Prognostic value of detecting recurrent ... - Human Pathology

Human Pathology (2006) 37, 272 – 282

www.elsevier.com/locate/humpath

Prognostic value of detecting recurrent glioblastoma multiforme in surgical specimens from patients after radiotherapy: should pathology evaluation alter treatment decisions? Tarik Tihan MD, PhDa,*, Justine Barletta MDa, Ian Parney MDb, Kathleen Lamborn PhDb, Penny K. Sneed MDc, Susan Chang MDb a

Department of Pathology, University of California at San Francisco, San Francisco, CA 94143-0511, USA Department of Neurological Surgery, University of California at San Francisco, San Francisco, CA 94143-0511, USA c Department of Radiation Oncology, University of California at San Francisco, San Francisco, CA 94143-0511, USA b

Received 12 September 2005; revised 16 November 2005; accepted 18 November 2005

Keywords: Glioblastoma multiforme; Malignant glioma; Postradiotherapy; Prognosis; Radionecrosis; Radiotherapy; Recurrent glioblastoma

Summary The prognostic significance of the histologic type and grade of gliomas at initial surgery is well established, but the value of histologic findings in resections after radiotherapy is unclear. Despite this uncertainty, pathologic interpretation of specimens after radiotherapy influences immediate treatment decisions. It is important to determine if, and to what extent, treatment decisions should be based on this information. We aimed to determine the prognostic value of pathologic evaluation in postradiation specimens from 54 patients with similar clinical features who underwent a second surgery for the treatment of radiologic worsening after external beam radiotherapy. We categorized the specimens from the second surgery as either recurrent tumor (category 1) or radionecrosis (category 2). Patients in category 1 had actively proliferating neoplasms with classical features of glioblastoma, whereas patients in category 2 had no evidence of tumor in their surgical specimens. Cases in which a clear-cut definition could not be made were labeled indeterminate (category 3). Despite the morphological evidence of tumor, there were no significant differences between categories 1 and 2 in any of the survival parameters tested. The only difference between groups was higher frequency of iodine 125 (125I) placement at second surgery in category 1 patients ( P b .028). Patients in category 1 with or without 125I treatment had similar survival characteristics. We conclude that histopathologic evaluation of postradiotherapy specimens was not helpful in predicting outcome or dictating further management. A comprehensive prospective study with advanced radiologic, pathologic, and molecular analyses may be more useful to determine prognostically valuable parameters. D 2006 Elsevier Inc. All rights reserved.

* Corresponding author. Department of Pathology, University of California at San Francisco Medical Center, San Francisco, CA 94143-0511, USA. E-mail address: [email protected] (T. Tihan). 0046-8177/$ – see front matter D 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.humpath.2005.11.010

Prognostic value of detecting recurrent glioblastoma multiforme

1. Introduction Glioblastoma multiforme (GBM) is the most common malignant glioma and is typically treated with surgery and subsequent radiation treatment. Because of its infiltrative nature, GBM is often incurable by surgery alone. Although radiation treatment prolongs survival of patients with GBM, it is not curative either [1,2]. Almost all patients with GBM have a progression of their disease after initial surgery and radiation treatment, and this is often reflected by the worsening of the radiologic appearance [3]. This worsening after radiation treatment can be due to tumor regrowth that is pathologically recognized as recurrent tumor, but it can also be due to radiation-induced injury, in other words, radionecrosis [4-7]. The symptoms of radionecrosis are similar to those of the recurrent tumor [8]. Radionecrosis is also difficult to distinguish from recurrence by conventional radioimaging techniques [4,9]. Recent studies reported that positron emission tomography [10-13] and magnetic reso-

Table 1

273

nance spectroscopy [14,15] may be useful in differentiating recurrent tumor from radionecrosis, but these modalities still fail to provide a clear-cut answer. Therefore, a second surgery to remove the radiologically abnormal tissue is often indicated to differentiate recurrent tumor from radionecrosis. In such cases, the surgical pathologist is asked to determine the cause of radiologic worsening as either recurrent tumor or radionecrosis. Subsequent treatment decisions are often influenced by this evaluation. Patients with recurrent tumor are often treated with further therapy, whereas adjuvant therapy can be withheld (at least initially) for patients with radionecrosis. Despite the practical role of the pathologic diagnosis in influencing subsequent treatment, such a practice has not been validated. The data on the predictive value of pathologic evaluation in the second surgeries are quite limited. Previous studies analyzed the value of stereotactic biopsies in the evaluation of recurrent tumor versus radionecrosis [5-7]. Because no study directly addressed this

Patient characteristics in categories and the group as a whole

Age

Sex Initial KPS

Location Extent of first surgery

Dose of initial EBRT Extent of second surgery

Initial boost radiotherapy

Time between first and second surgery KPS at second surgery

n (valid cases) Mean (median) SD Female Male n (valid cases) Median Range n (valid cases) R/L/bilateral n (valid cases) Gross total Subtotal Biopsy Unknown n (valid cases) Median n (valid cases) Gross total Subtotal Unknown n (valid cases) Yes 125 I implant Gamma knife n (valid cases) Median Range Median Range

