Association between Traumatic Brain Injury and the Subsequent Risk of Brain Cancer

Abstract

This population-based study in Taiwan aimed to investigate the risk of having a diagnosis of malignant brain tumors within 3 years after a traumatic brain injury (TBI). This study used data from the Traumatic Brain Injury Registry and the National Health Insurance Research Database. The study cohort comprised 5007 patients who had visited ambulatory care centers or had been hospitalized with a diagnosis of TBI between 2001 and 2002. The comparison cohort was 25,035 randomly selected enrollees. Each patient's brain cancer status was individually tracked for a 3-year period following their index date. Stratified Cox proportional hazards regressions were performed for analyses. During the 3 years of follow-up, nine patients in each cohort, both the TBI and the non-TBI cohort, were diagnosed with brain cancer. As compared to those patients without TBI, patients with TBI were more likely to receive a diagnosis of malignant brain tumors within the 3-year period following their index date: the incidence rate of malignant brain tumors was 6.28 (95% CI: 3.06–11.53) per 10,000 person-years in patients with TBI and 1.25 (95% CI: 0.61–2.29) per 10,000 person-years in patients without TBI. After adjusting for sociodemographic characteristics, the hazard of being diagnosed with malignant brain tumors during the 3-year follow-up period was 4.67 (95% CI: 1.84–11.83) times greater for those who sustained a TBI than for patients in the comparison cohort. In addition, we found an association between TBI severity and malignant brain tumor among patients with TBI (p=0.033). Our findings suggest a positive correlation between TBI and the relatively short-term development of malignant neoplasms of the brain.

Introduction

P rimary malignant brain tumors represent a heterogeneous group of devastating diseases (Wrensch et al., 2002) characterized by the uncontrolled proliferation and invasion of previously healthy tissues in the brain and spinal cord. This combination of invasion and proliferation can result in permanent brain damage and eventual death. Although a wide range of possible risk factors (e.g., smoking, occupation and industry, exposure to ionizing or non-ionizing radiation, infections, family history) has been examined and proposed over the past several decades, little consensus has been reached in terms of etiology (Bondy et al., 2008 ; Wrensch et al., 2002).

Although controversial, traumatic brain injury (TBI) is one of the relevant risk factors long proposed to predispose to brain tumors. Some case reports presented patients who had developed brain gliomas on the site of previous brain injuries (Anselmi et al., 2006; Henry and Rajshekhar; 2000; Zhou and Liu, 2010). Although these cases fulfilled the previously established causative criteria (Zulch, 1965), and seemed to unequivocally support a relationship between TBI and subsequent tumorigenesis, the unanimity of their results has not been matched in epidemiological studies, which have reported both positive and null associations (Burch et al., 1987; Carpenter et al., 1987; Hochberg et al., 1984; Hu et al., 1998; Inskip et al., 1995; Nygren et al., 2001; Preston-Martin et al., 1998, 1989; Schlehofer et al., 1992; Wrensch et al., 2000; Zampieri et al., 1994).

Whereas methodological obstacles made the interpretation of previous studies difficult, other factors may also have affected the internal validity of each study. Many previous studies relied on surveys to collect their data, and therefore would have been prone to recall bias, which was likely compounded by the small sample sizes. However, even studies using large-scale prospective designs reported similarly inconsistent results, with both null (Annegers et al., 1979; Nygren et al., 2001) and small positive (Inskip et al., 1998) associations between TBI and subsequent incidences of brain tumors being observed.

A link between TBI and the subsequent risk of brain tumor diagnosis cannot be ruled out on the basis of previous literature. Therefore, there is a need for further population-based epidemiological studies to make up for the relatively small prevalences of TBI and the small incidences of brain tumors to achieve necessary statistical rigor (Hochberg et al., 1984; Preston-Martin et al., 1998; Wrensch et al., 2002).

Additionally, international data suggest that the distribution of brain tumors and their relevant risk factors are likely to both range broadly geographically and differ greatly demographically (Davis, 2007; Hwang and Howng, 1992; Kepes et al., 1984). However, almost all studies (except for one case–control study conducted in China at the end of the last century [Hu et al., 1998]) investigating the risk of brain tumors in patients with TBI were from Western societies. No large-scale population-based study has yet been conducted in Asia.

