Recursive partitioning analysis of prognostic factors in WHO grade III glioma patients treated with radiotherapy or radiotherapy plus chemotherapy
Recursive partitioning analysis of prognostic factors in WHO grade III glioma patients treated with radiotherapy or radiotherapy plus chemotherapy
Chul-Kee Park 2
Se-Hoon Lee 1 3
Jung Ho Han 2
Chae-Yong Kim 2
Dong- Wan Kim 1 3
Sun Ha Paek 2 3
Dong Gyu Kim 2
Dae Seog Heo 1 3
Il Han Kim 0 3
Hee-Won Jung 2
0 Department of Radiation Oncology, Seoul National University College of Medicine, Seoul National University , Seoul 110-744 , Korea
1 Department of Internal Medicine, Seoul National University College of Medicine, Seoul National University , Seoul 110-744 , Korea
2 Department of Neurosurgery, Seoul National University College of Medicine, Seoul National University , Seoul 110-744 , Korea
3 Seoul National University Hospital Cancer Research Institute , Seoul 110-744 , Korea
Background: We evaluated the hierarchical risk groups for the estimated survival of WHO grade III glioma patients using recursive partitioning analysis (RPA). To our knowledge, this is the first study to address the results of RPA specifically for WHO grade III gliomas. Methods: A total of 133 patients with anaplastic astrocytoma (AA, n = 56), anaplastic oligodendroglioma (AO, n = 67), or anaplastic oligoastrocytoma (AOA, n = 10) were included in the study. These patients were treated with either radiotherapy alone or radiotherapy followed by PCV chemotherapy after surgery. Five prognostic factors, including histological subsets, age, performance status, extent of resection, and treatment modality were incorporated into the RPA. The final nodes of RPA were grouped according to their survival times, and the Kaplan-Meier graphs are presented as the final set of prognostic groups. Results: Four risk groups were defined based on the clinical prognostic factors excluding age, and split variables were all incorporated into the RPA. Survival analysis showed significant differences in mean survival between the different groups: 163.4 months (95% CI: 144.9-182.0), 109.5 months (86.7-132.4), 66.6 months (50.8-82.4), and 27.7 months (16.3-39.0), respectively, from the lowest to the highest risk group (p = 0.00). Conclusion: The present study shows that RPA grouping with clinical prognostic factors can successfully predict the survival of patients with WHO grade III glioma.
Anaplastic astrocytoma (AA), anaplastic
oligodendroglioma (AO), and anaplastic oligoastrocytoma (AOA) are
defined as the major histological categories of WHO grade
III gliomas, even though their classification based on the
known molecular biology information remains
controversial [1-4]. The relative survival rates at 5 years for AA
and AO are 29.4% and 45.2%, respectively. Although
the increase is modest, standardized radiotherapy or
chemotherapy has extended the survival period for
patients with high-grade gliomas, suggesting the possible
influence of prognostic factors such as age, performance
status, symptom duration, tumor resection, histological
type, and 1p/19q co-deletion [6-8]. However, most
randomized trials pooled both grades III or IV astrocytic
tumors and grade III oligodendroglial tumors as
malignant glioma[7,8]. The only exceptions are the studies on
PCV chemotherapy for AO[9,10]. A separate study of
WHO grade III glioma is needed because of the observed
difference in the outcome for grade IV glioblastoma using
the same treatment protocols[11,12]. Even in those
welldesigned studies employing consistent treatment
protocols, diverse biological behaviors and clinical outcomes
clearly exist for patients with AA, AO, and AOA. Therefore,
the prediction of survival outcomes for WHO grade III
glioma patients based on known clinical prognostic factors
attains importance with regard to the treatment plan.
Recursive partitioning analysis (RPA) classification, which
was initially described by the Radiation Therapy
Oncology Group (RTOG), is a useful tool that can divide
patients into homogenous groups based on the length of
survival. RPA has an advantage over the proportional
hazards model in identifying prognostic factors because it
makes fewer modeling assumptions and has an
established procedure that adapts to missing data through the
use of surrogate measures. Moreover, development of
high-speed computer systems has made this tedious work
easy for researchers. The RTOG presented results from
an RPA for all patients with malignant glioma (both
WHO grade III and IV) throughout the duration of their
clinical trials. Results obtained for re-analysis of
clinical trials using RPA, which focused only on glioblastoma
patients, were also reported.
