An estimated 24,560 malignant tumors of the brain and central nervous system are projected to occur in 2013 (7). Seventy percent of all adult malignant primary brain tumors are gliomas, that is, tumors arising from glial cells (27). The risk for glioma increases with age is approximately 40% greater for males than females and is greater in whites than blacks (39). Taller individuals may be at greater risk than those who are short (2,12). The only established modifiable risk factor for glioma is exposure to ionizing radiation (15), which accounts for only a small proportion of cases. Prognosis depends on age, tumor grade, histology, prior progressions, and performance status (28,40).
Studies to date reported only a weak, mostly nonsignificant effect of physical activity on glioma, including the Million Woman Study cohort (2) and the European Prospective Investigation into Cancer and Nutrition study (23). The National Institutes of Health–American Association of Retired Persons study of slightly more than 300,000 subjects found 35% lower glioma risk for subjects who recalled being physically active when they were between 15 and 18 yr old, but no significant risk reduction for activity later in life (24). Prognosis may be improved with exercise, that is, Ruden et al. (28) reported that median survival time was 68% greater in patients with recurrent malignant glioma who exercised ≥1.3 MET·h·d−1 versus less active patients. Performance status has also been associated with prognosis (40), and although physical activity is a strong determinant of performance status, their effects on prognosis are apparently independent (28).
This report tests whether exercise deceases the risk of brain cancer mortality prospectively in the National Walkers’ and Runners’ Health Study cohorts (32–38). These cohorts were specifically designed to maximize the statistical power to detect exercise–health associations. In addition to their large sample size, the broad range of energy expenditures, and the use of subjects knowledgeable of their exercise routines, their exercise energy expenditures were calculated from kilometers walked and run, which has been shown to be a superior metric to traditional time-based calculations (34–36). This is important because nondifferential errors in recall of physical activity are likely to bias results toward the null in most existing studies (21).
For these reasons, significant associations between exercise and brain cancer might be detected in the specialized cohorts of the current report, but not in general-purpose cohorts of primarily sedentary individuals reported by others.
MATERIALS AND METHODS
Mortality surveillance through December 31, 2008, was carried out by the National Death Index for three cohorts: the first and the second National Runners’ Health Study cohorts (NRHS-I and NRHS-II) and the National Walkers’ Health Study (NWHS). NRHS-I was recruited between 1991 and 1994 (primarily 1993), whereas NRHS-II and NWHS were recruited primarily between 1998 and 2001 (30–38). The three cohorts may be more accurately characterized as a single cohort that targeted the runners and walkers because all three used the same questionnaire (modified slightly for the different activities), the same sampling domain (subscription lists to running and walking publications and running and walking events), and the same survey staff and were funded by the same grants.
Participants completed baseline questionnaires on exercise, height, current and past body weight, diet, current and past cigarette use, and history of disease. The runners reported the usual miles run per week, and the walkers reported the usual miles walked per week and the usual pace (min·mile−1). These were used to estimate energy expenditure in terms of METs, where 1 MET is the energy expended sitting at rest (3.5 mL O2·kg−1·min−1) (13). In walkers, MET-hours per day walked was calculated by converting reported distance into duration (i.e., distance divided by miles per hour), which was then multiplied by the MET value for the reported pace (34,36). In runners, MET-hours per day run was calculated as kilometers run × 1.02 MET·h·km−1 (34,35). Previously, we have reported strong correlations between repeated questionnaires for self-reported running distance (r = 0.89) (32).
Education was solicited by requesting the participant to provide “years of education (e.g., HS = 12, BS or BA = 16, MS or MA = 18, PhD or MD = 20).” Height and weight were determined by asking the participant, “What is your current height (in inches, without shoes)?” and “What is your current weight (prepregnancy weight if pregnant)?” Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Intakes of meat and fruit were based on the questions “During an average week, how many servings of beef, lamb, or pork do you eat?” and “During an average week, how many pieces of fruit do you eat?” Correlations between these responses and values obtained from 4-d diet records in 110 men were r = 0.46 and r = 0.38 for consumptions of meat and fruit, respectively. Alcohol intake was estimated from the corresponding questions for 4-oz (112 mL) glasses of wine, 12-oz (336 mL) bottles of beer, and mixed drinks and liqueurs. Alcohol was computed as 10.8 g per 4-oz glass of wine, 13.2 g per 12 oz bottle of beer, and 15.1 g per mixed drink. The study protocol was approved by the University of California Berkeley committee for the protection of human subjects, and all subjects provided a signed statement of informed consent.