Therefore, we used Taiwan's nationwide population-based database to investigate the risk of being diagnosed with malignant brain tumors within 3 years of sustaining a TBI. As the occurrence of TBI may indicate an increased risk of brain cancer, the consideration of medium- to short-term follow-ups is well warranted to improve early detection and prognosis (Laws and Thapar, 1993).

Methods

Database

We conducted a cohort study with retrospectively collected data to evaluate the link between TBI and a subsequent diagnosis of brain cancer. In this study, we analyzed data sourced from three linked data sets. These three data sets included the Traumatic Brain Injury Registry, the National Health Insurance Research Database (NHIRD), and the Cause of Death data file. The Traumatic Brain Injury Registry is released by the Head & Spinal Cord Injury Research Group of the Taiwan Neurosurgical Society and has been used by numerous researchers to publish studies in internationally peer-reviewed journals. It comprises information drawn from the patient records of 56 major hospitals that make up 80% of all the TBI referral hospitals in Taiwan, and includes information regarding patient medical history, sex, age, head trauma condition and severity (Glasgow Coma Scale), date of hospital admission, presence or absence of intracranial hemorrhage and anatomical location of hemorrhage, injury type, treatment status, outcome (Glasgow Outcome Scale), and whether or not a helmet was worn. In Taiwan, TBI includes cases presenting with brain concussions, skull fractures, brain damage with clinically demonstrated neurological and cognitive deficits, post-traumatic amnesia, neurological sequelae during hospitalization, or any evidence of intracerebral hemorrhage. All the patients who died prior to arriving at the hospital were excluded from the TBI Registry.

The second data set used in this study was the NHIRD. It is derived from the Taiwan National Health Insurance (NHI) program and is provided to scientists in Taiwan for research purposes. The Taiwan NHI program was launched in 1995 and has maintained an enrollment rate of > 95%. As of 2005, the data set included all the medical claims data of ∼ 25,680,000 enrollees of the NHI program. This figure represents > 98% of Taiwan's total population. The Taiwan National Health Insurance Research Claim Database allows researchers to track all the medical services used by enrollees since the initiation of the NHI program.

The third data set adopted for this study was the Cause of Death data file. This data set is provided by the Taiwanese Department of Health (DOH) and covers the years spanning 2001 through 2005. It is mandatory for all deaths throughout Taiwan to be registered with the DOH. This register provides comprehensive information on decedents' sex, age, and marital and employment status, along with the municipality of residence, date and place of death, and diagnoses referring to the cause of death based upon the International Classification of Diseases, Ninth Revision codes (ICD-9-CM code).

The Taiwan NHIRD and the TBI Registry were linked by using each patient's individual identification number and birth date. This was performed by the Head & Spinal Cord Injury Research Group of the Taiwan Neurosurgical Society. The Taiwan Neurosurgical Society addressed all confidentiality protections by encrypting all personal identifiers and fully complying with the data regulations of the Bureau of NHI. Because these data sets consist of de-identified secondary data released for research purposes, this study was exempt from full review by the Taipei Medical University institutional review board (IRB).

Study sample

Our study features a study cohort and a comparison cohort. The study cohort was selected by first identifying all the patients who had ever visited ambulatory care centers (including emergency rooms and outpatient departments of hospitals and clinics) or had been hospitalized with a principal discharge diagnosis of TBI (ICD-9-CM codes 801–804 or 850–854) between January 1, 2001 and December 31, 2002 (n=6379). We excluded those patients who had ever been diagnosed with TBI prior to 2001 (n=125) in order to include only newly diagnosed cases. We also excluded patients who were < 18 years of age (n=1235) to limit the study sample to an adult population. Furthermore, we also excluded patients (n=12) who had been diagnosed with malignant neoplasms of the brain (ICD-9-CM codes 191 or 192) prior to their first ambulatory care visit or hospitalization for the treatment of TBI between 2001 and 2002, which was defined as their index date in this study. Because the NHIRD only allowed us to trace the use of medical services since 1996, we could not exclude subjects who had received a diagnosis of brain cancer prior to 1996. Ultimately, 5007 patients with TBI were included in the study cohort.

