurgery, Department of Surgical Specialties, Radiological Sciences and Public Health, University of Brescia;
  • 11 Department of Medicine, University of Udine;
  • 12 Department of Neurosurgery, Alessandro Manzoni Hospital, Lecco; and
  • 13 Department of Neurosurgery, Mater Olbia Hospital, Olbia, Italy
Free access

OBJECTIVE

Approximately half of glioblastoma (GBM) cases develop in geriatric patients, and this trend is destined to increase with the aging of the population. The optimal strategy for management of GBM in elderly patients remains controversial. The aim of this study was to assess the role of surgery in the elderly (≥ 65 years old) based on clinical, molecular, and imaging data routinely available in neurosurgical departments and to assess a prognostic survival score that could be helpful in stratifying the prognosis for elderly GBM patients.

METHODS

Clinical, radiological, surgical, and molecular data were retrospectively analyzed in 322 patients with GBM from 9 neurosurgical centers. Univariate and multivariate analyses were performed to identify predictors of survival. A random forest approach (classification and regression tree [CART] analysis) was utilized to create the prognostic survival score.

RESULTS

Survival analysis showed that overall survival (OS) was influenced by age as a continuous variable (p = 0.018), MGMT (p = 0.012), extent of resection (EOR; p = 0.002), and preoperative tumor growth pattern (evaluated with the preoperative T1/T2 MRI index; p = 0.002). CART analysis was used to create the prognostic survival score, forming six different survival groups on the basis of tumor volumetric, surgical, and molecular features. Terminal nodes with similar hazard ratios were grouped together to form a final diagram composed of five classes with different OSs (p < 0.0001). EOR was the most robust influencing factor in the algorithm hierarchy, while age appeared at the third node of the CART algorithm. The ability of the prognostic survival score to predict death was determined by a Harrell’s c-index of 0.75 (95% CI 0.76–0.81).

CONCLUSIONS

The CART algorithm provided a promising, thorough, and new clinical prognostic survival score for elderly surgical patients with GBM. The prognostic survival score can be useful to stratify survival risk in elderly GBM patients with different surgical, radiological, and molecular profiles, thus assisting physicians in daily clinical management. The preliminary model, however, requires validation with future prospective investigations. Practical recommendations for clinicians/surgeons would strengthen the quality of the study; e.g., surgery can be considered as a first therapeutic option in the workflow of elderly patients with GBM, especially when the preoperative estimated EOR is greater than 80%.

ABBREVIATIONS CART = classification and regression tree; CCI = Charlson Comorbidity Index; EGBM = elderly GBM; EOR = extent of resection; GBM = glioblastoma; HR = hazard ratio; KPS = Karnofsky Performance Scale; OS = overall survival; PFS = progression-free survival; RHR = relative HR.

OBJECTIVE

Approximately half of glioblastoma (GBM) cases develop in geriatric patients, and this trend is destined to increase with the aging of the population. The optimal strategy for management of GBM in elderly patients remains controversial. The aim of this study was to assess the role of surgery in the elderly (≥ 65 years old) based on clinical, molecular, and imaging data routinely available in neurosurgical departments and to assess a prognostic survival score that could be helpful in stratifying the prognosis for elderly GBM patients.

METHODS

Clinical, radiological, surgical, and molecular data were retrospectively analyzed in 322 patients with GBM from 9 neurosurgical centers. Univariate and multivariate analyses were performed to identify predictors of survival. A random forest approach (classification and regression tree [CART] analysis) was utilized to create the prognostic survival score.

RESULTS

Survival analysis showed that overall survival (OS) was influenced by age as a continuous variable (p = 0.018), MGMT (p = 0.012), extent of resection (EOR; p = 0.002), and preoperative tumor growth pattern (evaluated with the preoperative T1/T2 MRI index; p = 0.002). CART analysis was used to create the prognostic survival score, forming six different survival groups on the basis of tumor volumetric, surgical, and molecular features. Terminal nodes with similar hazard ratios were grouped together to form a final diagram composed of five classes with different OSs (p < 0.0001). EOR was the most robust influencing factor in the algorithm hierarchy, while age appeared at the third node of the CART algorithm. The ability of the prognostic survival score to predict death was determined by a Harrell’s c-index of 0.75 (95% CI 0.76–0.81).

CONCLUSIONS

The CART algorithm provided a promising, thorough, and new clinical prognostic survival score for elderly surgical patients with GBM. The prognostic survival score can be useful to stratify survival risk in elderly GBM patients with different surgical, radiological, and molecular profiles, thus assisting physicians in daily clinical management. The preliminary model, however, requires validation with future prospective investigations. Practical recommendations for clinicians/surgeons would strengthen the quality of the study; e.g., surgery can be considered as a first therapeutic option in the workflow of elderly patients with GBM, especially when the preoperative estimated EOR is greater than 80%.

ABBREVIATIONS CART = classification and regression tree; CCI = Charlson Comorbidity Index; EGBM = elderly GBM; EOR = extent of resection; GBM = glioblastoma; HR = hazard ratio; KPS = Karnofsky Performance Scale; OS = overall survival; PFS = progression-free survival; RHR = relative HR.

The prognosis of glioblastoma (GBM) is universally poor, especially in elderly patients, in whom the median survival ranges from 4 to 9 months.1–9 Approximately half of GBM cases occur in geriatric patients, and this trend is destined to increase with the aging of the population.

The optimal strategy for management of GBM in elderly patients (EGBM) remains controversial, especially in regard to the effects of extent of resection (EOR) on survival outcomes.10–19 There is overwhelming evidence to suggest that survival and neurological function4 outcomes can be optimized through maximal safe resection in younger patients with GBM. However, many neurosurgeons tend to avoid aggressive surgical interventions in EGBM patients because of the probable increased risk of perioperative complications.1,17–19 Life expectancy, overall health status, and quality of life in the elderly, however, are all increasing globally, which makes a strong case for redefining the concept of “elderly” and reframing it in the context of GBM surgical management.

Considering that the incidence of GBM is higher within this expanding age group of the older population, it is of utmost importance to identify prognostic factors and effective therapeutic strategies for improving survival and quality of life.17,18 An increasing number of prognostic survival tools are being developed to combine clinical, radiological, and molecular variables in an all-inclusive risk stratification model.20–22 Given the importance of each individual factor, it is often difficult to establish how these interact with each other and how they impact prognosis in the complexity of clinical settings. Cox survival analysis generally detects risk factors without highlighting how their interactions or various combinations influence the prognosis. The algorithms and computational statistics have already demonstrated an excellent performance in outcome predictions for a wide range of conditions, thus paving the way for a personalized medicine model.23

In light of this evidence, a multiparametric model for prognosis was elaborated, inclusive of radiological, molecular, and surgical variables, to assess prognosis in postoperative EGBM cases prior to postoperative treatment.

Methods

The methods of this study were based on a previous study in which a scoring system for patients of all ages with GBM was elaborated.21

Study Population and Inclusion Criteria

A shared cooperative retrospective database of 322 adult patients surgically treated for newly diagnosed GBM between January 2015 and December 2018 was created. There is no generally agreed upon criterion for the definition of “older people.”17 To provide results that can be widely applied across countries, we used an age cutoff of 65 years old for defining older patients in the current research.

Patients were enrolled according to the following criteria: 1) age ≥ 65 years; 2) no previous surgery; 3) no preoperative chemo- or radiotherapy; 4) presurgical evaluation using the Charlson Comorbidity Index (CCI);24 5) objective evaluation of preoperative tumor volume on MR images in DICOM format based on postcontrast T1- and T2-weighted MRI sequences; 6) objective estimation of EOR on postcontrast T1-weighted MRI sequences; 7) revision of histopathological specimens using the new 2016 WHO classification of tumors of the CNS; and 8) MGMT promoter methylation and IDH1/IDH2 mutation status assessment. Exclusion criteria included needle biopsy, incomplete imaging data, follow-up interval, and multifocal tumors.

Volumetric Analysis

The neuroradiological tumor growth pattern, expressed by the preoperative T1/T2 MRI index, and EOR were computed as previously described. Briefly, the achieved EOR in each case was objectively evaluated using pre- and postoperative MR images (DICOM format), based on the contrast area of postcontrast T1-weighted MRI sequences, using the following formula: (preoperative tumor volume − postoperative tumor volume)/preoperative tumor volume. With the aim of evaluating the role of tumor growth pattern on overall survival (OS), the preoperative MRI index was assessed as follows: T1/T2 = preoperative volumetric tumor volume on postcontrast T1-weighted images/preoperative volumetric tumor volume on T2-weighted images.21,25

Statistical Analysis

Categorical variables were reported as percentages and continuous variables were reported as means ± standard deviations or medians and ranges as appropriate, according to the data distribution. Normality of the continuous variables was tested using the Shapiro-Wilk test.

