Revisiting Anaplastic Astrocytomas I: An Expansive Growth Pattern is Associated with a Better Prognosis

Transcription

1 JOURNAL OF MAGNETIC RESONANCE IMAGING 28: (2008) Original Research Revisiting Anaplastic Astrocytomas I: An Expansive Growth Pattern is Associated with a Better Prognosis Hugh D. Moulding, MD, 1 David P. Friedman, MD, 2 Mark Curtis, MD, PhD, 3 Lawrence Kenyon, MD, PhD, 3 Adam E. Flanders, MD, 2 Sun Ha Paek, MD, PhD, 4 and David W. Andrews, MD 1, * Purpose: To study whether anaplastic astrocytomas that are nonenhancing and/or well-circumscribed (expansive) are associated with a better prognosis. Materials and Methods: We retrospectively identified 59 patients with pathologically confirmed World Health Organizaiton (WHO) grade III anaplastic astrocytoma who underwent craniotomy at our institution from 1995 through We assessed prognostic variables including age, enhancement (EAA 34 patients) vs. nonenhancement (NEAA 25 patients), MR growth patterns (expansive [28 patients] vs. mixed/infiltrative [31 patients]), recursive partitioning analysis (RPA) class, resection extent, and addition of chemotherapy. Primary outcome measure was survival. Results: Kaplan-Meier curves showed improved survival in NEAA, expansive tumors, and RPA 1 class patients. Within RPA class I patients, expansive growth pattern remained a significant advantage in survival time. Examining extent of resection also showed that patients with gross total resections (GTR) had a better prognosis. A multivariate (Cox proportional hazards) analysis showed that patient age and expansive tumor phenotype affected outcome, whereas RPA class, enhancement, and GTR did not. Conclusion: Circumscribed growth in histologically proven anaplastic astrocytoma, which has not been emphasized in past studies, has a considerable survival advantage. 1 Department of Neurological Surgery, Thomas Jefferson University, Philadelphia, Pennsylvania. 2 Department of Radiology, Thomas Jefferson University, Philadelphia, Pennsylvania. 3 Department of Pathology, Thomas Jefferson University, Philadelphia, Pennsylvania. 4 Department of Neurosurgery, Seoul National University (SHP), Seoul, Korea. *Address reprint requests to: D.W.A., Department of Neurological Surgery, Thomas Jefferson University, 909 Walnut Street, Suite 200, Philadelphia, PA david.andrews@jefferson.edu Received March 21, 2008; Accepted August 13, 2008 DOI /jmri Published online in Wiley InterScience ( Key Words: anaplastic astrocytoma; glioma; growth pattern; gadolinium enhancement prognosis J. Magn. Reson. Imaging 2008;28: Wiley-Liss, Inc. THE HISTOLOGICAL GRADES of astrocytoma defined by the World Health Organization (WHO) has been of critical value for understanding the prognosis of patients with these astrocytic tumors. For example, both WHO grade III and IV malignant astrocytomas have resisted treatment with surgery, radiation, and chemotherapy (1), with typical ranges of survival ranging from 1 to 5 years for anaplastic astrocytomas and 1 year to 18 months for glioblastomas (2 6). However, this focus on histologic criteria has overshadowed the pattern of growth. Scherer (7), and later Daumas-Duport (8), described forms of tumor growth, and discussed how these tumors vary in how they interact with surrounding parenchyma. Some tumors are expansive, primarily pushing the surrounding neuropil aside, while the more infiltrative and invasive tumors, on the other hand, show a wide margin of microscopic invasion into the surrounding parenchyma. MRI has superseded computerized tomography (CT) scans, which had replaced gross morphological consideration, for detecting and characterizing gliomas. However, these advanced techniques have not been as influential as histologic grade when determining prognosis. Still, although contrast enhancement (EAA) is widely considered a predictor of malignancy, recent studies have documented nonenhancing (NEAA) anaplastic astrocytomas, which are otherwise histologically indistinguishable from enhancing anaplastic astrocytomas (9). Whether or not lack of enhancement in anaplastic astrocytomas has an impact on survival has been a topic of debate (10 12), but one recent study noted the importance of NEAA as a prognostic indicator of survivor in WHO grade III gliomas (13). In the current analysis, we retrospectively review our experience regarding the diagnosis and treatment of 2008 Wiley-Liss, Inc. 1311

2 1312 Moulding et al. Figure 1. Example of NEAA with expansive growth pattern. (A) Gadolinium-enhanced T1- weighted axial MRI; (B) axial proton density MRI reveals same tumor volume. anaplastic astrocytomas, analyzing the relative contributions of enhancement characteristics, radiographic growth patterns, and treatment prognostic variables to the survival of these patients. PATIENTS AND METHODS Patient Population After approval of the institutional review board, clinical data were collected and analyzed in this study. Eligibility criteria included the following: all surgery for resection and histopathologic diagnoses were performed at our institution; diagnosis based on WHO criteria for grade III astrocytomas in each case with documentation of multiple mitoses, increased cellularity, and nuclear atypia; all patients received one course of radiation therapy after surgery with or without chemotherapy and followed up with contrast-enhanced brain magnetic resonance imaging (MRI) pre- and postoperatively. We excluded patients who had undergone prior surgeries for resection elsewhere; however, we did include two patients who underwent outside needle biopsy within a month of craniotomy for resection done at our institution. Also excluded were patients who had postoperative chemotherapy without radiation therapy, focal boost radiotherapy, CT imaging only, or an MRI without contrast enhancement. We also assessed patients for