The underlying and contributing (entity axis) causes of death were obtained from the National Death Index mortality surveillance through December 31, 2008. Possible matches between cohort members and the National Death Index database needed to agree on at least one of the following conditions: 1) social security number; 2) exact month and ±1 yr of birth and first and last names; 3) exact month and ±1 yr of birth and first name, middle initial(s), and last name; 4) exact month and day of birth and first and last names; 5) exact month and day of birth and first name, middle initial(s), and last name; or 6) exact month and year of birth and first and last names with the father’s surname on the National Death Index record. Agreement on last names was based on exact spelling or common misspellings. Multiple records were submitted for each participant to cover potential name variations (e.g., nicknames and previous married names). The National Death Index assigned probability scores to potential matches, and high probability score matches were further reviewed by survey staff for acceptance while blinded to exercise level and other variables that could influence mortality.
Cox proportional hazard analyses (STATA version 11.1; StataCorp, College Station, TX) were used to test whether brain cancer deaths (International Classification of Disease version 9 code 191 and version 10 code C71) were significantly related to MET-hours per day run or walked and other risk factors when adjusted for sex, baseline age (age and age2), education, winter birth, being white, and cohort effects. Results are presented as hazard ratios (HR) and their percent reductions in the risk (calculated as 100 [HR − 1]) for three categories of running or walking energy expenditure: 1) falling short or achieving the current physical activity recommendations for health (<750 MET·min·wk−1 =1.8 MET·h·d−1), 2) exceeding the recommendations by 1- to 2-fold (1.8–3.5 MET·h·d−1), and 3) exceeding the recommendations by ≥2-fold (≥3.6 MET·h·d−1) (13). The trend for increased brain cancer risk for winter versus summer births was tested by recoding the birth month as follows: February = 1, March = 5/6, April = 2/3, May = 1/2, June = 1/3, July = 1/6, August = 0, September = 1/6, October = 1/3, November = 1/2, December = 2/3, and January = 5/6. This coding of the birth month represents a linear change in risk between the highest (February) and lowest (August) risk months as identified in the article by Brenner et al. (4). All analyses were verified using logistic regression analyses (also STATA version 11.1), which are not dependent on the proportional hazard assumption.
Of the 153,420 subjects with complete data and eligible for analyses, 18 were excluded for reporting a previous diagnosis of brain cancer on their baseline questionnaires. The remaining sample included 110 deaths that had brain cancer listed as the underlying cause during the 11.7 ± 3.1 yr follow-up (mean ± SD). Table 1 presents their sample characteristics. Social security number was provided by 79.8% of those who ran or walked <1.8 MET·h·d−1, 84.8% of those who ran or walked 1.8–3.5 MET·h·d−1, and 88.7% of those who ran or walked ≥3.6 MET·h·d−1.
Cox proportional hazard analyses
Runners and walkers were combined because the brain cancer risk reduction did not differ significantly between MET-hours per day run and MET-hours per day walked (P = 0.66). When adjusted for sex, age, race, education, and cohort effects, the risk for brain cancer mortality was 43.2% lower for those who walked or ran 1.8–3.6 MET·h·d−1 (95% confidence interval [CI] = 2.6%–66.8% lower, P = 0.04) and 39.8% lower for those who walked or ran ≥3.6 MET·h·d−1 (95% CI = 0.0%–64.0% lower, P = 0.05) when compared with <1.8 MET·h·d−1. These results provide little evidence for additional risk reduction higher than 1.8 MET·h·d−1, although there may be limited statistical power to detect any additional improvement. Table 2 shows that when pooled, the runners and walkers who expended ≥1.8 MET·h·d−1 had a 42.5% lower risk of brain cancer mortality for the entire sample and 40.0% lower risk when three deaths that occurred within 1 yr of the baseline survey were excluded (95% CI = 1.3%–62.4%, P = 0.04). The risk reduction was weakened slightly when race, education, type of activity, and cohort effects were disregarded (P = 0.07) and was somewhat stronger for subjects ≥50 yr of age versus younger subjects at baseline.