We evaluated the prognostic factors in the distinct group
of newly diagnosed WHO grade III patients who were
treated with radiotherapy or radiotherapy plus
chemotherapy (PCV regimen) after surgery, in an attempt to
predict survival outcomes by RPA. To our knowledge, this is
the first study to address the RPA results focused
specifically on WHO grade III gliomas.
A total of 133 newly diagnosed AA, AO, and AOA patients
were included in this single-center retrospective study.
These patients were treated at the Seoul National
University Hospital between January 1990 and December 2004,
according to the baseline protocol, either with
radiotherapy alone or with radiotherapy plus PCV chemotherapy.
Histological diagnosis was re-evaluated according to the
WHO 2000 classification. Patient data were collected
according to the guidance specifications approved by the
institutional review board and included information
contained within the hospital charts and radiological studies.
Clinical data such as age, performance status, extent of
resection, and primary treatment modality after surgery
were collected. Data that were unavailable in the medical
records due to follow-up loss were obtained via a
telephone interview with the patient or, if the patient was
deceased, with his or her relatives with their permission.
The survival time was measured from the date of surgery
to the date of the patient's death. Patients who were alive
were classified as censored observations at the time of the
last follow-up. Variables selected for prognosis analysis
were those that were determined as significant based on
previous reports[7-9,11,15]. Age was classified as 50 and
above or under 50. Performance status was scored
according to the Eastern Cooperative Oncology Group (ECOG)
scales. The extent of resection was categorized as a
complete resection or incomplete resection (including
biopsy only) based on the immediate post-operative
imaging findings. Residual enhancing lesion on
T1enhanced images or measurable high signal intensity
lesion on T2 images without enhancement of magnetic
resonance images were considered residual lesion.
Histological diagnosis and base-line treatment protocols were
also included in the analysis.
The Kaplan-Meyer method was used to estimate the
overall survival distributions. The log-rank test (level of
significance = 0.05) was used to test the differences in the
overall survival distributions with respect to prognostic
variables. A Cox proportional hazards model (level of
significance = 0.05) was used to adjust for covariates. These
analyses were performed using SPSS ver 12.0. During
RPA, free software (R version 2.6.2; rpart package version
3.1.39-1, http://www.r-project.org/) was used for the
recursive decision tree creation with the split criteria of p
< 0.01 in the log-rank test. The final nodes were grouped
according to their survival times, and the Kaplan-Meier
graphs are presented as the final set of prognostic groups.
The baseline clinical data of the study population are
summarized in Table 1. Among 133 patients, 56 patients
had a histological diagnosis of AA, 67 had AO, and 10 had
AOA. The mean follow-up period of whole population
was 88.8 months. The age distributions at the time of
diagnosis, performance status, and baseline treatment
were comparable among the groups defined by the
histoECOG: Eastern Cooperative Oncology Group
PCV: procarbazine, lomustine, and vincristine
logical diagnosis. A predilection for incomplete resection
of tumors was apparent in the AA group due to their
diffuse infiltrating nature of growth. Radiotherapy alone was
part of the baseline treatment protocol before the
mid1990s (n = 67), and 6 to 12 cycles of PCV (procarbazine,
lomustine, and vincristine) chemotherapy were added
after radiotherapy beginning in the late-1990s (n = 66). In
all patients, radiotherapy was initiated within 6 weeks
after the surgery with a total mean dose of 59.4 Gy, 1.8 Gy
per fraction with five fractions per week. The target
volume included the residual tumor volume or surgical
cavity and surrounding edema with a margin of 3 cm. Seven
patients could not complete the planned dose of
radiotherapy due to intolerance. PCV chemotherapy started
within 4 weeks after the end of RT. Each cycle consisted of
lomustine 110 mg/m2 orally on day 1, procarbazine 60
mg/m2 orally on days 8 to 21, and vincristine 1.4 mg/m2
intravenous on days 8 and 29. Cycles were to be repeated
every 6 weeks. The mean number of cycles completed per
patient was 5.8. Due to the intolerance, 42% of patients
were treated with less than 6 cycles of PCV.
