Preoperative prognostic classification system for ... - Semantic Scholar

J Neurosurg 109:817–824, 109:000–000, 2008

Preoperative prognostic classification system for hemispheric low-grade gliomas in adults Clinical article EDWARD F. CHANG, M.D., JUSTIN S. SMITH, M.D., PH.D., SUSAN M. CHANG, M.D., KATHLEEN R. LAMBORN, PH.D., MICHAEL D. PRADOS, M.D., NICHOLAS BUTOWSKI, M.D., NICHOLAS M. BARBARO, M.D., ANDREW T. PARSA, M.D., PH.D., MITCHEL S. BERGER, M.D., AND MICHAEL M. MCDERMOTT, M.D. Brain Tumor Research Center, Department of Neurological Surgery, University of California, San Francisco, California Object. Hemispheric low-grade gliomas (LGGs) have an unpredictable progression and overall survival (OS) profile. As a result, the objective in the present study was to design a preoperative scoring system to prognosticate long-term outcomes in patients with LGGs. Methods. The authors conducted a retrospective review with long-term follow-up of 281 adults harboring hemispheric LGGs (World Health Organization Grade II lesions). Clinical and radiographic data were collected and analyzed to identify preoperative predictors of OS, progression-free survival (PFS), and extent of resection (EOR). These variables were used to devise a prognostic scoring system. Results. The 5-year estimated survival probability was 0.86. Multivariate Cox proportional hazards modeling demonstrated that 4 factors were associated with lower OS: presumed eloquent location (hazard ratio [HR] 4.12, 95% confidence interval [CI] 1.71–10.42), Karnofsky Performance Scale score ≤ 80 (HR 3.53, 95% CI 1.56–8.00), patient age > 50 years (HR 1.96, 95% CI 1.47–3.77), and tumor diameter > 4 cm (HR 3.43, 95% CI 1.43–8.06). A scoring system calculated from the sum of these factors (range 0–4) demonstrated risk stratification across study groups, with the following 5-year cumulative survival estimates: Scores 0–1, OS = 0.97, PFS = 0.76; Score 2, OS = 0.81, PFS = 0.49; and Scores 3–4, OS = 0.56, PFS = 0.18 (p < 0.001 for both OS and PFS, log-rank test). This proposed scoring system demonstrated a high degree of interscorer reliability (kappa = 0.86). Four illustrative cases are described. Conclusions. The authors propose a simple and reliable scoring system that can be used to preoperatively prognosticate the degree of lesion resectability, PFS, and OS in patients with LGGs. The application of a standardized scoring system for LGGs should improve clinical decision-making and allow physicians to reliably predict patient outcome at the time of the original imaging-based diagnosis. (DOI: 10.3171/JNS/2008/109/11/0817)

KEY WORDS progression

• •

extent of resection • scoring • survival

P

UBLISHED survival estimates for patients with LGGs range from 3 to > 20 years.2,7,8,12,14,16,17,20,23 The typically slow growth of these neoplasms requires extensive follow-up, which has limited the number of studies in the MR imaging era that have adequately addressed long-term outcomes. Consequently, there are few current

Abbreviations used in this paper: CI = confidence interval; EORTC = European Organization for Research and Treatment of Cancer; GTR = gross-total resection; HR = hazard ratio; KPS = Karnofsky Performance Scale; LGG = low-grade glioma; OR = odds ratio; OS = overall survival; PFS = progression-free survival; STR = subtotal resection; UCSF = University of California, San Francisco.

J. Neurosurg. / Volume 109 / November 2008

low-grade glioma

•

prognosis

•

evidence-based standards for guiding the medical and surgical treatment of patients with LGGs. Although some studies have addressed patient clinical status and tumor imaging features, they have not taken into consideration other factors that can affect the resectability of LGGs. Tumor invasion of functionally eloquent areas of the brain and diffuse lesion borders, for example, are critical factors that affect whether radical removal or GTR is feasible. The lack of studies addressing these issues has led to controversy regarding the appropriate management of LGGs. Some clinicians have advocated simple observation, whereas others have recommended maximal resec817

E. F. Chang et al. tion.3,9 The precise roles of surgery and radiation therapy, although both are commonly practiced, are unclear and unlikely to be resolved unless pretreatment-related factors are clearly identified. To address these questions, we analyzed an institutional series of patients with LGGs to identify prognosticators of survival and tumor progression to create a practical and simple preoperative clinicoradiographic grading system for LGGs. The primary aim of this study was to devise a system for clinicians to guide therapy at the time of presentation of a patient with a nonenhancing, infiltrative, low-grade glial neoplasm.