Recurrent tumor (category 1)

No recurrent tumor (category 2)

Undetermined (category 3)

Overall

P

31 48.29 (50) 12.0 10 21 24 90 60-90 31 18:13:00 31 4 (13%) 20 (65%) 3 (10%) 4 (13%) 20 5940 31 6 (19%) 21 (68%) 4 (13%) 27 7 (26%) 4 (15%) 3 (11%) 31 7.4 3-22 90 70-100

15 51.8 (52) 11.9 5 10 13 90 70-100 15 10:05:00 15 4 (27%) 6 (40%) 3 (20%) 2 (13%) 11 5940 15 3 (20%) 12 (80%) 0 14 7 (50%) 5 (36%) 2 (14%) 15 7.4 4-19 80 70-90

8 52.38 (50) 14.2 2 6 6 85 80-90 8 5:02:01 8 1 (13%) 5 (63%) 2 (25%) 0 5 6000 8 0 7 (88%) 1 (12%) 7 2 (29%) 0 2 (29%) 8 3.7 3-10 75 40-90

54 49.87 (51.5) 12.2 17 37 43 90 60-100 54 33:20:01 54 9 31 8 6 36 5940 54 9 (17%) 40 (74%) 5 (9%) 48 17 (35%) 9 (19%) 7 (16%) 54 7.3 3-22 85 40-100

NS

NOTE. NS implies P values greater than .1 for comparisons between categories 1 and 2. Abbreviations: R, right; L, left; EBRT, external beam radiotherapy.

NS

NS

NS NS

NS

NS

NS

274 issue in the resection material, we do not know the prognostic significance of differentiating recurrent tumor from radionecrosis in postradiation specimens from patients with GBM. It is imperative to determine the predictive value of pathologic diagnoses at surgery after radiotherapy to consider the impact of pathologic evaluation on subsequent treatment decisions. This study aims to evaluate the prognostic significance of pathologic findings in the second (postradiation treatment) surgical specimens from patients previously diagnosed with GBM. From a large group of patients with GBM, we identified all those who received external beam radiotherapy after the first surgery, followed by a second subtotal or total resection for the treatment of radiologic worsening. We identified patients with evidence of high-grade tumor from those demonstrating radionecrosis. The categories were then compared in terms of clinical outcome to determine whether finding pathologic evidence of recurrent tumor is indicative of a worse prognosis.

2. Materials and methods 2.1. Patient selection and clinical characteristics Appropriate permission was obtained from the institutional review board (Committee for Human Research— CHR approval no. H41175-22167). All patients with GBM who underwent a second surgery for radiologic disease progression at University of California at San Francisco, San Francisco, Calif, between 1989 and 2002 were identified by a search of the pathology database system (Co-Path Plus, Cerner DHT, Inc, Kansas City, Mo). All cases with paired specimens from the first and second surgery were selected for the study. Cases were included in the study if the patient had an initial diagnosis of GBM and had received external beam radiation therapy after the first surgery. Cases were excluded if they had a previous diagnosis of a low-grade glioma. Cases were also excluded if the patient did not undergo surgery as the first treatment of radiologic worsening subsequent to radiotherapy. Finally, all cases without clinical information or pathology material were excluded. Clinical information was compiled from computerized patient files and from patient charts. For each case, patient age at diagnosis and at second surgery, sex, Karnofsky performance score at initial (iKPS) and at second surgery, and tumor location were recorded (Table 1). The treatment course from diagnosis to death was also recorded. This included the initial external beam radiation dose, as well as any additional therapies such as radiosurgery, iodine 125 (125I) brachytherapy, chemotherapy regimens, and additional surgeries. The extent of initial surgery was recorded as gross total resection, subtotal resection (at least 90% removal), or biopsy based on immediate postoperative imaging. Second surgeries were either gross total or subtotal

T. Tihan et al.

Fig. 1 Timeline of events recorded for the study between initial surgery and death.

resections. The dates of treatments, clinical and radiologic recurrences, and death were recorded. Fig. 1 demonstrates the timeline of events throughout the follow-up period.

