|Home | About | Journals | Submit | Contact Us | Français|
The purpose of this study was to evaluate long-term survival in patients with nonpilocytic low-grade gliomas (LGGs). Records of 314 adult patients with nonpilocytic LGGs diagnosed between 1960 and 1992 at the Mayo Clinic, Rochester, Minnesota, were retrospectively reviewed. The Kaplan-Meier method estimated progression-free survival (PFS) and overall survival (OS). Median age at diagnosis was 36 years. Median follow-up was 13.6 years. Operative pathology revealed pure astrocytoma in 181 patients (58%), oligoastrocytoma in 99 (31%), and oligodendroglioma in 34 (11%). Gross total resection (GTR) was achieved in 41 patients (13%), radical subtotal resection (rSTR) in 33 (11%), subtotal resection in 130 (41%), and biopsy only in 110 (35%). Median OS was 6.9 years (range, 1 month–38.5 years). Adverse prognostic factors for OS identified by multivariate analysis were tumor size 5 cm or larger, pure astrocytoma histology, Kernohan grade 2, undergoing less than rSTR, and presentation with sensory motor symptoms. Statistically significant adverse prognostic factors for PFS by multivariate analysis were only tumor size 5 cm or larger and undergoing less than rSTR. In patients who underwent less than rSTR, radiotherapy (RT) was associated with improved OS and PFS. A substantial proportion of patients have a good long-term prognosis after GTR and rSTR, with nearly half of patients free of recurrence 10 years after diagnosis. Postoperative RT was associated with improved OS and PFS and is recommended for patients after subtotal resection or biopsy.
Low-grade gliomas (LGGs) are primary brain tumors classified as grade I and II by the WHO grading system.1 Treatment goals for patients with LGG include prolongation of both progression-free survival (PFS) and overall survival (OS). Even with prospective trials,2–5 many questions remain about management of LGG, including the role for and extent of surgical resection and the optimal timing of radiotherapy (RT), either immediately postoperatively or as a salvage strategy at progression. We analyzed a cohort of consecutive patients with nonpilocytic WHO grade II LGG to assess long-term outcomes and impact of tumor, patient, and treatment characteristics.
We used the Mayo Clinic tumor registry to perform a retrospective review of 314 consecutive patients aged 18 years or older with newly diagnosed, nonpilocytic LGG who presented between 1960 and 1992. To be included, biopsy confirmation and pathologic review by the Mayo Clinic Department of Pathology was required, as was a detailed operative report. Patients who had tumors in the optic tract or lower brainstem were excluded. Data were retrieved concerning patient presentation, extent of surgical resection, histologic type, adjuvant therapy, and other pertinent prognostic factors as well as type of recurrence, PFS, and OS.
No attempt was made to define an epicenter if multiple sites of brain were involved, but all involved sites were recorded. The extent of resection was assessed by the operative report, neurosurgeon’s impression, and postoperative imaging when available. A surgical procedure was designated as radical subtotal resection (rSTR) first, when the operative report described “radical subtotal resection”; second, when gross total resection (GTR) was clearly the goal of the operation but minimal known tumor was left in situ; or third, when imaging reports indicated small, questionable amounts of residual tumor after GTR.
The label “low grade” is assigned to tumors of astrocytic lineage solely on the basis of their microscopic appearance. The numerical grade assigned to a given tumor, however, can vary depending on which grading system is used (Table 1). Therefore, it is important to specify the grading system with which a tumor has been labeled. In the present analysis, all patients were classified by histologic type and grade according to WHO criteria and the Kernohan grading system.6–8 The Kernohan system allows for labeling of well-differentiated tumors as grade 1, thus dividing the WHO grade II tumors, which are recognized to display a spectrum of nuclear features and mitotic activity, into more favorable and less favorable groups. Because pilocytic tumors were excluded, no WHO grade I tumors were reported in this analysis.
Recurrence information was collected using all available data in the medical record. Specifically, imaging, clinical signs and symptoms, pathologic information, or initiation of any additional intervention such as surgery, RT, or chemotherapy were used to declare progression.
This study was approved by the Mayo Clinic Institutional Review Board. Patients who refused to allow their data to be used for medical research were excluded from this analysis.