We likewise selected the comparison cohort from the NHIRD. We first excluded all the beneficiaries who had ever visited an ambulatory care center or had been hospitalized with a diagnosis of TBI between 1996 and 2005. We additionally excluded patients who were < 18 years of age. We then randomly selected 25,035 beneficiaries (three for every patient with TBI) matched with the study cohort in terms of sex, age (<30, 30–39, 40–49, 50–59, 60–69, 70–79, and >79), and the year of index date by using SAS surveyselect program. A ratio of 3:1 was adopted, as it is generally advised to include no more than four controls per case. Little statistical power is gained by further increasing this ratio (Hennekens and Buring, 1987; Rothman, 1986). We also ascertained that none of the selected patients in the comparison cohort had ever received any diagnosis of brain cancer since 1996 prior to their index date, which was defined as their first use of medical care occurring during the index year.

We used this retrospectively collected data to individually trace each patient for a 3-year period prospectively starting from their index date, to identify those who were subsequently diagnosed with brain cancer (ICD-9-CM codes 191 or 192) during the follow-up period. Patients who died from non-brain cancer causes during the 3-year follow-up period were omitted from the regression models conducted in this study. Of the sampled patients, 2339 died from non-brain cancer causes: 461 from the study cohort (9.2% of study cohort) and 1878 from the comparison cohort (7.5% of the comparison cohort). In this study we have evaluated the assumption of independent removal of cases and found that the removal of both cases and controls was independent of survival time.

Statistical analysis

We used the SAS statistical package (SAS System for Windows, Version 8.2, Cary NC) to perform the statistical analyses in this study. We used the log-rank test to examine the difference in 3-year brain cancer-free survival rates between the two cohorts. Furthermore, stratified Cox proportional hazards regressions (stratified by sex, age group, propensity score, and the year of index health care use) were performed to compare the 3-year brain cancer-free survival rates between the two cohorts after removing the subjects who died from non-brain cancer causes during the study period and adjusting for monthly income, geographic region (Northern, Central, Eastern, and Southern Taiwan), and the urbanization level of the community in which the patient resided. We calculated a propensity score for each patient. A propensity score was initially used to balance demographic characteristics, which were distributed unequally between the study and comparison cohort. All the demographic characteristics available were entered into a multivariable logistic regression model as predictors to calculate each patient's expected probability of being included in the study group. Patients were subsequently grouped into deciles based on their propensity score. The regression model was then stratified by propensity score in deciles to ensure that, within each stratum, comparisons were made for patients with a similar expected probability of being in the study group and, to a large extent, a similar distribution of confounders. We also tested our data and found it to meet the proportionality assumption (survival curves for two strata [study cohort and comparison cohort]) and have hazard functions that are proportional over time. Finally, χ² tests were performed to examine the association between clinical characteristics at baseline (e.g., Glasgow Coma Scale, cerebral hemorrhage) and the presence of brain cancer among patients who sustained a TBI. A two-sided p-value of < 0.05 was considered statistically significant in this study.

Results

Table 1 shows the distribution of demographic characteristics of the sampled patients stratified by the presence of TBI. The mean age for the sampled patients was 43.5 years with a standard deviation of 19.2 years. After matching for sex and age group, patients with TBI had a greater tendency to have a monthly income between NT$15,841 (NT is an abbreviation for “New Taiwan” dollar) and NT$25,000 and to reside in the communities that were located in the southern part of Taiwan (p<0.001) and in the least urbanized region (p<0.001), than did patients without TBI.

Table 1.

Demographic Characteristics for Patients with Traumatic Brain Injury between 2001 and 2005 and Comparison Patients in Taiwan (n=30,042)

	Patients with TBI n=5007		Comparison patients n=25,035
Variable	Total no.	Column %	Total no.	Column %	p value
Gender					> 0.999
Male	2536	50.7	12,680	50.7
Female	2471	49.3	12,355	49.3
Age (years)					> 0.999
<30	1718	34.3	8590	34.3
30–39	775	15.5	3875	15.5
40–49	791	15.8	3955	15.8
50–59	592	11.8	2960	11.8
60–69	499	10.0	2495	10.0
70–79	403	8.1	2015	8.1
>79	229	4.6	1145	4.6
Monthly income					<0.001
0	1868	37.3	10,180	40.7
NT$1–15,840	831	16.6	3692	14.7
NT$15,841–25,000	1751	35.0	7551	30.2
≥NT$25,001	557	11.1	3612	14.4
Geographic region					<0.001
Northern	2014	40.2	11,814	47.2
Central	1198	23.9	6065	24.2
Southern	1661	33.2	6564	26.2
Eastern	134	2.7	592	2.4
Urbanization level					<0.001
1 (most urbanized)	1258	25.1	7553	30.2
2	1518	30.3	7159	28.6
3	884	17.7	4271	17.1
4	691	13.8	3414	13.6
5 (least urbanized)	656	13.1	2638	10.5

TBI, traumatic brain injury.