Methods We conducted a retrospective review with long-term follow-up of 281 adult patients harboring hemispheric infiltrative LGGs surgically treated at the UCSF between January 1989 and June 2005. Patients were excluded if they had undergone any prior resection of the tumor, except a previous biopsy procedure performed as part of a diagnostic workup that led to eventual surgical removal. All included patients had histologically confirmed Grade II low-grade infiltrative gliomas (World Health Organization Grade II).10 Patients with pilocytic or gemistocytic astrocytomas were excluded. Given that the study was focused on preoperative prognostication, only patients with non–contrast enhancing lesions were included for analysis. All research activities were approved by the UCSF institutional review board for human research. All preoperative clinical and radiographic features were derived from hospital charts. Clinical variables were recorded, including patient age, sex, presence of seizures and other symptoms on presentation, and KPS score. Preoperative MR images were reviewed while the reviewers remained blinded to patient outcome. Digital caliper measurements of the maximum tumor diameter (in any dimension) were performed using FLAIR or T2weighted sequences. Whether tumors had discrete, welldelineated borders as opposed to diffuse, infiltrative borders was also qualitatively assessed. Diffuse tumors had a more grossly infiltrative appearance, with irregular fingerlike borders, in some cases crossing sulci. The well-delineated tumors had more regular borders that respected local anatomical boundaries. Tumor location and anatomy were carefully documented, including whether the lesion had infiltrated presumed eloquent brain areas (Fig. 1). The presumed eloquent areas consisted of the sensorimotor strip (precentral and postcentral gyri), dominant hemisphere perisylvian language areas (superior temporal, inferior frontal, and inferior parietal areas), basal ganglia/internal capsule, thalamus, and calcarine visual cortex. It is worth emphasizing the fact that this designation of presumed eloquence is based on preoperative diagnostic anatomical MR imaging, not functional imaging or intraoperative functional mapping. Surgical Procedures

During the study period, 8 neurosurgeons used a variety of surgical methods for these lesions, ranging from a simple biopsy procedure for diagnosis alone to maximal 818

FIG. 1. Illustration showing presumed eloquent areas for UCSF LGG score.

tolerated resection sparing functional brain areas. Some cases involved both pre- and intraoperative functional mapping when tumors were located in or directly adjacent to eloquent brain areas. Intraoperative mapping techniques included awake language and/or motor mapping. Almost all cases involved intraoperative frameless navigation. Outcome Measures and Follow-Up

The main outcome measures were OS and PFS. Progression was defined as an unequivocal increase in the FLAIR or T2-weighted signal abnormality and/or newly detected areas of contrast enhancement on follow-up MR imaging compared with the baseline postoperative MR image obtained within 3 days after surgery. A secondary outcome measure, the extent of resection, was determined externally by neurooncologists, who used a qualitative nonvolumetric comparison of pre- and postoperative MR images. Gross-total resection was confirmed when all preoperatively demonstrated FLAIR or T2 abnormalities were absent on postoperative images. Subtotal resection was therefore any resection that was not gross total. J. Neurosurg. / Volume 109 / November 2008

Prognostic scoring system for low-grade glioma Follow-up information was obtained primarily through chart review, telephone interview, and the National Death Index archives. Telephone follow-up was stopped on February 15, 2007. The follow-up period for OS analysis was defined as the interval between the date of surgery and the date of death or the date the patient was last known to be alive. The follow-up duration for progression analysis was defined as the interval between the date of surgery and the date of the MR image on which the first progression event was detected or the date of the last known MR image without evidence of disease progression (whichever was first). Devising a Scoring System

A primary objective in this study was to devise a simple and reliable scoring system for LGGs. The predictors of the main outcome measures—OS and PFS—were used to determine which factors to incorporate in the scoring system. Continuous variables were dichotomized to facilitate scoring. To evaluate interobserver variability among the different raters, a random subset of 200 patients was selected for blinded scoring (M.M.M. and E.F.C.) based on patient clinical history and MR images. Statistical Analysis

Statistical analysis of potential prognostic factors for the main outcome measures, time-to-death, and time-toprogression was performed using Cox proportional hazards modeling. A logistic regression was used to evaluate potential predictors of GTR. Factors that resulted in a probability value < 0.05 on univariate analysis were entered into multivariate analysis in a backward stepwise fashion. In cases in which multiple cutpoints for a single variable met the criterion, the case with the greatest OR or HR was selected for inclusion in the multivariate analysis. Kaplan–Meier estimates were generated to illustrate the time-to-event curves. A Cohen kappa analysis was used to determine interobserver reliability.

