An international case-control study of adult glioma and meningioma: the role of head trauma.
ABSTRACT Increased brain tumour risk after head trauma suggested by case reports and clinical series has been previously studied epidemiologically with mixed results. An international multicentre case-control study investigated the role of head trauma from injury or sports participation in adult brain tumour risk.
In all, 1178 glioma and 330 meningioma cases were individually or frequency matched to 2236 controls. Only exposures that occurred at least 5 years before diagnosis and head injuries that received medical attention were considered.
Risk for ever having experienced a head injury was highest for male meningiomas (odds ratio [OR] = 1.5, 95% confidence interval [CI] : 0.9-2.6) but was lower for 'serious' injuries, i.e. those causing loss of consciousness, loss of memory or hospitalization (OR = 1.2, 95% CI: 0.6-2.3). Among male meningiomas, latency of 15 to 24 years significantly increased risk (OR = 5.4, 95% CI: 1.7-16.6), and risk was elevated among those who participated in sports most correlated with head injury (OR = 1.9, 95% CI: 0.7-5.3). Odds ratios were lower for male gliomas (OR = 1.2, 95% CI : 0.9-1.5 for any injury; OR = 1.1, 95% CI: 0.7-1.6 for serious injuries) and in females in general.
Evidence for elevated brain tumour risk after head trauma was strongest for meningiomas in men. Findings related to sports should be interpreted cautiously due to cultural variability in our data and our lack of complete data on physical exercise in general which appeared to be protective.
- SourceAvailable from: PubMed Central[show abstract] [hide abstract]
ABSTRACT: For a long time, head injury has been considered as a possible causative factor for later development of brain tumors. However, the actual role of previous head trauma in the pathogenesis of intracranial tumors is still a matter of debate, also due to the possible medico-legal implications. Some authors have suggested several criteria for establishing a possible causal relationship between the aforementioned factors. We report a case of a left posterior paraventricular high-grade glioma which developed 20 years after a posttraumatic hematoma occurring in the same area. This case is reported in detail and the relevant literature is reviewed.Case Reports in Oncology 01/2013; 6(2):403-9.
- [show abstract] [hide abstract]
ABSTRACT: This Letter to the Editor is a response to the article entitled: "Association between traumatic brain injury and the subsequent risk of brain cancer" by Chen et al., published in the May 2012 issue of Journal of Neurotrauma.Journal of neurotrauma 05/2012; 29(13):2365-6. · 4.25 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Background Alcohol is capable of traversing the blood-brain barrier and is thus a possible risk factor for brain cancer. Several epidemiological studies have been published on the issue, a number of those during recent years, with inconsistent findings.Materials and methodsWe performed a systematic literature search in the Medline and EMBASE databases. We found a total of 19 studies providing risk estimates for total alcohol or specific alcoholic beverages. Pooled estimates of the relative risks (RR) and 95% confidence intervals (CI) were calculated using random-effects models.ResultsThe pooled RR of brain cancer for alcohol drinkers versus non-drinkers was 0.97 (95% CI 0.82-1.15; based on 12 studies). Moderate (<2 drinks/day) and heavy alcohol drinkers had RRs of 1.01 (95% CI 0.81-1.25) and 1.35 (95% CI 0.85-2.15), respectively. With reference to specific alcoholic beverages, the RRs were 1.01 (95% CI 0.70-1.48) for wine, 0.96 (95% CI 0.82-1.12) for beer, and 1.20 (95% CI 1.01-1.42) for spirit consumption. The RRs for drinkers versus non-drinkers were 0.93 (95% CI 0.81-1.07) for glioma and 0.71 (95% CI 0.45-1.12) for meningioma.Conclusions Alcohol drinking does not appear to be associated with adult brain cancer, though a potential effect of high doses deserves further study.Annals of Oncology 10/2012; · 7.38 Impact Factor
C International Epldcmlological Association 1998 Printed in Great Britain
International Journal of Epidemiology 1998;27:579-586
An international case-control study
of adult glioma and meningioma:
the role of head trauma
Susan Preston-Martin,a Janice M Pogoda,b Brigitte Schlehofer,c Maria Blettner,c Geoffrey R Howe,d
Philip Ryan,e F Menegoz/ Graham G Giles,8 Ylva Rodvall,h N W Choi,1 Julian Little* and Annie Arslank
Background Increased brain tumour risk after head trauma suggested by case reports and
clinical series has been previously studied epidemiologically with mixed results.