OS and progression-free survival (PFS) were estimated using the Kaplan-Meier approach. The association between variables and survival distribution was tested using univariate and multivariate Cox proportional hazard models (after verification of proportional hazard assumptions). Patients with unknown survival were censored as of their last scan date. The variables we considered for univariate analysis were age, sex, Karnofsky Performance Scale (KPS) score, preoperative tumor volume computed on postcontrast T1- and T2-weighted MR images, tumor location, tumor side, EOR, postoperative adjuvant protocol used, IDH1/2 mutation, MGMT methylation status, and Ki-67. EOR was modeled as both a continuous and an ordinal variable (≤ 79%, 80%–89%, 90%–99%, and 100%) in univariate analysis to ensure consistency with previous studies that focused on the impact of glioma resection in terms of volumes.

In the univariate Cox regression, the preoperative T1/T2 MRI index was initially analyzed as a continuous variable. To better understand the variable’s association pattern, the Cox regression model was then applied to the quintiles for this variable. Subsequently, the variable was dichotomized using a cutoff we identified at the quintile that showed a significant hazard ratio (HR). The variables that were significantly associated in the univariate model (p < 0.05) were included in the multivariate regression model, according to the stepwise-backward selection method. All statistical analyses were performed by Stata/IC (version 13.0, StataCorp LP).

Classification and Regression Tree Method

To determine subgroup patients with different clinical prognoses, we used the decision tree model with the classification and regression tree (CART) method.21,26 This method is a machine learning model composed of hierarchical decision rules involving optimal cutoff values that recursively split independent factors into different groups. The groups of individuals are called nodes and form a branch node tree. Terminal nodes are groups of individuals that cannot be further subdivided on the basis of the established parameters (minimum size of subgroup, minimum number of events, and maximum p value required) to proceed in further subdivisions. In our study, nodes were required to have a minimum size of 20 patients, a minimum of 10 events, and a maximum p value of 0.05. The significant variables in the univariate analysis were considered to generate the model. Once the regression tree was generated, the nodes of the terminal branches were pruned (aggregated) on the basis of their relative HRs (RHRs) to obtain final groups with homogeneous mortality risk. The final groups were converted into a prognostic survival score ordered according to their RHRs.

Differences in terms of OS probability among the score categories were investigated using univariate Cox regression analysis. The performance of the prognostic survival score in predicting time to death was estimated using Harrell’s c-index.27 All statistical analyses were performed by Stata/IC (version 13.0, StataCorp LP).

Results

Survival Risk Factors

Table 1 lists the various features of the EGBM patients included in the study. The 1- and 2-year OS and PFS rates for the cohort were 42.07% and 14.89% (OS rates) and 24.8% and 9.64% (PFS rates), respectively. Univariate analysis indicated significantly improved OS in EGBM cases with the following features: young age (p = 0.035), high EOR (p < 0.0001), methylation of the MGMT promoter (p = 0.002), presence of low residual tumor (p < 0.0001), no corpus callosum involvement (0.040), and low preoperative T1/T2 MRI index (p < 0.0001). In the final model, variables with significant univariate analysis p values were included. Age (p = 0.018), tumor involvement of the corpus callosum (p = 0.023), preoperative T1/T2 MRI index (p = 0.002), EOR (p = 0.002), and MGMT methylation status (p = 0.012) were found to be independent survival risk factors (Table 2).

TABLE 1.

Clinical, radiological, molecular, surgical, and follow-up characteristics of the study population

ParameterValue
No. of pts322
Mean age ± SD, yrs72.28 ± 4.86
Sex, n (%)
 Female137 (42.55)
 Male185 (57.45)
Side, n (%)
 Lt170 (52.79)
 Rt152 (47.21)
Tumor site, n (%)
 Precentral112 (34.78)
 Postcentral100 (31.06)
 Temporal + insular110 (34.16)
Clinical presentation, n (%)
 No deficits14 (4.35)
 Nonspecific symptoms (headache, nausea, vomiting, disorientation)108 (33.54)
 Motor deficits76 (23.6)
 Sensory deficits7 (2.17)
 Visual/speech deficits64 (19.88)
 Seizures53 (16.46)
Median preop KPS score (range)90 (60–100)
Preop, n (%)
 Diabetes32 (9.94)
 Solid tumor25 (7.77)
 Chronic pulmonary disease24 (7.45)
 Peptic ulcer disease15 (4.66)
 Previous myocardial infarction12 (3.73)
 Chronic kidney disease9 (2.79)
 Cerebrovascular disease8 (2.48)
 Mild liver disease6 (1.86)
 Lymphoma/leukemia5 (1.55)
 Connective tissue disease3 (0.93)
 Metastatic tumor in recent history2 (0.62)
Preop CCI, n (%)*
 0150 (50.17)
 182 (27.42)
 252 (17.39)
 ≥315 (5.01)
Radiological features
 Ependymal involvement, yes vs no114 (35.4%) vs 208 (64.6%)
 Corpus callosum involvement, yes vs no96 (29.81%) vs 226 (70.19%)
 Necrotic-cystic component, yes vs no248 (77.02%) vs 74 (22.98%)
 Midline shift, yes vs no151 (46.89%) vs 171 (53.11%)
 Median preop tumor volume on postcontrast T1-weighted images (range), cm331.45 (0.39–197.7)
 Median preop tumor volume on T2-weighted images (range), cm356.5 (0.52–231)
Preop T1/T2 MRI index, n (%)
 <0.73185 (57.45)
 ≥0.73137 (42.55)
Median residual tumor (range), cm31.344 (0–191)
Median EOR (range), continuous variable95 (35–100)
EOR, n (%), categorical variable
 100%125 (38.82)
 90%–99%92 (28.57)
 80%–89%43 (13.35)
 ≤79%62 (19.25)
Biological features
MGMT met (yes vs no)155 (54.96%) vs 127 (45.04%)
IDH1/2 mutation (yes vs no)11 (3.62%) vs 293 (96.38%)
 Median Ki-67 % (range)25 (3–90)
Two-gene model, n (%)§
MGMT met & IDH1/2 mutation6 (2.16)
MGMT met & IDH1/2 wt149 (53.60)
MGMT unmet & IDH1/2 mutation5 (1.80)
MGMT unmet & IDH1/2 wt118 (42.45)
Median hospitalization (range), days8 (5–14)
Postop course, n (%)
 No deficits152 (47.2)
 Nonspecific postop symptoms (headache, nausea, vomiting, disorientation)44 (13.66)
 Motor deficits73 (22.67)
 Sensory deficits3 (0.93)
 Visual/speech deficits46 (14.29)
 Seizures4 (1.24)
Postop protocol, n (%)
 Stupp protocol250 (77.64)
 CT or RT alone50 (15.53)
 No adjuvant treatment22 (6.83)
6-month follow-up, n (%)
 No deficits176 (73.64)
 Motor deficits35 (14.64)
 Sensory deficits8 (3.35)
 Visual/speech deficits18 (7.53)
 Seizures2 (0.84)
Median preop KPS score (range)90 (60–100)
OS (alive vs dead)77 (23.91%) vs 245 (76.09%)
OS at 1-yr follow-up42.07%
OS at 2-yr follow-up14.89%
PFS (no recurrence vs recurrence)52 (16.15%) vs 270 (83.85%)
PFS at 1-year follow-up24.8%
PFS at 2-year follow-up9.64%
OS assessed by age
 65–69 yrs44.20%
 70–74 yrs46.16%
 ≥75 yrs28.57%
OS assessed by EOR
 100%55.03%
 90%–99%44.96%
 80%–89%35.32%
 ≤79%16.44%

CT = chemotherapy; met = methylation; pts = patients; RT = radiation therapy; wt = wild-type.

Cohort = 299 patients.

Cohort = 282 patients.

Cohort = 304 patients.

Cohort = 278 patients.

Cohort = 239 patients.

TABLE 2.