Results Study Population Baseline patient demographics and tumor characteristics are described in Table 1. The median age was 38.2 years (range 18–72 years). Eighty-seven percent of patients presented with seizures. Forty-three patients (15.3%) had a KPS score of 100, with no tumor-related symptoms and lesions that had been discovered as an incidental finding during evaluation for minor head trauma or another disease such as a pituitary mass. The majority of patients (79.7%) had a preoperative KPS score of 90, with minor symptoms such as a presenting seizure or persistent headaches. Fourteen patients (5%) had a KPS score ≤ 80. Tumors were well distributed throughout the supratentorial space, predominantly involving the frontal lobe (72%), temporal lobe (34%), parietal lobe (16%), insula (27%), and occipital lobe (2.5%). An eloquent location was observed in 174 cases (61.9%). Diffuse tumor borders were observed in 35% of patients. The median maximum tumor diameter was 4.6 cm (range 1.4–11 cm). The pathological subtypes of World Health Organization Grade II tumors were as follows: asJ. Neurosurg. / Volume 109 / November 2008

TABLE 1 Summary of characteristics in 281 adults harboring LGGs Characteristic

age in yrs 60 51–60 41–50 31–40 18–30 age at op in yrs median range female sex KPS score 100 90 80 70 history of seizures lesion location frontal temporal parietal insula occipital lt side involved eloquent brain involved diffuse tumor borders max tumor diameter in cm median range tumor diameter in cm 6.0 5.1–6.0 4.1–5.0 3.1–4.0 2.1–3.0 2

No. (%)

8 (2.8) 28 (10.0) 82 (29.2) 108 (38.4) 55 (19.6) 38.2 18–72 110 (39.1) 43 (15.3) 224 (79.7) 13 (4.63) 1 (0.35) 245 (87.2) 201 (72) 96 (34) 44 (15.6) 77 (27.4) 7 (2.5) 148 (52.6) 174 (61.9) 99 (35.2) 4.6 1.4–11 74 (26.4) 46 (16.4) 67 (23.9) 40 (14.3) 41 (14.6) 13 (4.6)

trocytomas, 129 patients (46%); oligodendrogliomas, 112 patients (40%); and oligoastrocytomas, 49 patients (17%). Radiation was administered to 113 patients, and chemotherapy was given to 105. Predictors of OS

Sixty-six deaths occurred. The median follow-up for those still alive was 5.8 years. The Kaplan–Meier median estimated duration of survival for the entire population was 12.0 years (range 0.2–16 years), with a 5-year survival probability of 0.86. Univariate predictors of a lower rate of survival included older age, lower KPS score, eloquent tumor location, greater lesion diameter, and diffuse tumor borders. Temporal and parietal lobe involvement, in contrast, appeared to be favorable predictors on univariate analysis (Table 2). The KPS score was analyzed dichotomously as ≤ 90 or ≤ 80. Multivariate Cox proportional hazards modeling revealed that 4 factors were associated with a shorter OS: presumed eloquent lesion location (HR 4.12, 95% CI 1.71–10.42), KPS score ≤ 80 (HR 3.53, 95% CI 1.56–8.00), age > 50 years (HR 1.96, 95% CI 1.47–3.77), and lesion diameter > 4 cm (HR 3.43, 95% CI 1.43–8.06; Table 3 and Fig. 1). The maximum lesion diameter was dichotomized at 4 cm in the multivariate analysis because it was the size 819

E. F. Chang et al. TABLE 2 Univariate predictors of OS, PFS, and extent of resection OS

STR

Factor

p Value

HR (95% CI)

p Value

HR (95% CI)

p Value

OR (95% CI)

age (yrs) 30 40 50 60 KPS score 90 80 seizures female sex eloquent brain involved max tumor diameter in cm 6 5 4 3 2 diffuse tumor borders tumor location frontal temporal parietal insula occipital