Thirteen possible prognostic factors were analyzed for their association with both OS and PFS on univariate analysis. These variables (age, sex, midline/bilateral involvement, size, histology, extent of resection, RT, chemotherapy, sensory motor symptoms at presentation, supratentorial location, Kernohan grade, enhancement on computed tomography [CT] and/or magnetic resonance imaging [MRI]) were chosen because of their prognostic importance in prior studies.3–5,9–29
PFS and OS curves were calculated using the Kaplan-Meier method30 and compared with the log-rank test.31 Multivariate analysis for both recurrence and survival was performed using the Cox proportional hazards model, including variables that were statistically significant (p < .05) on univariate analysis.32 Deaths without documented recurrence were considered censored observations for recurrence at time of death. This method assumes that all patients who had unknown disease status at the time of death or last follow-up died of disease. All tests were 2 tailed. The χ2 test was used to analyze prognostic factors with the likelihood of surgical extent or postoperative RT use.
Patient and tumor characteristics are cataloged in Table 2. Tumor characteristics were identifed from the operative and radiographic report (CT, MRI, angiography) as well as electroencephalograms. Because the starting point of the study interval (1960) predated the modern neuroimaging era (1975), CT data were available for 155 patients (49%) and MRI data were available for 61 patients (19%). Enhancement on imaging was evident for 75 patients (48%) with CT information and 17 patients (28%) with MRI information. Of note, after 1975, CT data were available for 68% of patients.
The majority of patients (301/314; 96%) had supratentorial tumors, most commonly involving the frontal lobe (163 patients [52%]) and temporal lobe (126 patients [40%]).
GTR was achieved in 41 patients (13%), rSTR in 33 (11%), subtotal resection (STR) in 130 (41%), and biopsy only in 110 (35%).
Histologic subtype for the entire group showed pure astrocytoma in 181 tumors (58%), mixed oligoastrocytoma in 99 (31%), and oligodendroglioma in 34 (11%). Operative pathology revealed that of the 314 WHO grade II tumors, 285 (91%) were Kernohan grade 2. As pilocytic tumors were excluded, the majority of Kernohan grade 2 tumors were pure astrocytomas (18/29 [62%]).
Eighty-three patients (26%) were observed and did not receive adjuvant therapy in the immediate postoperative period. Postoperative RT was given to 231 patients (74%) and chemotherapy was given to 21 patients (7%); of the latter, 16 (5%) received chemotherapy in conjunction with RT and 5 (2%) received chemotherapy alone. Twelve patients received carmustine and nine received combination therapy with procarbazine, lomustine, and vincristine.
The total dose, calculated at midplane for opposed treatments or at isocenter for nonopposed treatments, ranged from 28.5 Gy to 70.2 Gy, with a median dose of 54 Gy delivered in a median of 28 fractions over a median 40 days. The three most common total doses delivered were 50 or 50.4 Gy (83 patients [26%]), 54 Gy (31 patients [10%]), and 60 Gy (50 patients [16%]). Whole-brain RT (WBRT) was administered to 53 patients (17%), WBRT plus partial brain boost to 41 patients (13%), and a partial brain only field to 220 patients (70%).
Overall, the median follow-up was 13.6 years. Recurrence was documented in 174 patients who had a median time to progression of 7.2 years and a median PFS of 5.0 years. The 5-, 10-, and 15-year PFS rates were 50%, 27%, and 17%, respectively (Fig. 1).
Adverse prognostic factors identified on univariate analysis for PFS included age 40 years or older (p < .001), presentation with sensory motor symptoms (p = .004), tumor size 5 cm or larger (p < .001), Kernohan grade 2 disease (p = .01), astrocytoma histology (p = .02), and undergoing less than an rSTR (p < .001). Neither postoperative RT nor chemotherapy was associated with PFS (p = .92 and p = .47, respectively). On multivariate analysis (Table 3), only size 5 cm or larger and undergoing less than an rSTR remained statistically significant (both p < .001).
We analyzed extent of resection using four categories of surgical treatment (GTR, rSTR, STR, and biopsy only). Patients who underwent at least rSTR had improved PFS (p = .001). When comparing GTR with rSTR, no significant differences were noted in PFS. These patients (GTR/rSTR) were considered to have more aggressive surgical resection and were combined for the remainder of this analysis. Similarly, no difference in PFS was noted between patients who underwent biopsy only or STR. Patients who underwent more limited surgery (STR/biopsy only) were also combined for the remainder of the analysis. Extent of resection was strongly associated with PFS outcome. The 10-year PFS was 47% vs. 21% for those who underwent GTR/rSTR vs. STR/biopsy only (p < .001). A χ2 analysis was performed to assess factors associated with extent of resection. Patients with unfavorable characteristics such as 5 cm or larger (p = .001), sensory motor symptoms (p = .05), or CT enhancement (p = .02) were less likely to undergo aggressive surgery (GTR/rSTR).