An international multicentre case-control study investigated the role of head
trauma from injury or sports participation in adult brain tumour risk.
Methods In all, 1178 glioma and 330 meningioma cases were individually or frequency
matched to 2236 controls. Only exposures that occurred at least 5 years before
diagnosis and head injuries that received medical attention were considered.
Results Risk for ever having experienced a head injury was highest for male meningio-
mas (odds ratio [OR] = 1.5, 95% confidence interval [CI] : 0.9-2.6) but was lower
for 'serious' injuries, i.e. those causing loss of consciousness, loss of memory or
hospitalization (OR = 1.2, 95% CI: 0.6-2.3). Among male meningiomas, latency
of 15 to 24 years significantly increased risk (OR = 5.4, 95% CI : 1.7-16.6), and
risk was elevated among those who participated in sports most correlated with
head injury (OR = 1.9, 95% CI: 0.7-5.3). Odds ratios were lower for male gliomas
(OR= 1.2, 95% CI: 0.9-1.5 for any injury; OR = 1.1,95% CI: 0.7-1.6 for serious
injuries) and in females in general.
Conclusions Evidence for elevated brain tumour risk after head trauma was strongest for
meningiomas in men. Findings related to sports should be interpreted cautiously
due to cultural variability in our data and our lack of complete data on physical
exercise in general which appeared to be protective.
Keywords Head injuries, brain injuries, brain neoplasms, sports medicine
Accepted 2 December 1997
Head trauma as a risk factor for adult brain tumour has been
a controversial topic in medicine for over a century. Although
early case reports and clinical series are often vague and difficult
to evaluate,1 some are quite striking, such as the case reported
in which a piece of wire was extracted from the centre of a
1 University of Southern California (USC), Department of Preventive
Medicine, USC/Norris Comprehensive Cancer Center, 1441 Eastlake
Avenue, Suites 4412- 4413, MS 44, Los Angeles, CA, USA 90033-0800.
b Statology, Truckee, CA, USA.
c German Cancer Research Centre. Heidelberg, Germany.
d Columbia University School of Public Health, New York, NY, USA.
e University of Adelaide, Adelaide, Australia.
' Cancer Registry of bere, Grenoble, France.
8 Victorian Cancer Registry, Victoria, Australia.
h Institute for Environmental Medicine, Stockholm, Sweden.
1 Manitoba Cancer Treatment and Research Foundation, Manitoba, Canada.
J University of Aberdeen Medical School, Aberdeen, Great Britain.
k Internationa] Agency for Research on Cancer, Lyon, France.
meningioma, presumably driven in 20 years earlier by an ex-
plosion in which the patient had been involved.2 In their classic
study, Cushing and Eisenhard reported a history of head trauma
in one-third of their brain tumour patients which led them to
postulate a relationship between head trauma and subsequent
brain tumour development.3
Few epidemiological studies have analysed head trauma and
brain tumour risk. Case-control studies have found meningio-
mas to be associated with serious head injuries or head injuries
requiring medical attention,4"6 and, in men, with increasing
number of such injuries.6 In the only study that included data
related to contact sports, increased risk of meningioma was ob-
served among boxers. Other case-control studies have found
no relationship between serious head injury and meningioma
but included very few cases.7'8 One case-control study of
glioblastoma reported a relative risk of 10.6 for severe head
injury over age 15 but had several methodological limitations,
including the fact that it was not population-based.9 Several
other studies of gliomas in adults have not observed a head
by guest on July 12, 2011
580 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
trauma association,6-7'10"12 and studies of head trauma and
childhood brain tumours, which are primarily gliomas, have
produced conflicting results.13"17 Two additional adult brain
tumour studies that investigated head trauma are part of the
international collaborative study and are thus not detailed
We analysed data from a multicentre international case-
control study of adult brain tumour involving 1509 cases from
eight study centres. Specifically, we tested the hypotheses that
adult glioma and meningioma risk is related to a history of (1)
medically-treated head injuries, and (2) participation in sports
that may result in a serious head blow or repetitive head trauma
(such as boxing).
