Atopy and Risk = of Brain Tumors: A Meta-analysis

http://www.medscape.com/ =

To Print: Click = your browser's=20 PRINT button.

NOTE: To view = the article=20 with Web enhancements, go to:
http://www.medscape.c= om/viewarticle/566922

Atopy and Risk of Brain Tumors: A = Meta-analysis

Eleni Linos; Tim Raine; Alvaro Alonso; = Dominique=20 Michaud

J Natl Cancer Inst. 2007;99(20):1544-1550.=20 =A92007 Oxford University Press

Posted 12/11/2007

Abstract and Introduction

Abstract

Background: Glioma is a rapidly progressive disease, and = little=20 is known about its etiology. Atopic diseases are on the rise in = western=20 populations, with increasing interest on their long-term health=20 consequences. An inverse association between atopy and the risk of = glioma=20 has been observed. We carried out a meta-analysis of studies = examining the=20 association between atopic disease and risk of glioma and=20 meningioma.
Methods: In an electronic literature search = of the=20 MEDLINE, ISI Web of Science, and EMBASE databases from 1979 = through=20 February 2007, we identified case-control and cohort studies = quantifying=20 associations between a history of asthma, eczema, or hay fever or = allergy=20 and a medically confirmed diagnosis of glioma or meningioma. We = performed=20 meta-analysis by pooling studies according to the inverse of their = variances. We evaluated publication bias using funnel plot and = sensitivity=20 analyses.
Results: A total of eight observational = studies were=20 included, with a total of 3450 patients diagnosed with glioma and = 1070=20 patients with meningioma. A history of atopic disease was = inversely=20 related to risk of glioma. The pooled relative risks (RRs) of = glioma=20 comparing those with a history of an atopic condition with those = with no=20 history of that condition were 0.61 (95% confidence interval [CI] = =3D 0.55=20 to 0.67) for allergy, 0.68 (95% CI =3D 0.58 to 0.80) for asthma, = and 0.69=20 (95% CI =3D 0.58 to 0.82) for eczema. Proxy reporting was unlikely = to=20 explain the association because the pooled relative risk estimate = from=20 studies without proxy reporting remained inverse and statistically = significant (RR =3D 0.66, 95% CI =3D 0.58 to 0.75). Publication = bias was also=20 an unlikely explanation for the inverse association because the=20 association persisted in a sensitivity analysis and the funnel = plot was=20 symmetric. No overall statistically significant association was = noted for=20 atopy and meningioma, although the information on this disease was = limited=20 and heterogeneous.
Conclusions: We observed a strong = inverse=20 relationship between atopic disease and glioma that is unlikely to = be=20 explained by methodologic bias alone.

Introduction

Atopic disease is a hyperactive immune state that is = traditionally=20 characterized by T-cell production of high levels of cytokines = such as=20 interleukin (IL)-4 and IL-13, which belong to the T helper 2=20 (T_H2) cytokine subclass, and an associated shift away = from=20 T_H1 responses. Atopic diseases, including asthma, = eczema, and=20 allergy, have been related to lower risk of primary glioma in = numerous=20 epidemiologic studies.^[1-9] The association between = atopic=20 disease and meningioma has been less = consistent.^[2-5-7-8-10]=20 The absence of a consistent relationship between atopy and = meningioma=20 compared with that observed between atopy and glioma has been = attributed=20 to differences in disease etiology, smaller study sizes with = insufficient=20 power, or differences in proxy reporting, which is less common in = studies=20 of meningioma than in studies of glioma.