The risk for brain cancer mortality was also 4.13-fold greater for whites than nonwhites (95% CI = 1.02- to 16.80-fold, P = 0.05) and 1.91-fold greater for winter births (February vs August, 95% CI = 1.01- to 3.59-fold, P = 0.05) but was unrelated to height (HR = 0.36 per meter, 95% CI = 0.03–5.09, P = 0.45), BMI (HR = 1.00 per kilogram per square meter, 95% CI = 0.95–1.06, P = 0.87), medications for hypertension (HR = 1.04, 95% CI = 0.56–1.92, P = 0.11) or diabetes (HR = 0.51, 95% CI = 0.07–3.68, P = 0.80), or intakes of red meat (HR = 0.86 per serving per day, 95% CI = 0.49–1.50, P = 0.59), fruit (HR = 0.90 per piece per day, 95% CI = 0.76–1.07, P = 0.23), or alcohol (HR = 1.00 per gram per day, 95% CI = 0.99–1.00, P = 0.95). The adjustment for these additional variables did not eliminate the significantly lower risk for brain cancer mortality for ≥1.8 MET·h·d−1 versus <1.8 MET·h·d−1 run or walked, including adjustment for BMI (Table 2). The reduction in brain cancer mortality for ≥1.8 versus <1.8 MET·h·d−1 also remained significant when the data were restricted to subjects who provided their social security numbers (41.4% risk reduction, 95% CI = 1.9%–65.0%, P = 0.04).
Verification using logistic regression analyses
The results cited previously are entirely consistent with those from the less-restrictive logistic regression analyses that adjusted for follow-up duration, that is, the adjusted odds for brain cancer mortality were as follows: 1) 43.2% lower for those who walked or ran 1.8–3.5 MET·h·d−1 (95% CI = 2.4%–66.9%, P = 0.04), 2) 40.3% lower for those who walked or ran ≥3.6 MET·h·d−1 (95% CI = 0.0%–64.5%, P = 0.05), 3) 4.1-fold greater for whites than nonwhites (95% CI = 1.0- to 16.7-fold, P = 0.05), and 3) 1.9-fold greater for winter than summer births (February vs August, 95% CI = 1.0- to 3.6-fold, P = 0.05). Brain cancer mortality was unrelated to height (P = 0.44), BMI (P = 0.85), hypertension medication (P = 0.94), diabetes medications (P = 0.46), and intakes of red meat (P = 0.59), fruit (P = 0.24), or alcohol (P = 0.96).
Our analyses demonstrate a significant, inverse association between baseline exercise and brain cancer mortality. Specifically, the risk for brain cancer mortality was 43.2% lower for those who ran or walked 1.8–3.6 MET·h·d−1 (equivalent to 12- to 25-km running or 19- to 37-km brisk walking per week) and was 39.8% lower for those who ran or walked >3.6 MET·h·d−1 compared with those less active (<1.8 MET·h·d−1). We combined not meeting the exercise recommendations (i.e., <1.07 MET·h·d−1) with achieving the recommendations (i.e., 1.07–1.8 MET·h·d−1) because their HRs were similar and because of the need to increase statistical power. A greater proportion of more-active subjects provided their social security numbers to assist with follow-up, suggesting a possible bias favoring more complete mortality surveillance in the most active subjects. The lower risk for brain cancer mortality with greater exercise was observed despite this bias and was also shown to be significant when the analyses were restricted to subjects who provided social security numbers.
Our analyses also provide additional evidence that glioma or brain tumor mortality risk is greater for whites versus nonwhites (39) and for winter versus summer births (9), even when adjusted for education.
Gliomas are heterogeneous, and their etiologies presumably diverse. In vitro studies suggest a role of the insulin-like growth factor 1 (IGF-1) signaling pathway in glioma proliferation and progression (1,30,31), where overexpression of IGF-1 receptors in glioma cells correlates with histopathologic grade and proliferation index (14), and the disruption of the IGF-1 pathway causes tumor regression (10). Different physical activities appear to affect IGF-1 differently, which may also explain the significant reduction in brain cancer mortality risk with running and walking (reported here), but not total activity (used by other studies). Running and walking are endurance-type exercises. Resting IGF-I concentrations decrease after short-term endurance training (25) and increase after short-term resistance training (3). Running and walking might also reduce glioma risk by reducing the amount of bioavailable IGF-1. Endurance exercise training is reported to increase basal IGFBP-1 concentrations, an important inhibitor of IGF-I bioactivity, while decreasing free and total IGF-I concentrations (25). Immediately after running a marathon, circulating IGFBP-I levels are reported to increase 12-fold while insulin remains unchanged (18). Coincidentally perhaps, the risks for breast, colorectal, prostate, and lung cancers are increased 1.2- to 5-fold in association with elevated IGF-I (29), and the risks for each have been purported to decreased with exercise (19).