their overall functional status utilizing the Karnofsky Performance Scale (KPS) (14). This widely applied scale assigns a functional score (100 to 0 in increments of 10) based on a combination of neurologic signs and symptoms, independence in activities of daily living, and level or required medical services. We excluded patients with a KPS of 70, the threshold below which patients need assistance with daily living tasks. We also excluded patients who were diagnosed with WHO grade III tumors with an oligodendroglial component, as they have a separate and improved prognosis (15). A minimum of 6 months of follow-up was required. Recursive partitioning analysis (RPA) is a nonparametric statistical technique that has been utilized to identify important pretreatment prognostic variables into distinct prognostic classes. RPA class was determined for each patient (4). Pre- and Postoperative MRI, Extent of Resection All pre- and postoperative MRI scans were reviewed with a single neuroradiologist who was aware of the WHO Grade III diagnosis but otherwise was blinded to the clinical outcome and specific details of the pathology. Preoperative lesions were categorized as having either no tumor enhancement, trace/ punctuate enhancement, or obvious enhancement. We categorized no or trace enhancement as nonenhancement (NEAA), and conspicuously enhancing tumors as EAA. In our attempt to define tumor margins, we adapted glioma growth pattern classifications originally set forth as anatomic descriptions (7,8) to typical appearances we noted on MRI scans. All patients were injected with Magnevist at a standard dose of 0.1 mmol/kg, and all patients were scanned in 1.5 T magnets. The T1- and T2-weighted images were standard spin-echo pulse sequences (5-mm slice thickness with a 2.5-mm gap). The FLAIR images were a 5-mm slice thickness with a 1-mm gap. We used MPRAGE images routinely on localization studies but did not use these images to determine either enhancement or growth patterns, in the latter case because we sought better image registration with images of similar slice thickness. Typical pulse sequences were as follows: T1-weighted images at TR /TE 8 12, T2-weighted images at TR /TE , and FLAIR images at TR /TE Neither spectroscopy nor perfusion studies were routinely performed because all patients were scanned for intraoperative image-guidance purposes. Also, these techniques were not available in the earlier phase of this study period. For the majority of assessments, T1-weighted images were compared to FLAIR images; earlier outside studies in some cases only included T2- weighted images for comparison. Growth type I or expansive tumors corresponded to a clearly identifiable hypointense T1-weighted nidus with or without enhancing characteristics, associated with distinct, identical margins on FLAIR or T2-weighted sequence with the same nidus volume (Fig. 1). These tumors corresponded to what had anatomically been described as expansive and sharply circumscribed growth with little adjacent brain invasion. Growth type

3 Prognosis for Grade III Glioma Subgroups 1313 Figure 2. Example of NEAA with mixed growth pattern. (A) Gadolinium-enhanced T1- weighted axial MRI; (B) axial FLAIR MRI reveals larger tumor volume. II, or mixed growth pattern, tumors combined a clearly identifiable EAA or NEAA nidus with a variable rim of FLAIR or T2-weighted sequence extending beyond the T1-weighted nidus (Fig. 2). These tumors corresponded to what had anatomically been described as a mixed expansive-infiltrative growth pattern by Daumas-Duport et al. (8). Growth type III, or infiltrative, tumors corresponded to purely infiltrating tumors without a distinct nidus, with or without contrast EAA, and widespread involvement of contiguous brain on T2- weighted, proton density, or FLAIR sequences (Fig. 3). All patients underwent MRI-based image-guided tumor resections. MRI scans were obtained within 72 h of craniotomy to assess any residual area of contrast enhancement (16). Although NEAA preoperatively, gadolinium was routinely administered for postoperative NEAA patients as well. Extent of resection was designated according to modifications of a previously published series (17). For EAA, gross total resection (GTR) was defined as no residual enhancement on gadolinium-enhanced T1-weighted postoperative MRI scans. For either EAA or NEAA, GTR also involved resection of more than 95% of the preoperative T1-weighted hypointense/t2-weighted hyperintense tumor volume. Subtotal resection (STR) removed 85% to 95% of preoperative T2-weighted tumor volume. Patients with partial resections (PR) had 85% reduction in T2-weighted volume after surgery. In cases where tumors were reresected, tumors were by convention not restaged by WHO criteria due to the artifacts created by radiation. Histopathologic review with quantitation of proliferative index All specimens were analyzed by one of two neuropathologists and included interpretations of permanent sections stained with hematoxylin and eosin and cut at 4 thickness. For all primary brain tumor specimens, histopathologic review routinely included the processing and histologic evaluation of all tissue received. Fragmented tissue did not create artifacts that would affect grading of the tumor. As mentioned above, recurrent tumors were not restaged due to artifacts created by conformal radiation. All specimens were rereviewed by one neuropathologist according to the revised WHO 2000 criteria (18), which have not been altered in the recently published 2007 criteria (19). The presence of necrosis or microvascular proliferation was interpreted as a WHO Grade IV lesion and was excluded from fur- Figure 3. Example of NEAA with infiltrative growth pattern. (A) Gadolinium-enhanced T1-weighted axial MRI; (B) axial T2-weighted MRI reveals broader, infiltrating tumor volume.