0.012 0.2275 0.1518 0.001 0.0393 0.001 0.0569

0.0001 0.1079 0.9827

0.0001 0.0027 0.1261 0.0006

0.0001 0.0039 0.5185

0.0001

1.03 (1.01–1.06) 1.51 (0.33–2.98) 1.43 (0.43–2.35) 3.52 (0.16–6.32) 3.42 (0.09–11.1) 0.85 (0.80–0.90) 3.09 (0.97–9.67) 8.33 (3.92–17.5) 1.48 (0.78–3.2) 0.994 (0.60–1.65) 7.81 (3.05–18.18) 1.21 (1.07–1.37) 1.55 (0.88–2.74) 2.42 (1.45–4.03) 7.09 (3.16–15.87) 3.04 (1.42–6.49) 1.59 (0.39–6.53) 3.18 (1.98–8.00)

0.26 0.05 0.22 0.99 0.83 0.0027 0.7341 0.0196 0.75 0.85

0.0001 0.0002 0.0003 0.0021 0.0012 0.0067 0.0915 0.0611

0.99 (0.97–1.01) 1.48 (1.01–2.21) 1.27 (0.89–1.80) 0.99 (0.52–1.91) 0.89 (0.28–2.79) 0.93 (0.88–0.97) 1.58 (0.52–2.05) 3.57 (1.83–6.06) 0.97 (0.68–1.40) 0.97 (0.68–1.37) 2.53 (1.71–3.76) 1.18 (1.08–1.29) 1.98 (1.37–2.86) 1.72 (1.22–2.44) 1.90 (1.29–2.81) 1.92 (1.20–3.07) 1.37 (0.76–5.59) 1.39 (0.98–1.95)

0.503 0.5206 0.2334 0.3061 0.638 0.001 0.002 0.9733 0.0113 0.48

0.0001

0.0001

0.0001

0.0001

0.0001

0.0001 0.03

0.0001

1.01 (0.98–1.04) 0.81 (0.42–1.54) 1.36 (0.82–2.77) 1.52 (0.68–3.38) 1.47 (0.29–7.46) 0.88 (0.83–0.94) 3.60 (1.85–7.07) 0.58 (0.25–1.34) 2.47 (1.23–4.98) 1.19 (0.72–1.97) 11.92 (6.60–21.53) 1.74 (1.46–2.08) 6.67 (2.92–15.24) 5.28 (2.90–9.62) 4.91 (2.86–8.41) 4.05 (2.18–7.51) 3.51 (1.11–11.03) 4.53 (2.59–7.93)

0.9797 0.18 0.0145 0.14 0.62

0.98 (0.67–1.44) 0.78 (0.55–1.12) 0.58 (0.38–0.89) 0.76 (0.53–1.10) 0.80 (0.33–1.95)

0.88

0.0001 0.023

0.0001 0.71

0.96 (0.56–1.65) 4.98 (2.63–9.41) 0.49 (0.24–0.90) 19.09 (5.82–62.59) 1.37 (0.26–7.16)

0.856 0.029 0.0012 0.955 0.984

1.05 (0.61–1.82) 0.58 (0.35–0.97) 0.42 (0.24–0.71) 0.65 (0.38–1.09) 1.01 (0.25–4.16)

cutoff associated with the greatest HR (7.09) on univariate analysis. If the maximum diameter was used instead of the dichotomous variable, it did not appear to be significant when entered as a continuous variable in the backward stepwise regression. Notably, only 17 deaths occurred among the 148 patients with tumors based in noneloquent areas of the brain. Predictors of PFS

One hundred thirty-four progression events were observed. The median follow-up was 4.9 years in patients who did not have tumor progression. The Kaplan–Meier estimated median time to radiographically demonstrated progression was 3.57 years (range 0.3–13 years). Univariate analysis showed that eloquent brain involvement, paTABLE 3 Multivariate predictors of OS, tumor progression, and extent of resection Factor

OS eloquent brain involved KPS score 80 patient age 50 yrs tumor diameter 4 cm tumor progression eloquent brain involved STR eloquent brain involved diffuse borders temporal lobe location max tumor diameter insular location parietal lobe location

820

PFS

p Value

0.0018 0.0025 0.0412 0.0014

HR (95% CI)

4.12 (1.71–10.42) 3.53 (1.56–8.00) 1.96 (1.47–3.77) 3.43 (1.43–8.06)

0.0001

2.53 (1.71–3.75)