Several mechanisms have been proposed to explain the = association=20 between atopic disease and glioma. The findings may reflect a true = causal=20 effect of the heightened immune function seen in atopy on tumor=20 development. For example, one hypothesis is that the hyperreactive = state=20 of the immune system in atopic individuals may result in increased = immune=20 surveillance and limit abnormal cell proliferation. Animal studies = support=20 a role for cytokines involved in atopy on brain tumor=20 protection,^[11-12] and biomarkers of asthma are = inversely=20 related to glioma development.^[13-14]

Alternatively, the inverse associations observed may be = noncausal and=20 may arise from methodologic biases inherent in epidemiologic study = design.=20 These biases include possible selection bias due to inappropriate = choice=20 of controls or low control response rates, measurement error from=20 self-reported allergy assessment, reverse causation, or = confounding by=20 unmeasured causes. Furthermore, a high frequency of exposure = ascertainment=20 by proxy may lead to bias, especially if proxies systematically=20 underreport allergic disease.^[8] However, to fully = explain the=20 associations observed with glioma, at least some of these biases = would=20 have to be of substantial magnitude and to extend to multiple = published=20 studies.

We conducted a systematic literature review and meta-analysis = of the=20 association between atopic diseases and incidence of glioma and = meningioma=20 among international populations. In this article we summarize the=20 scientific evidence from observational studies and quantify the = magnitude=20 of the observed effect.

Methods

Search Strategy and Study Selection

We searched MEDLINE, EMBASE, and ISI Web of Science from = January 1979=20 through February 2007 using combinations of the search terms = glioma,=20 meningioma, brain tumor, atop*, asthma, eczema, and allerg*. To be = eligible for inclusion in this meta-analysis, publications had to = include=20 original data from case-control or cohort studies, to report on a=20 medically confirmed outcome of primary glioma or meningioma, and = to=20 present an odds ratio or relative risk (RR) quantifying the = association=20 between atopic diseases including allergy, asthma, or eczema and = primary=20 glioma. We initially identified three cohort studies [reported in = one=20 publication]^[8] and nine case-control=20 studies^[1-7-9-12] that met our criteria and appeared to = be=20 relevant in assessing our hypothesis ( Table 1 ). One of the cohort studies in = Schwartzbaum et=20 al.^[8] combined varying exposures, including a hospital = discharge diagnosis of diabetes or immunologic or allergic = disease, and=20 was therefore not included in any further analyses. We selected = the most=20 common definition of exposure among studies for adequate = comparability.=20 Allergy, asthma, and eczema were each evaluated as dichotomous = exposures=20 (yes or no) reflecting self-reported history of any versus no such = condition. If data from a study appeared in more than one = publication, we=20 used the most recent publication for this analysis. Information on = study=20 design, participant characteristics, measurement of allergic = disease,=20 adjustment for potential confounders, percentage of study = participants who=20 reported exposure by proxy, and estimates of associations were = extracted=20 by two independent investigators (E. Linos and D. Michaud).

Statistical Analysis

Study-specific log relative risks (cohort studies) and log odds = ratios=20 (case-control studies) were weighted by the inverse of their = variances to=20 obtain a pooled relative risk estimate and its 95% confidence = interval=20 (CI). Odds ratios were considered to be estimates of relative = risks.=20 Random-effect models were used to statistically pool all estimates = because=20 in the absence of heterogeneity the random effects model exactly = equals=20 the fixed-effect model. We assessed heterogeneity among studies = using the=20 Q test statistic of DerSimonian and Laird.^[15]=20 Statistical evaluation of publication bias is difficult with few, = small=20 studies.^[16] We therefore assessed potential = publication bias=20 using sensitivity analysis and funnel plot symmetry^[17] = and=20 calculated fail-safe numbers using a weighted = method.^[18] The=20 natural logarithm of the relative risk in each study was plotted = against=20 the inverse of its variance to obtain a funnel plot. A fail-safe = number=20 reflects the number of null studies of average precision that = would need=20 to be added to the meta-analysis to make the overall effect=20 non-statistically significant.^[18-19] If the number of = studies=20 that is required to render the association non-statistically = significant=20 exceeds 5n + 10, where n is the original number of = studies,=20 publication bias is extremely unlikely.^[18] All = analyses were=20 conducted with STATA v. 9.2.^[20] All tests of = statistical=20 significance were two-sided. Our report adhered to the = Meta-analysis of=20 Observational Studies in Epidemiology checklist for meta-analyses = of=20 observational research.^[21]