4 1314 Moulding et al. ther analysis. As generally practiced at our institution, we established more than three mitoses/mm 2 or approximately three mitoses in 10 microscopic high-powered fields (400 magnification) as a threshold value for a grade III diagnosis. Patients with less than three mitoses/mm 2 were excluded, as such infrequent numbers of mitoses have been shown to be closer in prognosis to WHO grade II gliomas if no other WHO criteria were evident (18,20). As part of a routine assessment, immunostaining was performed with Kiel antibody number 67 (Ki-67) monoclonal antibody. To determine proliferative activity, immunohistochemical estimation of the Ki-67 labeling index (LI) was performed and analyzed. Statistical Analyses All data were entered into a statistical spreadsheet (JMP IN 5.1). Group comparisons were performed as appropriate with a two-tailed t-test (unequal variances), chi-square contingency square analysis, or Cox proportional hazard modeling. Multivariate analysis was performed using variables that were significant during bivariate analysis. Significance was established for all testing at P RESULTS Patient Characteristics, Histology, and Surgical Outcomes After institutional review board approval, we identified 66 WHO grade III astrocytomas operated between September 1995 and March of 2006, 59 of whom met eligibility criteria. All were primary malignancies without antecedent history of progression from a lower grade glioma. Those excluded were one patient for a tumor with oligodendoglial components and two patients with tumors manifesting necrosis on pathology rereview. One patient had a previous outside hospital resection, and three patients were excluded for insufficient or incomplete follow-up. Patient characteristics are featured in Tables 1 and 2. By WHO 2000 criteria (19,21), the most frequent two features that established the diagnosis of grade III tumor included pleomorphism and presence of multiple mitoses. Permanent section diagnosis revealed characteristic features of anaplastic astrocytoma for all cases analyzed. Twenty-five patients were identified with preoperative MRI scans that revealed either no or trace preoperative gadolinium tumor enhancement, and were grouped as NEAA, while 34 patients with conspicuously enhancing tumors with were assigned to EAA (Table 1). All NEAA patients were operated without further observation, except for one patient who presented with visual scintillations and was followed for 5 years prior to craniotomy, which was recommended when the tumor revealed interval growth. We also categorized tumors according to preoperative MRI growth pattern appearances and noted 28 growth type I, or expansive tumors, 31 growth type II/III, or mixed/infiltrating tumors (Table 2). Representative cases of NEAAs exhibiting each growth pattern are featured in Figures 1 3. It should be noted, as well, that striking intratumoral regional heterogeneity was noted in some cases as shown in Figure 4. Of interest, 20 of 28 (71.4%) growth type I tumors were NEAA, while only 5 of 31 or 16.1% of growth type II and III tumors were NEAA. Conversely, 8 of 28 (28.6%) growth type I tumors were EAA, while 26 of 31 or 83.9% of growth type II and III tumors were EAA. These proportions were significantly different (P ). Other patient characteristics reflected tumor groupings. EAA vs. NEAA mean ages approached significance (48.3 for EAA and 40.8 for NEAA, P ), while the expansive tumor vs. mixed/infiltrative group ages were significantly different (40.8 vs. 49.0, P ). Ki-67 index mean was significantly different between NEAA and EAA (7.78% vs %, P ), but was not significantly different between expansive vs. mixed/infiltrative tumors (8.81% vs %, P ). Extent of resection was significantly different between the expansive and mixed/infiltrative groups (P ), but not between the two enhancement groups (P ). No significant differences were seen for gender, RPA class, tumor location, tumor laterality, treatment with chemotherapy, and later reoperation for recurrence. All patients received one course of radiation therapy after an initial surgical resection, and 37 patients had additional treatment with chemotherapy. Chemotherapy was not systematically administered according to a protocol, but rather as the result of individual clinical judgment of the treating neurooncologist. The types of chemotherapy administered are delineated in Tables 1 and 2. Survival Analysis Median survival time (MST) for the entire cohort of patients was 355 weeks. Analysis of subgroups revealed significant differences between NEAA (MST not reached) vs. EAA (MST 63 weeks) (P , Cox model; Fig. 5A). Also significantly different was the survival of patients with expansive vs. mixed/infiltrative tumors (MST not reached vs. MST of 57 weeks, P , Cox model; Fig. 5B). When the NEAA, EAA, expansive, and mixed/infiltrative traits were used to further subdivide patients into one of four cohorts (Fig. 5C), there was statistical significance