The significant reduction in brain cancer mortality could also be related metabolic processes affecting blood pressure, plasma triglyceride concentrations, type 2 diabetes, or insulin resistance. High-grade gliomas have been associated with elevated diastolic blood pressure and plasma triglyceride concentrations (8). In addition, glioma prognosis is made worst by hyperglycemia (6,22), type 2 diabetes, and obesity (5). Elevated insulin concentrations may increase glioma risk because of insulin’s promitotic properties and by increasing free (unbound) IGF-1 by binding competitively to IGF-1 binding proteins (11,16). Running lowers blood pressure, plasma triglyceride concentrations, body weight, and fasting plasma glucose concentrations (32,33).
Low-grade gliomas (WHO grade 2 astrocytomas, oligodendrogliomas, and oligoastrocytomas) occur in young adults between the ages of 30 and 45 yr and are characterized by continuous slow growth and death 5–15 yr after onset (27). Histological classification was not available for our sample; however, low-grade gliomas are expected to represent a greater proportion of younger than older brain cancer deaths. Although running or walking ≥1.8 MET·h·d−1 was associated with significantly lower brain cancer mortality in subjects >50 yr but not younger subjects, there was no significant interaction between exercise and baseline age and, thus, no evidence to suggest that exercise affects brain cancer mortality differently for high versus low grade glioma. The longer survival for low-grade glioma also provides the opportunity for death due to other causes; however, only three deaths listed brain cancers as a contributing cause of death, and their inclusion did not affect the analyses (results not displayed).
Several important caveats warrant consideration. There were only 110 brain cancer deaths, and additional follow-up is required to more firmly establish this potentially important benefit of exercise. The small number of brain cancer deaths meant that even a 40% risk reduction achieved only a P = 0.05 level of significance. As only fatal brain cancers were studied, these analyses cannot distinguish an etiologic from a prognostic effect of exercise. Some diagnoses of brain cancer as the underlying cause may actually be metastatic rather than primary brain tumors, given that the former are 10-fold more common (20). Running, walking, and other baseline variables were self-reported from the participants’ baseline questionnaires. Exercise levels and other subject characteristics could have changed before the onset of brain cancer. In addition, the effects of other (nonrunning and nonwalking) exercises on brain cancer mortality could not be assessed because walking and other exercise were not collected in the original runners cohort, which represented 55% of the brain cancer death. Thus, these results pertain exclusively to the subjects’ primary exercise: walking in the walkers and running in the runners. Because the exercise performed was self-selected, it is not known whether exercise caused the reduction in brain cancer mortality, or if persons with less susceptibility to brain cancer chose to exercise. Although self-reported running and walking distances have been shown 1) to be highly reproducible in repeated questionnaires (32); 2) to show stronger relationships to body weight and the prevalence of hypertension, high cholesterol, diabetes than time-based running and walking measurements (34–36); and 3) to be predictive of multiple diseases prospectively (32,33,37,38), they have not been validated using objectively measured distance. Vital status was known only from the National Death Index, and therefore, some subjects who have died are likely to be misclassified as alive.
In conclusion, these analyses may provide the best evidence to date that regular exercise reduces brain cancer risk prospectively. They suggest a benefit to exercising at a greater dose (e.g., running 12–25 km·wk−1) than currently recommended (e.g., running 7.3–12 km·wk−1). Although we did not find a greater benefit for running over walking, a substantially greater proportion of by runners (88.8%) than walkers (52.1%) managed to exercise ≥1.8 MET·h·d−1, which might simply reflect the practical advantage of achieving the same amount of exercise in half the time. It has been suggested that 40% brain cancer patients already meet national guidelines for physical activity (17); however, this is probably an overestimate because: 1) physical activity questionnaires that record frequency, intensity, and time spent being physically active substantially overestimate the proportion who meet the guidelines (26), and 2) the survey’s 28% response rate was probably somewhat enriched with exercisers. Although our analyses cannot test whether exercise specifically improves survival in brain cancer patients, it is not unreasonable to expect that if physical activity decreases the risk of incident glioma, it might also extend survival, as has been reported (28).