Postoperative RT did not affect PFS overall. However, by χ2 analysis, use of RT was statistically significantly associated with factors shown to be adverse. RT was preferentially delivered to patients with more aggressive histologies such as astrocytoma (p = .001) and Kernohan grade 2 pathology (p < .001) and to those who underwent less aggressive surgical procedures such as STR/biopsy only (p < .001).
The median OS was 6.9 years (range, 1 month–38.5 years), representing 250 deaths. The 10- and 15-year OS rates were 36% and 23%, respectively (Fig. 1). One hundred forty-four patients (46%) died from or with disease, 102 (32%) died from unknown or undocumented causes, 4 (1%) died without known evidence of disease, and 42 (13%) were alive without evidence of disease at last follow-up.
Adverse prognostic factors identified on univariate analysis included age 40 years or older (p = .008), size 5 cm or larger (p < .001), astrocytoma histology (p = .008), undergoing less than an rSTR (p < .001), Kernohan grade 2 disease (p = .004), presentation with sensory motor symptoms (p < .001), supratentorial location (p = .007), and enhancement on CT (p = .03). Neither postoperative RT nor chemotherapy affected OS (p = .96 and p = .87, respectively). Again, when extent of resection was analyzed, those who underwent at least an rSTR had improved OS (p < .001). The 10-year OS was 57% vs. 30% for those who underwent GTR/rSTR vs. STR/biopsy alone (p < .001).
On multivariate analysis (Table 3), statistical significance remained for size 5 cm or larger (p = .05), pure astrocytoma histology (p = .05), Kernohan grade 2 disease (p = .05), presentation with sensory motor symptoms (p = .001), and undergoing less than an rSTR (p = .03). The use of RT did not show a statistically significant association on multivariate analysis for OS (risk ratio [RR], 0.75; 95% confidence interval [CI], 0.62–1.1; p = .34). Because CT data were unavailable for 159 patients, we conducted an exploratory multivariate analysis without using CT enhancement as a potential interacting factor. Similar results were seen with this analysis in that size 5 cm or larger, pure astrocytoma histology, Kernohan grade 2 disease, presentation with sensory motor symptoms, undergoing less than an rSTR, and supratentorial location were significantly associated with a decline in OS (all p values < .05). In addition, use of postoperative RT was associated with a statistically significant improvement in OS (RR, 0.81; 95% CI, 0.69–0.96; p = .01).
Overall, patients treated with STR/biopsy only combined with RT had intermediate rates of 10-year PFS and OS (Fig. 2). Specifically, 10-year OS was 34% for STR/biopsy alone combined with RT, compared with 11% for STR/biopsy alone and 57% for GTR/rSTR with or without RT (Fig. 2B). A separate analysis was done for patients who underwent STR/biopsy alone. For this group, the median OS was 5.4 years and PFS was 4.2 years.
On univariate analysis, adverse prognostic factors for PFS included size 5 cm or larger (p = .004), age 40 years or older (p = .005), presentation with sensory motor symptoms (p = .02), and lack of postoperative RT (p < .001). The median PFS was 5.0 years for those who received RT and 2.69 years for those who did not (Fig. 2C).
On univariate analysis, adverse prognostic factors for OS included age 40 years or older (p = .05), presentation with sensory motor symptoms (p = .02), enhancement on CT scan (p = .01), size 5 cm or larger (p = .002), astrocytoma histology (p = .002), and lack of postoperative RT (p < .001). The median OS were 6.3 years for those who received RT and 3.2 years for those who did not (Fig. 2D).
On multivariate analysis, age 40 years or older (p = .002), presentation with sensory motor symptoms (p = .01), and use of postoperative RT (RR, 0.67; 95% CI, 0.58–0.94; p = .04) remained statistically significantly associated with OS.
Median survival ranged from 5.1 to 9.5 years, depending on histologic type (p = .008). Patients with oligoastrocytomas experienced intermediate outcomes compared with those with astrocytomas (lower survival) and oligodendrogliomas (higher survival). Specifically, survival at 15 years was 18% for patients with astrocytoma, 25% for those with oligoastrocytoma, and 41% for those with oligodendroglioma.