2.2. Pathologic evaluation All available pathology material was evaluated by 2 pathologists (J. Barletta, T. Tihan) blinded to the time of surgery and the original diagnoses. Specimens from the initial, second, and subsequent surgeries were grouped randomly. Well-recognized features of radiation treatment were recorded [16]. These included coagulative necrosis without pseudopalisading, vascular necrosis, vascular hyalinization, reactive vascular changes, dystrophic calcification, perivascular chronic inflammation, and reactive atypia (gliosis with atypia, Fig. 2). All features were recorded as absent, focal (demonstrated in b50% of the entire specimen), moderate (present in 50%-75% of the specimen), or extensive (involving N75% of the specimen). In addition, the presence of viable tumor with radiation effect and viable tumor with or without high-grade features was recorded using the same semiquantitative scale. Tumor with highgrade features was defined as viable malignant glioma with mitotic figures, bona fide vascular proliferation, or necrosis with pseudopalisading. In other words, tumor sections with high-grade features appeared in every way identical to unirradiated malignant gliomas. Tumor with treatment effect was considered when scattered atypical cells with extensive degenerative changes were encountered. Once the pathology review was complete, the specimens from the second surgeries were assigned to category 1, 2, or 3 (Table 2). 2.2.1. Category 1: recurrent tumor Specimens in this category demonstrated features of a high-grade glioma with mitoses. In some cases, there was only minimal evidence of radiation effect. Specimens with vascular proliferation or necrosis with pseudopalisading that were bona fide features of the actively growing tumor rather than the result of radiation effect were also considered to be in this category (Fig. 3A,B). 2.2.2. Category 2: no recurrent tumor Specimens in this category demonstrated features of radiation treatment, but no evidence of viable tumor. Cases with scattered atypical cells with extensive degenerative


275

Fig. 2 The features of radiation effect. A, Coagulative necrosis without pseudopalisading and vascular necrosis. B, Vascular hyalinization. C, Reactive vascular change. D, Calcium deposition. E, Perivascular chronic inflammation. F, Gliosis with atypia.

changes, tumor with radiation effect, were still included in this category (Fig. 3C, D). 2.2.3. Category 3: undetermined This category was primarily used for cases in which a clear-cut distinction of a recurrent tumor could not be made. Specimens in this category demonstrated foci of viable tumor cells without appreciable radiation effect or without high-grade features (Fig. 3E, F). In addition, there was clear histologic evidence of radiation effect elsewhere in the

resection. Also included in this category were cases with pathologic material too scant to make a final histologic categorization. In the absence of a radiation history, these tumors could have been reported as low-grade glioma.

2.3. Statistical analysis The main goal of the statistical analysis was to compare outcomes of patients in category 1 versus those in category 2. The statistical analyses were performed by using either

276

T. Tihan et al.

Table 2

Pathological characteristics in categories and the group as a whole

n (valid cases) Gliosis with atypia Calcium deposition Perivascular inflammation Necrosis without pseudopalisading Vascular hyalinization Vascular necrosis Reactive vascular changes Nonviable tumor with treatment effect Viable tumor without high-grade features* Viable tumor with high-grade features**




Overall

31 30 (96.8%) 8 (25.8%) 10 (32.3%) 29 (93.5%) 30 (96.8%) 29 (93.5%) 29 (93.5%) 30 (96.8%) 0 (0%) 31 (100%)

15 13 (86.7%) 5 (33.3%) 7 (46.7%) 14 (93.3%) 13 (86.7%) 13 (86.7%) 12 (80%) 10 (67.7%) 0 (0%) 0 (0%)

8 7 2 2 7 7 6 6 8 6 0

54 50 15 19 50 50 48 47 48 6 31

(87.5%) (25.0%) (25.0%) (87.5%) (87.5%) (75.0%) (75.0%) (100%) (75.0%) (0%)

* Exclusion criterion from categories 1 and 2. ** Inclusion criterion for category 1.

the SPSS Advanced Statistics Package for Windows, release 11.0.1 (SPSS Inc, Chicago, Ill) or the SAS system for Windows (SAS Institute, Cary, NC). The former was used to compare all 3 categories and perform Kaplan-Meier survival analysis. Both packages were also used to compare category 1 with category 2 variables. Analysis of variance was used to test for differences among categories in the following variables: patient age at first surgery, iKPS, age at second surgery, Karnofsky performance score (KPS) at second surgery, radiation dose, time between first and second surgery in months, and time between first surgery and second recurrence in months. Pearson v 2 test was used to compare all nominal variables (tumor location, sex, extent of surgery, type of chemotherapy, change in chemotherapy, permanent 125I seed implants, additional surgery, radiotherapy, or chemotherapy anytime after the second recurrence). KaplanMeier survival analysis was used to estimate survival times with 95% CIs. Progression-free survival (PFS) is defined as the time between the second surgery and the subsequent radiologic or clinical relapse, whereas survival after second surgery (SFSS) is the time between second surgery and death, and overall survival (OS) is defined as the time between initial surgery and death. Comparison of outcomes was made among 3 categories using the logrank test. In all statistical analyses and tables, a P value of greater than .1 was recorded as nonsignificant (NS). P values less than .1 were reported, and statistical significance was considered to be P b .05.