Data originated from a multicentre international case-control
study conducted to investigate risk factors for primary adult
brain tumour, specifically glioma and meningioma. Investigators
from eight centres in six countries (Adelaide and Melbourne,
Australia; Grenoble, France; Heidelberg, Germany; Toronto and
Winnipeg, Canada; Stockholm, Sweden; and Los Angeles, Cali-
fornia, USA) collaborated to develop the international protocol,
design a standardized questionnaire, and make decisions re-
garding study design, field work, and analysis. The study was
coordinated by the International Agency for Research on Cancer
(Lyon, France), where data from all centres were compiled
and merged into combined data sets. In all, 729 male cases from
seven centres and 779 female cases from eight centres were
included. Inclusion periods varied slightly by study centre, with
an average selection period of 2 years. The range of diagnosis
ages was 20-79 years; most cases were diagnosed in the late
1980s, however diagnosis years ranged from 1984 to 1992. Both
individual (Grenoble, Los Angeles, Melbourne, Stockholm, Win-
nipeg) and frequency (Adelaide, Heidelberg, Toronto) matching
by age and gender was used. Some centres also matched on race
or geographical region. At seven centres (all but Stockholm),
proxy respondents were used for index subjects unavailable for
interview (325 cases, 65 controls).
Information was sought only on medically treated head injuries.
A subgroup, 'serious injuries,' was defined as medically-treated
injuries causing loss of consciousness, loss of memory or requir-
ing hospitalization. Indicator variables representing three ex-
posure periods (5-14, 15-24, and ^25 years before diagnosis)
were created for the analysis of latency; subjects who had
experienced more than one injury could be exposed in more
than one period. Although questionnaires were similar for all
centres, questions about sports participation varied substantially
by centre. Each centre compiled a list of sports common to the
geographical region (therefore each centre's list of sports was
different), and these sports were explicitly asked about; how-
ever some centres also recorded responses on additional sports
participation. For analysis purposes, subjects were considered to
have sports participation exposure if, at least 5 years before
diagnosis, they participated in any sport on that centre's list or
if they participated in another sport that was then added to the
centre's list. Correlation with head injuries (not necessarily
caused by the sport) was computed for each sport among those
subjects (both cases and controls) with exposure data available
for that sport; sports were then trichotomized by degree of
correlation with head injuries (high = r &0.10, moderate = 0.01
< r < 0.10, low = r =s0.01). Education level was used as a meas-
ure of socioeconomic status (SES) according to a seven-point
scale: 7 = college degree; 6 = some college; 5 = technical train-
ing, apprenticeship, or adult classes; 4 = high school graduate; 3
= some high school; 2 = 7-9 years of schooling; 1 = <7 years of
schooling. Subjects who first had a head or neck x-ray at least
5 years before diagnosis were considered to be exposed to head
Maximum likelihood estimates of odds ratios (OR) and 95%
confidence intervals (CI) were computed using both conditional
and, to minimize the problem of missing data within strata,
unconditional logistic regression. For individually matched
studies, strata for conditional analyses were defined by matched
sets; for frequency matched studies, strata were denned by
centre, gender and 5-year age groups. Unconditional analyses
were stratified by centre, gender, and 5-year age groups, and all
controls were used for tumour-specific analyses. Since estimates
were similar using both methods, only results from conditional
analyses are reported. Risk estimates and CI from random
effects models20 (with centre as the random effect) are reported
for exposure effects that signicantly differed by centre, based on
the likelihood ratio tests;21 otherwise, results from fixed effects
models are reported. Correlations between participation in
individual sports and history of head injury were calculated
using a modification of the Cramer coefficient,22 denoted < 1 > 2
in this paper. The three head injury latency indicators were used
in a multiple logistic regression to assess their independent
effects on risk. Similarly, multiple logistic regression was used to
test for independent effects of participation in each of three
categories of sports grouped by correlation with head injuries,
in this paper, multiple logistic regression (involving a single
outcome) is referred to as 'multivariate analysis.' Hypothesis
testing was two-sided.
For analyses involving head injuries occurring ^25 years
before diagnosis, subjects <25 years old at reference date were
excluded. Also, subjects with no opportunity to respond that
they had participated in a particular sport because the sport was
not on their centre's list were excluded from analyses involving
that sport. All analyses excluded exposures occurring within
the 5-year period before diagnosis. Subjects with missing data
(mostly cases) were excluded from analyses involving those
data. Other methods of handling missing data were considered
(defining as unexposed, treating as a separate exposure group)
but had no effect on results. Analyses were performed both
including and excluding proxy data.