0.0001 0.01 0.0447 0.0034 0.0035 0.017

9.20 (4.57–18.52) 3.12 (1.29–7.78) 2.35 (1.02–5.44) 1.36 (1.12–1.68) 6.93 (1.89–25.46) 0.32 (0.13–0.77)

tient age (> 30 years only), KPS score, and greater maximum lesion diameter were associated with a worse rate of PFS (p < 0.05, HR > 1). In contrast, parietal lobe involvement was associated with improved PFS (Table 2). Multivariate Cox proportional hazards modeling demonstrated that a presumed eloquent tumor location was the only predictor associated with lowered PFS (HR 2.53, 95% CI 1.71–3.75; Table 3). Predictors of STR and GTR

Gross-total resection was achieved in 92 patients (33%), and STR in 189 patients (67%). Univariate logistic regression analysis showed that a history of seizures, an eloquent lesion location, tumor diameter, diffuse tumor borders, and temporal and insular lobe involvement were associated with STR. Parietal lobe involvement was associated with a greater likelihood of GTR (Table 2). Multivariate logistic regression showed that 4 factors increased the odds of STR: presumed eloquent tumor location (OR 9.20, 95% CI 4.57–18.52), diffuse lesion borders (OR 3.12, 95% CI 1.29–7.78), temporal lobe involvement (OR 2.35, 95% CI 1.02–5.44), insular lobe involvement (OR 6.93, 95% CI 1.89–25.4), and greater maximum tumor diameter (OR 1.36, 95% CI 1.12–1.68). In contrast, parietal lobe tumors appeared to have a relatively increased likelihood of GTR (OR 0.32, 95% CI 0.13–0.77; Table 3). Evaluation of a Proposed Preoperative Scoring System

A scoring system was created based on our determination of the prognostic factors for OS. In the scoring system, 1 point was assigned for each factor present. The final score is the total number of points assigned. The final 4 factors chosen were as follows: 1) presumed eloquence J. Neurosurg. / Volume 109 / November 2008

Prognostic scoring system for low-grade glioma involved (Y/N), 2) KPS score ≤ 80 (Y/N), 3) patient age > 50 years (Y/N), and 4) tumor diameter > 4 cm (Y/N; Table 4). The scoring system was then applied to our data set. Highly significant differences were observed for OS and progression by using the UCSF LGG scoring system (p < 0.001, log-rank test), effectively stratifying the study population into low (Scores 0–1), intermediate (Score 2), and high risk (Scores 3–4) groups. Scores of 0 and 1 were combined, as were Scores of 3 and 4, because there were few patients in each category and only minimal differences between the groups (between 0/1 and 2: p < 0.0001, log-rank test; between 2 and 3/4: p < 0.001, log-rank test). The 5-year cumulative OS probabilities were as follows: Scores 0–1 = 0.97, Score 2 = 0.81, and Scores 3–4 = 0.56. The 5-year PFS probability estimates were as follows: Scores 0–1 = 0.76, Score 2 = 0.49, and Scores 3–4 = 0.18 (Fig. 2 and Table 5). Four examples of the scoring system application are shown in Fig. 3. To determine the interobserver reliability of the grading system, 200 randomly selected patient profiles and MR images were blindly scored by a neurosurgeon attending (M.M.M.) and resident (E.F.C.). A high concordance was found with an overall kappa value of 0.86 (Table 6). The eloquent location variable accounted for the majority of the variance (kappa = 0.82).

Discussion We devised a preoperative scoring system for longterm prognostication in patients with hemispheric LGGs. Four variables (areas of eloquence, patient age > 50 years, KPS score ≤ 80, and lesion diameter > 4 cm) were predictive of survival on multivariate analysis and were therefore used for the scoring system. The main advantages of the proposed scoring system are its simplicity and reliability. The results highlight the predictive value of preoperative clinical and radiographic factors, some of which affect the resectability of LGGs. The application of this scoring system will, we hope, aid clinicians in providing prognostic information to patients at presentation during surgical planning and will allow standardization of surgical and adjuvant therapy clinical trials. Several of the predictors in this study have been identified in other series. Patient age, for example, has been the most well-established predictor of survival in several multivariate analyses.2,12–14,17,20,21 Whereas others have shown a linear relationship between patient age and survival, we observed that patients > 50 years old had a specifically increased risk of death. The cause of this