between the groups (Cox model, P ). The MST for mixed/ infiltrative EAA was 61 weeks, for mixed/infiltrative NEAA it was 320 weeks, and for both expansive AA subtypes the MST was not reached. When these subgroups were compared, the NEAA expansive group was significantly different than the EAA mixed/expansive group (P , Cox model). Also significant were the NEAA expansive group vs. the NEAA mixed/ infiltrative group (P , Cox model), and the EAA expansive vs. EAA mixed/infiltrative group (P , Cox model). No other subcomparisons were significant. RPA class I patients had a significant survival advantage when compared to all other RPA class patients with a MST not yet reached compared to an MST of 45 weeks (P , Cox model; Fig. 6A). When each RPA class was individually compared, RPA class I patients had a survival advantage against RPA class II patients (P , Cox model) and RPA class IV patients (P

5 Prognosis for Grade III Glioma Subgroups 1315 Table 1 Patient Characteristics by Tumor Enhancement NEAA EAA All P-value N (%) 25 (42.4%) 34 (57.6%) 59(100 %) Age: mean SD median Gender Male Female Mean follow-up in weeks * Range (min. of 6 m F/U for survivors) RPA class I II III IV MRI growth pattern < Type I (expansive) Type II/III(mixed/infiltrative) Tumor location Frontal Temporal Parietal Occipital Cerebellar Thalamus Tumor laterality Right Left Midline Extent of resection GTR STR PR Chemotherapy Yes Temodar i.v. BCNU Gliadel PCV NS No Later reoperation Yes No Ki-67 Index mean SD % % % Ki-67 Index median (available for 55 pts) 6.70% 10.00% 7.88% Ki-67 Kiel antibody number 67; pts patients; EAA enhancement; NEAA nonenhancement; GTR gross total resections; RPA recursive partitioning analysis; PR partial resection; F/U follow-up. *P-values in bold are statistically significant , Cox model). There was only one RPA class III patient, nullifying any comparisons to other RPA classes. No other RPA class survival lengths were significantly different from one another (Cox models). The survival advantage seen in RPA class I patients, which are, by definition, younger patients, as well as the well-documented survival advantage for younger patients with malignant gliomas (10,22), created concern that the survival advantage seen in patients with NEAA tumors (Fig. 5A) or expansive tumors (Fig. 5B) was strictly a feature of age. One way we attempted to address this was to isolate all RPA class I patients (which by definition are less than 50 years old) and reexamine the variables of enhancement and growth pattern (the other being a multivariate analysis, see below). Kaplan-Meier curves for this reanalysis are shown in Figures 6B and C. Enhancement status was approached but was not found to be significant for RPA class I patients analyzed in this way (Fig. 6B, P , Cox model), whereas growth pattern remained a significant factor for RPA class I patients (Fig. 6C, P , Cox model). Extent of resection was significantly associated with survival for the entire cohort (Fig. 7A, P , Cox model). MST for GTR has not been reached, the MST for STR was 355 weeks, and for partial resection was 32 weeks. Within the patients with NEAA, resection extent was not a significant factor in survival (Fig. 7B, n 25, P , Cox model), but for EAA, resection extent was significant (Fig. 7C, N 34, P , Cox

6 1316 Moulding et al. Table 2 Patient Characteristics by Tumor Growth Pattern Type I (expansive) Type II/III (mixed/infiltrative) P-value N (%) 28 (47.5%) 31 (52.5%) Age: mean SD * median Gender Male Female Follow-up in weeks (mean) < Median not reached 57 Range (min of 6 m F/U for survivors) RPA class I II 4 6 III 0 1 IV 3 10 Tumor location Frontal Temporal 8 12 Parietal 4 4 Occipital 0 2 Cerebellar 1 1 Thalamus 0 1 Tumor laterality Right Left Midline 1 0 Extent of resection GTR 11 8 STR 15 9 PR 2 14 Chemotherapy Yes Temodar 13 9 i.v. BCNU 1 3 Gliadel 1 1 PCV 3 4 NS 0 2 No Later reoperation Yes 5 12 No Ki-67 Index mean SD % % Ki-67 Index median (available for 56 pts) 6.73% 8.88% See Table 1 for abbreviations. *P-values in bold are statistically significant. model). Within the patients with expansive tumors, resection extent was not significant (Fig. 7D, n 28, P , Cox model), but for mixed/infiltrative growth pattern tumors, resection extent was significant for survival (Fig. 7E, n 31, P , Cox model). Within patients with GTR, expansive tumor phenotype showed a survival advantage that approached significance (Fig. 7F, n 19, P ). This survival advantage was more pronounced in those patients that underwent a STR (Fig. 7G, n 24, P ). Chemotherapy did not add prognostic significance between any subgroups not already discussed above, nor did it create significance within radiographic subgroups (data not shown). Multivariate analysis (MVA) of prognostic, treatment variables, and tumor radiographic features is summarized in Table 3. Of interest, the most significant variable was patient age (P 0.001), followed by expansive tumor type (P 0.042). To determine if age obscured the variable of RPA class within the multivariate analysis (as they very closely coincide), a separate MVA without age was performed. In this model, RPA class I was significant (P 0.020), and expansive growth pattern also remained significant (P 0.022). No other changes in significance occurred in this additional MVA. Variables that were not significant in either multivariate analysis were MRI enhancement, Ki-67 index, and resection extent. DISCUSSION Historically, the prognosis for gliomas has been based largely on histologic criteria. The introduction of the Kernohan grading system in 1949 marked the beginning of an era in which the classification of central nervous system (CNS) tumors was a reflection of bio-