Distributions of tumour morphology and topography by major
tumour type are shown in Table 1. Distributions by centre, gen-
der, age group, and education for each of the two major types
are shown in Table 2. Cases were less educated than controls;
this was a significant trend for male gliomas (P = 0.007), which
contrasts with a clear trend of increasing incidence of male
by guest on July 12, 2011
HEAD TRAUMA AND BRAIN TUMOURS 581
Table 1 Distribution of tumour classifications, multicentre international case-control study of adult brain tumours, 1984-1991
Subependymal giant cell astrocytoma
Choroid plexus papilloma
Ependymoma, anaplastic type
Astrocytoma, anaplastic type
Giant cell glioblastoma
Glioblastoma with sarcomatous component
Oligodendroglioma, analplastic type
Other parts of the brain
glioma (as well as all histologies combined) with increasing SES
observed in Los Angeles County. Since education, as a measure
of SES, is likely to be related to medically-treated head injuries
as well as sports participation, it was considered a potential
confounder in all analyses.
Univariate results of medically-treated head injury analyses
are presented in Table 3. Although male meningioma cases
were 1.5 times more likely than controls to have sustained a
medically-treated injury, this difference was not statistically sig-
nificant and risk decreased to 1.15 when restricted to serious
head injuries. For gliomas, OR for any injury and for serious
injuries were not significantly different from 1.0. For female
gliomas, OR significantly varied by study centre (P = 0.02). In
general, cases were more likely than controls to have
experienced more than one injury (OR = 1.4, 95% CI: 1.0-2.0
for gliomas and meningiomas, males and females combined).
Considering all subjects, men were twice as likely as women to
have ever had any head injury (31% versus 16%) or a serious
injury (19% versus 9%), and among women with head injuries,
brain tumour risk was not significantly elevated. Age at first
injury (child versus adult) had no effect on risk (data not shown).
In a multivariate analysis of latency (i.e. length of period from
injury to tumour diagnosis), the only significant increase in risk
was among male meningioma cases, who were five times more
likely than controls to have sustained a head injury 15-24 yean
before diagnosis (P = 0.004; Table 4). There were no significant
increases in risk for any latency period for gliomas or for female
Adjusting for education did not appreciably affect risk es-
timates. Head x-ray exposure was also considered for confound-
ing effects; however, in these data, controls were more likely to
have been exposed to head x-rays, even after controlling for
education. Thus, head x-ray exposure was not a positive con-
founder. Excluding proxy data had little effect on risk estimates.
Weak inverse associations between risk of brain tumour and
sports participation *5 years before diagnosis were observed
(Table 5); this association was statistically significant for gliomas
in men. Level of involvement, measured by total hours of
participation, had no effect on risk (not shown). In a multi-
variate analysis of participation in sports 'groups' defined by
by guest on July 12, 2011
582 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 2 Demographics by tumour type, multicemre International case-control study of adult brain tumours, 1984-1991
High school graduate
Some high school
7-9 years schooling
<7 years schooling
* Number of controls available for tumour-specific analyses (all controls from centres with frequency-matched design, individudally-matched controls from
centres with individually-matched designs).
b Includes technical training and apprenticeships.
Table 3 Medically treated head injuries at least 5 years before diagnosis, conditional unlvariate analysis, multlcentre international case-control
study of adult brain tumours, 1984-1991
(%) No. controls (%) OR (95% CI) (%) No. controls (%) OR (95% CI)
Ever any Injury
No. of Injuries
1 Based on random effects model; P - 0.02 for the test of erposure/centre interaction using the likelihood ratio test;
follows: Adelaide » 0.99 (0.44-2.22), Grenoble = 2.00 (0 50-8.00), Heidelberg = 0.37 (0.14-0.99), Los Angeles =
(1.02-3.44). Stockholm = 1.60 (0.52-4.89), Toronto- 1.77 (0.52-6.05). Winnipeg - 0.82 (0.28-2.41).
b Loss of consciousness, loss of memory, or hospltaliration required.