3. Results 3.1. Patient characteristics We identified 906 surgical specimens from 748 patients with the diagnosis of GBM at the database of

the Department of Pathology, University of California at San Francisco, between 1989 and 2002. Among those, 112 surgical specimens from 54 patients met the inclusion criteria (see bMaterials and methodsQ section). We have excluded 694 patients for a variety of reasons: In 491 patients, pathology material was not sufficient or was not available for review (outside consultations, initial surgery performed elsewhere). In 93 patients, there was a prior diagnosis of a lower-grade glioma or a mixed glioma. In 79 patients, there was no clear information on radiotherapy, and in 31 patients, the clinical information was insufficient. The median age at diagnosis of cases included in the study was 51 years (range, 21-78 years). There were 17 women and 37 men. All 54 tumors were in the supratentorial region: 33 in the right, and 20 in the left hemisphere. In one patient, the tumor involved the corpus callosum and both frontal lobes. The median iKPS for all patients at first surgery was 90.

3.2. Initial treatment The extent of surgery for the initial procedure is presented in Table 1. The extent of resection was not recorded for 6 of the patients. All patients received external beam radiotherapy, and all but 2 patients had an external beam dose of at least 58.7 Gy. In addition, 7 patients received local gamma knife radiosurgery boost with 14 to 18 Gy, and 9 patients received temporary intratumoral 125I brachytherapy with 38.5 to 60.7 Gy at 35 to 60 cGy/h along with interstitial hyperthermia in 5 cases. The exact details of radiotherapy could not be elucidated in another 8 patients. All patients for whom radiation records were available were treated with partial brain radiotherapy except for one patient who underwent 12.6 Gy to the whole brain before fields were coned down to tumor location. Thirty-eight patients received 59.4 to 60 Gy in 30 or 33 fractions. Other radiation dose fractionation schemes included 61.2 Gy in 34 fractions (3 patients), 62.2 Gy in 37 fractions, 70.4 Gy in 44 fractions,


277

Fig. 3 Typical histologic features for categories. A and B, Category 1 specimen. C and D, Category 2 specimen. E and F, Category 3 specimen.

58.7 Gy in 33 fractions, 53 Gy in 23 fractions, and 40 Gy in 20 fractions. Thirty patients received chemotherapy as a part of their initial treatment plan. Chemotherapy information was not available for 4 patients. The chemotherapy that patients received at first surgery included a range of agents and doses. The drugs included BCNU, hydroxyurea, temozolo-

mide, tamoxifen, topotecan, carboplatin, difluoromethylornithine, Marimastat, procarbazine, vincristine.

3.3. Second surgery All patients underwent a second surgery after a median period of 7.3 months after initial surgery. None of the patients had any form of salvage radiation or chemotherapy

278 Table 3

T. Tihan et al. Subsequent treatment after second surgery in categories and the group as a whole

All forms of salvage radiotherapy

Subsequent chemotherapy Subsequent surgery

n (valid cases) No Yes 125 I implant at second surgery Gamma knife after second surgery Further radiotherapy After second surgery After second recurrence n (valid cases) No Yes Unknown




Overall

P

27 5 22 17 1

14 7 7 1 2

7 3 4 2 0

48 15 33 20 3

NS

4 16 18 31 18 8 5

4 6 9 15 9 3 3

2 7 6 8 6 1 1

10 29 33 54 33 12 9

.002 NS NS NS NS

NOTE. NS implies P values greater than .1 for comparisons between categories 1 and 2.

before the second surgery. In all cases, a radiologic worsening was recorded after radiotherapy, and the second surgery was performed to remove the radiologic abnormality. The median KPS for all patients at the time of this second surgery was 85 (range, 40-100). All patients underwent resection (gross total or subtotal) in an attempt to remove the radiologically abnormal area, and resection was evaluated by postoperative scans.

3.4. Subsequent treatment At the time of second surgery, 20 patients received permanent low-activity intratumoral 125I brachytherapy. Two other patients received radiosurgery less than a month after second surgery to treat smaller additional lesions distant from the original tumor location. Twenty-nine patients received chemotherapy after second surgery: 10 of these started receiving chemotherapy, 14 began receiving a different regimen, 3 continued to receive the same regimen, and the type of regimen could not be determined in the remaining 2 patients. Twenty-two patients did not receive postoperative chemotherapy: 12 stopped receiving their

Table 4

PFS

SFSS

OS

preoperative regimen, and the remaining 10 continued to be managed without chemotherapy. We were not able to obtain postoperative chemotherapy information for 5 patients (including 2 with unknown chemotherapy regimens) because the treatments were continued elsewhere. The agents used were more diverse than the list for the initial chemotherapy and included cisplatin and thalidomide as well as all mentioned above. The results of the subsequent treatment for the categories are presented in Table 3. Forty-eight of the 54 patients had further clinical or radiologic evidence of worsening at a median period of 3.3 months after second surgery. Between the time of this second recurrence and death, 12 of these underwent additional surgery, 33 patients received additional chemotherapy, and 10 had salvage radiotherapy 2.7 to 36.5 months after the second surgery (Table 3). The median follow-up period from initial diagnosis was 18.3 months (range, 7-83 months). At the end of the followup period, 49 of the patients were dead because of the disease, 2 were alive with their disease, and 3 were lost to follow-up after 19, 30, and 41 months. The values for PFS, SFSS, and OS for the whole group are presented in Table 4.