Results

A total of 12 studies of glioma and atopic disease were = identified in=20 our literature search ( Table 1 ). Data from three case-control=20 studies^[2-4] were included in a subsequent larger = international=20 analysis;^[5] therefore, the smaller studies were = excluded from=20 our pooled analysis. As already noted, one of the three cohort = studies in=20 Schwartzbaum et al.^[8] was excluded because of problems = with=20 exposure definition. Our final analysis included a total of 53223=20 participants and 3450 cases of glioma. Studies were conducted in = seven=20 countries, including Australia, Canada, France, Germany, Sweden, = the=20 United Kingdom, and the United States, between 1977 and 2004. All = studies=20 used medically confirmed primary glioma as the outcome and = self-reported=20 atopic disease as the exposure, although four = studies^[6-7-9-13]=20 asked specifically about medically diagnosed atopic disease. = Potential=20 confounders that were adjusted for in each study are shown in Table 2 . All studies adjusted for age and = sex, and some=20 studies controlled for additional variables, including, race, = location,=20 and socioeconomic status, which are potential confounders. We did = not find=20 substantial evidence of heterogeneity among the studies of allergy = and=20 glioma (P =3D .51). The study-specific risk ratios ranged = from 0.45=20 to 1.09 (Fig. 1). In the pooled analysis, those who reported a = history of=20 any allergy had a lower risk of glioma than those who reported no = history=20 of allergy (pooled RR =3D 0.61, 95% CI =3D 0.55 to 0.67, = P<.001).=20 Six studies^[5-9-12] reported risks of glioma with = respect to a=20 history of asthma or eczema specifically. We did not detect = statistically=20 significant between-study heterogeneity for associations of glioma = with=20 asthma (P =3D 0.72) or with eczema (P =3D 0.82). The = pooled RR=20 for glioma among those reporting a history of asthma compared with = no such=20 history was 0.68 (95% CI =3D 0.58 to 0.80, P<.001) (Fig. = 2). A=20 similar inverse association with glioma was noted for a history of = eczema=20 (RR =3D 0.69, 95% CI =3D 0.58 to 0.82, P<.001).=20

Figure 1.
Study-specific and pooled relative risk (RR) and 95% = confidence=20 intervals (CIs) of the association between allergy and glioma. = The size=20 of the shaded boxes is proportional to the sample size, and the = diamond=20 represents the pooled effect. The overall pooled RR =3D 0.61 = (95% CI =3D=20 0.55 to 0.67, P<.001).

Figure 2.
Study-specific and pooled relative risk (RR) and 95% = confidence=20 intervals (CIs) of the association between asthma and eczema and = glioma.=20 The size of the shaded boxes is proportional to the sample size, = and the=20 diamond represents the pooled effect. (A) Asthma. The = overall=20 pooled RR =3D 0.68 (95% CI =3D 0.58 to 0.80, P<.001). = (B)=20 Eczema. The overall pooled RR =3D 0.69 (95% CI 0.58 to 0.82,=20 P<.001).

Proxy Reporting

The participation of proxies is common in studies of = progressive=20 disease, and if reporting by proxy is systematically different = from=20 individual patient reporting it can lead to bias. Three = case-control=20 studies^[4-6-9] provided relative risk estimates of the=20 association between allergy and glioma with no or minimal data = collection=20 by proxy. Two of these studies^[4-6] reported relative = risks=20 among the subsets of cases for whom self-reported information was=20 available, and because the other study^[9] did not have = ethical=20 permission to approach proxies, 98% of its cases were = self-reported.=20 Neither of the two cohort studies^[8] described any = reporting by=20 proxy. The pooled relative risk estimate for all five studies = without any=20 proxy reporting remained inverse and statistically significant (RR = =3D 0.66,=20 95% CI =3D 0.58 to 0.75).