The median overall survival of the entire population was
59.0 months (95% confidence interval = 38.0 through
80.0). Estimated survival rates at 1-, 2-, 5-, and 10-years
were 84.9%, 72.8%, 49.1%, and 35.5%, respectively.
Survival was significantly better in radiotherapy plus PCV
group (mean 118.0 months; 95% confidence interval =
100.7 through 135.4, median not reached) than
radiotherapy only group (mean 57.7 months; 95% confidence
interval = 41.7 through 73.8, median 29.0 months; 95%
confidence interval = 19.2 through 38.9) (p = 0.00).
All possible prognostic variables determined as significant
affected the overall survival in the univariate analysis. The
results of the univariate analysis on median survival are as
follows; age (65 months in < 50 years vs 19 months in
50 years), performance status (53 months in ECOG grade
1 vs 13 months in ECOG grade 2 (survivals showed
stratification with ECOG grade and showed significant split
between grade 1 and 2)), histology (29 months in AA vs
37 months in AOA vs 79 months in AO), and extent of
resection (median survival not reached in complete
resection vs 19 months in incomplete resection). The results of
the Cox proportional-hazard analysis using these
prognostic variables showed that the prolonged overall
survival was independently affected by young age, good
performance status, histological diagnosis of AO,
complete resection, or addition of adjuvant PCV
chemotherapy after radiotherapy (Table 2).
Using the above mentioned significant prognostic
variables, a recursive decision tree comprising 132 patients was
created after exclusion of one patient during the analysis
due to short survival time (less than 1 month). A total of
8 terminal nodes were produced in 6 splits (Figure 1).
Among the variables included in the RPA, age was omitted
from the split criteria according to the order of priority.
Based on the median survival time of the terminal nodes,
we were able to categorize them into four groups (Table
3). Survival analysis using the Kaplan-Meyer and log-rank
test confirmed the significant differences among groups
(p = 0.00, Figure 2). Independent of the histological
diagnosis, patients treated with radiotherapy plus PCV
chemotherapy after complete resection, or after incomplete
resection but with the best performance status (ECOG
grade 0), can expect the longest survival (Group A). Using
the present results, survival probabilities can be estimated
based on post-surgical clinical settings.
Table 2: Multivariate Cox proportional-hazards results for the prognostic value of variables related to the survival of WHO grade III
glioma patients (n = 131).
ECOG: Eastern Cooperative Oncology Group
AO: anaplastic oligodendroglioma
PCV: procarbazine, lomustine, and vincristine
Although AA, AO, and AOA are grouped as the same
histological grade by the WHO classification system, among
them exist diverse biological behaviors linked to clinical
outcome, even within the same histological diagnosis.
However, straightforward comparative analyses of the
prognostic factors among the WHO grade III gliomas have
been reported infrequently. In the present study,
hierarchical stratification of prognostic variables, such as
performance status, histological diagnosis, extent of
95% confidence interval
resection, and adjuvant chemotherapy after radiotherapy
among the WHO grade III gliomas could be deduced
successfully using the RPA method. Based on the results of
the present study, several conclusions could be drawn: (1)
good complete resection is the most important prognostic
factor for performance status, (2) the oligodendroglial
component of the tumor favors better prognosis, and (3)
PCV chemotherapy may be beneficial for certain groups of
patients. These findings are not novel; however, using the
risk group splits according to the given condition of the
FDiegcuisrieon1tree constructed by recursive partitioning analysis
Decision tree constructed by recursive partitioning analysis. Terminal nodes ( ) are categorized into 4 groups based
on their median survival times.
SFuigrvuirveal 2plot of the risk groups defined in Table 3
Survival plot of the risk groups defined in Table 3. Kaplan-Meyer analysis and the log-rank test revealed significant
differences among the groups (p = 0.00).
patients, we can estimate survival based on the chosen
According to various studies, age is an invariably
important prognostic factor of performance status for WHO
grade III glioma[8,13,17,18]. We also observed significant
prognostic values for performance status in the present
analysis. However, it is important to note that age was
excluded as a factor during the RPA, although it was a
significant variable in the univariate analysis.