Predictors of OS in univariate and multivariate analyses

Univariate AnalysisMultivariate Analysis
VariableHR95% CIp ValueHR95% CIp Value
Age, yrs1.0281.001–1.0550.0351.0401.006–1.0750.018
Sex
 Male1
 Female1.1610.901–1.4970.247
Side
 Lt1
 Rt1.1690.908–1.5060.224
Tumor site
 Precentral1
 Retrocentral0.9870.723–1.3480.938
 Temporal + insular1.0260.758–1.3890.866
Preop CCI (0 vs ≥1)0.8620.664–1.1190.265
Preop KPS score1.1120.905–1.1100.574
Radiological features
 Ependymal involvement (yes vs no)1.1320.900–1.4250.286
 Corpus callosum involvement (yes vs no)1.3080.998–1.7130.0401.4491.051–1.9980.023
 Necrotic-cystic component (yes vs no)0.9800.727–1.3220.897
 Midline shift (yes vs no)1.0570.822–1.3600.664
 Preop tumor volume on postcontrast T1-weighted images, cm31.0000.998–1.0020.500
 Preop tumor volume on T2-weighted images, cm30.9940.992–0.997<0.00010.9970.993–1.0000.123
 Preop T1/T2 MRI index5.4083.144–9.301<0.00013.2061.537–6.6850.002
 Residual tumor, cm31.0161.009–1.022<0.00011.0180.997–1.0400.081
EOR (continuous variable)0.9830.978–0.987<0.00010.9850.977–0.9940.002
EOR (categorical variable)
 100%1
 90%–99%1.2960.944–1.7990.108
 80%–89%1.7741.190–2.6440.005
 ≤79%3.4242.412–4.861<0.0001
Biological features
MGMT methylation (yes vs no)0.6450.490–0.8490.0020.6780.500–0.9190.012
IDH1/2 mutation (yes vs no)0.4760.223–1.0140.0550.6580.287–1.5120.325
 Ki-671.0020.995–1.0100.456
Two-gene model
MGMT met & IDH1/2 mut1
MGMT met & IDH1/2 wt1.7400.639–4.7360.278
MGMT unmet & IDH1/2 mut1.1830.264–5.2980.826
MGMT unmet & IDH1/2 wt2.7881.020–7.6180.045

Boldface type represents statistically significant results (p < 0.05).

CART Model

The CART analysis was applied to elaborate a promising, thorough, and new clinical prognostic survival score for EGBM patients based on surgical, neuroradiological, and molecular determinants. Specifically, the model generation is based on 3 phases of analysis as reported in our previous study:21 1) the Kaplan-Meier approach was used to identify the most important survival factors; 2) a decision tree algorithm was applied to stratify OS in different prognosis groups; and 3) the prognostic survival score was computed. In detail, the CART model derives from independent predictor factors detected by the univariate analysis (age, preoperative tumor T1/T2 MRI index, tumor involvement of corpus callosum on preoperative MRI, EOR, MGMT methylation status, residual tumor evidenced on postcontrast T1-weighted MR images, and IDH1/2 mutation status).

First, the CART analysis was performed on 250 cases that met all the selection criteria, leading to the definition of 6 terminal nodes. Terminal nodes with similar RHRs were grouped together to form a final diagram composed of 5 classes with different OSs (p < 0.0001), which were used to create the prognostic survival score. Specifically, patients belonging to scores 1, 2, 3, 4, and 5 had RHR values of ≤ 0.40, 0.57–0.73, 1.37, 1.87, and > 2.60, respectively (Fig. 1A). The score performance in predicting death was defined by a Harrell’s c-index of 0.75 (95% CI 0.76–0.81). Subsequently, to investigate the impact of IDH1/2 mutation on OS, the CART model was applied to IDH1/2 wild-type EGBM patients (239 cases). A score from 1 to 4 was obtained from the 4 terminal nodes (Harrell’s c-index of 0.74, 95% CI 0.69–0.78; Fig. 1B). The 1-year estimated OS was computed for each score category (Tables 3 and 4). Overall, to facilitate the visualization of the survival analysis stratified by the score groups resulting from the CART models, Kaplan-Meier curves were generated (Fig. 2).

FIG. 1.
FIG. 1.

Random forest, CART algorithm. The CART algorithm provides a graphic visualization of the interaction between risk factors detected by Cox survival analysis. In each hierarchical node, the study population is split according to the presence (green) or not (red) of the variable able to influence prognosis. A: CART model performed on 250 EGBM cases that met all the selection criteria. A score from 1 to 5 was assigned to the 6 terminal nodes thus defined based on the RHR. Patients belonging to the subgroup with RHR = 0.73 and to the subgroup with RHR = 0.55 were joined to create a single group, labeled “score 2.” This was done considering the small sample size of each subgroup and similarities in RHR values. B: CART model applied to IDH1/2 wild-type EGBM patients (239 cases). A score from 1 to 4 was obtained from the 4 terminal nodes.

TABLE 3.

Estimated OS at 12, 18, and 24 months in EGBM patients (n = 250) according to the CART score

Estimated OS (%)
Score*VariableHR95% CIp Value12 mos18 mos24 mos
1EOR >80%, preop T1/T2 MRI index <0.73, residual tumor <1 cm3, MGMT met176.7356.3044.45
2EOR >80%, preop T1/T2 MRI index <0.73, residual tumor <1 cm3, MGMT unmet1.6260.985–2.6850.05869.0048.5315.25
EOR >80%, preop T1/T2 MRI index <0.73, residual tumor >1 cm3, age <70 yrs
3EOR >80%, preop T1/T2 MRI index <0.73, residual tumor >1 cm3, age >70 yrs3.2901.875–5.7750.00040.7412.076.04
4EOR >80%, preop T1/T2 MRI index >0.734.7432.829–7.9540.00015.803.95
5EOR <80%7.5294.712–12.0310.00015.213.04

Boldface type represents statistically significant results (p < 0.05).

A survival score from 1 to 5 was defined based on CART analysis.

TABLE 4.

Estimated OS at 12, 18, and 24 months in EGBM patients with IDH1/2 wild-type (n = 239) according to the CART score

Estimated OS (%)
Score*VariableHR95% CIp Value12 mos18 mos24 mos
1EOR >82%, preop T1/T2 MRI index <0.73, residual tumor <1 cm3172.9053.3832.52
2EOR >82%, preop T1/T2 MRI index <0.73, residual tumor >1 cm31.9141.216–3.0130.00549.7222.9011.45
3EOR >82%, preop T1/T2 MRI index >0.733.6222.300–5.703<0.000116.254.06
4EOR <82%5.7723.875–8.596<0.000115.21

Boldface type represents statistically significant results (p < 0.05).

A survival score from 1 to 4 was defined based on CART analysis.

FIG. 2.
FIG. 2.

OS stratified by CART analysis prognostic score. Kaplan-Meier curves display the OS of EGBM patients according to the prognostic scores elaborated by the CART model. A: Survival stratified by the score from 250 EGBM cases that met the selection criteria. B: Survival stratified by the score from 239 EGBM IDH1/2 wild-type cases.

Discussion

In this retrospective investigation based on 322 elderly cases with newly diagnosed GBM, OS was analyzed based on the stratification of clinical, radiological, and molecular variables. The key findings for consideration were as follows: 1) age, volumetric tumor MRI pattern (expressed by the preoperative T1/T2 MRI index), EOR, and MGMT methylation status were confirmed as independent survival predictors on multivariate Cox regression analysis; 2) a novel prognostic score for EGBM surgical patients was assessed by CART analysis; and 3) surgery can be considered as a first therapeutic option in the workflow of EGBM patients, especially when the preoperative estimated EOR is greater than 80%.

GBM in the Elderly

In the last decade, the role of surgery has been shown to be the primary option in GBM management, especially in light of recent literature demonstrating a survival benefit associated with a greater EOR.8,21,25,28–36 The incidence of intracranial tumors in elderly individuals is increasing due to aging of the population and increasing life expectancy.37

In 2017, according to Eurostat, the European Union population over the age of 65 years has an additional 20 years of life expectancy, and the percentage of people older than 80 years is expected to more than double in the coming decades, thus unavoidably changing the shape of the age pyramid. These age-associated demographic trends are already having a significant impact with regard to modifications in the disease epidemiology and management of neurosurgical care. Unfortunately, EGBM patients tend to have a drastically reduced survival compared with their younger counterparts.1,38,39 This could partially be explained by unfavorable tumor biology, performance status, comorbidities, treatment toxicity, trend toward less aggressive treatment, etc. Taking into account the poor prognosis and progressive increase in the incidence of GBM in the elderly population, investigation of treatment efficiency is of significant interest in the management of these patients10,11,13–19 (Table 5, Fig. 3).

TABLE 5.