Clinical outcomes in categories and the group as a whole

n (valid cases) Median 95% CI n (valid cases) Median 95% CI n (valid cases) Median 95% CI




P

27 4.0 3-5 28 9.3 7-13 28 18.3 16-22

13 2.2 2-3 13 8.5 8-10 13 17.6 11-28

8 2.9 2-3 8 8.2 4-10 8 14.9 5-21

NS

NOTE. NS implies P values greater than .1 for comparisons between categories 1 and 2.

NS

NS


Fig. 4

279

Kaplan-Meier survival curve for PFS (A), SFSS (B), and OS (C) for categories 1 to 3 and all patients.

Fig. 4 demonstrates the Kaplan-Meier analysis of PFS and OS.

3.5. Pathologic analysis and clinical features of categories 3.5.1. Category 1: recurrent tumor (31 cases) By definition, tumors in this category displayed malignant glioma with high-grade features. All tumors had several mitotic tumor cells, in other words, recurrent tumor. In 24 of the cases (77.4%), the recurrent tumor was focal, comprising between 25% and 50% of the surgical specimen. The extent of recurrent tumor was intermediate (50%-75%) in 5 specimens (16.1%) and extensive (N75%) in 2 cases (6.5%). In all but 1 case, there were scattered atypical cells with extensive degenerative changes in areas separate from the viable tumor. Several histologic changes consistent with radiation treatment were seen in all cases. Specifically, 30 cases had vascular hyalinization, 30 had gliosis with atypia, 29 had vascular necrosis, 29 had reactive vascular changes including telangiectasia, and 29 had focal or intermediate geographic necrosis. Eight of the 31 tumors demonstrated focal dystrophic calcifications, and there were focal perivascular lymphoplasmacytic infiltrates in 10 cases. The clinical variables and information before the second surgery with median values are given in Table 1. Three of the patient with category 1 received gamma knife radiation boost, and 4 patients received temporary 125I brachytherapy boost during or immediately after radiotherapy as part of their treatment plan. Fifteen patients received chemotherapy after first surgery for the same purpose. The chemotherapy agents included BCNU (6 patients), hydroxyurea (5 patients), temozolomide, carboplatin, Marimastat, and topotecan (1 patient each). The initial chemotherapy agents were not recorded for 4 patients. Treatment subsequent to second surgery is presented in Table 3. Twenty-seven of the 31 category 1 patients had radiologic or clinical evidence of disease progression after second surgery at a median period of 4 months. Subsequent treatments after second recurrence

are presented in Table 3. Chemotherapy agents used for patients in this category after second surgery were BCNU (6 patients), temozolomide (4 patients), procarbazine, CCNU, and vincristine regimen (2 patients), cisplatin, and tamoxifen (1 patient each). Type of chemotherapy was not recorded for 4 cases. These 4 cases were different than the patients in which chemotherapy information was not available. Overall, a total of 21 patients (68%) received 125I brachytherapy during the course of their disease. Among category 1 patients, there was no difference in OS, PFS, and SFSS between patients who received 125I brachytherapy and those who did not receive 125I brachytherapy at second surgery. At the end of the follow-up period, 28 of the patients had died of a disease, and 3 patients were lost to follow-up after 19, 30, and 41 months. 3.5.2. Category 2: no recurrent tumor (15 cases) The patients in this category, by definition, had neither tumor with high-grade features nor low-grade tumor without evidence of treatment effect. Ten of the 15 tumors had focal tumor cells with extensive degenerative changes. All specimens demonstrated several histologic features indicative of radiation treatment. Geographic necrosis was present in almost all cases, typically accompanied by vascular necrosis, gliosis with atypia, and vascular hyalinization. Reactive vascular changes including telangiectasia were observed in 12 cases, whereas dystrophic calcification was present in 5 of the 15 cases, and perivascular lymphoplasmacytic infiltrates were found in 7 of the cases. Two patients in this group received local gamma knife radiosurgery boost, and 5 had 125I brachytherapy as a part of their initial treatment plan. Twelve patients received initial adjuvant chemotherapy that included hydroxyurea (4 patients), BCNU (1 patient), temozolomide (3 patients), tamoxifen (2 patients), and PCV regimen (1 patient). Treatment regimen could not be identified for 1 patient. Overall, a total of 7 patients received 125I (47%) brachytherapy anytime during the course of their disease. Thirteen of 15 patients had clinical or radiologic evidence of disease progression at a median period of 2.2 months. Subsequent treatment modalities are

280 presented in Table 3. Thirteen patients in this category were dead at the end of follow-up period. 3.5.3. Category 3: undetermined (8 cases) Two of the specimens in this category demonstrated histologic features of radiation treatment but were too limited for better characterization. The remaining 6 specimens demonstrated foci of low-grade infiltrating tumor in which clear evidence of radiation effect was not seen. In other areas, these 6 specimens demonstrated geographic necrosis, vascular hyalinization, vascular necrosis, and gliosis with atypia. Calcification was identified in 2 cases, and perivascular lymphoplasmacytic infiltrates were found in 2 other cases. In most instances, designation of category 3 implied an uncertainty on the part of the neuropathologists to accurately place the case in either of the other 2 categories. This was done in an attempt to keep the first 2 categories more uniform in terms of their histologic features. All patients had clinical or radiologic evidence of disease progression at a median interval of 2.9 months. The median follow-up period for this category was 15 months (14.9 F 17.6 [median F SD]). All patients in this category were dead at the end of the follow-up period.

3.6. Comparison of categories Because the goal of the study was to determine whether recurrent category identified by pathologic evaluation predicts a worse outcome, the comparisons were made primarily between categories 1 and 2 patients, who had similar mean age, iKPS, and extent of resection at initial surgery (Table 1). We failed to demonstrate any difference in any of the outcome parameters between categories 1 and 2 (Table 4). We also wanted to ensure that patients in category 3 were not significantly different such that had they been included as either one of the other categories, they could have altered our results. When all 3 categories were analyzed together, there was a statistical difference in terms of KPS at the second surgery ( P = .004) because of the lower values in category 3, with one patient at KPS 40. There were no significant differences among the 3 categories in terms of side (left/ right), location (lobes involved), sex, extent of first surgery, mode and dose of external beam radiation, gamma knife radiosurgery or 125I brachytherapy boost, extent of second surgery, additional surgery, radiotherapy or chemotherapy after second surgery, PFS, SFSS, or OS. The power analysis of categories suggested that the sample size was not large enough to detect differences shorter than 6 and 4 months of the observed mean values for OS and SFSS, respectively, but was acceptable for detecting smaller differences in PFS (~2 months). The only statistically significant difference between categories 1 and 2 was the percentage of patients receiving low-activity permanent 125I brachytherapy after the second surgery (17 of 31 versus 1 of 15, P = .002). Seventeen of 31 patients in category 1 received 125I brachytherapy with the

T. Tihan et al. second surgery in contrast with only 1 of 15 patients in category 2 ( P = .002). Analysis of change of chemotherapy after second surgery was confounded by the diversity of chemotherapeutic regimens used. There were insignificant differences in terms of chemotherapy regimens among categories in any segment of the follow-up period. None of these differences reached a P value less than .1 in analyses. Slightly more category 2 patients received chemotherapy after the second surgery (52% in category 1 versus 40% in category 2), whereas fewer patients in category 1 had initial chemotherapy (48% in category 1 versus 73% in category 2). When the entire follow-up period was considered, these insignificant differences diminished further. All patients received a form of alkylating agent during the follow-up period. Tests for equality of survival distributions for categories demonstrated no significant differences in terms of OS (log-rank P = .675), SFSS (log-rank P = .911), or PFS (log-rank P = .269).

4. Discussion In this study, we have attempted to determine whether recognizing a recurrent GBM in postradiotherapy specimens was associated with a worse patient outcome. The obvious implication of this analysis is that had there been such a significant difference, use of appropriate reporting terminology would have to be selected carefully. The specimens obtained from reexcisions were considered to have recurrent tumor when we were able to recognize all the histologic elements of a high-grade astrocytoma, or no recurrent tumor when we could only identify histologic alterations due to radiotherapy (radionecrosis). We included the category 3 in our analysis to ensure that a statistically significant group was not left out. Based on this categorization, we expected the recurrent tumor category to behave more aggressively and recur earlier than no recurrent tumor category. There was precedence to this presumption because a study using stereotactic biopsies to stratify tumors in the same fashion found that biopsies showing radionecrosis had a 4-fold increased survival compared with biopsies showing recurrent tumor [7]. If valid, the assumption that patients with GBM with recurrent tumor have a worse prognosis than patients with radionecrosis should support the decision to treat patients with tumor recurrence more aggressively. Without evidence of survival differences, it may not be reasonable to distinguish one group from another when deciding treatment. It is also possible, yet less likely to argue, that either the patients in category 2 (no recurrent tumor) may be delayed in receiving potentially beneficial treatment or the patients in category 1 (recurrent tumor) receive a more aggressive treatment without improved benefit. The findings of this study suggest that the outcome of patients with histologically recurrent tumor is not worse than that of patients with histologic evidence of radionecrosis alone. Progression-free survival, which is a better criterion to