Analysis of the extent of resection for WHO grade III
gliomas was based on the available data collected in an
uncontrolled study setting. The present analysis has
pitfalls, such as that superficial, small, well-demarcated
tumors tend to undergo complete resection, whereas
deep-seated, extensive, diffuse tumors are more likely to
undergo biopsy only or incomplete resection[7,8].
Evidence for the prognostic impact of the extent of resection
on AA remains sparse; however, randomized studies
investigating a combination of multiple treatment
Number of patients
ities support the beneficial impact of complete resection
for AO and AOA[9,10].
Better prognosis of oligodendroglial tumors over
astrocytic tumors is also supported by previous studies.
This evidence was further investigated, and the underlying
genetic signature, such as the 1p/19q co-deletion in
oligodendroglial tumors, was found to be responsible for the
favorable prognosis. Moreover, the diffuse nature of AA
that precludes complete resection might have affected the
outcome. It is a limitation of the present study that we
could not include any molecular markers into RPA due to
unavailability of appropriate tissue samples to carry on
Radiotherapy and PCV chemotherapy, along with other
alkylating agents such as nitrosoureas and temozolomide,
are still considered to be the major options for the
management of AA, AO, and AOA patients. Nonetheless, the
role of radiotherapy in the treatment of AO/AOA and that
of PCV chemotherapy for AA still requires further
investigation [21-24]. Although the results of the present study
show a beneficial effect of PCV chemotherapy on the
outcome of the multivariate analysis, the possibility of a
selection bias remains, since patients with poor
performance status were excluded from further chemotherapy.
Another noteworthy finding of the present study is the
evidence for the diversity of the prognosis of AA, AO, and
AOA patients, despite the best available treatment. The
estimated survival ranged from good (group A) to bad
(group D, such as glioblastoma) status. The results imply
that AA, AO, and AOA patients with the best performance
status, or with completely extirpated tumors, can be
maintain prolonged survival without any recurrence with
radiotherapy followed by PCV chemotherapy.
There are many clinical studies employing RPA to define
risk groups[14,25-28]. RPA is a robust tool for the
stratification of prognostic factors and for the identification of a
homogenous group of patients for a given disease and
treatment strategy. However, there are limitations to RPA
application. For example, it is a post-hoc test, and no
predictions can be made using the final splits, since
prognostic factors are selected by chance. If multiple variables
that are highly correlated exist, the selection of factors may
vary. Despite these limitations, the RPA method
allows for a clear distinction between the WHO grade III
glioma patients risk groups. The advantage of this study is
that only those patients with WHO grade III glioma who
received the best available treatment were included in the
analysis. It is more likely that the novel anticancer agent
such as temozolomide will take over the mainstream of
WHO grade III glioma management sooner or later.
However, we believe that the systemized analysis for classic
management has its own significance because it can be
The present study shows that RPA grouping can
successfully predict the survival of patients with WHO grade III
glioma. Performance status, extent of resection,
histological diagnosis, and treatment modality are the major
determinants of patients' survival. These results may provide a
tool for the collection of baseline data for further
investigation of treatment modalities in the different risk groups
The authors declare that they have no competing interests.
CKP designed the study and wrote the draft and
manuscript. SHL, JHH, and CYK were involved in the
interpretation of the results. DWK, SHP, DGK, DSH, IHK, and
HWJ managed the patients and reviewed the manuscript.
All authors read and approved the final manuscript.
This study was supported by a grant of the Seoul National University
Hospital Research Fund and the statistical analysis was supported by the Seoul
National University Hospital Medical Research Collaborating Center.
1. Louis DN , Ohgaki H , Wiestler OD , Cavenee WK , Burger PC , Jouvet A , Scheithauer BW , Kleihues P : The 2007 WHO classification of tumours of the central nervous system . Acta Neuropathol 2007 , 114 ( 2 ): 97 - 109 .
2. Kleihues P , Louis DN , Scheithauer BW , Rorke LB , Reifenberger G , Burger PC , Cavenee WK : The WHO classification of tumors of the nervous system . J Neuropathol Exp Neurol 2002 , 61 ( 3 ): 215 - 225 . discussion 226 - 219
3. Kleihues P , Burger PC , Scheithauer BW : The new WHO classification of brain tumours . Brain Pathol 1993 , 3 ( 3 ): 255 - 268 .
4. Scheithauer BW , Fuller GN , VandenBerg SR : The 2007 WHO classification of tumors of the nervous system: controversies in surgical neuropathology . Brain Pathol 2008 , 18 ( 3 ): 307 - 316 .
5. CBTRUS ( 2008 ). Statistical Report: Primary Brain Tumors in the United States , 2000 - 2004 . Published by the Central Brain Tumor Registry of the United States . .
6. Compostella A , Tosoni A , Blatt V , Franceschi E , Brandes AA : Prognostic factors for anaplastic astrocytomas . J Neurooncol 2007 , 81 ( 3 ): 295 - 303 .
7. Stupp R , Reni M , Gatta G , Mazza E , Vecht C : Anaplastic astrocytoma in adults . Crit Rev Oncol Hematol 2007 , 63 ( 1 ): 72 - 80 .
8. Bent MJ van den: Anaplastic oligodendroglioma and oligoastrocytoma . Neurol Clin 2007 , 25 ( 4 ): 1089 - 1109 . ix-x
9. Cairncross G , Berkey B , Shaw E , Jenkins R , Scheithauer B , Brachman D , Buckner J , Fink K , Souhami L , Laperierre N , Mehta M , Curran W : Phase III trial of chemotherapy plus radiotherapy compared with radiotherapy alone for pure and mixed anaplastic oligodendroglioma: Intergroup Radiation Therapy Oncology Group Trial 9402 . J Clin Oncol 2006 , 24 ( 18 ): 2707 - 2714 .
10. Bent MJ van den, Carpentier AF , Brandes AA , Sanson M , Taphoorn MJ , Bernsen HJ , Frenay M , Tijssen CC , Grisold W , Sipos L , HaaxmaReiche H , Kros JM , van Kouwenhoven MC , Vecht CJ , Allgeier A , Lacombe D , Gorlia T : Adjuvant procarbazine, lomustine, and vincristine improves progression-free survival but not overall survival in newly diagnosed anaplastic oligodendrogliomas and oligoastrocytomas: a randomized European Organisation for Research and Treatment of Cancer phase III trial . J Clin Oncol 2006 , 24 ( 18 ): 2715 - 2722 .
11. Levin VA , Silver P , Hannigan J , Wara WM , Gutin PH , Davis RL , Wilson CB : Superiority of post-radiotherapy adjuvant chemotherapy with CCNU, procarbazine, and vincristine (PCV) over BCNU for anaplastic gliomas: NCOG 6G61 final report . Int J Radiat Oncol Biol Phys 1990 , 18 ( 2 ): 321 - 324 .
12. Hildebrand J , Sahmoud T , Mignolet F , Brucher JM , Afra D : Adjuvant therapy with dibromodulcitol and BCNU increases survival of adults with malignant gliomas . EORTC Brain Tumor Group. Neurology 1994 , 44 ( 8 ): 1479 - 1483 .
13. Curran WJ Jr, Scott CB , Horton J , Nelson JS , Weinstein AS , Fischbach AJ , Chang CH , Rotman M , Asbell SO , Krisch RE , Nelson DF : Recursive partitioning analysis of prognostic factors in three Radiation Therapy Oncology Group malignant glioma trials . J Natl Cancer Inst 1993 , 85 ( 9 ): 704 - 710 .
14. Lamborn KR , Chang SM , Prados MD : Prognostic factors for survival of patients with glioblastoma: recursive partitioning analysis . Neuro Oncol 2004 , 6 ( 3 ): 227 - 235 .
15. Bent MJ van den, Taphoorn MJ , Brandes AA , Menten J , Stupp R , Frenay M , Chinot O , Kros JM , Rijt CC van der, Vecht Ch J , Allgeier A , Gorlia T : Phase II study of first-line chemotherapy with temozolomide in recurrent oligodendroglial tumors: the European Organization for Research and Treatment of Cancer Brain Tumor Group Study 26971 . J Clin Oncol 2003 , 21 ( 13 ): 2525 - 2528 .