centre-specific OR (95% CI) were as
0 46 (0 22-0.96). Melbourne = 1.88
by guest on July 12, 2011
HEAD TRAUMA AND BRAIN TUMOURS 583
Table 4 Medically treated head injuries at least 5 years before diagnosis, multivariate conditional analysis of latency periods, multicentre
international study of adult brain tumours, 1984-1991
Years since Injury*
(%) No. controls (%)
(95% CI) (%) No. controls (%) OR (95% CI)
' Groups are not mutually exclusive; subjects <25 years old at reference are excluded; reference group is subjects who did not sustain a head Injury requiring
medical treatment > 5 years prior to diagnosis.
Table 5 Sports participation at least 5 years before diagnosis, conditional analysis, multicentre international case-control study of adult brain
Rugby or football1*
' A subject is exposed if he reported partidparion in any sport on that centre's list or if he partiapated In another sport that was then added to the centre's list.
b American, Australian rules or Gaelic football
degree of correlation with head injuries (Table 6), no significant
risk relationships were observed; however, risk was elevated for
male meningioma cases who participated in sports most highly
correlated with head injuries (Table 7). A much lower elevated
risk was evident for female glioma cases who participated in
sports highly correlated with head injuries. Education was not a
confounder. Unconditional analyses and the exclusion of proxies
produced similar results, although the multivariate sports group
analysis was difficult to evaluate due to sparse data.
The evidence that head injuries are a risk factor for brain tumour
is strongest for meningiomas. Furthermore, almost all anecdotal
cases have involved meningiomas.24 In the present study, risk
of ever having had a head injury was highest for meningiomas
in males, particularly when injuries were sustained 15-24 years
before diagnosis. No such increased risk was observed for female
meningiomas or for either gender when analyses were restricted
to serious injuries (the subset of medically-treated injuries caus-
ing loss of consciousness, loss of memory or requiring hospital-
ization), suggesting recall bias as a possible explanation for the
findings relating to any injury. Yet if recall bias occurred, we
might expect elevated risks regardless of tumour type since it is
not likely that lay people are aware of the meningioma
association. In our data, the only potential risk factor that
resulted in elevated risk for all tumour types was number of
injuries among males. In studies of differential recall of
exposures related to birth outcome, it has been shown that
spurious inferences occur only under extreme conditions and
that concerns about recall bias are overrated.25'26 Another
possible reason for a lack of an association for serious injuries is
that we excluded (by the way respondents were queried) all
injuries that were not treated medically. In an earlier study of
meningiomas in men in Los Angeles County, a sizeable pro-
portion of injuries that had caused loss of consciousness or a
permanent scar had not been treated medically.5
The interval between head injury and tumour diagnosis
noted in published case reports has varied widely and ranges
from 1 to 67 years. Our finding of a strong association of men-
ingioma with trauma 15-24 years before diagnosis is consistent
with a previous study of male meningioma cases in Los Angeles
County, in which risk was significantly associated with head
injuries experienced 3>20 years before diagnosis.6 Latency periods
of this length are feasible for meningiomas, which are usually
benign, slow-growing tumours.27
by guest on July 12, 2011
584 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Table 6 Correlations between reported head injuries and individual sports, multicentre international case-control study of adult brain tumours,
Australian rules football
a l l
* Solid lines represent cutpoints for categorical analysis.
b G = Grenoble, H • Heidelberg, S « Stockholm, W » Winnipeg, T - Toronto, A - Adelaide, M = Melbourne, L - Los Angeles.
c Not all subjects within centres were asked about sports on the centres' lists (some sports were added to lists in the course of interviewing).
d Measure of correlation between two dichoiomous variables.
Problems inherent in analysing risk factors for male menin-
gioma include the rarity of the disease, the low incidence in
males, and the fact that many registries still do not indude benign
brain tumours. Despite data from six studies conducted in five
different countries (Melbourne and Los Angeles did not indude
meningiomas), exposure prevalence was not high enough to
enable certain analyses or to achieve statistical significance for
the relative risks observed. For example, risk of sustaining
more than one injury was highest among male meningiomas
(OR = 1.7) but was based on only five cases (95% CI: 0.6-5.0).