Median dose of 54 Gy was chosen as a cut-off to assess for dose response. When patients who received RT were compared, there was a trend toward improved OS for those who received 54 Gy or more vs. those who did not (10-year OS 39% vs. 24%; p = .06). For patients who underwent STR/biopsy only, there was no significant difference in outcome for those treated with 54 Gy or more (10-year OS 36% vs. 23%; p = .14).
One hundred seventy-four patients experienced a presumed recurrence. Symptoms were present in 158 patients (91%) at the time of recurrence, and 141 (81%) had documented radiographic disease recurrence. Ninety-two patients (53%) underwent a biopsy, and 76 patients (44%) underwent at least an STR at the time of recurrence. Pathologic data were available for 151 patients whose specimens were reviewed by Mayo Clinic. In 19 (13%) of 151 patients, radiation necrosis was noted in the specimen, and in 4 (3%) of 151 specimens, neither tumor nor necrosis was identified. Thus, these 23 patients were not counted as having recurrences in our statistical analysis. In 66 (44%) of 151 patients, the tumor had progressed to a higher grade than at diagnosis. The remaining 62 patients (41%) had pathologic findings consistent with their initial diagnosis. Thirty patients had RT for recurrence and 83 patients received chemotherapy. Median survival after recurrence was 1.2 years.
The indolent natural course of LGGs has resulted in uncertainties and controversies regarding the role, timing, and technique of both surgery and RT. Subjects of debate include more aggressive vs. more limited resection such as biopsy (or even observation), timing of RT as adjuvant therapy immediately after surgery vs. delayed until recurrence, and the role for chemotherapy. Long-term follow-up is paramount to make such evaluations of PFS and OS. With a median follow-up of 13.6 years, the current study is among the largest with such long-term follow-up. Additionally, this study is unique in accomplishing a separate analysis of patients undergoing STR and biopsy alone. Table 4 outlines data from the present series alongside results of the European Organisation for Research and Treatment of Cancer (EORTC) protocol 22845 and other randomized trials.2–5
Surgery at the time of diagnosis provides tissue diagnosis in addition to a potential therapeutic debulking benefit. Similar to other investigations, our study displayed improved outcomes for patients who underwent more aggressive resections. However, GTR is often not possible without serious risk of neurologic injury because of tumor location or infiltration. We found that aggressive resections were more likely to have been performed in patients with more favorable tumor characteristics, such as size smaller than 5 cm, lack of enhancement on CT scan, and lack of sensory motor symptoms. To date, no randomized trials have specifically compared “up-front” surgery with a more conservative approach of delayed surgery. However, considering both retrospective and prospective data, many neurosurgeons favor maximally safe resection.3,4,10,12,16,22,27,29,33 Although not randomized, three prospective studies correlated aggressive surgery (GTR or near GTR) and improved prognosis.3,4,34 Additionally, an intraoperative MRI study found that patients undergoing STR experienced 1.4 and 4.9 times the risk of recurrence and death, respectively, compared with GTR (12). Another recent study, using MRI volumetric analysis, correlated resection of 90% or more with improved OS and PFS (27).
Retrospective data are inconsistent relative to the effect of postoperative RT. Due to perceived toxicities with RT and the disease’s indolent nature,35,36 some advocate delaying RT until there is evidence of progression, symptoms, or high-grade transformation.14,17,37–48 EORTC 22845 evaluated the timing of RT in a phase 3 trial of immediate RT (54 Gy) or observation until progression. With follow-up just over 7 years, postoperative RT significantly prolonged PFS (median, 5.3 vs. 3.4 years) without affecting OS (median, 7.4 vs. 7.2 years).2,5 These results are inconsistent with the present study’s findings of improved OS on multivariate analysis but no impact on PFS for adjuvant RT. In addition, in the present study, the benefit of postoperative RT was most clearly seen, in terms of improved OS and PFS, in patients undergoing more limited resections.
The most likely explanation for differences in PFS outcomes between studies is the imbalance in patient groups who did or did not receive postoperative RT. In the present study, treatment decisions were not randomized, and as noted, RT was preferentially delivered to higher-risk patients, such as those with more aggressive tumors and, most important, after less aggressive resections. The strong effect the extent of resection had on OS and PFS in this study likely confounds the PFS data. An alternate interpretation of the current data is that RT can improve outcomes in high-risk patients to approximate the PFS outcomes seen in more favorable risk categories. The differences in OS outcomes between the studies may be attributable to the significantly longer follow-up in this study than in EORTC 22845 (13.6 vs. 7.7 years). Furthermore, in the EORTC trial, 44% of patients underwent GTR (>90% resection), a much higher percentage than in this report, where only 23% of patients underwent GTR/rSTR. In light of our findings, it is therefore possible that the EORTC trial was not able to show a survival benefit because it included a large percentage of patients at lower risk of progression and death who were less likely to benefit from RT. This is further supported by the current results in which the patients who underwent more limited resections benefited most, in terms of OS and PFS, from postoperative RT (Fig. 2A,B).