The incidence of being diagnosed with malignant brain tumors during the 3-year follow-up period after the index date is shown in Table 2. During the 3 years of follow-up, nine patients each in both TBI and non-TBI cohorts were diagnosed with brain cancer. As compared with those patients without TBI, patients with TBI were more likely to receive a diagnosis of malignant brain tumors within the 3-year period following their index date: the incidence rate of malignant brain tumors was 6.28 (95% CI: 3.06–11.53) per 10,000 person-years in patients with TBI and 1.25 (95% CI: 0.61–2.29) per 10,000 person-years in patients without TBI. We performed the log-rank test, which suggested that patients with TBI had significantly lower 3-year brain cancer-free survival rates when compared with patients without TBI (χ² value:14.413; p<0.001).

Table 2.

Crude and Adjusted Hazard Ratios for Brain Cancer During the 3-Year Follow-Up Period for Patients with Traumatic Brain Injury and Comparison Patients in Taiwan (n=30,042)

	Total sample	Patients with TBI n=5007	Comparison patients n=25,035
Presence of brain cancer	No. %	No. %	No. %
Yes	18 (0.06)	9 (0.18)	9 (0.04)
Incidence rate per 10,000 person-years (95% CI)	2.08 (1.27-3.22)	6.28 (3.06-11.53)	1.25 (0.61-2.29)
Crude HR (95% CI)		4.78^*** (1.89-12.12)	1.00
Adjusted HR (95% CI)		4.67^** (1.84-11.83)	1.00

p<0.01; ^*** p<0.001.

Hazard ratio (HR)was calculated by using stratified Cox proportional regression (stratified on sex, age group, propensity score, and the year of index date) with cases omitted if individuals died from non-brain-cancer causes during the 3-year follow-up period. Adjustments are made for patient's geographical location, urbanization level, and monthly income.

TBI, traumatic brain injury.

Table 2 further presents the crude and adjusted hazard ratios (HR) for malignant brain tumors within the 3-year study period between the cohorts. Stratified Cox proportional hazards regressions (stratified on sex, age group, propensity score, and the year of index date) revealed that the HR for brain cancer within the study period for patients with TBI was 4.78 (95% CI: 1.89–12.12, p<0.001) compared with patients without TBI. After removing subjects who died from non-brain cancer causes and adjusting for monthly income, geographic region, and urbanization level of the community in which the patient resided, the HR for receiving a diagnosis of malignant brain tumors during the 3-year follow-up period following an index date for patients with TBI was 4.67 (95% CI: 1.84–11.83, p<0.01) times the risk of patients without TBI.

Table 3 presents the clinical characteristics of the sampled TBI patients who subsequently were diagnosed with malignant brain tumors during the 3-year follow-up period. We found an association between TBI severity and malignant brain tumor among patients with TBI (p=0.033). In addition, there were significant differences in rates of malignant brain tumors among patients classified by cerebral hemorrhage, skull fracture, loss of consciousness, subarachnoid hemorrhage, and subdural hematoma.

Table 3.

Clinical Characteristics at Baseline for the Sampled Patients with Traumatic Brain Injury in Taiwan, 2001–2002 (n=5007)

	Presence of brain cancer
	Yes		No
Variable	Total no.	Row %	Total no.	Row %	p value
Glasgow Coma Scale					0.033
Severe (3–8)	3	0.45	668	99.55
Moderate (9–12)	1	0.95	104	99.05
Mild (13–15)	5	0.12	4226	99.88
Cerebral hemorrhage					0.010
Yes	2	0.89	223	99.11
No	7	0.15	4775	99.85
Skull fracture					0.043
Yes	1	1.05	94	98.95
No	8	0.16	4904	99.84
Loss of consciousness					< 0.001
Yes	3	1.18	252	98.82
No	6	0.13	4746	99.87
Subarachnoid hemorrhage					< 0.001
Yes	2	1.74	113	98.26
No	7	0.14	4885	99.86
Epidural hematoma					–
Yes	–		35	100
No	9	0.18	4963	99.82
Subdural hematoma					0.015
Yes	1	1.39	71	98.61
No	8	0.16	4927	99.84

Discussion

To the best of our knowledge, this is the first large-scale population-based study conducted investigating the relationship between TBI and malignant brain tumors in an Asian society. Our findings suggested that the occurrence of TBI was independently associated with a nearly 4.67-fold increase in the risk of being diagnosed with malignant brain tumors during a 3-year follow-up period in adults ≥18 years of age after adjusting for sociodemographic characteristics. Furthermore, we found an association between TBI severity and malignant brain tumor among patients with TBI.