Assessing the effect of head trauma that may result from
sports partidpation proved difficult, mainly because of cultural
differences in popularity and definitions of different sports. Not
only did each centre use a different sports list, depending on the
area's most popular sports, but different activities may share a
common sports name among regions. For example, 'football'
can be synonomous with 'soccer,' depending on geographical
region. Furthermore, some centres asked about sports In gen-
eral while others restricted the questioning to 'contact' sports,
which assumes that contact sports are most likely to result in
by guest on July 12, 2011
HEAD TRAUMA AND BRAIN TUMOURS 585
Tible 7 Participation in sports at least 5 years before diagnosis grouped by degree of correlation with reported head injuries, conditional
multivanate analysis, multicentre international case-control study of adult brain tumours, 1984-1991
(%) No. controls (%) OR (95% CI) (%) No. controls (%) OR (95% CI)
1 Not mutually exclusive; based on correlation (high = Oj »0.1, medium = 0.01 < fl>2 < 0.10, low = 4>2 <0 01) with head injuries. Number of centres with
sports in high, medium, and low groups are 8, 7, and 5, respectively. Reference group is subjects that did not participate in any sport on the centre's list and
did not report participation in any other sport that was then added to the centre's list.
head injuries. This premise is probably false, the most notable
exception being horseback riding. We were therefore forced to
use data from the case-control study itself to correlate head
injuries with specific sports, and our data did not link the head
injuries reported with sports participation. Also, because some
centres asked only about contact sports, we were unable to con-
trol for the potential confounding effect of general physical
activity. Finally, the only method available for evaluating level
of participation was total hours of participation rather than a
skill level, such as professional or amateur, which may more
accurately reflect intensity of participation. Given these weak-
nesses, a notable observation was that, among male meningiomas
only, risk was elevated (although not significant) for participation
in sports most correlated with head injuries.
The fact that epidemiological studies, including this one,
have not shown a convincing causal relationship between head
trauma and brain tumour development may reflect the defici-
encies of studies investigating this association and the fact that
the association, if one exists, is not a direct one. Experimental
data have shown that trauma can act as a cocarcinogen in the
presence of an initiating carcinogen.28"33 For example, among
offspring of pregnant rats injected with ethylnitrosourea, a sig-
nificant increase in gliomas in the region of a brain injury induced
at one month of age was observed compared to a non-injured
control group;33 it was hypothesized that cells damaged by the
initiating carcinogen proliferated as a natural result of the trauma,
leading to tumour formation. Future studies of head trauma
and brain tumour risk should consider potential initiators of
carcinogenesis, such as nitrite from cured meats, as modifiers of
the trauma effect on risk of brain tumour.
Our findings suggest that an association between head
trauma and brain tumour risk cannot be ruled out and should
therefore be further studied. Study designs that consider pos-
sible effect modifiers, that minimize recall bias and, for analyses
of sports-related head injuries, that effectively control protect-
ive effects of exercise are essential. Large study populations
are also imperative because of the relatively low prevalence of
head injuries and the low incidence of brain tumours, particu-
larly meningiomas for which increased risk was observed in
1 Dunsmore RH, Roberts M. Trauma as a cause of brain tumour:
a medicolegal dilemma. Conn Med 1974;38:521-23.
2Reinhardt G. Trauma-Fremdkrper-Hirngeschwulst. Munchen Med
3 Cushing H, Eisenhardt L. Meningiomas. Springfield, 111: Charles C
4 Preston-Martin S, Paganini-Hill A, Henderson BE, Pike MC, Wood C.
Case-control study of intracranial meningiomas in women in Los
Angeles County. J Natl Cancer Inst 1980:65:67-73.
5 Preston-Martin S, Yu MC, Henderson BE, Roberts C. Risk factors
for meningiomas in men in Los Angeles County. J Natl Cancer Inst
6 Preston-Martin S, Mack W, Henderson BE. Risk factors for gliomas
and meningiomas in males in Los Angeles County. Cancer Res 1989;
7 Choi NW, Schuman LM, Gullen WH. Epidemiology ol primary central
nervous system neoplasms. II. Case-control study. Am J Epidemiol
8 Carpenter AV, Flanders WD, Frome EL, Cole P, Fry SA. Brain cancer
and nonoccupational risk factors: a case-control study among workers
at two nuclear facilities. Am J Public Health 1987,77:1180-82.