Publication Bias

To assess the possibility of publication bias, we recalculated = the=20 pooled relative risk estimates for the association of glioma with = allergy=20 under the following extreme assumptions in a sensitivity analysis: = 1) a=20 similar number of unpublished studies have been performed; 2) = these have=20 analogous sample size and number of cases; but 3) they observed = null=20 results with relative risks of 1. Under these assumptions, the=20 statistically significant inverse association between allergy and = glioma=20 persisted (RR =3D 0.78, 95% CI =3D 0.73 to 0.84, = P<.001). In=20 addition, a funnel plot of the results of studies examining the=20 association between allergy and glioma (Fig. 3) was symmetric, = which=20 implies that publication bias was unlikely. Finally, in the = fail-safe=20 analysis^[18] the number of null studies that would need = to be=20 added to render the association of allergy with glioma = non-statistically=20 significant was 333, which exceeds the criterion of 45 (5n = +=20 10).

Figure 3.
Funnel plot of studies on allergy and risk of glioma. The = natural=20 logarithm of the relative risk is plotted on the horizontal = axis, and=20 the estimate of precision (1/variance) is plotted on the = vertical axis.=20 Each point represents a published study of allergy and=20 glioma.

Meningioma

Of the 12 studies identified originally, four=20 case-control^[2-5-7-9] and two cohort^[8] = studies=20 reported on the association between a history of allergy and risk = of=20 meningioma. Because the data from one case-control = study^[2]=20 were subsequently included in a larger international = study,^[5]=20 our pooled estimate does not include the smaller study. Medically=20 confirmed cases of primary meningioma were recorded in all studies = included (total N =3D 1070). Exposure was assessed by = questionnaire or=20 interview, as described above. The random-effects pooled estimate = of these=20 studies was not statistically significantly different from 1 (RR = =3D 0.93,=20 95% CI =3D 0.72 to 1.19) (Fig. 4). The studies reporting on = meningioma and=20 allergy were somewhat heterogeneous in their results (P =3D = .09). In=20 a subanalysis that excluded the study with the largest relative=20 risk,^[8] which was the most likely cause of this = heterogeneity,=20 the pooled risk ratio showed an inverse association between = meningioma and=20 allergy that was of borderline statistical significance (RR =3D = 0.84, 95% CI=20 =3D 0.72 to 0.98, P =3D .029).

Figure 4.
Study-specific and pooled relative risks (RRs) and 95% = confidence=20 intervals for the association between allergy and meningioma. = The=20 overall pooled RR =3D 0.93 (95% CI =3D 0.72 to 1.19, P = =3D.57). When=20 the single study contributing the most to heterogeneity [i.e., = Cohort=20 II]^[8] is excluded, the pooled RR =3D 0.84 (95% CI = =3D 0.72 to=20 0.98, P =3D .029).

Discussion

The results of this meta-analysis suggest an inverse = association=20 between atopy and glioma; specifically, the risk of glioma was 40% = lower=20 among those with a history of allergy than those with no such = history, 30%=20 lower among those with a history of eczema, and 30% lower among = those with=20 a history of asthma. This inverse association with allergy = persisted when=20 the analysis was restricted to studies without proxy responders = and when=20 accounting for publication bias. No overall association was noted = for=20 allergy and meningioma, although a smaller number of cases of this = disease=20 were available.

The atopic diseases that were included in this analysis = encompass a=20 range of airway, cutaneous, gastrointestinal, and vascular = pathologies=20 that were self-reported as eczema, asthma, or allergy. In the = context of=20 such variable definitions of atopic disease, it is striking that = the=20 associations of these various pathologies with risk of glioma were = strong=20 and consistent across studies. Furthermore, these results are = consistent=20 with the finding that IgE, which is a biomarker of atopic allergy = whose=20 measurement is unaffected by individual recall and reporting, was = also=20 inversely associated with glioma. A case-control study that = included 228=20 subjects with available serum samples reported an odds ratio of = 0.37 (95%=20 CI =3D 0.22 to 0.64) comparing those with high total serum IgE = levels to=20 those with normal levels.^[14]