7 Prognosis for Grade III Glioma Subgroups 1317 Figure 4. Hematoxylin and eosin 400 photomicrograph of NEAA with two separate regions of resection: left panel reveals low grade astrocytoma; right panel reveals anaplastic astrocytoma with pleomorphism, hypercellularity, and mitosis (black arrow). logical behavior, and provided a basis for the development of treatment protocols (23). Since that time, there have been modifications to this grading system, most notably by the WHO, based on carefully documented correlations between histologic features and survival (24). By the WHO grading system, revised in 2000 (18) and most recently revised in 2007 (19), accretion of distinct abnormal histologic features has been utilized to assign histologic grade. This grading system has been useful in establishing a correlation between biological behavior, and ultimately, prognosis. Malignant transformation has always been regarded as the transition from grade II to grade III tumors. The crucial distinction between grade II and grade III astrocytomas is based on the detection of mitoses. A solitary mitosis is often found only after a thorough and often tedious search through multiple permanent sections (25). Giannini et al. (20) found that the prognosis of grade III tumors with solitary mitosis resembled that of grade II tumors. However, the reliance upon histologic criteria may be at the expense of anatomic, and in the modern context, radiographic, growth patterns. Scherer (7) revealed three salient patterns of glioma growth in 1940 expansive, mixed, and infiltrative each with prognostic significance. Catherine Daumas-Duport (8) later reported a similar classification scheme. In her system, type I tumors involved a tumor tissue nidus without surrounding parenchymal tumor cell infiltration; type II tumors included a nidus with parenchymal tumor cell infiltration and type III tumors included diffuse infiltration without a nidus. Examples of type I tumors included WHO grade I tumors (e.g., gangliogliomas, pilocytic astrocytomas, and xanthoastrocytomas). Type II tumors were typically malignant gliomas, but examples of low-grade gliomas were included (WHO grades II IV) and type III tumors were regarded as low- and intermediate-grade gliomas, or rarely grade IV gliomas. With these established anatomic correlations, we have used modern MR imaging techniques to adapt Scherer s and Daumas-Duport s prior anatomic classification systems into three common MR radiographic patterns of glioma growth. Drawing from Scherer s earlier classification, we use the nomenclature established by Daumas-Duport, namely, growth type I for expansive tumors, growth type II for mixed tumors, and growth type III for infiltrative tumors. Of great interest, we have noted all three patterns of growth within WHO grade III astrocytomas. When assessing preoperative MRI scans, we noted a predominant association of growth type I tumors with NEAA as opposed to growth type II/III tumors, which are more frequently seen in the EAA group (Table 1). Because they both involve some degree of radiographically demonstrable infiltration, we have grouped together the growth type II and III tumors in our analysis. Using these classifications, we identified a subgroup of patients with anaplastic astrocytomas who have a prognosis far better than expected for WHO grade III gliomas. For greater stringency, we excluded patients with a diagnosis of anaplastic astrocytoma based on less than three mitoses/mm 2. A recent analysis has also shown that grade III tumors with an oligodendroglial component have a better prognosis than pure grade III tumors (15), and mixed tumors were therefore excluded from this analysis. Although there exists other factors such as Ki-67 index, lack of tumor enhancement (13), extent of resection, and RPA class that have been previously discussed in the literature (10 12,26 28), in our study only younger patient age and expansive growth pattern are independent and significant variables that signify improved prognosis in a multivariate analysis model. This raises an interesting question: where does the transition to malignancy truly exist for astrocytomas? Classical teaching has been based on the histologic grading system, with transformation to malignancy occurring between WHO grade II to III glioma grades. However, in this study, we show a number of histologically confirmed WHO grade III astrocytomas that act more like WHO grade II tumors in regard to survival prognosis (29). What remains unclear is whether or not NEAA grade III astrocytomas yield the same prognosis as EAA grade