Review of the literature in elderly patients with GBM

Authors & YearNo. of PtsAge (yrs)ResultsConclusions
Han et al., 202014
 Metanalysis including retrospective observational studies10,815>60GTR associated w/ a significant improvement in OS compared w/ STR (HR 0.70, 95% CI 0.64–0.77); elderly pts who underwent GTR showed lower risk of 3-mo (OR 0.47, 95% CI 0.24–0.93), 6-mo (OR 0.38, 95% CI 0.26–0.56), 9-mo (OR 0.35, 95% CI 0.25–0.49), & 1-yr (OR 0.40, 95% CI 0.29–0.56) mortalityGTR seems to be more effective than STR in achieving longer survival in elderly pts w/ high-grade glioma
Lombardi et al., 201916
 Retrospective cohort study113>65According to CGA, 35% of pts were classified as fit, 30% as vulnerable, & 35% as frail; median OS was 16.5 mos (95% CI 14.6–18.2 mos), 12.1 mos (95% CI 8.1–16.1 mos), & 10.3 mos (95% CI 8.8–11.8 mos) for fit, vulnerable, & frail pts, respectively (p = 0.1); median PFS was 11.2 mos (95% CI 6.07–16.4 mos), 7.7 mos (95% CI 4.6–10.7 mos), & 7.1 mos (95% CI 5.7–8.4 mos) for fit, vulnerable, & frail pts, respectively (p = 0.2)CGA held prognostic significance in elderly pts w/ GBM; it is likely that CGA fit pts are those who would benefit from combined treatment w/ RT-CT
Cohen-Inbar, 201919
 Cohort study20,705>75A stepwise decrease in GTR attained as a function of age (36% at 18–44 yrs vs 24% at 75 yrs, p < 0.001)EGBM pts who undergo a GTR rather than biopsy alone have improved PFS & OS; a greater EOR appears to correlate w/ an incremental OS benefit in the elderly population, similar to younger pts
 Prospective study30>65GTR carries reduced risks for mortality & a modest benefit in terms of functional independence & OS (171 vs 85 days, p = 0.035)
 Retrospective study80>65Negative prognostic factors for EGBM pts undergoing GTR, including a preop KPS score <80, COPD, presenting motor/ language/cognitive deficit, & tumor largest diameter >4 cm
Minniti et al., 201917
 Systematic review & meta-analysis12,607>60OS was 5.71 mos (95% CI 5.04–6.36 mos) in biopsy, 8.68 mos (95% CI 7.87–9.48 mos) in STR, & 14.04 mos (95% CI 12.8–15.2 mos) in GTRMaximal degrees of tumor removal when the operative option is indicated, regardless of age, preventing new permanent neurological deficits & maintaining good quality of life
 RCT30>65Median survival times of 171 & 85 days after resection or biopsy, respectively (p = 0.035)
Asmaa et al., 201810
 Cohort study20,705>75GTR decreased in a stepwise manner as a function of pt age (from 36% [18–44 yrs] to 24% [>75 yrs], p < 0.001); GTR had a 2- to 3-fold increase in OSA continuous linear increase in survival after combination of different modalities w/ best outcomes observed in pts undergoing aggressive resection followed by adjuvant CRT combined w/ TMZ; molecular stratification & analysis of MGMT methylation may help to identify pts who may particularly benefit from CT
 Retrospective cohort study274>6521.9% complications after resection, w/ a rate of neurological complications of 7.7%
 Retrospective cohort study124>65KPS score <80 was of negative prognostic value (p < 0.006), STR or GTR was associated w/ significantly improved OS (median 11.0 & 15.0 mos, p < 0.02) compared w/ partial resection or biopsy (both 4.0 mos)
 RCT30>65Pts undergoing resection had significantly longer survival time (171 vs 85 days, p = 0.03) compared w/ those having biopsy
 Systematic review & meta-analysis12,607>60Increase in PFS & OS among pts who had undergone either STR or GTR compared w/ biopsy; no change in mortality & morbidity & improvements in KPS score in pts undergoing GTR compared w/ STR
Okada et al., 201718
 RCT30>65Surgical removal of the tumor prolonged survival by 2.8 times more than biopsy (median OS 171 days after the craniotomy vs 85 days after the biopsy)Surgery is aimed at achieving maximal cytoreduction of the tumor; resection as much as possible is associated w/ favorable prognosis even in elderly pts w/ GBM; in addition to histological diagnosis, information regarding molecular markers, such as IDH mutation & MGMT methylation status, is necessary to develop a treatment strategy
 Systematic review & meta-analysis12,607>60Resection was superior to biopsy in OS (mean difference 3.88 mos, 95% CI 2.14–5.62 mos, p < 0.001), PFS, postoperative KPS score, & mortality; GTR was significantly superior to STR in terms of OS (mean difference 3.77 mos, 95% CI 2.26–5.29 mos, p < 0.001), PFS, & postop KPS score
 Retrospective cohort study206>70Survival advantage of GTR compared w/ the biopsy (OS 10.7 vs 2.8 mos)
 Retrospective cohort study124>65Survival advantage of GTR compared w/ the biopsy (OS 15 vs 5.6 mos)
 Retrospective cohort study237>65Survival advantage of GTR compared w/ STR (OS 17.7 vs 16.1 mos)
 Retrospective cohort study120>65Survival advantage of GTR compared w/ STR (OS 14.1 vs 9.6 mos)
Braun & Ahluwalia, 201711
 Retrospective study80>65Median OS of 5.7 mos in the surgery group vs 4.0 mos in the biopsy groupA maximal safe resection is recommended, after stratifying for periop risk factors including performance status & medical comorbidities
 Retrospective study146>65Improved median OS of 17.7 mos in GTR compared w/ 4.0 mos in the biopsy group
 Cohort study20,705>75GTR in the group of pts ≥75 yrs was associated w/ an OR of 0.5 compared w/ pts 18–44 yrs old; GTR had a longer OS than those who underwent STR
Halani et al., 201713
 Retrospective study146>65Improved median OS of 17.7 mos in GTR compared w/ 4.0 mos in the biopsy groupTreatment options available that not only improve OS, but also do not necessarily compromise quality of life; aim for GTR of tumor in elderly pts, regardless of age; if GTR is not possible, STR results in improved OS compared w/ those pts who undergo operative biopsy alone (median OS: GTR 15 mos vs STR 9 mos vs biopsy 4 mos); pts >65 yrs have an added benefit from combined TMZ & RT after GTR (median OS: no adjuvant treatment 2.0 mos vs RT alone 4.0 mos vs CMT 5.0 mos vs CT alone 8.0 mos vs Stupp protocol 18.0 mos)
 Retrospective cohort study120>65Survival advantage of GTR compared w/ STR (OS 14.1 vs 9.6 mos)
 Retrospective cohort study124>65Survival advantage of GTR compared w/ the biopsy (OS 15 vs 4 mos)
 Retrospective cohort study237>65Survival advantage of GTR compared w/ STR (OS 17.7 vs 16.1 mos)
 Retrospective cohort study120>65GTR had longer survival compared w/ STR (14.1 vs 9.6 mos, p = 0.038); pts >75 yrs had worse survival than younger (median OS: 7.9 vs 15.1 mos, p < 0.0001); for EOR, age >75 yrs (HR 1.06, 95% CI 1.02–1.10) was associated w/ worse outcomes, & greater KPS scores (HR 0.97, 95% CI 0.95–0.99) were associated w/ better prognosis
 Retrospective study58>80Median OS was 4.2 mos; factors associated w/ greater OS included KPS score >90 (p < 0.05)
Jordan et al., 201615
 Retrospective study80>65Median OS of 5.7 mos in the surgery group vs 4.0 mos in the biopsy groupWith acceptable perioperative risk stratification according to overall health & performance status, greater EOR may allow for prolonged survival as well as larger samples of tissue for genetic & molecular markers; involved-field RT is also an effective modality of therapy for elderly pts; TMZ given either alone or adjuvant to RT is associated w/ improved survival in elderly pts
 Retrospective cohort study168>60GTR + RT & re-resection of recurrent tumor each provided independent prognostic benefits in older group
 RCT30>65Pts undergoing resection had significantly longer survival time (171 vs 85 days, p = 0.03) compared w/ those having biopsy
 Retrospective cohort study124>65Survival advantage of GTR compared w/ the biopsy (OS 15 vs 4 mos)
 Retrospective study103>65PFS & OS improved w/ increasing EOR; RT & TMZ after surgery were also associated w/ improved PFS & OS

CGA = comprehensive geriatric assessment; CMT = combined modality radiochemotherapy; COPD = chronic obstructive pulmonary disease; CRT = chemo- and radiotherapy; GTR = gross-total resection; RCT = randomized controlled trial; STR = subtotal resection; TMZ = temozolomide.