Despite their variety, the different manifestations of atopy = are all=20 characterized by the production of T_H2 cytokines, = including=20 IL-4 and IL-13. Single-nucleotide polymorphisms in the gene coding = for the=20 α-chain of the IL-4 receptor, which is common to both the = IL-4 and IL-13=20 signaling pathways, and in the gene for IL-13 itself have been = associated=20 with an increased risk of asthma.^[22-24] Intriguingly, = these=20 same polymorphisms were also inversely associated with glioma=20 susceptibility in one study.^[13] It is possible that=20 T_H2 cytokines may help protect against glioma through = their=20 regulation of humoral immunity, including the production of B-cell = antibodies. Although the brain is regarded as an "immunologically=20 privileged" site,^[25] peritumor inflammation alters the = properties of the blood-brain barrier and allows infiltration by = cells of=20 the peripheral immune system.^[26-27] Thus, in addition = to=20 humoral immunity, cell-mediated immunity may also be involved in = the=20 response to gliomas.^[28]

Considerable experimental evidence, including evidence from = rodent=20 models, supports a role for IL-4 in the immune response to = gliomas. For=20 example, immunization of glioma-bearing rats with tumor cells=20 overexpressing IL-4 led to rejection of an otherwise fatal = progressive=20 glioma.^[11] Likewise, neural progenitor cells = engineered to=20 overexpress IL-4 produced regression of induced gliomas in both = mice and=20 rats.^[29] Gliomas are known to express transforming = growth=20 factor β,^[30] and this cytokine may in fact = function in synergy=20 with IL-4 to drive local T_H1 responses.^[31] = Indeed,=20 the protective responses seen in rats immunized with = IL-4-overexpressing=20 gliomas were associated with the production of high levels of the=20 T_H1 effector cytokine interferon gamma by = tumor-infiltrating=20 CD4⁺ T cells.^[32] Furthermore, it is = possible that=20 IL-4 and other T_H2 cytokines have protective actions = through a=20 range of individual pleiotropic effects that are not directly = related to=20 the T_H1/T_H2 paradigm, which sees = T_H1 and=20 T_H2 cytokines as driving opposing=97at times=20 antagonistic=97immunologic pathways. Such effects might include = the=20 suppression of angiogenesis.^[33] In addition, IL-4 has = been=20 shown to inhibit DNA synthesis in low-grade astrocytoma cell lines = but not=20 in more aggressive tumor cell lines,^[34] and similar = results=20 have been demonstrated for IL-13,^[35] raising the = possibility=20 that these cytokines may act to disrupt tumor progression rather = than=20 initiation.

If T_H1 responses are also important for prevention = or=20 rejection of gliomas and the apparent protection is related to = increased=20 immune reactivity rather than to a T_H1/T_H2=20 dichotomy, then the apparent protection from glioma associated = with atopy=20 should also be seen with other immunopathologies, including some=20 autoimmune conditions that are regarded as T_H1 in = nature, such=20 as type 1 diabetes or multiple sclerosis. In this regard, the = study of=20 Brenner et al.^[7] is informative in that it showed an = inverse=20 association between a history of either autoimmunity or atopy with = glioma=20 that was strongest in those reporting a history of = both.^[7]=20 Future studies seeking to clarify the nature of the relationship = between=20 immune status and glioma should also address = T_H1-mediated=20 pathologies, as well as the more thoroughly studied T_H2 = responses.

Meningioma is a disease with distinct pathology and progression = to that=20 of glioma. The dissimilar results in relation to atopic disease = could=20 therefore be due to differences in tumor etiology. If the inverse=20 association noted for atopy is a function of elevated immune = reactivity=20 acting to limit tumor progression rather than incidence, we might = expect=20 an inverse association between atopy and all brain tumors, = especially=20 among more rapidly progressive cancers that might trigger stronger = inflammatory responses. The findings of Schwartzbaum et = al.^[8]=20 suggest a stronger inverse association between allergy status and=20 high-grade, but not low-grade, glioma and hence support this = hypothesis.=20 In this context, the slowly progressive nature of meningioma may = elicit a=20 milder immune response, explaining the weaker association observed = here.=20 Although the few studies on meningioma associations with atopy = showed=20 heterogeneous results, we identified one study that contributed to = most of=20 the heterogeneity. Exposure assessment may have been less accurate = in this=20 study because it was the only investigation to measure atopy using = a=20 single question. Interestingly, when this study was excluded the = pooled=20 estimate for the remaining homogeneous studies became inverse and=20 statistically significant. Moreover, a history of autoimmunity has = been=20 associated with a lower risk of meningioma as well as of=20 glioma,^[7] emphasizing the potential importance of a = vigorous=20 immune response in protection from any intracranial tumor. = However, more=20 studies are needed to confirm this hypothesis.