This research was supported by grant HL094717 from the National Heart, Lung, and Blood Institute and was conducted at the Ernest Orlando Lawrence Berkeley National Laboratory (Department of Energy DE-AC03-76SF00098 to the University of California). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The author has declared that no competing interests exist.
Results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. Belfiore A, Frasca F, Pandini G, Sciacca L, Vigneri R. Insulin receptor isoforms and insulin receptor/insulin-like growth factor receptor hybrids in physiology and disease. Endocr Rev
. 2009; 30: 586–623.
2. Benson VS, Pirie K, Green J, Casabonne D, Beral V. Lifestyle factors and primary glioma and meningioma tumours in the Million Women Study cohort. Br J Cancer
. 2008; 99: 185–90.
3. Borst SE, De Hoyos DV, Garzarella L, et al. Effects of resistance training on insulin-like growth factor-I and IGF binding proteins. Med Sci Sports Exerc
. 2001; 33 (4): 648–53.
4. Brenner AV, Linet MS, Shapiro WR, et al. Season of birth and risk of brain tumors in adults. Neurology
. 2004; 63: 276–81.
5. Chambless LB, Parker SL, Hassam-Malani L, McGirt MJ, Thompson RC. Type 2 diabetes mellitus and obesity are independent risk factors for poor outcome in patients with high-grade glioma. J Neurooncol
. 2012; 106: 383–9.
6. Derr RL, Ye X, Islas MU, Desideri S, Saudek CD, Grossman SA. Association between hyperglycemia and survival in patients with newly diagnosed glioblastoma. J Clin Oncol
. 2009; 27: 1082–6.
7. Dolecek TA, Propp JM, Stroup NE, Kruchko C. CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2005–2009. Neuro Oncol
. 2012; 14 (5 Suppl): v1–49.
8. Edlinger M, Strohmaier S, Jonsson H, et al. Blood pressure and other metabolic syndrome factors and risk of brain tumour in the large population-based Me-Can cohort study. J Hypertens
. 2012; 30: 290–6.
9. Efird JT. Season of birth and risk for adult onset glioma. Int J Environ Res Public Health
. 2010; 7: 1913–36.
10. Gariboldi MB, Ravizza R, Monti E. The IGFR1 inhibitor NVP-AEW541 disrupts a pro-survival and pro-angiogenic IGFSTAT-HIF1 pathway in human glioblastoma cells. Biochem Pharmacol
. 2010; 80: 455–62.
11. Giovannucci E. Nutrition, insulin, insulin-like growth factors and cancer. Horm Metab Res
. 2003; 35: 694–704.
12. Gunnell D, Okasha M, Smith GD, Oliver SE, Sandhu J, Holly JM. Height, leg length, and cancer risk: a systematic review. Epidemiol Rev
. 2001; 23: 313–42.
13. Haskell WL, Lee IM, Pate RR, et al. Physical activity
and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Med Sci Sports Exerc
. 2007; 39 (8): 1423–34.
14. Hirano H, Lopes MB, Laws ER, et al. Insulin-like growth factor 1 content and pattern of expression correlates with histopathologic grade in diffusely infiltrating astrocytomas. Neuro Oncol
. 1999; 1: 109–19.
15. Inskip PD, Linet MS, Heineman EF. Etiology of brain tumors in adults. Epidemiol Rev
. 1995; 17: 382–414.
16. Ish-Shalom D, Christoffersen CT, Vorwerk P, et al. Mitogenic properties of insulin and insulin analogues mediated by the insulin receptor. Diabetologia
. 1997; 40 (2 Suppl): S25–S31.
17. Jones LW, Guill B, Keir ST, et al. Patterns of exercise across the cancer trajectory in brain tumor patients. Cancer
. 2006; 106: 2224–32.
18. Koistinen H, Koistinen R, Selenius L, Ylikorkala Q, Seppälä M. Effect of marathon run on serum IGF-I and IGF-binding protein 1 and 3 levels. J Appl Physiol
. 1996; 80: 760–4.