Articles are grouped according to the final results that our review research included (Fig. 3). The 9 review studies reported are based on a total of 34 studies.

FIG. 3.
FIG. 3.

Systematic review and meta-analysis flowchart. The figure shows the different decisional phases regarding the inclusion and exclusion of the papers reviewed in the studies based on the role of surgery in EGBM patients.

Survival Analysis in EGBM

This retrospective investigation supports the widely known role of age, tumor preoperative MRI index, EOR, and MGMT methylation status as independent predictors of survival. Our results confirmed the poorer prognosis for EGBM patients with increasing age.1,6,40 When stratifying the survival results according to age intervals, we found that 1-year OSs in subgroups of patients who were 65–69, 70–74, 75–79, and ≥ 80 years old were 42.20%, 46.16%, 28.57%, and 4.26%, respectively.

Although several investigations have found that extensive resection is associated with longer survival in EGBM patients, aggressive surgery remains a controversial issue, mainly due to concerns over the balance between treatment benefits and side effects based on age, comorbidity conditions, and supposed different tumor biologies.10,11,13,14,16,17,19 With regard to the role of EOR, several retrospective investigations showed that greater EOR appears to correlate with an incremental OS benefit in the elderly population, similar to younger patients.10,11,13,14,16,17,19

Despite previous investigations recognizing EOR as an independent survival predictor, volumetric data were analyzed as qualitative and not quantitative variables. In this study, we reported the quantitative data and identified a threshold value capable of discriminating the survival benefit. The volumetric analysis showed 1-year survival rates of 55.03%, 44.96%, 35.32%, and 16.44%, when the EOR was 100%, 90%–99%, 80%–89%, and ≤ 79%, respectively. In addition, infiltration of the corpus callosum caused a worse prognosis (p = 0.023) as an indirect measure of the possibility of obtaining radical resection in consideration of the vast tumor infiltration.

Our study also noted the role of MGMT even among EGBM patients (p = 0.002), which was consistent with other investigations.16,17 Concerning the radiological data, our results confirmed the prognostic survival value of the preoperative T1/T2 MRI index, clarifying its role in predicting a more aggressive biological behavior in those cases with a value close to 1 (p = 0.002). In view of the wide heterogeneity of GBM, combining next-generation sequence analysis and assessments of MRI texture analysis parameters could further clarify the role of this volumetric index.21,41

Interestingly, unlike other studies, no correlation between KPS score (p = 0.254) and CCI score (p = 0.574) and OS was found.1,37 This could be explained by the fact that in the present investigation, patients treated surgically had high KPS scores with chronic yet stabilized comorbidities, while those with relevant morbidities were excluded in the preoperative anesthesiological evaluation. Patients with a preoperative KPS score < 60 or chronic uncontrolled diseases underwent only needle biopsy and, consequently, were excluded from this investigation.

Flanigan et al.1 reported that patients with a CCI score ≥ 1 were associated with decreased survival (p = 0.018). The difference obtained in identifying CCI score as a survival prognostic factor in EGBM patients could be due to the presence or absence of chronic comorbidities in addition to the specific therapeutic controls and treatments. This suggests the need for more adequate scales to assess the severity of comorbidities, in addition to other chronic pathologies, in patients with GBM.

Overall, these results emphasize the importance of a multidisciplinary approach for a careful evaluation of surgical options in EGBM patients.

CART Analysis

The principal novelty of this study is the elaboration of a prognostic and integrated score based on the CART analysis. This approach identifies the strongest prognostic variables and generates a creative and integrative visualization chart that illustrates the results in an easily interpretable tree structure from a large number of data. The model is based on the influence of interactions of various parameters and not on single ones, which permits the creation of a model that closely reflects the disease complexity. In this investigation, the CART analysis based on the 250 cases provided 6 terminal nodes, the RHRs of which were used to generate the prognostic score with the purpose of facilitating the survival stratification before patients were discharged postoperatively.

Patients belonging to the subgroups with RHR = 0.73 and RHR = 0.55 were joined to create a single group labeled “score 2.” This was done because of the small sample size of each subgroup and similarities in RHR values. Patients with scores of 1 or 2 had better survival, with 1-year estimated OSs of 76.73% and 69.00%, respectively. The worst survival was for patients with scores from 3 to 5, with 1-year estimated OS after surgery ranging between 40.74% and 15.21% after surgery.

It is important to highlight that the EOR was the most robust influencing factor in the algorithm hierarchy, while age appeared at the third node of the CART algorithm, thus strengthening the role of surgery also in EGBM and performing patients (who have a high preoperative KPS score), when surgical planning allowed us to preoperatively estimate an EOR > 80%.6,17,42 In a previous investigation, Flanigan et al. elaborated a risk prognostic score in EGBM patients based on variables identified using the multivariate stepwise analysis (age, EOR, preoperative weakness, tumor size, and CCI).1 A point designation was then given to each factor and points were totaled for each patient, considering only the presence of the variables and ruling out their interactions. Our model thus provides for a rapid and accurate assessment of survival prognosis after surgery and relies on concrete parameters rather than on a subjective metric.

It is widely reported in the literature that the IDH1/2 mutation is associated with a better prognosis in GBM patients in terms of disease-free survival and progression. IDH1/2 mutant GBMs represent less than 10% of all GBMs and show different genetic, epigenetic, and clinical features compared with the 1/2 wild-type counterpart.43 In this investigation, the IDH1/2 mutation was detected in only 3.4% of cases (11 patients), with a similar distribution within each score class identified by CART analysis (3, 1, 2, 3, and 2 IDH1/2 EGBM patients for scores 1, 2, 3, 4, and 5, respectively). The equal portioning of the IDH1/2 mutation may determine that its positive survival impact is equally distributed within the different score groups identified by CART analysis.

Analyzing only IDH1/2 wild-type EGBM patients, we excluded age and MGMT methylation status from the model, while the EOR, residual tumor, and preoperative neuroradiological tumor growth pattern, expressed by the preoperative T1/T2 MRI index, were confirmed as the fundamental nodes in the prognostic model. Generally, IDH-mutant gliomas are younger, and in these patients GBM is more likely to derive from a low-grade glioma and therefore also carries the IDH mutation.44 This investigation highlights that EOR is the main prognostic factor in EGBM patients, giving rise to the first split node in both generated CART models (Fig. 1). This may suggest that the OS is influenced by other numerous mutations, such as mutations in ATRX, CIC, EGFR, FUBP1, NOTCH1, PTEN, H3F3A, IDH1/2, PIK3CA, and BRAF, and amplifications in EGFR or MDM2; copy number alterations of chromosomes 1p, 7, 10, and 19q are involved in glioma genesis and tumor progression.41,45

Limitations and Future Directions

A limitation of our study is that we included only patients with resectable GBM based on clinical and radiological criteria (high preoperative KPS score, controlled comorbidities, and/or high chances of achieving a large EOR). Patients who underwent needle biopsy, who did not undergo resection, were thus excluded. An additional limitation is that comorbidities were not adequately discussed (heart disease, cancer, anticoagulation, etc.), because only patients with controlled mild or moderate comorbidities were considered for surgery. It is thus important to have stringent clinical selection criteria in EGBM patients to select which may benefit from surgery so that underlying comorbidities do not have a direct impact on surgical outcomes.

Despite the inherent limitations of the retrospective nature of this study, the prognostic score elaborated for EGBMs could be useful in a day-to-day clinical environment. It could also prove to be useful after surgery and previous oncological treatments, to discuss prognosis and draw future prospective clinical trials. Patients with better OS showed a better PFS and lower score.21,45 The prognostic survival score assessed in this investigation can thus be considered an indirect measure of tumor progression.

An additional limitation is represented by the heterogeneous treatment at tumor recurrence. Each patient underwent individualized management at tumor progression. It is well known that to improve the prediction models, salvage treatment information should be updated in the analysis at the time of tumor progression. Moreover, the study lacks details about functional recovery time and neurocognitive outcomes according to variable levels of resection. To overcome this drawback, future prospective multicenter studies based on larger cohorts with longer follow-up periods need to include time-dependent analysis.