Despite the intriguing findings of this meta-analysis, it has = several=20 important limitations. One is that brain tumors are rare and = glioma is a=20 rapidly progressive disease, making it difficult to obtain = high-quality=20 information in epidemiologic investigation. To date, most of the=20 information on this topic comes from case-control studies, which = permit=20 larger number of cases than prospective cohorts of rare diseases = but are=20 limited by retrospective ascertainment of exposure and do not = provide a=20 clear definition of the temporal relation between exposure and = disease.=20 The possibility of selection bias was evaluated by considering the = selection of controls and control response rates (range =3D = 43%-87% in=20 included studies; data not shown). Control selection was generally = appropriate in all studies included, making selection bias an = unlikely=20 explanation for the consistent inverse association.

In addition, measurement error associated with exposure is of=20 particular concern when using self-reported measures. Many studies = included in our meta-analysis tried to elicit a history of=20 physician-diagnosed or medically treated atopy in an attempt to = reduce=20 exposure misclassification.^[1-6-9-13] The reliability = of=20 self-reported measures of allergies was addressed in some studies, = which=20 quoted similar rates of disease in the control subjects or = underlying=20 cohort and in the general population.^[6-8] Moreover, = the=20 validity of self-reported asthma measures from questionnaires = compared=20 with the gold standard of clinical diagnosis is reasonably high; = average=20 sensitivity is 68%, and average specificity is 94%.^[36] = Therefore, measurement error is less problematic with this subtype = of=20 atopic disease.

A further difficulty in drawing causal interpretations from=20 observational data occurs if the temporal order of exposure = preceding=20 outcome is not clearly defined and reverse causation cannot be = excluded.=20 Brain tumors are known to depress immune = function,^[26-37] and=20 the latency period of glioma is not fully established. Therefore, = it is=20 possible that both IgE levels and allergic reactions reported by = patients=20 with glioma are influenced by subclinical cancer. Some studies = have=20 addressed this issue by obtaining information about allergy many = years=20 preceding glioma diagnosis^[2-5-13] and by collecting = data on=20 overall duration, latency, and age of onset.^[5-7-9] = However,=20 because inverse associations were also observed when allergies = developed=20 at an early age or when they were present many years before brain = cancer=20 diagnosis,^[5-9] reverse causation is unlikely to be = responsible=20 for the overall findings.

We also considered publication bias and systematic differences = due to=20 reporting by proxy as potential alternative explanations for these = findings. Our sensitivity analyses and funnel plot imply that = these biases=20 are unlikely to fully account for the strong and consistent = associations=20 observed.

Future prospective studies may be able to clarify the = temporality of=20 this relationship by confirming that atopic disease precedes = cancer=20 incidence. However, the lack of adequate data on history and = timing of=20 atopic conditions in most cohort questionnaires limits exposure=20 assessment. Instead, blood markers of atopy, including serum IgE = collected=20 before diagnosis, may be a more appropriate way to examine this=20 association. Moreover, information from multiple cohort studies = will=20 likely need to be pooled to allow for adequate numbers of = cases.

In summary, a history of atopic disease is inversely associated = with=20 risk of glioma. These results are consistent across many = geographic=20 settings, study designs, and different atopic diseases and may = reflect a=20 protective effect of the immunologic milieu associated with atopic = allergy=20 on tumor growth. Furthermore, some data exist to support the = biologically=20 plausible proposition that the inverse association of atopy with = glioma=20 reflects a broader association between heightened immunity and=20 intracranial tumors. Methodologic limitations are unlikely to = explain the=20 full magnitude of the observed effect, but there remains a need = for=20 further prospective studies that can more clearly evaluate the = temporal=20 relationship between atopy and brain tumors.