There has been a debate in the literature in following patients who previously sustained a TBI over time for further development of brain tumors. In specifically examining the risk of malignant neoplasm of the brain, our positive findings were consistent with a large-scale epidemiological investigation implemented in Denmark (Inskip et al., 1998). Using nationwide registries of hospital discharges and incident cancers, Inskip and associates evaluated a cohort of 228,055 Danish residents. They reported that during the following 15 years, TBI might contribute a small increase in the overall risk of brain tumors, especially during the first year after TBI. Although increased early detection may help explain elevated risks of brain tumors in the first year after TBI, parallel decrease was not observed in cases in the following years. In an international case–control study with 1509 adult brain tumors cases, risk of experiencing a TBI was highest for meningiomas in men (OR: 1.5; 95% CI: 0.9–2.6), especially among those with a 15- to 24-year latency (OR: 5.4; 95% CI:1.7–16.6) (Preston-Martin et al., 1998). Other case reports also presented patients with post-traumatic brain neoplasms who met all the required criteria for a potential causal association between TBI and the subsequent occurrence of brain tumors (Anselmi et al., 2006; Henry & Rajshekhar, 2000; Magnavita et al., 2003).

Despite the presence of positive findings, an almost equivalent weight of evidence displayed a null association. In a comparison of personal histories of TBI among 476 subjects who were newly diagnosed with glioma and their controls in San Francisco, California, head injuries requiring medical attention did not significantly contribute to the risk of subsequent adult glioma (Wrensch et al., 2000). Using registries in Sweden, a large population-based cohort study identified a similar null association between TBI and the risk of primary brain tumors, both overall and in subgroup analysis by tumor subtypes (Nygren et al., 2001). In general, the discrepancies in results between ours and the previous inconsistent studies may be the result of methodological variations (e.g., various study designs, grouping of brain tumors, sample size, years to follow-up) or the differences in diagnostic radiology application after TBI (as ionizing radiation is one of the few well-evaluated risk factors for brain tumors).

It is worth concern that the symptoms of early brain tumors (e.g., seizures, ataxia, or aphasia) might contribute to the fall or accident that caused a TBI (Wrensch et al., 2002). However, patients who had been diagnosed with malignant neoplasms of the brain, either before or during the diagnostic assessment at the time of TBI occurrence, were excluded in our study. Furthermore, medical examinations must fulfill the diagnostic criteria stipulated by the NHI before being prescribed.

Our study also identified significant differences in the rates of brain cancer among patients classified by certain clinical characteristics at baseline (e.g., TBI grade, cerebral hemorrhage, skull fracture, loss of consciousness, subarachnoid hemorrhage). Nevertheless, the interpretation of these results should be performed with caution, as they were obtained from a small sample size (only nine patients with malignant brain tumor) with restricted statistical power. Future studies will be needed to investigate and compare the effects of TBI-related clinical characteristics to better understand the clinical significance of these findings.

As brain cancer is most treatable in its earlier stages (Laws and Thapar, 1993) awareness of the association between TBI and brain cancer is critical for both early detection and patient health. Therefore, physicians should be alert to the association detected between TBI and malignant brain tumors within a relatively short follow-up period. Appropriate adherence to screening and regular medical care following TBI could help early detection of important symptoms of brain cancer (e.g., headaches and numbness) and prompt appropriate diagnostic testing and treatment. These procedures may greatly enhance the chance of identifying malignant brain tumors in their earlier, most curable stages (Laws and Thapar, 1993).