Future studies are needed to further assess the numerous molecular and genomic markers that may prove to be of clinical interest in GBM. In addition, radiological features should be included in prospective future clinical studies considering the growing importance of radiogenomics. Considering the heterogeneity of GBM, texture features from multiparametric MRI and next-generation sequence analysis could prove to be of assistance in managing these patients. Computed prognostic scores may prove to be useful in a day-to-day clinical setting and in research to provide more thorough assessments in future prospective clinical trials. In conclusion, the survival score could be useful when deciding and discussing prognosis to better address the entire management of EGBM patients.

Conclusions

Elderly patients with GBM typically carry a poor prognosis. There are no gold standards or widespread guidelines to be applied to this group of patients, and optimal strategy and management remain debatable. Advanced age alone should not necessarily preclude optimal resection followed by adjuvant radiation and chemotherapy. Our study showed that prediction models can be used to generate a promising, thorough, and new clinical prognostic score for EGBM surgical patients to guide clinicians in the decision-making process. Thorough evaluation and selection of EGBM patients may lead to favorable survival benefit.

Acknowledgments

We would like to thank Mark Zeppieri, MD, PhD, for the scientific and English revision of the manuscript.

Disclosures

The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

Author Contributions

Conception and design: Ius, Sabatino. Acquisition of data: Ius, Somma, Altieri, Angileri, Certo, Cofano, Della Pepa, La Rocca, Panciani, Pignotti, Spena, Sabatino. Analysis and interpretation of data: Ius, Pignotti, Sabatino, Somma. Drafting the article: Ius, Somma, Pignotti, Sabatino. Critically revising the article: all authors. Reviewed submitted version of manuscript: Ius, Somma, Altieri, Angileri, Barbagallo, Cappabianca, Certo, Cofano, D’Elia, Della Pepa, Fontanella, Germanò, Garbossa, Isola, La Rocca, Maiuri, Olivi, Panciani, Pignotti, Skrap, Spena, Sabatino. Approved the final version of the manuscript on behalf of all authors: Ius. Statistical analysis: Ius, Isola. Administrative/technical/material support: Ius, Pignotti. Study supervision: Ius, Angileri, Barbagallo, Cappabianca, Esposito, Fontanella, Germanò, Garbossa, Olivi, Skrap, Sabatino.

References

  • 1

    Flanigan PM, Jahangiri A, Kuang R, Developing an algorithm for optimizing care of elderly patients with glioblastoma. Neurosurgery. 2018;82(1):6475.

    • Search Google Scholar
    • Export Citation
  • 2

    Chaichana KL, Chaichana KK, Olivi A, Surgical outcomes for older patients with glioblastoma multiforme: preoperative factors associated with decreased survival. Clinical article. J Neurosurg. 2011;114(3):587594.

    • Search Google Scholar
    • Export Citation
  • 3

    Iwamoto FM, Cooper AR, Reiner AS, Glioblastoma in the elderly: the Memorial Sloan-Kettering Cancer Center Experience (1997–2007). Cancer. 2009;115(16):37583766.

    • Search Google Scholar
    • Export Citation
  • 4

    Babu R, Komisarow JM, Agarwal VJ, Glioblastoma in the elderly: the effect of aggressive and modern therapies on survival. J Neurosurg. 2016;124(4):9981007.

    • Search Google Scholar
    • Export Citation
  • 5

    Scott JG, Suh JH, Elson P, Aggressive treatment is appropriate for glioblastoma multiforme patients 70 years old or older: a retrospective review of 206 cases. Neuro Oncol. 2011;13(4):428436.

    • Search Google Scholar
    • Export Citation
  • 6

    Tanaka S, Meyer FB, Buckner JC, Presentation, management, and outcome of newly diagnosed glioblastoma in elderly patients. J Neurosurg. 2013;118(4):786798.

    • Search Google Scholar
    • Export Citation
  • 7

    Álvarez de Eulate-Beramendi S, Álvarez-Vega MA, Balbin M, Prognostic factors and survival study in high-grade glioma in the elderly. Br J Neurosurg. 2016;30(3):330336.

    • Search Google Scholar
    • Export Citation
  • 8

    Molinaro AM, Hervey-Jumper S, Morshed RA, Association of maximal extent of resection of contrast-enhanced and non-contrast-enhanced tumor with survival within molecular subgroups of patients with newly diagnosed glioblastoma. JAMA Oncol. 2020;6(4):495503.

    • Search Google Scholar
    • Export Citation
  • 9

    Mariniello G, Peca C, De Caro MDB, Glioblastoma in the elderly: the impact of advanced age on treatment and survival. J Neurol Surg A Cent Eur Neurosurg. 2014;75(4):276281.

    • Search Google Scholar
    • Export Citation
  • 10

    Asmaa A, Dixit S, Rowland-Hill C, Management of elderly patients with glioblastoma-multiforme-a systematic review. Br J Radiol. 2018;91(1088):20170271.

    • Search Google Scholar
    • Export Citation
  • 11

    Braun K, Ahluwalia MS. Treatment of glioblastoma in older adults. Curr Oncol Rep. 2017;19(12):81.

  • 12

    Cohen-Inbar O. Geriatric brain tumor management part I: meningioma. J Clin Neurosci. 2019;67:59.

  • 13

    Halani SH, Babu R, Adamson DC. Management of glioblastoma multiforme in elderly patients: a review of the literature. World Neurosurg. 2017;105:5362.

    • Search Google Scholar
    • Export Citation
  • 14

    Han Q, Liang H, Cheng P, gross total vs. subtotal resection on survival outcomes in elderly patients with high-grade glioma: a systematic review and meta-analysis. Front Oncol. 2020;10:151.

    • Search Google Scholar
    • Export Citation
  • 15

    Jordan JT, Gerstner ER, Batchelor TT, Glioblastoma care in the elderly. Cancer. 2016;122(2):189197.

  • 16

    Lombardi G, Bergo E, Caccese M, Validation of the comprehensive geriatric assessment as a predictor of mortality in elderly glioblastoma patients. Cancers (Basel). 2019;11(10):E1509.

    • Search Google Scholar
    • Export Citation
  • 17

    Minniti G, Lombardi G, Paolini S. Glioblastoma in elderly patients: current management and future perspectives. Cancers (Basel). 2019;11(3):E336.

    • Search Google Scholar
    • Export Citation
  • 18

    Okada M, Miyake K, Tamiya T. Glioblastoma treatment in the elderly. Neurol Med Chir (Tokyo). 2017;57(12):667676.

  • 19

    Cohen-Inbar O. Geriatric brain tumor management part II: Glioblastoma multiforme. J Clin Neurosci. 2019;67:14.

  • 20

    Ganggayah MD, Taib NA, Har YC, Predicting factors for survival of breast cancer patients using machine learning techniques. BMC Med Inform Decis Mak. 2019;19(1):48.

    • Search Google Scholar
    • Export Citation
  • 21

    Ius T, Pignotti F, Della Pepa GM, A novel comprehensive clinical stratification model to refine prognosis of glioblastoma patients undergoing surgical resection. Cancers (Basel). 2020;12(2):E386.

    • Search Google Scholar
    • Export Citation
  • 22

    Chen W, Xie X, Wang J, A comparative study of logistic model tree, random forest, and classification and regression tree models for spatial prediction of landslide susceptibility. Catena. 2017;151:147160.

    • Search Google Scholar
    • Export Citation
  • 23

    Senders JT, Staples PC, Karhade AV, Machine learning and neurosurgical outcome prediction: a systematic review. World Neurosurg. 2018;109:476486.e1.

    • Search Google Scholar
    • Export Citation
  • 24

    Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol. 1994;47(11):12451251.

  • 25

    Ius T, Pignotti F, Della Pepa GM, Glioblastoma: from volumetric analysis to molecular predictors. J Neurosurg Sci. Published online February 4, 2020. doi:10.23736/S0390-5616.20.04850-X

    • Search Google Scholar
    • Export Citation
  • 26

    Yohannes Y, Hoddinott J. Classification and Regression Trees: An Introduction. International Food Policy Research Institute; 1999.

  • 27

    Schmid M, Wright MN, Ziegler A. On the use of Harrell’s C for clinical risk prediction via random survival forests. Expert Syst Appl. 2016;63(C):450459.

    • Search Google Scholar
    • Export Citation
  • 28

    Li YM, Suki D, Hess K, Sawaya R. The influence of maximum safe resection of glioblastoma on survival in 1229 patients: Can we do better than gross-total resection? J Neurosurg. 2016;124(4):977988.