Table 1. Epidemiologic Studies of Atopy and Risk of=20 Glioma*

First author, date = (reference) Country Design Case patients/control subjects Type of control subjects (% response = rate) % proxy reporting of case patients = (control=20 subjects) Exposure assessment

Schoemaker, 2006 (9)⁼⁸⁶ United Kingdom Case-control 965/1716 Population (45%) 2% Interview; diagnosis of hay fever, asthma, or eczema and = timing

Schwartzbaum, 2005 (13)⁼⁸⁶ Sweden Case-control 174/633 Population (66%) 9% Interview; diagnosis of asthma, hay fever, eczema, and=20 aller-gies over the prior 10 years

Schwartzbaum, 2003 (8)⁼⁸⁶ Sweden Cohort I 37/14535 Twins born 1886-1925 0 Questionnaire; ever asthma, fever, allergic rhinitis, or = eczema

Schwartzbaum, 2003 (8)⁼⁸⁶ Sweden Cohort II 42/29573 Twins born 1926-1958 0 Single question on ever croup, asthma, hay fever, or=20 eczema

Schwartzbaum, 2003 (8)⁼⁸⁷ Sweden Cohort III 68/52067 Both twin cohorts 0 Hospital discharge records: asthma, eczema, = autoimmu-nity, or=20 diabetes

Brenner, 2002 (7)⁼⁸⁶ United States Case-control 489/799 Noncancer hospital (86%) 24% (4%) Interview; physician diagnosis of systemic lupus, = eczema, hay=20 fever, or asthma

Wiemels, 2002 (6)⁼⁸⁶ United States Case-control 405/402 Population (74%) 34% Interview; full allergy history

Schlehofer, 1999 (5)⁼⁸⁶ Six countries^=A7 Case-control 1178/1987 Population (ns) 27% (4%) Interview and self-administered SEARCH questionnaire; = history=20 of allergies 2 years before tumor

Cicuttini, 1997 (4)^=B6 Australia Case-control 416/422 Population (43%) 43.7% Interview; SEARCH questionnaire

Ryan, 1992 (2)^=B6 Australia Case-control 110/416 Population (63%) 25% (0.5%) Interview; SEARCH questionnaire

Schlehofer, 1992 (3)^=B6 Germany Case-control 226/418 Population (72%) 4% Interview; SEARCH questionnaire

Hochberg, 1990 (1)⁼⁸⁶ United States Case-control 160/128 Friends of cases (87%) 20% (2%) Self-administered questionnaire; ever-treated=20 allergies

*SEARCH =3D glioma, brain tumor, atop*, asthma, eczema and=20 allergy*.
⁼⁸⁶Included in final = analysis.
⁼⁸⁷Not=20 included in final analysis because exposure definition was not = as=20 precise as in other studies.
^=A7Australia, Canada, = France,=20 Germany, Sweden, United States.
^||Not=20 specified.
^=B6Not included individually in our = meta-analysis=20 because the data were also part of pooled analysis [reference = (5)].

Table 2. Study-specific and Pooled RR and 95% CIs for = Atopic=20 Disease and Glioma*

First author, = date=20 (reference) RR (95% CI) for history = of Possible confounders = adjusted for in=20 analysis

Any allergy Asthma Eczema

Schoemaker, 2006 (9) 0.63 (0.53-0.76) 0.71 (0.54-0.92) 0.74 (0.56-0.97) Age (5-year groups), sex, location, interview year, SES=20 (Townsend index)

Schwartzbaum, 2005 (13)⁼⁸⁶ 0.64 (0.33-1.25) 0.67 (0.43-1.05) Age, sex

Schwartzbaum, 2003 (8)⁼⁸⁶ 0.45 (0.19-1.07) 0.52 (0.19-1.48) Age, sex

Schwartzbaum, 2003 (8)⁼⁸⁶ 1.09 (0.48-2.48)