19. Kushi LH, Doyle C, McCullough M, et al. American Cancer Society Guidelines on nutrition and physical activity
for cancer prevention
: reducing the risk of cancer with healthy food choices and physical activity
. CA Cancer J Clin
. 2012; 62: 30–67.
20. Langley RR, Fidler IJ. The biology of brain metastasis. Clin Chem
. 2013; 59: 180–9.
21. Matthews CE, Moore SC, George SM, Sampson J, Bowles HR. Improving self-reports of active and sedentary behaviors in large epidemiologic studies. Exerc Sport Sci Rev
. 2012; 40: 118–26.
22. McGirt MJ, Chaichana KL, Gathinji M, et al. Persistent outpatient hyperglycemia is independently associate with decreased survival after primary resection of malignant brain astrocytomas. Neurosurgery
. 2008; 62: 286–9.
23. Michaud DS, Bové G, Gallo V, et al. Anthropometric measures, physical activity
, and risk of glioma and meningioma in a large prospective cohort study. Cancer Prev Res (Phila)
. 2011; 4: 1385–92.
24. Moore SC, Rajaraman P, Dubrow R, et al. Height, body mass index, and physical activity
in relation to glioma risk. Cancer Res
. 2009; 69: 8349–55.
25. Nemet D, Connolly PH, Pontello-Pescatello AM, et al. Negative energy balance plays a major role in the IGF-I response to exercise training. J Appl Physiol
. 2004; 96: 276–82.
26. Prince SA, Adamo KB, Hamel ME, et al. A comparison of direct versus self-report measures for assessing physical activity
in adults: a systematic review. Int J Behav Nutr Phys Act
. 2008; 5: 56.
27. Ricard D, Idbaih A, Ducray F, Lahutte M, Hoang-Xuan K, Delattre JY. Primary brain tumours in adults. Lancet
. 2012; 379 (9830): 1984–96.
28. Ruden E, Reardon DA, Coan AD, et al. Exercise behavior, functional capacity, and survival in adults with malignant recurrent glioma. J Clin Oncol
. 2011; 29: 2918–23.
29. Ryan PD, Goss PE. The emerging role of the insulin-like growth factor pathway as a therapeutic target in cancer. Oncologist
. 2008; 13: 16–24.
30. Trojan J, Cloix JF, Ardourel MY, Chatel M, Anthony DD. Insulin-like growth factor type 1 biology and targeting in malignant gliomas. Neuroscience
. 2007; 145: 795–811.
31. Wang H, Wang H, Shen W, et al. Insulin-like growth factor-binding protein 2 enhances GBM invasion by activating invasion-enhancing genes. Cancer Res
. 2003; 63: 4315–21.
32. Williams PT. Relationship of distance run per week to coronary heart disease risk factors in 8283 male runners. The National Runners’ Health Study. Arch Intern Med
. 1997; 157: 191–8.
33. Williams PT. Vigorous exercise
, fitness and incident hypertension, high cholesterol, and diabetes. Med Sci Sports Exerc
. 2008; 40 (6): 998–1006.
34. Williams PT. Advantage of distance- versus time-based estimates of walking in predicting adiposity. Med Sci Sports Exerc
. 2012; 44 (9): 1728–37.
35. Williams PT. Non-exchangeability of running vs. other exercise in their association with adiposity, and its implications for public health recommendations. PLoS One
. 2012; 7: e36360.
36. Williams PT. Distance walked and run as improved metrics over time-based energy estimation in epidemiological studies and prevention
; evidence from medication use. PLoS One
. 2012; 7: e41906.
37. Williams PT. Reduced risk of incident kidney cancer from walking and running. Med Sci Sports Exerc
38. Williams PT, Thompson PD. Walking versus running for hypertension, cholesterol, and diabetes mellitus risk reduction. Arterioscler Thromb Vasc Biol
. 2013; 33: 1085–91.
39. Wrensch M, Minn Y, Chew T, Bondy M, Berger MS. Epidemiology of primary brain tumors: current concepts and review of the literature. Neuro Oncol
. 2002; 4: 278–99.
40. Wu W, Lamborn KR, Buckner JC, et al. Joint NCCTG and NABTC prognostic factors analysis for high-grade recurrent glioma. Neuro Oncol
. 2010; 12: 164–72.