    • Search Google Scholar
    • Export Citation
  • 29

    Lacroix M, Abi-Said D, Fourney DR, A multivariate analysis of 416 patients with glioblastoma multiforme: prognosis, extent of resection, and survival. J Neurosurg. 2001;95(2):190198.

    • Search Google Scholar
    • Export Citation
  • 30

    Somma T, Ius T, Certo F, From the champion to the team: new treatment paradigms in contemporary neurosurgery. World Neurosurg. 2019;131:141148.

    • Search Google Scholar
    • Export Citation
  • 31

    Ius T, Cesselli D, Isola M, Combining clinical and molecular data to predict the benefits of carmustine wafers in newly diagnosed high-grade gliomas. Curr Treat Options Neurol. 2018;20(2):3.

    • Search Google Scholar
    • Export Citation
  • 32

    Chaichana KL, Cabrera-Aldana EE, Jusue-Torres I, When gross total resection of a glioblastoma is possible, how much resection should be achieved? World Neurosurg. 2014;82(1-2):e257e265.

    • Search Google Scholar
    • Export Citation
  • 33

    Gessler F, Bernstock JD, Braczynski A, Surgery for glioblastoma in light of molecular markers: impact of resection and MGMT promoter methylation in newly diagnosed IDH-1 wild-type glioblastomas. Neurosurgery. 2019;84(1):190197.

    • Search Google Scholar
    • Export Citation
  • 34

    Illic R, Somma T, Savic D, A survival analysis with identification of prognostic factors in a series of 110 patients with newly diagnosed glioblastoma before and after introduction of the Stupp regimen: a single-center observational study. World Neurosurg. 2017;104:581588.

    • Search Google Scholar
    • Export Citation
  • 35

    Altieri R, Zenga F, Fontanella MM, Glioma surgery: technological advances to achieve a maximal safe resection. Surg Technol Int. 2015;27:297302.

    • Search Google Scholar
    • Export Citation
  • 36

    Altieri R, Zenga F, Ducati A, Tumor location and patient age predict biological signatures of high-grade gliomas. Neurosurg Rev. 2018;41(2):599604.

    • Search Google Scholar
    • Export Citation
  • 37

    Cloney M, D’Amico R, Lebovic J, Frailty in geriatric glioblastoma patients: a predictor of operative morbidity and outcome. World Neurosurg. 2016;89:362367.

    • Search Google Scholar
    • Export Citation
  • 38

    Biau J, Chautard E, De Schlichting E, Radiotherapy plus temozolomide in elderly patients with glioblastoma: a “real-life” report. Radiat Oncol. 2017;12(1):197.

    • Search Google Scholar
    • Export Citation
  • 39

    Arvold ND, Cefalu M, Wang Y, Comparative effectiveness of radiotherapy with vs. without temozolomide in older patients with glioblastoma. J Neurooncol. 2017;131(2):301311.

    • Search Google Scholar
    • Export Citation
  • 40

    Sherrod BA, Gamboa NT, Wilkerson C, Effect of patient age on glioblastoma perioperative treatment costs: a value driven outcome database analysis. J Neurooncol. 2019;143(3):465473.

    • Search Google Scholar
    • Export Citation
  • 41

    Yang Y, Yan LF, Zhang X, Optimizing texture retrieving model for multimodal MR image-based support vector machine for classifying glioma. J Magn Reson Imaging. 2019;49(5):12631274.

    • Search Google Scholar
    • Export Citation
  • 42

    Cunha MLVD, Esmeraldo ACS, Henriques LAW, Elderly patients with glioblastoma: the impact of the surgical resection extent on survival. Rev Assoc Med Bras (1992). 2019;65(7):937945.

    • Search Google Scholar
    • Export Citation
  • 43

    Li KK-W, Shi Z-F, Malta TM, Identification of subsets of IDH-mutant glioblastomas with distinct epigenetic and copy number alterations and stratified clinical risks. Neurooncol Adv. 2019;1(1):vdz015.

    • Search Google Scholar
    • Export Citation
  • 44

    Louis DN, Perry A, Reifenberger G, The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary. Acta Neuropathol. 2016;131(6):803820.

    • Search Google Scholar
    • Export Citation
  • 45

    Gorlia T, Stupp R, Brandes AA, New prognostic factors and calculators for outcome prediction in patients with recurrent glioblastoma: a pooled analysis of EORTC Brain Tumour Group phase I and II clinical trials. Eur J Cancer. 2012;48(8):11761184.

    • Search Google Scholar
    • Export Citation

If the inline PDF is not rendering correctly, you can download the PDF file here.

Contributor Notes

Correspondence Tamara Ius: Santa Maria della Misericordia University Hospital, Udine, Italy. tamara.ius@gmail.com.

INCLUDE WHEN CITING DOI: 10.3171/2020.7.FOCUS20420.

Disclosures The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

  • View in gallery

    Random forest, CART algorithm. The CART algorithm provides a graphic visualization of the interaction between risk factors detected by Cox survival analysis. In each hierarchical node, the study population is split according to the presence (green) or not (red) of the variable able to influence prognosis. A: CART model performed on 250 EGBM cases that met all the selection criteria. A score from 1 to 5 was assigned to the 6 terminal nodes thus defined based on the RHR. Patients belonging to the subgroup with RHR = 0.73 and to the subgroup with RHR = 0.55 were joined to create a single group, labeled “score 2.” This was done considering the small sample size of each subgroup and similarities in RHR values. B: CART model applied to IDH1/2 wild-type EGBM patients (239 cases). A score from 1 to 4 was obtained from the 4 terminal nodes.

  • View in gallery

    OS stratified by CART analysis prognostic score. Kaplan-Meier curves display the OS of EGBM patients according to the prognostic scores elaborated by the CART model. A: Survival stratified by the score from 250 EGBM cases that met the selection criteria. B: Survival stratified by the score from 239 EGBM IDH1/2 wild-type cases.

  • View in gallery

    Systematic review and meta-analysis flowchart. The figure shows the different decisional phases regarding the inclusion and exclusion of the papers reviewed in the studies based on the role of surgery in EGBM patients.

  • 1

    Flanigan PM, Jahangiri A, Kuang R, Developing an algorithm for optimizing care of elderly patients with glioblastoma. Neurosurgery. 2018;82(1):6475.

    • Search Google Scholar
    • Export Citation
  • 2

    Chaichana KL, Chaichana KK, Olivi A, Surgical outcomes for older patients with glioblastoma multiforme: preoperative factors associated with decreased survival. Clinical article. J Neurosurg. 2011;114(3):587594.

    • Search Google Scholar
    • Export Citation
  • 3

    Iwamoto FM, Cooper AR, Reiner AS, Glioblastoma in the elderly: the Memorial Sloan-Kettering Cancer Center Experience (1997–2007). Cancer. 2009;115(16):37583766.

    • Search Google Scholar
    • Export Citation
  • 4

    Babu R, Komisarow JM, Agarwal VJ, Glioblastoma in the elderly: the effect of aggressive and modern therapies on survival. J Neurosurg. 2016;124(4):9981007.

    • Search Google Scholar
    • Export Citation
  • 5

    Scott JG, Suh JH, Elson P, Aggressive treatment is appropriate for glioblastoma multiforme patients 70 years old or older: a retrospective review of 206 cases. Neuro Oncol. 2011;13(4):428436.

    • Search Google Scholar
    • Export Citation
  • 6

    Tanaka S, Meyer FB, Buckner JC, Presentation, management, and outcome of newly diagnosed glioblastoma in elderly patients. J Neurosurg. 2013;118(4):786798.

    • Search Google Scholar
    • Export Citation
  • 7

    Álvarez de Eulate-Beramendi S, Álvarez-Vega MA, Balbin M, Prognostic factors and survival study in high-grade glioma in the elderly. Br J Neurosurg. 2016;30(3):330336.

    • Search Google Scholar
    • Export Citation
  • 8

    Molinaro AM, Hervey-Jumper S, Morshed RA, Association of maximal extent of resection of contrast-enhanced and non-contrast-enhanced tumor with survival within molecular subgroups of patients with newly diagnosed glioblastoma. JAMA Oncol. 2020;6(4):495503.

    • Search Google Scholar
    • Export Citation
  • 9

    Mariniello G, Peca C, De Caro MDB, Glioblastoma in the elderly: the impact of advanced age on treatment and survival. J Neurol Surg A Cent Eur Neurosurg. 2014;75(4):276281.