!"#$ $! ## % # &#' () *


FIG. 2. Graphs showing long-term OS (upper) and PFS (lower) in patients after resection of LGG. Cum. = Cumulative.

association is unclear but may involve a more malignant underlying biology related to an advanced age1 as well as natural life expectancy, higher incidence of other diseases, and general vulnerability to illness. Studies that have documented the cause of death in patients with LGGs are likely to better address this issue. A poor KPS score is another well-documented predictor of survival in other studies.4,11,13–15 Although KPS is a fairly nonspecific indicator of functional status, our analysis confirms its relevance to long-term prognostication. The majority of patients had a KPS score of 90 or 100 at the time of surgery, but a minority of patients with a KPS score of ≤ 80 had a distinctly worse rate of survival. In our patient population, medically refractory uncontrolled seizures and neurological deficits accounted for almost all the cases in which the KPS score was < 90. Fortunately, an advanced age and poor functional status are risks factors that apply to the minority of patients with LGGs. Whereas age and KPS score are important clinical predictors, we found that radiographically identified 821

E. F. Chang et al.

! ' ( ' ' &($ )

! ' ( ' ' &($

" & &

#$

#$

" & &

"

#

%

$"

# ##

"%

#

"

%

# ## " &

%

%

$#

"

$ % #

#

"

&

% $ %

&"

$#

"

"

"$ $# &

%&

#

$

&

& &

$

"&

"

&

$# " $

"

&"

$ #

"

$

$ &

* + !!' ,+! -

features of LGGs were also critical and in fact are more central in predicting PFS. Tumor size and eloquent brain involvement were both strong predictors of STR, and thus it is not surprising that these factors also affected PFS and long-term OS. Tumor size has been identified as an important predictor in several studies.2,3,8,11 Tumor size appeared to linearly predict survival in the present series of patients, but a cutoff at 4 cm was designated mainly to separate bigger from smaller tumors in general. Gross-total resection is more difficult to achieve with larger infiltrative tumors,3 and these lesions reflect a more extensive disease burden. To our knowledge, this is the first study which has demonstrated the long-term significance of an eloquent tumor location on patient survival. Because LGGs typically grow slowly and often spare neural function as they infiltrate eloquent brain areas, resection in many cases can be dangerous and limited, especially when performed without functional mapping. Recent studies involving magnetic source imaging have confirmed functional activity within LGGs.5,18,19 In our series, more than half of the patients had tumors that involved areas located in, or directly adjacent to, presumed eloquent regions. Among the 148 patients with tumors located outside presumed eloquent areas, we observed only 17 deaths, which highlights the strong predictive value of presumed eloquence. Low-grade gliomas located in eloquent brain regions can cause more neurological impairment and preclude complete debulking of the tumor burden. Other imaging features of these tumors were evaluated in devising our scoring system. The presence of discrete tumor borders on the T2-weighted MR images was not a prognostic factor for either time to progression or survival. It was the senior author’s (M.M.M.) impression that a discrete border would make complete resection more likely regardless of an eloquent or noneloquent location. This proved not to be the case. The prospective multicenter Phase III EORTC trials 22844 and 22845, which were focused on the role of radiation therapy dose effects and timing, respectively, also generated similar predictors.8,17,23 Although differences in 822

radiation dose and timing did not appear to affect survival, several pretreatment clinical and imaging factors were identified as important for long-term survival (older age, presence of preoperative neurological deficits, largest tumor diameter, and tumor crossing the midline). Our study confirmed some of these predictors, but there were some notable differences in methodology and findings. The median survival in the EORTC trial was ~ 7.3 years, whereas the median survival in our series was considerably longer at 12.0 years. The reason for this discrepancy is unclear but may reflect advances in early diagnosis and surgical management; the EORTC trials were initiated almost a decade before the present study. Furthermore, CT as opposed to MR imaging was the primary modality used in the EORTC trials. Only 3 patients in our series had LGGs crossing the midline and were therefore not analyzed (too few patients for analysis). More importantly, the effect of an eloquent tumor location on patient survival was absent from the EORTC trial and other previous studies. In the present retrospective study, however, issues of treatment selection bias could not be accounted for. Although our scoring system is simple and easy to apply, it is only meant to guide prognosis and clinical management and research in the preoperative setting. Several advantages in this study included the following: 1) all patients underwent imaging studies after the widespread adoption of MR imaging; 2) a uniform histological diagnosis was conducted by an independent centralized pathology review; 3) telephone interviews and MR imaging reviews were conducted to ensure the best follow-up data possible. Nonetheless, other known prognosticators such as the 1p-19q deletion from tissue specimens6,22 and the actual extent of resection based on postoperative imaging are likely to improve the prediction of long-term survival. More accurate modeling could be accomplished by using actual patient ages and tumor sizes for prognostication as opposed to simple cutoffs. This scoring system should be validated at other institutions. We will attempt to validate the findings of this study by having other institutions apply the scoring system to patients with LGGs. J. Neurosurg. / Volume 109 / November 2008