8 1318 Moulding et al. III tumors. Although contrast enhancement has been associated with malignancy, lack of enhancement has also been recognized as a feature of WHO grade III gliomas, but has generally been regarded as a radiographic exception. Lack of enhancement in anaplastic astrocytomas was first noted in a series of CT scans (30) and more recently with MRI scans (9,13,31,32). More recently, Tortosa et al. (10) demonstrated contrast enhancement within anaplastic astrocytoma as a negative prognostic indicator, as did a study by Brandes et al. (12). On the other hand, a recent study by Keles et al. (11) did not find preoperative gadolinium enhancement to be predictive of survival in patients with anaplastic astrocytoma, but did find that a larger burden of postoperative contrast enhancement did portend a poorer outcome. Adding to this body of literature, our study found NEEA within anaplastic astrocytomas of positive Figure 5. Kaplan-Meyer curves showing survival by radiographic characteristics. (A) Survival by presence or absence of tumor enhancement, P (B) Survival for patients with either expansive vs. mixed/infiltrative tumor growth patterns, P (C) Survival curves for the four different subgroups of enhancement and growth pattern characteristics, P Figure 6. Kaplan-Meyer curves by RPA class. (A) Survival of all patients by RPA class, P (B) Survival of RPA class I patients subgrouped by presence or absence of tumor enhancement, P (C) Survival of RPA class I patients subgrouped by expansive or mixed/infiltrative growth patterns, P

9 Prognosis for Grade III Glioma Subgroups 1319 Figure 7. Kaplan-Meyer curves by resection extent. (A) Survival of all patients grouped by resection extent, GTR gross total resection, STR subtotal resection, PR partial resection, P (B) Survival of NEAA patients grouped by resection extent, P (C) Survival of EAA patients grouped by resection extent, P (D) Survival of patients with expansive tumors grouped by resection extent, P (E) Survival of patients with mixed/infiltrative tumors grouped by resection extent, P (F) Survival of patients with GTR grouped by expansive or mixed/infiltrative growth patterns, P (G) Survival of patients with STR grouped by expansive or mixed/infiltrative growth patterns, P prognostic significance for survival, which was statistically significant by log rank, but its significance was lost in multivariate analysis. Our survival data are remarkably similar to those of Pope et al. (13), who reported a median survival of EAA patients of 99 weeks and a mean survival in the NEAA patients of 252 weeks compared with, in the current study, a median survival of 80 weeks and a mean survival of 249 weeks for the EAA and NEAA groups, respectively. Based on the regional heterogeneity of gliomas, it has been our surgical practice to proceed with aggressive resections when possible to minimize sampling error and maximize the therapeutic benefit of surgery. In an analysis by Coons et al. (33), a systematic examination of 353 regions in 18 gliomas was performed. Of note, 80% of regions assessed in grade III tumors revealed anaplastic features, whereas 20% revealed only lowgrade features, and 44% of regions in diagnosed grade IV tumors fulfilled WHO criteria, whereas 56% of regions revealed anaplastic features. In our series, an example of grade II III regional heterogeneities is featured in patient 16 (Fig. 4). Although we did not see grade III IV heterogeneities, these potential regional heterogeneities raise the concern that the survival advantage seen with the favorable radiographic subgroups may simply reflect a comparison to less favorable groups in which grade IV tumors are assigned a grade III level due to sampling error. Although this concern is impossible to refute, our review reflects a median survival of 80 weeks for all enhancing tumors, which is approximately 30% longer than the median survival of published outcomes for grade IV tumors (34). Another treatment variable that was significant in our bivariate analysis was the extent of the surgical resection. This has been shown in previous studies to be significant for anaplastic astrocytoma survival (11,16,35 37). Our review also reflected comparable aggressive resections in 10 of 34 EAA cases vs. NEAA cases (9 of 25) and 8 of 31 mixed/infiltrative tumors vs. expansive tumors (11 of 28) (Table 1). If we assume that GTR yields the most accurate histologic grading and a better prognosis, we still noted a survival advantage in the expansive group compared to the mixed/infiltrative group, which approached statistical significance when the GTR subgroup was analyzed (P , log rank). Also, with a median follow-up