Prognostic value of detecting recurrent glioblastoma multiforme determine the predictive value of pathologic evaluation, was not statistically different between categories 1 and 2. The survival values obtained for the whole group were comparable to the results published in the literature [3,17,18]. We were able to extract 2 time points, in which reliable KPS values were recorded, at the initial and second surgery. Karnofsky performance score values after second surgery were either not recorded or recorded at various times after surgery, making comparisons difficult. Nevertheless, there was no statistical difference in PFS and SFSS, even if there might be differences in KPS after second surgery. We believe that there is no evidence to suggest such a difference among categories. It is possible that KPS may be affected by different factors, yet equally, among categories. It is critical to determine what would be the minimum nontrivial effect size, that is, what is the shortest period that we should consider important between categories 1 and 2. We have postulated that it would have been possible to demonstrate a significant survival difference with these numbers if the comparison was based on initial pathologic diagnosis (eg, low-grade versus high-grade glioma). If it were fair to compare the low-grade histology of category 2 and the high-grade histology of category 1 with the low- and high-grade histology in initial pathology samples, it would certainly be fair to expect a difference in survival probability in excess of 300% of mean survival value of high-grade tumors. Furthermore, if our results were to parallel the findings of McGirt et al [7], we would have detected a ~4fold difference in survival between categories 1 and 2. Nevertheless, our study is limited by the number of patients in each category, and it does not provide sufficient power to determine differences shorter than 2 months in PFS and shorter than 4 to 6 months in SFSS and OS. It is also important to consider that patients who undergo a second surgery after radiation represent a selective population. This is evident from the number of cases eligible in this study among more than 900 patient with glioblastomas. Because the patients in this study do not represent a random sample, it is not possible to make inferences about the results for all glioblastoma patients who may be treated differently and who do not undergo a second surgical procedure. Nevertheless, the issue of recurrent tumor versus radionecrosis is exactly relevant to this group of patients who will present to the surgical pathologist for a decision of btreat or not to treatQ based on pathology interpretation. Our results differ from the findings of McGirt et al who evaluated the utility of stereotactic biopsy in differentiating recurrent malignant astrocytomas (WHO grades 3 and 4) from radiation-induced lesions using a group of 114 patients with grade 3 or 4 astrocytomas. These authors found a more than 4-fold difference in survival between patients diagnosed with radionecrosis (median survival, 27 months) and patients who were diagnosed with recurrent malignant glioma (median survival, 7 months). They also found that patients with lesions occurring more than 5 months after

281

radiotherapy, which were the result of radiation injury, had an increased survival time as compared with patients harboring recurrent astrocytoma [7]. Similarly, Forsyth et al [6] reported that the presence of radiation-induced necrosis in stereotactic biopsies from 51 patients with treated gliomas correlated with a better prognosis. These investigations differed from the current study in 3 important ways. First, both of these studies evaluated the prognostic significance of stereotactic biopsies, whereas the current study focused on patients who had undergone resections of their lesion. It is quite possible that the choice of surgical procedure (biopsy versus resection) may be due to practical issues (eg, deep seated versus superficial location) that can introduce a selection bias, preventing an objective comparison of these studies with our results. Second, one can presume that even at the second surgery, survival is positively correlated with extent of resection of patients with GBM [17,19,20]. Resection of recurrent malignant glioma has also been shown to increase the survival of patients with malignant glioma [21]. Third, McGirt et al evaluated specimens from patients originally diagnosed with both grades 3 and 4 astrocytomas as well as patients with 1 or 2 prior resections, and Forsyth et al evaluated specimens from patients with a range of histologic types, whereas we selected only patients with grade 4 astrocytomas in the initial specimens. Inclusion of different tumor grades is another possible explanation for the different results obtained in their study. Our study also underscores the challenges in attempting to identify homogenous patient populations from which clear deductions of treatment efficacy can be made. The multiplicity of treatments for any given patient highlights the inability to control GBMs even with treatments that are considered effective. Subsequently, it becomes difficult to attribute any change in outcome to single intervention or analytical parameter in a retrospective study such as ours. Unlike the initial pathologic evaluation that is less complicated and less compounded by other intervening events, pathologic evaluation of the subsequent specimens does not carry the same demonstrable prognostic value. One other limiting issue in the collection of the retrospective data is the difficulty in determining quality of life issues, which were not among the goals of this study. Issues such as the need to reduce the dose of steroids or the effects of cerebral edema may influence surgical decisions and may be more important than the goal of improving OS. Nevertheless, because there were no significant differences in 2 KPS values among groups, we can assume that each category was affected by a variety of factors to an equal extent, even if the factors influencing KPS may be different. Despite the confounding factors and the challenges in the interpretation of the data, one issue seems to be clear: We cannot suggest that pathologic interpretation of postradiotherapy resection specimens should alter subsequent treatment of patients with GBM. It is often possible to

282 demonstrate survival differences using initial pathologic type and grade of gliomas, but it has not been possible to show any outcome difference among our categories. The simple practical implication of this study is that pathologic evaluation of a postradiotherapy specimen in a patient does not seem to matter in terms of the patient’s survival. It is, however, possible that more advanced serial imaging studies such as those reported recently may determine tumors that really recur, giving us a better measure of outcome and better stratification of patients [22]. Future studies should accurately reveal the rationale for treatment choices and collection of the data to better understand the true significance of pathologic interpretation. There is a great need to design prospective studies in which clinical, pathologic, and especially functional radioimaging parameters are measured in tandem to attempt a better estimate of subsequent patient outcome.