Prognostic scoring system for low-grade glioma ameter > 4 cm. The total score was inversely proportional to predicted survival. We also evaluated prognosticators for PFS and extent of resection. The application of a standardized scoring system for LGGs should improve clinical decision-making and help guide future research efforts. Acknowledgment We thank Kenny Chung for his help in preparing the manuscript. Disclaimer The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

References

FIG. 3. Four example applications of the UCSF LGG scoring system. A: Score 1: A T2-weighted MR image obtained in a 35-year-old man who had presented with a generalized tonicclonic seizure, revealing a discrete, well-delineated, 3.2-cm right frontal hyperintense mass, with no abnormal enhancement. B: Score 2: A T2-weighted MR image obtained in a 34-year-old woman with a 3-month history of progressively worsening numbness in her left hand, showing a diffuse, ill-defined focus of T2 prolongation in the right frontal lobe measuring 2.2 cm in maximum diameter. This lesion was considered to be in a region of eloquence because it was situated directly adjacent to and overlying the precentral gyrus (motor strip), at the expected location of the hand representation. C: Score 3: An MR image obtained in a 22-year-old woman with persistent headaches for > 2 years, demonstrating an 8.5-cm left temporal nonenhancing lesion with heterogeneous T2 prolongation and involving presumed eloquent perisylvian language areas. D: Score 4: An MR image obtained in a 59-year-old man who had presented with a history of complex partial seizures during the past 4 years, revealing T2 prolongation in the left insular region involving the medial and anterior temporal lobe, inferior frontal cortex, and basal ganglia regions and measuring 5.2 cm in maximum diameter.

Conclusions In this study, we derived a simple and reliable preoperative scoring system for the long-term OS of patients with LGGs. The LGG score is the sum of the presence of 4 factors: involvement of a presumed eloquent location, KPS score ≤ 80, patient age > 50 years, and maximal lesion di

" # $%&' ) '

(

! ! (! * (!


1. Barker FG II, Chang SM, Huhn SL, Davis RL, Gutin PH, McDermott MW, et al: Age and the risk of anaplasia in magnetic resonance-nonenhancing supratentorial cerebral tumors. Cancer 80:936–941, 1997 2. Bauman G, Lote K, Larson D, Stalpers L, Leighton C, Fisher B, et al: Pretreatment factors predict overall survival for patients with low-grade glioma: a recursive partitioning analysis. Int J Radiat Oncol Biol Phys 45:923–929, 1999 3. 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 74:1784–1791, 1994 4. Eyre HJ, Crowley JJ, Townsend JJ, Eltringham JR, Morantz RA, Schulman SF, et al: A randomized trial of radiotherapy versus radiotherapy plus CCNU for incompletely resected lowgrade gliomas: a Southwest Oncology Group study. J Neurosurg 78:909–914, 1993 5. Ganslandt O, Fahlbusch R, Nimsky C, Kober H, Moller M, Steinmeier R, et al: Functional neuronavigation with magnetoencephalography: outcome in 50 patients with lesions around the motor cortex. Neurosurg Focus 6(3):E3, 1999 6. Hamlat A, Saikali S, Chaperon J, Carsin-Nicol B, Calve ML, Lesimple T, et al: Proposal of a scoring scale as a survival predictor in intracranial oligodendrogliomas. J Neurooncol 79:159–168, 2006 7. Janny P, Cure H, Mohr M, Heldt N, Kwiatkowski F, Lemaire JJ, et al: Low grade supratentorial astrocytomas. Management and prognostic factors. Cancer 73:1937–1945, 1994 8. Karim AB, Maat B, Hatlevoll R, Menten J, Rutten EH, Thomas DG, et al: A randomized trial on dose-response in radiation therapy of low-grade cerebral glioma: European Organization for Research and Treatment of Cancer (EORTC) Study 22844. Int J Radiat Oncol Biol Phys 36:549–556, 1996 9. Keles GE, Lamborn KR, Berger MS: Low-grade hemispheric gliomas in adults: a critical review of extent of resection as a factor influencing outcome. J Neurosurg 95:735–745, 2001 10. Kleihues P, Louis DN, Scheithauer BW, Rorke LB, Reifenberger G, Burger PC, et al: The WHO classification of tumors of the nervous system. J Neuropathol Exp Neurol 61:215–219, 2002 11. Kreth FW, Faist M, Rossner R, Volk B, Ostertag CB: Supratentorial World Health Organization Grade 2 astrocytomas and oligoastrocytomas. A new pattern of prognostic factors. Cancer 79:370–379, 1997 12. Laws ER Jr, Taylor WF, Clifton MB, Okazaki H: Neurosurgical management of low-grade astrocytoma of the cerebral hemispheres. J Neurosurg 61:665–673, 1984 13. Leighton C, Fisher B, Bauman G, Depiero S, Stitt L, MacDonald D, et al: Supratentorial low-grade glioma in adults: an analysis