of 342 weeks, the EAA subgroup with GTR has not reached a median survival (N 10; Fig. 7c), which compares favorably to a median survival of 13.6 months in a comparable cohort of grade IV tumors with GTR (38). Our results also show that gross total resection was more likely with expansive tumor phenotypes. Although it seems intuitive that an expansive tumor would be more amenable to complete surgical resection, it also raises the question as to which is the significant factor impacting survival. From our multivariate analysis, extent of tumor resection did not remain significant, while tumor growth pattern did. When the data were divided into GTR and STR subgroups (Fig. 7f and g), we still noted a survival advantage in the expansive group. The RPA classes used were based on a statistical technique called recursive partitioning analysis (4): among 26 pretreatment characteristics identified in

10 1320 Moulding et al. Figure 7 (Continued) 1578 patients from three previous RTOG trials, age of 50 and Karnofsky performance status of 70 were important variables defining RPA classes as homogeneous survival groups. When we assessed survival by RPA class, we noted a significant survival difference between RPA class 1 and class 2 and/or class 4 assignments. Interestingly, RPA class itself was not significant in multivariate analysis, although this may be due to close similarities between RPA class and age. When reanalyzed without age as a variable, RPA class was significant on multivariate analysis. Of note, however, Table 3 Multivariate Analysis of Variables Affecting Survival for Entire Cohort Variable Chi-square P-value Hazards ratio Age 0.001* Expansive growth pattern Average Ki-67 index Resection extent GTR vs. STR STR vs. PR No enhancement on MRI RPA class I See Table 1 for abbreviations. *P-values in bold are statistically significant. is that in either model, expansive tumor growth pattern remained significant. Tumor cell growth and invasion implies complex interactions between malignant cells, glial cells, neuronal cells, and endothelial cells. Future studies should explore molecular mechanisms related to glioma invasiveness or contrast enhancement, perhaps using microarrays to examine large scale gene expression of these radiographic phenotypes (39). In summary, these data support and reaffirm the prognostic importance of NEAA as a prognostic variable in grade III gliomas. The more important variable, however, appears to be an expansive growth pattern, which we have noted is often associated with NEAA radiographic phenotypes, which may be linked by common or similar patterns of gene expression. If supported by further study, these findings could require a modification of the WHO grading system to include subclassifications of grade III tumors based on growth characteristics. The transition to malignancy may actually occur between subgroups in grade III, not in the histologic transition from grade II to grade III criteria. ACKNOWLEDGMENT The authors wish to thank Mitchell Maltenfort, PhD, for his assistance with statistical analysis.

11 Prognosis for Grade III Glioma Subgroups 1321 REFERENCES 1. Andrews DW, Resnicoff M, Flanders AE, et al. Results of a pilot study involving the use of an antisense oligodeoxynucleotide directed against the insulin-like growth factor type I receptor in malignant astrocytomas. J Clin Oncol 2001;19: Burger PC, Vogel FS, Green SB, Strike TA. Glioblastoma multiforme and anaplastic astrocytoma. Pathologic criteria and prognostic implications. Cancer 1985;56: Fischbach AJ, Martz KL, Nelson JS, et al. Long-term survival in treated anaplastic astrocytomas. A report of combined RTOG/ ECOG studies. Am J Clin Oncol 1991;14: Curran WJ Jr, Scott CB, Horton J, et al. Recursive partitioning analysis of prognostic factors in three Radiation Therapy Oncology Group malignant glioma trials. J Natl Cancer Inst 1993;85: Lacroix M, Abi-Said D, Fourney DR, et al. A multivariate analysis of 416 patients with glioblastoma multiforme: prognosis, extent of resection, and survival. J Neurosurg 2001;95: Laws ER, Parney IF, Huang W, et al. Survival following surgery and prognostic factors for recently diagnosed malignant glioma: data from the Glioma Outcomes Project. J Neurosurg 2003;99: Scherer S. The forms of growth in glioma and their practical significance. Brain 1940;63: Daumas-Duport C, Monsaingeon V, Szenthe L, Szikla G. Serial stereotactic biopsies: a double histological code of gliomas according to malignancy and 3-D configuration, as an aid to therapeutic decision and assessment of results. Appl Neurophysiol 1982;45: Ginsberg LE, Fuller GN, Hashmi M, Leeds NE, Schomer DF. The significance of lack of MR contrast enhancement of supratentorial brain tumors in adults: histopathological evaluation of a series. Surg Neurol 1998;49: Tortosa A, Vinolas N, Villa S, et al. Prognostic implication of clinical, radiologic, and pathologic features in patients with anaplastic gliomas. Cancer 2003;97: Keles GE, Chang EF, Lamborn