Our study has several methodological strengths. First, records of patients with TBI and those in the comparison cohort were retrieved from the Traumatic Brain Injury Registry and the NHIRD claims data sets. As > 98% of the Taiwanese population is currently enrolled in the NHI program in Taiwan, this use of a highly representative nationwide population-based data set with continuous healthcare utilization records may decrease the effect of recall and selection bias. Problems pertaining to non-response and loss to follow-up were also minimized. Second, with the use of a nationwide health insurance claims data set, longitudinal records for a large sample of patients are available to appropriately examine the association between two relative rare events (i.e., the relatively low prevalences of TBI and the even lower incidences of brain tumors (Preston-Martin et al., 1998; Zhou and Liu, 2010;). Third, the identification of malignant brain tumors in our study is valid and definite. In Taiwan's NHI program, biopsy and histological verification are required before a precise diagnosis of carcinoma can be made and treatment can proceed.

In spite of the abovementioned strengths, several limitations should be noted. First, lack of active surveillance during the follow-up period in both the TBI and non-TBI cohorts might contribute to non-differential misclassification and bias our results toward the null. Furthermore, the study design of our investigation could not rule out the possibility that the increased risk of being diagnosed with a malignant brain tumor diagnosis subsequent to sustaining a TBI may be partly attributed to an ascertainment bias. More specifically, TBI patients might have residual impairments and therefore receive follow-up for their injuries (e.g., MRIs, CT scans) more frequently than non-TBI patients. This increased exposure to the medical system may have led to a higher detection rate of subsequent brain cancer than for controls who did not have any head/brain related conditions. If this study did suffer from an ascertainment bias, it would have resulted in an overestimation of the association between TBI and subsequent brain cancer. Therefore, the results of the current study should be interpreted with caution, and further studies accounting for the potential impact of ascertainment bias are needed.

Second, our study emphasized the importance of TBI as a signal for impending brain cancer within a relatively short follow-up time. Some malignant brain tumors may be characterized by longer latencies, and require more time to develop. Third, the lack of information about the external causes of TBI (e.g., motor vehicle crashes, falls, violence) and the histological information of carcinoma further made subgroup analysis impossible. We could not determine certain patient characteristics or lifestyle related factors (e.g., family history, diet, physical activity level, smoking, alcohol consumption, and body mass index) from the claims data set used in our study. However, little consensus has been reached in terms of any definite risk or etiological factors for brain tumors (Bondy et al., 2008; Wrensch et al., 2002), with no known factors having been established to cause any substantial confounding impact on the link between TBI and subsequent brain cancer (Nygren et al., 2001). Therefore, it is moreover possible that the generalization of these results may be limited to the population of Taiwan and similar Asian nations, as relevant risk factors for brain tumor are likely to range broadly by geographic region and demographic characteristics.

In conclusion, over a 3-year study period, patients with TBI were observed to have an increased risk of subsequent diagnosis of malignant neoplasms of the brain. It is the hope of the authors that this research will help improve diagnostic procedures to aid in the early detection of brain tumors, and, through early treatment, improve prognosis. More studies are needed to further elucidate the association between TBI and the heterogeneous groups of brain tumors before prevention, early identification, and better prognosis may further be made possible.

Footnotes

Acknowledgment

This study is based, in part, on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health, Taiwan and managed by the National Health Research Institutes. The interpretations and conclusions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health, or the National Health Research Institutes.

Author Disclosure Statement

No competing financial interests exist.

References

Annegers

J.F.

, Laws

E.R.

Jr. , Kurland

L.T.

, Grabow

J.D.

1979. Head trauma and subsequent brain tumors. Neurosurgery, 4:203–206.

Anselmi

, Vallisa

, Bertè

, Vanzo

, Cavanna

2006. Post-traumatic glioma: report of two cases. Tumori, 92:175–177.

Bondy

M.L.

, Scheurer

M.E.

, Malmer

, Barnholtz–Sloan

J.S.

, Davis

F.G.

, Il'yasova

, Kruchko

, McCarthy

B.J.

, Rajaraman

, Schwartzbaum

J.A.

, Sadetzki

, Schlehofer

, Tihan

, Wiemels

J.L.

, Wrensch

, Buffler

P.A.

Brain Tumor Epidemiology Consortium. 2008. Brain tumor epidemiology: consensus from the Brain Tumor Epidemiology Consortium. Cancer, 113:1953–1968.

Burch

J.D.

, Craib

K.J.

, Choi

B.C.

, Miller

A.B.

, Risch

H.A.

, Howe

G.R.