823

E. F. Chang et al.

14.

15.

16.

17.

18.

19. 20.

of prognostic factors and timing of radiation. J Clin Oncol 15: 1294–1301, 1997 Lote K, Egeland T, Hager B, Stenwig B, Skullerud K, BergJohnsen J, et al: Survival, prognostic factors, and therapeutic efficacy in low-grade glioma: a retrospective study in 379 patients. J Clin Oncol 15:3129–3140, 1997 Nicolato A, Gerosa MA, Fina P, Iuzzolino P, Giorgiutti F, Bricolo A: Prognostic factors in low-grade supratentorial astrocytomas: a uni-multivariate statistical analysis in 76 surgically treated adult patients. Surg Neurol 44:208–223, 1995 Piepmeier J, Christopher S, Spencer D, Byrne T, Kim J, Knisel JP, et al: Variations in the natural history and survival of patients with supratentorial low-grade astrocytomas. Neurosurgery 38:872–879, 1996 Pignatti F, van den Bent M, Curran D, Debruyne C, Sylvester R, Therasse P, et al: Prognostic factors for survival in adult patients with cerebral low-grade glioma. J Clin Oncol 20: 2076–2084, 2002 Schiffbauer H, Berger MS, Ferrari P, Freudenstein D, Rowley HA, Roberts TP: Preoperative magnetic source imaging for brain tumor surgery: a quantitative comparison with intraoperative sensory and motor mapping. J Neurosurg 97: 1333–1342, 2002 Schiffbauer H, Ferrari P, Rowley HA, Berger MS, Roberts TP: Functional activity within brain tumors: a magnetic source imaging study. Neurosurgery 49:1313–1321, 2001 Shaw E, Arusell R, Scheithauer B, O’Fallon J, O’Neill B, Di-

824

napoli R, et al: Prospective randomized trial of low- versus high-dose radiation therapy in adults with supratentorial lowgrade glioma: initial report of a North Central Cancer Treatment Group/Radiation Therapy Oncology Group/Eastern Cooperative Oncology Group study. J Clin Oncol 20:2267–2276, 2002 21. Shaw EG, Daumas-Duport C, Scheithauer BW, Gilbertson DT, O’Fallon JR, Earle JD, et al: Radiation therapy in the management of low-grade supratentorial astrocytomas. J Neurosurg 70:853–861, 1989 22. Smith JS, Perry A, Borell TJ, Lee HK, O’Fallon J, Hosek SM, et al: Alterations of chromosome arms 1p and 19q as predictors of survival in oligodendrogliomas, astrocytomas, and mixed oligoastrocytomas. J Clin Oncol 18:636–645, 2000 23. van den Bent MJ, Afra D, de Witte O, Ben Hassel M, Schraub S, Hoang-Xuan K, et al: Long-term efficacy of early versus delayed radiotherapy for low-grade astrocytoma and oligodendroglioma in adults: the EORTC 22845 randomised trial. Lancet 366:985–990, 2005 Manuscript submitted October 1, 2007. Accepted January 23, 2008. Address correspondence to: Edward F. Chang, M.D., Department of Neurological Surgery, University of California, San Francisco, 505 Parnassus Avenue, M779, San Francisco, California 94143. email: ChangEd@neurosurg.ucsf.edu.