KR, et al. Volumetric extent of resection and residual contrast enhancement on initial surgery as predictors of outcome in adult patients with hemispheric anaplastic astrocytoma. J Neurosurg 2006;105: Brandes AA, Nicolardi L, Tosoni A, et al. Survival following adjuvant PCV or temozolomide for anaplastic astrocytoma. Neuro-oncol 2006;8: Pope WB, Sayre J, Perlina A, Villablanca JP, Mischel PS, Cloughesy TF. MR imaging correlates of survival in patients with high-grade gliomas. AJNR Am J Neuroradiol 2005;26: Karnofsky D, Burchmal J. The clinical evaluation of chemotherapeutic agents in cancer. In: MacLeod C, ed. Evaluation of chemotherapeutic agents. Irvington: Columbia University Press, 1949; p Donahue B, Scott CB, Nelson JS, et al. Influence of an oligodendroglial component on the survival of patients with anaplastic astrocytomas: a report of Radiation Therapy Oncology Group Int J Radiat Oncol Biol Phys 1997;38: Albert FK, Forsting M, Sartor K, Adams HP, Kunze S. Early postoperative magnetic resonance imaging after resection of malignant glioma: objective evaluation of residual tumor and its influence on regrowth and prognosis. Neurosurgery 1994;34:45 60; discussion Sawaya R, Hammoud M, Schoppa D, et al. Neurosurgical outcomes in a modern series of 400 craniotomies for treatment of parenchymal tumors. Neurosurgery 1998;42: ; discussion Kleihues P, Louis DN, Scheithauer BW, et al. The WHO classification of tumors of the nervous system. J Neuropathol Exp Neurol 2002;61: ; discussion World Health Organization classification of tumours. Louis D, Ohgaki H, Weisler O, Cavenee W, editors. Lyon: International Agency for Research on Cancer, Giannini C, Scheithauer BW, Burger PC, et al. Cellular proliferation in pilocytic and diffuse astrocytomas. J Neuropathol Exp Neurol 1999;58: Kleihues P, Sobin LH. World Health Organization classification of tumors. Cancer 2000;88: Perry A, Jenkins RB, O Fallon JR, et al. Clinicopathologic study of 85 similarly treated patients with anaplastic astrocytic tumors. An analysis of DNA content (ploidy), cellular proliferation, and p53 expression. Cancer 1999;86: Gonzales M. The 2000 World Health Organization classification of tumours of the nervous system. J Clin Neurosci 2001;8: Kleihues P BP, Scheithauer BW. Histological typing of tumours of the central nervous system. International histological classification of tumours. Geneva: World Health Organization, 1993; p Sallinen PK, Sallinen SL, Helen PT, et al. Grading of diffusely infiltrating astrocytomas by quantitative histopathology, cell proliferation and image cytometric DNA analysis. Comparison of 133 tumours in the context of the WHO 1979 and WHO 1993 grading schemes. Neuropathol Appl Neurobiol 2000;26: Sneed PK, Prados MD, McDermott MW, et al. Large effect of age on the survival of patients with glioblastoma treated with radiotherapy and brachytherapy boost. Neurosurgery 1995;36: Burger PC, Green SB. Patient age, histologic features, and length of survival in patients with glioblastoma multiforme. Cancer 1987;59: Salmon I, Dewitte O, Pasteels JL, et al. Prognostic scoring in adult astrocytic tumors using patient age, histopathological grade, and DNA histogram type. J Neurosurg 1994;80: Wessels PH, Weber WE, Raven G, Ramaekers FC, Hopman AH, Twijnstra A. Supratentorial grade II astrocytoma: biological features and clinical course. Lancet Neurol 2003;2: Chamberlain MC, Murovic JA, Levin VA. Absence of contrast enhancement on CT brain scans of patients with supratentorial malignant gliomas. Neurology 1988;38: Kondziolka D, Lunsford LD, Martinez AJ. Unreliability of contemporary neurodiagnostic imaging in evaluating suspected adult supratentorial (low-grade) astrocytoma. J Neurosurg 1993;79: Mihara F, Numaguchi Y, Rothman M, Kristt D, Fiandaca M, Swallow L. Non-enhancing supratentorial malignant astrocytomas: MR features and possible mechanisms. Radiat Med 1995;13: Coons SW, Johnson PC. Regional heterogeneity in the DNA content of human gliomas. Cancer 1993;72: Galanis E, Buckner JC, Burch PA, et al. Phase II trial of nitrogen mustard, vincristine, and procarbazine in patients with recurrent glioma: North Central Cancer Treatment Group results. J Clin Oncol 1998;16: Ammirati M, Vick N, Liao YL, Ciric I, Mikhael M. Effect of the extent of surgical resection on survival and quality of life in patients with supratentorial glioblastomas and anaplastic astrocytomas. Neurosurgery 1987;21: Berger MS, Deliganis AV, Dobbins J, Keles GE. The effect of extent of resection on recurrence in patients with low grade cerebral hemisphere gliomas. Cancer 1994;74: Nazzaro JM, Neuwelt EA. The role of surgery in the management of supratentorial intermediate and high-grade astrocytomas in adults. J Neurosurg 1990;73: Hentschel SJ, Sawaya R. Optimizing outcomes with maximal surgical resection of malignant gliomas. Cancer Control 2003;10: Freije WA, Castro-Vargas FE, Fang Z, et al. Gene expression profiling of gliomas strongly predicts survival. Cancer Res 2004;64: