The clinical evolution of anaplastic glioma (anaplastic astrocytoma, oligodendroglioma, and oligoastrocytoma) is variable. Previous studies merged patients with anaplastic glioma and the much more common glioblastoma multiforme. Therefore, the conclusions on prognostic factors reflected in part the consequences of an analysis in a heterogeneous population.
To identify clinical, neuroradiologic, pathologic, and molecular factors with prognostic significance, we analyzed 95 treated patients with a histologic diagnosis of anaplastic glioma. Variables included age, gender, clinical manifestations at diagnosis (seizures, focal neurologic deficit, and cognitive changes), computed tomographic (CT) scan characteristics (diffuse, ring, and no enhancement), tumor location, extent of resection, histopathology, postoperative Karnofsky performance status (KPS) score, adjuvant chemotherapy, tumor response, proliferation index (Ki-67 expression), and p53, p16, pRb, and epidermal growth factor receptor immunohistochemical expression.
Ninety-five patients with a histologic diagnosis of anaplastic astrocytoma (73%), anaplastic oligoastrocytoma (16.6%), or anaplastic oligodendroglioma (10.4%) constituted the basis of this study. Median overall survival was 29 months. Multivariate analysis revealed that an age of 49 years or younger (P < 0.03), postoperative KPS score of 80 or higher (P < 0.007), absence of ring enhancement (P = 0.03), and a proliferation index of 5.1% or lower (P = 0.044) were independently associated with longer survival. The presence of an oligodendroglial component was associated with better prognosis in the univariate analysis (P = 0.009), although this lost power in the multivariate analysis.
Diffuse gliomas are one of the most common primary intracerebral neoplasms and they are classified according to their histologic characteristics. Whereas glioblastoma multiforme (World Health Organization [WHO] Grade IV astrocytoma) is associated with uniform poor survival rates, the clinical outcome varies considerably among patients with anaplastic glioma, ranging from rapid progression to prolonged survival.1–3 The most consistent prognostic variables in malignant gliomas are patient age, Karnofsky performance status (KPS) score, tumor grade, and treatment (extent of resection and postoperative radiation therapy).2, 4–7 However, clinical parameters do not fully account for the observed variation in survival rates. Therefore, additional indicators are needed to more accurately determine the prognosis of patients with anaplastic gliomas. Recent studies suggested that some biologic parameters, such as gene alterations and proliferative status, might also play a role in the prognosis of high-grade gliomas.3, 8–11 In addition, histologic parameters other than grade, such as the presence of an oligodendroglial component in anaplastic astrocytomas, also predict longer survival.12–14
Most of the above-mentioned studies referred to general series that mainly included glioblastoma multiformes, which makes it difficult to extrapolate the conclusions to the minority of patients with anaplastic gliomas. In fact, studies specifically focused on anaplastic glioma are limited to a few series.1, 3, 11, 13, 15 Most of them analyzed only clinical parameters or molecular factors involved in the tumor process. In addition, neuroradiologic features, as well as the molecules implicated in tumor proliferation (e.g., p16),16 have never been evaluated in this subset of patients.
In the current study, 95 patients with a histopathologic diagnosis of anaplastic glioma were analyzed to identify clinical, neuroradiologic, pathologic, and molecular parameters with prognostic significance for survival.
MATERIALS AND METHODS
This study comprised 95 consecutive adult patients with anaplastic glioma (WHO Grade III). They were diagnosed and treated between January 1993 and December 1999 at the Hospital Clínic de Barcelona and the Ciutat Sanitària i Universitària de Bellvitge. All the patients had histopathologic confirmation of anaplastic astrocytoma, anaplastic oligoastrocytoma, or anaplastic oligodendroglioma. Initial consensus meetings were held at the onset of the study before the independent review of the samples. In addition, histopathologic diagnosis was reevaluated by two independent neuropathologists (IF and TR) and classified according to the 2000 WHO classification.17 Cases with discrepancies were re-reviewed by both pathologists until a consensus was reached. A diagnosis of pure anaplastic astrocytoma was made when two or more of the following features were present: nuclear atypia, mitotic activity, or minimal endothelial proliferation. Mixed anaplastic gliomas were identified when tumors contained anaplastic features and geographically distinct astrocytic and oligodendroglial components.
All patients were managed according to a previously established common diagnostic and therapeutic protocol. This protocol included adjuvant radiation therapy (RT) and chemotherapy after surgical resection. The RT consisted of focal cranial irradiation with a margin on the order of 2–3 cm surrounding the tumor volume. A high-energy and a rigid immobilization system was used and a dose of 60 Gy was administered in standard daily fractions of 2 Gy. Chemotherapy consisted of either carmustine (BCNU; 200 mg/m2 every 8 weeks × six cycles) or the PCV regimen (procarbazine 60 mg/m2 on Days 8–21, CCNU 110 mg/m2 on Day 1, and vincristine 1.4 mg/m2 on Days 8 and 29) every 6 weeks × six cycles.
Neuroradiologic images were reviewed at the time of the current study. Because some patients did not receive preoperative magnetic resonance imaging (MRI) scans, only computed tomographic (CT) scans were analyzed. The absence or presence of diffuse (nonring) and ring enhancement was evaluated in the initial postcontrast CT scan. The extent of surgical resection was quantified on the postcontrast CT scan performed up to the fifth day after surgery. In patients without tumor enhancement, the grade of resection was assessed according to surgical opinion and the results of the CT and MRI scans performed after surgery. The KPS score was recorded after surgery at the time RT was initiated.
After treatment, all patients were followed in the outpatient clinic until death or last visit. Follow-up consisted of medical history, physical examination, and cranial MRI scan every 4 months or when a tumor recurrence was suspected. Demographic, clinical, therapeutic, and follow-up data were obtained from the database introduced in both medical centers in 1993.
Immunohistochemical studies for p53, p16, pRb, epidermal growth factor receptor (EGFR), and Ki-67 antibodies were done in 76 of 95 patients (80%). This subgroup of patients was representative of the whole series with no differences in their clinical and pathologic characteristics (data not shown). Immunohistochemistry was carried out following the streptavidin-biotin-peroxidase (LSAB) method (Dako, Carpinteria, CA). Briefly, 6-μm sections were obtained. After blocking endogenous peroxidase was added, the sections were boiled in citrate buffer (for p53, p16, pRb, and Ki-67 antibodies). For EGFR antibody, sections were previously incubated with pepsin reagent (Biomeda, Foster City, CA). All sections were then incubated with normal serum at room temperature for 2 hours and with one of the primary antibodies at 4 °C overnight. The mouse monoclonal antibody (MoAb) p53 (pAb1801; Oncogene Science, Manhasset, NY) was diluted 1:100. The mouse MoAb p16 (Ab-4; NeoMarkers, BioNova, Madrid, Spain) was used at a dilution of 1:50. The mouse pRb antibody (G3-245; Pharmingen, San Diego, CA) was diluted 1:500. The rabbit polyclonal EGFR antibody (Ab-4, Oncogene Sciences) and the mouse Ki-67 antibody (Dako) were used at a dilution of 1:25. The sections were then incubated with biotinylated antimouse and antirabbit IgG secondary antibody for 10 minutes, followed by LSAB for 10 minutes at room temperature. The peroxidase reaction was visualized with 0.05% diaminobenzidine (Sigma, St. Louis, MO) and 0.01% hydrogen peroxide. Sections were slightly counterstained with hematoxylin. Human colon carcinoma was used as a positive control for p53 immunohistochemistry and human skin tissue was used as positive control for EGFR immunohistochemical studies. Tonsil tissue served as a positive control for p16, pRb, and Ki-67 immunohistochemistry. Negative controls were performed by omitting the primary antibody. Both positive and negative controls were included in each assay.
Each slide stained for p53, p16, pRb, EGFR, and Ki-67 was individually reviewed and scored by two independent observers (AT and MB). Discrepancies in scoring between the two observers were resolved by additional review of the specimens and discussion between the reviewers until a consensus was achieved. Approximately 15–20 fields at × 400 magnification were analyzed per specimen. Scoring for p53 and p16 was done on a 4-point scale from 0 to 3. A score of 0 indicated no or rare occurrence of stained nuclei, 1 indicated that 1–5% of cells had positive staining, 2 indicated that 5–50% of cells stained positively, and 3 indicated that greater than 50% of cells had positive staining. For purposes of statistical analysis, a score of 0 was negative, whereas scores of 1, 2, and 3 were condensed to overexpression for p53 protein and normal immunoreactivity for p16 protein. Epidermal growth factor receptor and pRb immunohistochemistry was viewed as positive when tumor cells demonstrated cytoplasmic and plasma membrane immunoreactivity, and negative when no stained cells were present.
Proliferation index was evaluated by Ki-67 immunohistochemistry. Ki-67 scoring was accomplished by determining the percentage of positive nuclei from regions of maximal nuclear staining after counting 1000 tumor cells, or as many cells as possible in the case of small specimens at × 400 magnification. Cells were counted as Ki-67 positive if diffuse nuclear staining was present.
All immunohistochemical analyses were carried out blind to the clinical information.
The end point of this study was overall survival, which was measured from the date of surgical resection that confirmed the diagnosis of anaplastic glioma to the last follow-up visit or death. The univariate analysis was done by constructing probability curves according to the Kaplan–Meier method and comparing them by the log rank test. Twenty variables were included in this analysis: age, gender, clinical manifestations at diagnosis (seizures, focal neurologic deficit, and cognitive changes), neuroradiologic characteristics (diffuse, ring, and no enhancement), tumor location, extent of resection, hospital stay, histopathology, KPS score, adjuvant chemotherapy, tumor response, proliferation index (Ki-67 expression), and p53, p16, pRb, and EGFR immunohistochemical expression. For continuous variables, the cutoff level chosen was their median value. Variables achieving a P value of less than 0.05 in the univariate analysis were subsequently introduced in a multivariate stepwise proportional hazards regression analysis (Cox model) to identify the parameters independently associated with survival. Univariate and multivariate analyses were performed for both the whole series (n = 95) and the subset of patients for whom immunohistochemistry data were available (n = 76).
Statistical analysis was done in April 2001. All calculations were performed using the SPSS 9.0 software package (SPSS, Chicago, IL). Quantitative variables and length of follow-up were expressed as median and range.
Characteristics of Patients
The main clinical characteristics of the 95 patients are summarized in Table 1. Fifty-nine patients were men and the median age of the patients was 49 years. Contrast enhancement on CT scan was observed in 63 patients, 36 of whom had a ring pattern. Characteristics of patients according to their radiographic pattern are summarized in Table 2. The histologic diagnosis was anaplastic astrocytoma in 69 patients (72%), anaplastic oligoastrocytoma in 16 patients (17%), and anaplastic oligodendroglioma in 10 patients (11%). Concordance between both pathologists was higher than 90%. No differences in age, gender, symptoms at presentation, tumor location, extent of surgery, and KPS score were observed between tumors with and without an oligodendroglial component, except for ring enhancement, which was more frequently observed in pure anaplastic astrocytomas (P = 0.04).
Table 1. Characteristics of Patients with Anaplastic Glioma
For p16 immunohistochemistry data: (a) negative: no immunostaining; b) positive: normal expression. These figures reach statistical significance when comparing contrast ring enhancement with the other two (P = 0.01).
For p53 immunohistochemistry data: (a) normal: no protein accumulation; b) overexpression: protein accumulation.
For EGFR immunohistochemistry data: a) negative: no protein accumulation; b) overexpression: protein accumulation.
Eighty-seven patients received adjuvant RT. Of the remaining eight patients four did not receive RT because of a low postoperative KPS score and four had received previous RT for a low-grade astrocytoma. Seventy-six patients received adjuvant chemotherapy. Patients with anaplastic oligodendroglioma were treated with PCV (n = 7). Most patients with anaplastic astrocytoma who received chemotherapy were treated with BCNU (n = 44), except 9 who received PCV and 3 patients who were treated with carboplatin and cyclophosphamide as part of a therapeutic trial.18 Patients with mixed glioma were treated with either BCNU (n = 7) or PCV (n = 6). Nineteen patients did not receive adjuvant chemotherapy because of medical conditions or age (n = 13), rejection of treatment (n = 2), or inclusion in specific protocols (n = 4).
p53 overexpression was detected in 43 of 77 patients (56%) and was more frequent among patients with anaplastic astrocytomas than among patients with gliomas with an oligodendroglial component (66% vs. 30%, P = 0.004; Fig. 1A). Immunoreactivity to p16 was undetectable in 23 patients (30%). Absence of p16 immunoreactivity was found in 35% of patients with anaplastic astrocytomas and in 15% of patients with anaplastic oligodendrogliomas or oligoastrocytomas (P = 0.08; Fig. 1B). Overexpression of EGFR was detected in 25% of patients. All of them had anaplastic astrocytoma (Fig. 1C). Finally, lack of pRb expression was observed in 30% of patients, with no differences among histologic subtypes (Fig. 1D).
Median proliferation index measured by labeling of Ki-67 was 5.1% (range, 0.3–56%). This parameter was significantly associated with p53 overexpression (p53 overexpression, 10.8 ± 11.6% vs. p53 negative expression, 4.7 ± 5.2%; P = 0.003), whereas no correlation was observed for the other molecular markers (p16, pRb, and EGFR).
Predictive Factors of Survival
At the time of analysis, 38 of 95 patients (40%) remained alive, with a median follow-up of 52 months (range, 7–82 months). Four patients were lost to follow-up. The median overall survival was 29 months, with a 2 and 5-year probability of survival of 53% (95% confidence interval [CI]: 63.3–42.6) and 38% (95%CI: 48.9–27), respectively (Fig. 2).
Histologic diagnosis had significant influence on survival when all three subtypes were independently considered (median survival was 16.7 [95% CI: 11.8–21.5] months for patients with anaplastic astrocytoma, 37.1 [95% CI: 20.2–53.1] months for patients with anaplastic oligodendroglioma, and not reached for patients with anaplastic oligoastrocytoma; P = 0.009). Selective comparisons disclosed that this fact was mainly due to the differences between anaplastic astrocytoma and anaplastic oligoastrocytoma (P = 0.01), whereas no significant differences were observed between anaplastic astrocytoma and anaplastic oligodendroglioma (P = 0.20) or between anaplastic oligoastrocytoma and anaplastic oligodendroglioma (P = 0.26). Considering the low prevalence of tumors with an oligodendroglial component, either pure or mixed, and the similar long-term survival periods, these two groups of patients were subsequently evaluated as a single unit. In addition, neuroradiologic characteristics also had a significant influence on survival when all three subgroups (ring, diffuse, and nonenhancement) were considered (P = 0.04). No significant differences were observed between nonenhancement and nonring enhancement groups (P = 0.4).
As shown in Table 3, age, focal neurologic deficit at diagnosis, ring enhancement on CT scan, oligodendroglial component, and postoperative KPS score were identified as variables with prognostic influence on survival in the whole series of 95 patients in the univariate analysis. Treatment with chemotherapy failed to show prognostic value. When the analysis was repeated after excluding eight patients who did not receive RT, the results were not modified.
Table 3. Variables with Predictive Value in the Univariate Analysis of Survival in the Whole Series and in the Subgroup of Patients for Whom HC Data Were Available
When these variables were introduced in the Cox regression analysis, only age 49 years or younger (P < 0.03), absence of ring enhancement (P = 0.03), and a postoperative KPS score of 80 or higher (P < 0.007) were independently associated with survival (Table 4). The presence of an oligodendroglial component showed a tendency for better survival, but it did not achieve statistical significance as an independent prognostic variable in the multivariate analysis (P = 0.052).
Table 4. Variables with Independent Prognostic Value in the Multivariate Analysis in the Whole Series and in the Subgroup of Patients for Whom IHC Data Were Available
Immunohistochemistry for 76 patients revealed that a lack of p53 immunoreactivity (P = 0.02), positive p16 immunostaining (P = 0.01), and a proliferative index less than or equal to 5.1% (P < 0.0009) had prognostic influence on survival in the univariate analysis (Table 3). Neither EGFR overexpression (P = 0.2) nor a lack of pRb staining (P = 0.4) was predictive of survival. When p16, p53, and Ki-67 data were introduced in the multivariate analysis along with the above-mentioned clinical parameters, a proliferative index of 5.1% or less was identified as an independent prognostic factor (P = 0.044), along with the absence of ring enhancement (P = 0.043) and a postoperative KPS score greater than or equal to 80 (P = 0.041; Table 4).
The influence of predictive factors on survival was further evaluated by calculating an index of probability of survival to include the four independent prognostic variables (age, ring enhancement, postoperative KPS score, and proliferative index). To calculate this index, each variable was scored as 0 or 1 according to the absence or presence of the favorable category. Patients were classified into two groups. Group 1 patients (nonfavorable group) scored 0, 1, or 2 points and Group 2 patients (favorable group) scored 3 or 4 points. Kaplan–Meier analysis confirmed that this prognostic index allowed an accurate classification of patients with respect to their probability of survival. Indeed, Group 2 patients had a probability of remaining alive almost 20 times higher than patients from Group 1 (relative risk: 19.05 [95%CI: 5.78–62.73, P < 0.0001; Fig. 3).
The results of the current study suggest that age, KPS score, and CT ring enhancement are the most consistent predictive factors of survival in patients with anaplastic gliomas. In addition, proliferation index, established by labeling of Ki-67, also had prognostic influence in this series. Finally, as previously reported,12–14 the presence of an oligodendroglial component may also constitute a favorable prognostic factor, although this factor did not achieve statistical significance (P = 0.052) in the multivariate analysis, perhaps because of the small number of patients.
As in previous studies of patients with high-grade gliomas,1, 7, 11, 13, 15, 19, 20 age and postoperative KPS score were independent predictors of survival in patients with anaplastic gliomas. In addition, the current study found that the presence of ring enhancement on the CT scan was a negative predictor of survival. This association had only been observed in patients with anaplastic oligodendrogliomas21 and had never been evaluated in patients with anaplastic astrocytomas and mixed gliomas. A potential explanation for this association may be that these tumors were in fact glioblastoma multiforme and that the histologic specimen had not reflected the true nature of the tumor. Against this possibility is the lack of correlation between the presence of ring enhancement with the extent of resection, which would amplify the possibility of sampling error,22 and the lack of association with EGFR expression, a molecular feature more typical of glioblastoma multiforme.23 However, the association of ring enhancement with poor prognosis could indicate a more malignant subtype of anaplastic glioma, as suggested by the association with no p16 immunoreactivity (P = 0.01), a feature probably linked to more aggressive tumor behavior.
A second relevant finding of our study is that the proliferation index, measured by the Ki-67 index of labeling, was a strong independent predictor of survival. Tumor proliferation has been evaluated in previous studies of anaplastic gliomas, with controversial results.3, 15, 24 Discordant results are probably explained by variations in technique and interpretation. In the current study, the correlation observed between the proliferation index and abnormal p53 status may help to explain the value of this variable as a prognostic factor.
The extent of surgical resection or treatment with chemotherapy did not have an influence on survival. Extent of resection was identified as a prognostic factor in one of the largest series of anaplastic astrocytomas,4 but this finding was not reproduced in other studies.7, 25 Similarly, the role of adjuvant chemotherapy in anaplastic gliomas is also controversial except for the subset of patients with anaplastic oligodendroglioma and alterations of chromosome arms 1p and 19q.6, 21, 26–28 However, the number of patients in our study is probably not enough to identify the potential value of these two treatment modalities on survival. Only metaanalyses of large randomized trials have shown a beneficial effect of chemotherapy in malignant gliomas.29
Many investigations have focused on glioma behavior. Their goals were to identify molecular factors correlating with histologic features and to predict clinical outcome. Unfortunately, the results have been quite poor.8, 30–33 Allelic loss of chromosomes 1p and 19q, which predicts response to chemotherapy and survival in anaplastic oligodendroglioma, constitutes a meaningful exception.21 The current study also investigated whether some genes involved in tumor differentiation, proliferation, progression, and apoptosis had any influence on the prognosis of patients with anaplastic gliomas. In the absence of p53 overexpression, normal p16 immunoreactivity was significantly associated with higher survival in the univariate analysis, but this predictive value was lost in the multivariate analysis. However, it is known that p53 and p16 protein expression do not always correlate with mutations, loss of heterozygosity, or epigenetic changes in the corresponding genes.34 Therefore, the results of the current study do not definitively rule out a potential involvement of these genes in the prognosis of patients with anaplastic glioma, especially considering their statistical significance in the univariate analysis. Further studies evaluating both p53 and p16 at the DNA level are required to elucidate their putative role.
These results suggest that young age at the time of tumor diagnosis, lack of ring enhancement on CT scan, a good postoperative KPS score, and a low proliferation index are independent predictors of survival in patients with anaplastic glioma. The identification of these predictors may be important for the design of future clinical trials.
The authors thank A. Castells for statistical support and T. Yohannan for editorial assistance.