Because diagnostic and surgical techniques evolved over the period of the present evaluation, and in order to offer a more relevant comparison to the modernera EORTC and NCCTG trials (which both began enrolling patients in 1986), we analyzed the present data of the 95 patients (30%) who presented in 1986 or later. Median follow-up in this group was 12 years. Patients who presented in 1986 or later, compared with patients who presented before 1986, had improved OS (median, 9.2 vs. 6 years; p = .02) but not PFS (median, 5.3 vs. 4.4 years; p = .17). Of these 95 patients, 24 (25%) underwent GTR/rSTR. Similarly, of the 219 patients whose disease was diagnosed before 1986, 49 (22%) underwent GTR/rSTS. Therefore, it does not seem likely that the extent of surgical resection varied on the basis of decade of presentation when the EORTC and NCCTG trials’ time frames are used. Similar to the results of the entire study cohort, the extent of resection was strongly predictive of both OS and PFS in the cohort of patients whose disease was diagnosed in 1986 or later (univariate analysis data not shown). Use of RT was not associated with either OS or PFS in this modern cohort (univariate analysis data not shown). For the 71 patients whose disease was diagnosed in 1986 or later and who underwent more limited surgical procedures (biopsy/STR), postoperative RT showed a trend toward PFS on univariate analysis (5-year PFS, 51% vs. 13%; 10-year PFS, 21% vs. 11%). However, this difference was not statistically significant (p = .18). The use of postoperative RT was not associated with improved OS (p = .65) in this subgroup.
The disparity in these results, compared with those for the whole group, may be attributed to insufficient follow-up and patient numbers to detect a statistical difference. Furthermore, even though a similar percentage of patients underwent GTR/rSTR, improvements in imaging and surgical techniques that allow for earlier detection, more aggressive yet safe resection, and more accurate classification of surgical extent could certainly be affecting the data. Also, the limitations of referral bias are likely to be even more pronounced in the modern series with more high-risk patients being referred for RT. Because the OS in the more modern group is improved and extent of surgery is the strongest predictor for survival in our series, it is likely that all these factors contribute to the differences in results. Overall, the data presented herein from the subset of more modern patients are not unlike the findings of the EORTC trial for which postoperative RT showed improved PFS2,5 but no improvement in OS.
A minority of patients received chemotherapy at the time of diagnosis, combined with RT in only 5%, and 1% received chemotherapy alone. Chemotherapy was much more frequently given to patients at the time of recurrence (48% of recurrences). No strong evidence exists for routine use of postoperative chemotherapy, especially in light of two phase 3 trials that showed no significant improvement in OS or PFS.34,49 Temozolomide is a promising oral alkylating agent with proven benefit in high-grade gliomas.50,51 Thus, temozolomide has been used in some small trials to delay RT in patients with newly diagnosed LGG.52,53 Ongoing trials by the EORTC, the Eastern Cooperative Oncology Group, and the Radiation Therapy Oncology Group will evaluate temozolomide’s role for those deemed at high risk for early progression. Until these results are known, the current findings suggest careful consideration for the use of temozolomide as a means of delaying RT because it may have a detrimental effect on PFS and OS, especially in patients with minimal surgical resection.
In summary, analysis of a large series of patients with nonpilocytic WHO grade II LGG with long-term follow-up from a single institution indicates that for the majority of patients prognosis is poor (15-year OS 23%), especially recognizing that these patients are frequently young adults. However, there are subsets of patients who do significantly better, such as those undergoing an aggressive surgical resection. Our results favor performing a maximally safe resection, and in patients who have undergone STR/biopsy alone, we favor early use of RT, given the observed benefits in OS and PFS.
We acknowledge financial support by the Mayo SPORE in Brain Cancer (P50 CA108961; principal investigator, Brian P. O’Neill). Cores and shared resources were supported by a Cancer Center Support Grant (P30 CA15083; principal investigator, Robert B. Diasio) to Mayo Clinic Cancer Center.