1987. An exploratory case–control study of brain tumors in adults. J. Natl. Cancer Inst., 78:601–609.

Carpenter

A.V.

, Flanders

W.D.

, Frome

E.L.

, Cole

, Fry

S.A.

1987. Brain cancer and nonoccupational risk factors: a case–control study among workers at two nuclear facilities. Am. J. Public Health, 77:1180–1182.

Davis

F.S.

2007. Epidemiology of brain tumors. Expert Rev. Anticancer Ther., 7:S3–6.

Hennekens

C.H.

, Buring

J.E.

1987. Epidemiology in Medicine. Little, Brown and Company: Boston

Henry

P.T.

, Rajshekhar

2000. Post-traumatic malignant glioma: case report and review of the literature. Br. J. Neurosurg., 14:64–67.

Hochberg

, Toniolo

, Cole

1984. Head trauma and seizures as risk factors of glioblastoma. Neurology, 34:1511–1514.

10.

, Johnson

K.C.

, Mao

, Guo

, Zhao

, Jia

, Bi

, Huang

, Liu

1998. Risk factors for glioma in adults: a case–control study in northeast China. Cancer Detect. Prev., 22:100–108.

11.

Hwang

S.L.

, Howng

S.L.

1992. An analysis of brain tumors in south Taiwan. Kaohsiung J. Med. Sci., 8:656–664.

12.

Inskip

P.D.

, Linet

M.S.

, Heineman

E.F.

1995. Etiology of brain tumors in adults. Epidemiol. Rev., 17:382–414.

13.

Inskip

P.D.

, Mellemkjaer

, Gridley

, Olsen

J.H.

1998. Incidence of intracranial tumors following hospitalization for head injuries (Denmark) Cancer Causes Control, 9:109–116.

14.

Kepes

J.J.

, Chen

W.Y.

, Pang

L.C.

, Kepes

1984. Tumors of the central nervous system in Taiwan, Republic of China. Surg. Neurol., 22:149–156.

15.

Laws

E.R.

Jr. , Thapar

1993. Brain tumors. CA Cancer J. Clin., 43:263–671.

16.

Magnavita

, Placentino

R.A.

, Mei

, Ferraro

, Di Trapani

2003. Occupational head injury and subsequent glioma. Neurol. Sci., 24:31–33.

17.

Nygren

, Adami

, Ye

, Bellocco

, af Geijerstam

J.L.

, Borg

, Nyrén

2001. Primary brain tumors following traumatic brain injury—a population-based cohort study in Sweden. Cancer Causes Control, 12:733–737.

18.

Preston–Martin

, Mack

, Henderson

B.E.

1989. Risk factors for gliomas and meningiomas in males in Los Angeles County. Cancer Res., 49:6137–6143.

19.

Preston–Martin

, Pogoda

J.M.

, Schlehofer

, Blettner

, Howe

G.R.

, Ryan

, Menegoz

, Giles

G.G.

, Rodvall

, Choi

N.W.

, Little

, Arslan

1998. An international case–control study of adult glioma and meningioma: the role of head trauma. Int. J. Epidemiol., 27:579–586.

20.

Rothman

K.J.

1986. Modern Epidemiology: Little, Brown and Company: Boston.

21.

Schlehofer

, Blettner

, Becker

, Martinsohn

, Wahrendorf

1992. Medical risk factors and the development of brain tumors. Cancer, 69:2541–2547.

22.

Wrensch

, Miike

, Lee

, Neuhaus

2000. Are prior head injuries or diagnostic X-rays associated with glioma in adults? The effects of control selection bias. Neuroepidemiology, 19:234–244.

23.

Wrensch

, Minn

, Chew

, Bondy

, Berger

M.S.

2002. Epidemiology of primary brain tumors: current concepts and review of the literature. Neuro. Oncol., 4:278–299.

24.

Zampieri

, Meneghini

, Grigoletto

, Gerosa

, Licata

, Casentini

, Longatti

P.L.

, Padoan

, Mingrino

1994. Risk factors for cerebral glioma in adults: a case–control study in an Italian population. J. Neurooncol., 19:61–67.

25.

Zhou

, Liu

2010. Post-traumatic glioma: report of one case and review of the literature. Int. J. Med. Sci., 7:248–250.

26.

Zulch

K.J.

1965. Brain Tumors. Their Biology and Pathology, 2nd. Springer–Verlag Publisher: New York.