References [1] Walker MD, Green SB, Byar DP, et al. Randomized comparisons of radiotherapy and nitrosoureas for the treatment of malignant glioma after surgery. N Engl J Med 1980;303:1323 - 9. [2] Paszat L, Laperriere N, Groome P, Schulze K, Mackillop W, Holowaty E. A population-based study of glioblastoma multiforme. Int J Radiat Oncol Biol Phys 2001;51:100 - 7. [3] Levin VA, Wara WM, Davis RL, et al. Northern California Oncology Group protocol 6G91: response to treatment with radiation therapy and seven-drug chemotherapy in patients with glioblastoma multiforme. Cancer Treat Rep 1986;70:739 - 43. [4] Burger PC, Dubois PJ, Schold Jr SC, et al. Computerized tomographic and pathologic studies of the untreated, quiescent, and recurrent glioblastoma multiforme. J Neurosurg 1983;58:159 - 69. [5] Zamorano L, Katanick D, Dujovny M, Yakar D, Malik G, Ausman JI. Tumour recurrence vs radionecrosis: an indication for multitrajectory serial stereotactic biopsies. Acta Neurochir Suppl (Wien) 1989;46:90 - 3. [6] Forsyth PA, Kelly PJ, Cascino TL, et al. Radiation necrosis or glioma recurrence: is computer-assisted stereotactic biopsy useful? J Neurosurg 1995;82:436 - 44. [7] McGirt MJ, Bulsara KR, Cummings TJ, et al. Prognostic value of magnetic resonance imaging–guided stereotactic biopsy in the evaluation of recurrent malignant astrocytoma compared with a lesion due to radiation effect. J Neurosurg 2003;98:14 - 20.

T. Tihan et al. [8] Valk PE, Dillon WP. Radiation injury of the brain. AJNR Am J Neuroradiol 1991;12:45 - 62. [9] Dooms GC, Hecht S, Brant-Zawadzki M, Berthiaume Y, Norman D, Newton TH. Brain radiation lesions: MR imaging. Radiology 1986;158:149 - 55. [10] Valk PE, Budinger TF, Levin VA, Silver P, Gutin PH, Doyle WK. PET of malignant cerebral tumors after interstitial brachytherapy. Demonstration of metabolic activity and correlation with clinical outcome. J Neurosurg 1988;69:830 - 8. [11] Glantz MJ, Hoffman JM, Coleman RE, et al. Identification of early recurrence of primary central nervous system tumors by [18F]fluorodeoxyglucose positron emission tomography. Ann Neurol 1991;29: 347 - 55. [12] Langleben DD, Segall GM. PET in differentiation of recurrent brain tumor from radiation injury. J Nucl Med 2000;41:1861 - 7. [13] Chao ST, Suh JH, Raja S, Lee SY, Barnett G. The sensitivity and specificity of FDG PET in distinguishing recurrent brain tumor from radionecrosis in patients treated with stereotactic radiosurgery. Int J Cancer 2001;96:191 - 7. [14] Dowling C, Bollen AW, Noworolski SM, et al. Preoperative proton MR spectroscopic imaging of brain tumors: correlation with histopathologic analysis of resection specimens. AJNR Am J Neuroradiol 2001;22:604 - 12. [15] Rock JP, Hearshen D, Scarpace L, et al. Correlations between magnetic resonance spectroscopy and image-guided histopathology, with special attention to radiation necrosis. Neurosurgery 2002;51: 912 - 9 [discussion 919-20]. [16] Burger PC, Mahley Jr MS, Dudka L, Vogel FS. The morphologic effects of radiation administered therapeutically for intracranial gliomas: a postmortem study of 25 cases. Cancer 1979;44:1256 - 72. [17] Jeremic B, Grujicic D, Antunovic V, Djuric L, Stojanovic M, Shibamoto Y. Influence of extent of surgery and tumor location on treatment outcome of patients with glioblastoma multiforme treated with combined modality approach. J Neurooncol 1994;21:177 - 85. [18] Barker II FG, Prados MD, Chang SM, et al. Radiation response and survival time in patients with glioblastoma multiforme. J Neurosurg 1996;84:442 - 8. [19] 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;95:190 - 8. [20] Scott JN, Rewcastle NB, Brasher PM, et al. Which glioblastoma multiforme patient will become a long-term survivor? A populationbased study. Ann Neurol 1999;46:183 - 8. [21] Ammirati M, Galicich JH, Arbit E, Liao Y. Reoperation in the treatment of recurrent intracranial malignant gliomas. Neurosurgery 1987;21:607 - 14. [22] Smith JS, Cha S, Mayo MC, et al. Serial diffusion-weighted magnetic resonance imaging in cases of glioma: distinguishing tumor recurrence from postresection injury. J Neurosurg 2005;103:428 - 38.