16. Oken MM , Creech RH , Tormey DC , Horton J , Davis TE , McFadden ET , Carbone PP : Toxicity and response criteria of the Eastern Cooperative Oncology Group . Am J Clin Oncol 1982 , 5 ( 6 ): 649 - 655 .
17. Scott CB , Scarantino C , Urtasun R , Movsas B , Jones CU , Simpson JR , Fischbach AJ , Curran WJ Jr: Validation and predictive power of Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis classes for malignant glioma patients: a report using RTOG 90-06 . Int J Radiat Oncol Biol Phys 1998 , 40 ( 1 ): 51 - 55 .
18. Tortosa A , Vinolas N , Villa S , Verger E , Gil JM , Brell M , Caral L , Pujol T , Acebes JJ , Ribalta T , Ferrer I , Graus F : Prognostic implication of clinical, radiologic, and pathologic features in patients with anaplastic gliomas . Cancer 2003 , 97 ( 4 ): 1063 - 1071 .
19. Nomiya T , Nemoto K , Kumabe T , Takai Y , Yamada S : Prognostic significance of surgery and radiation therapy in cases of anaplastic astrocytoma: retrospective analysis of 170 cases . J Neurosurg 2007 , 106 ( 4 ): 575 - 581 .
20. Miller CR , Dunham CP , Scheithauer BW , Perry A : Significance of necrosis in grading of oligodendroglial neoplasms: a clinicopathologic and genetic study of newly diagnosed high-grade gliomas . J Clin Oncol 2006 , 24 ( 34 ): 5419 - 5426 .
21. Prados MD , Scott C , Curran WJ Jr, Nelson DF , Leibel S , Kramer S : Procarbazine, lomustine, and vincristine (PCV) chemotherapy for anaplastic astrocytoma: A retrospective review of radiation therapy oncology group protocols comparing survival with carmustine or PCV adjuvant chemotherapy . J Clin Oncol 1999 , 17 ( 11 ): 3389 - 3395 .
22. The Medical Research Council Brain Tumor Working Party: 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 ( 2 ): 509 - 518 .
23. Celli P , Nofrone I , Palma L , Cantore G , Fortuna A : Cerebral oligodendroglioma: prognostic factors and life history . Neurosurgery 1994 , 35 ( 6 ): 1018 - 1034 . discussion 1034 - 1015
24. Nijjar TS , Simpson WJ , Gadalla T , McCartney M : Oligodendroglioma . The Princess Margaret Hospital experience (1958- 1984) . Cancer 1993 , 71 ( 12 ): 4002 - 4006 .
25. Gulbas H , Erkal HS , Serin M : The use of recursive partitioning analysis grouping in patients with brain metastases from non-small-cell lung cancer . Jpn J Clin Oncol 2006 , 36 ( 4 ): 193 - 196 .
26. Gaspar L , Scott C , Rotman M , Asbell S , Phillips T , Wasserman T , McKenna WG , Byhardt R : Recursive partitioning analysis (RPA) of prognostic factors in three Radiation Therapy Oncology Group (RTOG) brain metastases trials . Int J Radiat Oncol Biol Phys 1997 , 37 ( 4 ): 745 - 751 .
27. Mirimanoff RO , Gorlia T , Mason W , Bent MJ Van den, Kortmann RD , Fisher B , Reni M , Brandes AA , Curschmann J , Villa S , Cairncross G , Allgeier A , Lacombe D , Stupp R : Radiotherapy and temozolomide for newly diagnosed glioblastoma: recursive partitioning analysis of the EORTC 26981/22981-NCIC CE3 phase III randomized trial . J Clin Oncol 2006 , 24 ( 16 ): 2563 - 2569 .
28. Chow E , Abdolell M , Panzarella T , Harris K , Bezjak A , Warde P , Tannock I : Recursive partitioning analysis of prognostic factors for survival in patients with advanced cancer . Int J Radiat Oncol Biol Phys 2009 , 73 ( 4 ): 1169 - 1176 .