9 Hochberg F, Toniolo P, Cole P. Head trauma and seizures as risk factors
of glioblastoma. Neurology 1984:34:1511-14.
10Burch JD, Cralb KJP, Choi BCK, Miller AB, Risch HA, Howe GR. An
exploratory case-control study of brain tumours in adults. J Natl
Cancer Inst 1987:78:601-09.
1' Codd MB, Kurland LT, O'Fallon M, Beard CM, Cascino TL. Case-
control study of neuroepithelial tumours in Rochester, Minnesota,
1950-1977. Ncuroepidemiology 1990,9:17-26.
12Zampieri P, Meneghini F, Grigoletto F et al. Risk factors for cerebral
glioma in adults: a case-control study in an Italian population.
13 Preston-Martin S, Yu MC, Benton B, Henderson BE. N-nitroso
compounds and childhood brain tumours: a case-control study. Cancer
14 Howe GR, Burch JD, ChiareUi AM, Risch HA, Choi BCK. An explor-
atory case-control study of brain tumours in children. Cancer Res 1989;
by guest on July 12, 2011
586 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
"Kuljten RR, Bunln GR, Nass CC, Meadows AT. Gestational and fam-
ilial risk factors for childhood astrocytoma: results of a case-control
study. Cancer Res 1990:50:2608-12.
16 Bunln GR. Buckley JD, Boesel CP, Rorke LB, Meadows AT. Risk
factors for astrocytlc glioma and primitive neuroectodermal tumour
of the brain in young children: a report from the Children's Cancer
Group. Cancer Epidemiol Biomarkers Prev 1994:3:197-204.
17McCredle M Malsonneuve P, Boyle P. Perinatal and early postnatal
risk factors for malignant brain tumours In New South Wales children.
IntJ Cancer 1994:56:11-15.
18 Ryan P, Lee MW, North JB, McMlchael AJ. Risk factors for tumours of
the brain and meninges: results from the Adelaide adult brain tumour
study. IntJ Cancer 1992:51:20-27.
"sdilehofer B, Blettner M, Becker N, Martinsohn C, Wahrendorf J.
Medical risk factors and the development of brain tumours. Cancer
20 DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin
21 Breslow NE, Day NE. Statistical Methods in Cancer Research. Vol. I. The
Analysis of Case-Control Studies. IARC Scientific Publication 32. Lyon:
22 Cramer H. Mathematical Methods of Statistics. Princeton, NJ: Princeton
University Press, 1946.
23 Zar JH. Biostatistical Analysis. Second edn. Englewood Cliffs, NJ: Prentice-
Hall, Inc., 1984.
24 Bamett GH, Chou SM, Bay JW. Posttraumatic intracranial meningioma:
a case report and review of the literature. Ncurosurgery 1986:18:75—78.
25 Drews CD, Kraus JF, Greenland S. Recall bias in a case-control study
of sudden infant death syndrome. Int J Epidemiol 1990,19:405-11.
26 Khoury MJ, James LM, Erickson JD. On the use of affected controls
to address recall bias in case-control studies of birth defects. Teratology
27 Kamiryo T, Matsunaga T, Fujisawa H, Ito H, Aoki H. Brain tumours in
Yamaguchi prefecture: incidence through four years. No To Shinkei
28 Deelman HT. The part played by injury and repair in the development
of cancer; with remarks on the growth of experimental cancers.
29 Rous P, Kidd JG. Conditional neoplasms and subthreshold neoplastic
states: a study of the tar tumours of rabbits. J Exp Med 1941:73:365-90.
30 Haran-Ghera N, Tralnin N, Fiore-Donati L, Berenblum IA. A possible
two-stage mechanism in rhabdomyosareoma induction In rats. Br J
31 Glinos AD, Bucher NLR, Aub JC. The effect of liver regeneration on
tumour formation in rats fed 4-dimethyl-amino-aK>benzene. J Exp
32 Ball JK, Field WEH, Roe FJC, Walters M. The carcinogenic and co-
cardnogenlc effects of paraffin was pellets and glass beads in the
mouse bladder. Br J Urol 1964:36:225-37.
33 Morantz RA, Shain W. Trauma and brain tumours: an experimental
study. Neurosurgery 1978:3:181-86.
by guest on July 12, 2011