    • Search Google Scholar
    • Export Citation
  • 10

    Asmaa A, Dixit S, Rowland-Hill C, Management of elderly patients with glioblastoma-multiforme-a systematic review. Br J Radiol. 2018;91(1088):20170271.

    • Search Google Scholar
    • Export Citation
  • 11

    Braun K, Ahluwalia MS. Treatment of glioblastoma in older adults. Curr Oncol Rep. 2017;19(12):81.

  • 12

    Cohen-Inbar O. Geriatric brain tumor management part I: meningioma. J Clin Neurosci. 2019;67:59.

  • 13

    Halani SH, Babu R, Adamson DC. Management of glioblastoma multiforme in elderly patients: a review of the literature. World Neurosurg. 2017;105:5362.

    • Search Google Scholar
    • Export Citation
  • 14

    Han Q, Liang H, Cheng P, gross total vs. subtotal resection on survival outcomes in elderly patients with high-grade glioma: a systematic review and meta-analysis. Front Oncol. 2020;10:151.

    • Search Google Scholar
    • Export Citation
  • 15

    Jordan JT, Gerstner ER, Batchelor TT, Glioblastoma care in the elderly. Cancer. 2016;122(2):189197.

  • 16

    Lombardi G, Bergo E, Caccese M, Validation of the comprehensive geriatric assessment as a predictor of mortality in elderly glioblastoma patients. Cancers (Basel). 2019;11(10):E1509.

    • Search Google Scholar
    • Export Citation
  • 17

    Minniti G, Lombardi G, Paolini S. Glioblastoma in elderly patients: current management and future perspectives. Cancers (Basel). 2019;11(3):E336.

    • Search Google Scholar
    • Export Citation
  • 18

    Okada M, Miyake K, Tamiya T. Glioblastoma treatment in the elderly. Neurol Med Chir (Tokyo). 2017;57(12):667676.

  • 19

    Cohen-Inbar O. Geriatric brain tumor management part II: Glioblastoma multiforme. J Clin Neurosci. 2019;67:14.

  • 20

    Ganggayah MD, Taib NA, Har YC, Predicting factors for survival of breast cancer patients using machine learning techniques. BMC Med Inform Decis Mak. 2019;19(1):48.

    • Search Google Scholar
    • Export Citation
  • 21

    Ius T, Pignotti F, Della Pepa GM, A novel comprehensive clinical stratification model to refine prognosis of glioblastoma patients undergoing surgical resection. Cancers (Basel). 2020;12(2):E386.

    • Search Google Scholar
    • Export Citation
  • 22

    Chen W, Xie X, Wang J, A comparative study of logistic model tree, random forest, and classification and regression tree models for spatial prediction of landslide susceptibility. Catena. 2017;151:147160.

    • Search Google Scholar
    • Export Citation
  • 23

    Senders JT, Staples PC, Karhade AV, Machine learning and neurosurgical outcome prediction: a systematic review. World Neurosurg. 2018;109:476486.e1.

    • Search Google Scholar
    • Export Citation
  • 24

    Charlson M, Szatrowski TP, Peterson J, Gold J. Validation of a combined comorbidity index. J Clin Epidemiol. 1994;47(11):12451251.

  • 25

    Ius T, Pignotti F, Della Pepa GM, Glioblastoma: from volumetric analysis to molecular predictors. J Neurosurg Sci. Published online February 4, 2020. doi:10.23736/S0390-5616.20.04850-X

    • Search Google Scholar
    • Export Citation
  • 26

    Yohannes Y, Hoddinott J. Classification and Regression Trees: An Introduction. International Food Policy Research Institute; 1999.

  • 27

    Schmid M, Wright MN, Ziegler A. On the use of Harrell’s C for clinical risk prediction via random survival forests. Expert Syst Appl. 2016;63(C):450459.

    • Search Google Scholar
    • Export Citation
  • 28

    Li YM, Suki D, Hess K, Sawaya R. The influence of maximum safe resection of glioblastoma on survival in 1229 patients: Can we do better than gross-total resection? J Neurosurg. 2016;124(4):977988.

    • Search Google Scholar
    • Export Citation
  • 29

    Lacroix M, Abi-Said D, Fourney DR, A multivariate analysis of 416 patients with glioblastoma multiforme: prognosis, extent of resection, and survival. J Neurosurg. 2001;95(2):190198.

    • Search Google Scholar
    • Export Citation
  • 30

    Somma T, Ius T, Certo F, From the champion to the team: new treatment paradigms in contemporary neurosurgery. World Neurosurg. 2019;131:141148.

    • Search Google Scholar
    • Export Citation
  • 31

    Ius T, Cesselli D, Isola M, Combining clinical and molecular data to predict the benefits of carmustine wafers in newly diagnosed high-grade gliomas. Curr Treat Options Neurol. 2018;20(2):3.

    • Search Google Scholar
    • Export Citation
  • 32

    Chaichana KL, Cabrera-Aldana EE, Jusue-Torres I, When gross total resection of a glioblastoma is possible, how much resection should be achieved? World Neurosurg. 2014;82(1-2):e257e265.

    • Search Google Scholar
    • Export Citation
  • 33

    Gessler F, Bernstock JD, Braczynski A, Surgery for glioblastoma in light of molecular markers: impact of resection and MGMT promoter methylation in newly diagnosed IDH-1 wild-type glioblastomas. Neurosurgery. 2019;84(1):190197.

    • Search Google Scholar
    • Export Citation
  • 34

    Illic R, Somma T, Savic D, A survival analysis with identification of prognostic factors in a series of 110 patients with newly diagnosed glioblastoma before and after introduction of the Stupp regimen: a single-center observational study. World Neurosurg. 2017;104:581588.

    • Search Google Scholar
    • Export Citation
  • 35

    Altieri R, Zenga F, Fontanella MM, Glioma surgery: technological advances to achieve a maximal safe resection. Surg Technol Int. 2015;27:297302.

    • Search Google Scholar
    • Export Citation
  • 36

    Altieri R, Zenga F, Ducati A, Tumor location and patient age predict biological signatures of high-grade gliomas. Neurosurg Rev. 2018;41(2):599604.

    • Search Google Scholar
    • Export Citation
  • 37

    Cloney M, D’Amico R, Lebovic J, Frailty in geriatric glioblastoma patients: a predictor of operative morbidity and outcome. World Neurosurg. 2016;89:362367.

    • Search Google Scholar
    • Export Citation
  • 38

    Biau J, Chautard E, De Schlichting E, Radiotherapy plus temozolomide in elderly patients with glioblastoma: a “real-life” report. Radiat Oncol. 2017;12(1):197.

    • Search Google Scholar
    • Export Citation
  • 39

    Arvold ND, Cefalu M, Wang Y, Comparative effectiveness of radiotherapy with vs. without temozolomide in older patients with glioblastoma. J Neurooncol. 2017;131(2):301311.

    • Search Google Scholar
    • Export Citation
  • 40

    Sherrod BA, Gamboa NT, Wilkerson C, Effect of patient age on glioblastoma perioperative treatment costs: a value driven outcome database analysis. J Neurooncol. 2019;143(3):465473.

    • Search Google Scholar
    • Export Citation
  • 41

    Yang Y, Yan LF, Zhang X, Optimizing texture retrieving model for multimodal MR image-based support vector machine for classifying glioma. J Magn Reson Imaging. 2019;49(5):12631274.

    • Search Google Scholar
    • Export Citation
  • 42

    Cunha MLVD, Esmeraldo ACS, Henriques LAW, Elderly patients with glioblastoma: the impact of the surgical resection extent on survival. Rev Assoc Med Bras (1992). 2019;65(7):937945.

    • Search Google Scholar
    • Export Citation
  • 43

    Li KK-W, Shi Z-F, Malta TM, Identification of subsets of IDH-mutant glioblastomas with distinct epigenetic and copy number alterations and stratified clinical risks. Neurooncol Adv. 2019;1(1):vdz015.

    • Search Google Scholar
    • Export Citation
  • 44

    Louis DN, Perry A, Reifenberger G, The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary. Acta Neuropathol. 2016;131(6):803820.

    • Search Google Scholar
    • Export Citation
  • 45

    Gorlia T, Stupp R, Brandes AA, New prognostic factors and calculators for outcome prediction in patients with recurrent glioblastoma: a pooled analysis of EORTC Brain Tumour Group phase I and II clinical trials. Eur J Cancer. 2012;48(8):11761184.

    • Search Google Scholar
    • Export Citation

Metrics

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 209 209 209
PDF Downloads 128 128 128
EPUB Downloads 0 0 0