Schwartzbaum, 2003 (8) 0.46 (0.14-1.48)

Brenner, 2002 (7)⁼⁸⁶ 0.67 (0.52-0.86) 0.63 (0.43-0.92) 0.76 (0.45-1.27) Age (10-year groups), sex, race, location (distance from = hospital), hospital

Wiemels, 2002 (6)⁼⁸⁶ 0.47 (0.33-0.67) 0.57 (0.38-0.86) Age, sex, race (white/nonwhite)

Schlehofer, 1999 (5)⁼⁸⁶ 0.59 (0.49-0.71) 0.75 (0.55-1.03) 0.64 (0.47-0.86) Age (5-year group), sex, location

Cicuttini, 1997 (4) 0.8 (0.5-1.4) 0.8 (0.5-1.4) 0.9 (0.5-1.4) Age (5-year group), sex

Ryan, 1992 (2) 0.54 (0.33-0.89) 0.40 (0.14-1.15) 0.22 (0.05-0.92) Age (2-year group), sex, location (postcode)

Schlehofer, 1992 (3) 0.7 (0.5-1.0)

Hochberg, 1990 (1)⁼⁸⁶ 0.6 (0.4-1.0) Age, sex

Pooled estimates 0.61 (0.55-0.67) 0.68 (0.58-0.80) 0.69 (0.58-0.82)

*Blank cells appear when the specific exposure-disease = association=20 was not evaluated in that particular study. CI =3D confidence = interval; RR=20 =3D relative risk; SES =3D socioeconomic = status.
⁼⁸⁶Studies=20 included in final pooled analysis. The remaining studies were = not=20 included because of insufficiently detailed exposure definition = (8) or=20 because the data (2-4) were also part of pooled analysis = [reference=20 (5)].

References

Hochberg F, Toniolo P, Cole P, Salcman M. = Nonoccupational risk=20 indicators of glioblastoma in adults. J Neurooncol (1990) = 8:55-60.
Ryan P, Lee MW, North B, McMichael AJ. Risk factors = for tumors=20 of the brain and meninges: results from the Adelaide Adult = Brain=20 Tumor Study. Int J Cancer (1992) 51:20-7.
Schlehofer B, Blettner M, Becker N, Martinsohn C, = Wahrendorf=20 J. Medical risk factors and the development of brain = tumors.=20 Cancer (1992) 69:2541-7.
Cicuttini FM, Hurley SF, Forbes A, Donnan GA, Salzberg = M,=20 Giles GG, et al. Association of adult glioma with medical=20 conditions, family and reproductive history. Int J Cancer = (1997)=20 71:203-7.
Schlehofer B, Blettner M, Preston-Martin S, Niehoff D, = Wahrendorf J, Arslan A, et al. Role of medical history in = brain=20 tumour development. Results from the international adult = brain=20 tumour study. Int J Cancer (1999) 82:155-60.
Wiemels JL, Wiencke JK, Sison JD, Miike R, McMillan A, = Wrensch=20 M. History of allergies among adults with glioma and = controls. Int=20 J Cancer (2002) 98:609-15.
Brenner AV, Linet MS, Fine HA, Shapiro WR, Selker RG, = Black=20 PM, et al. History of allergies and autoimmune diseases = and risk=20 of brain tumors in adults. Int J Cancer (2002) 99:252-9.
Schwartzbaum J, Jonsson F, Ahlbom A, Preston-Martin S, = Lonn S,=20 Soderberg KC, et al. Cohort studies of association between = self-reported allergic conditions, immune-related = diagnoses and=20 glioma and meningioma risk. Int J Cancer (2003) 106:423-8.
Schoemaker MJ, Swerdlow AJ, Hepworth SJ, McKinney PA, = van=20 Tongeren M, Muir KR. History of allergies and risk of = glioma in=20 adults. Int J Cancer (2006) 119:2165-72.
Schoemaker MJ, Swerdlow AJ, Hepworth SJ, van Tongeren = M, Muir=20 KR, McKinney PA. History of allergic disease and risk of=20 meningioma. Am J Epidemiol (2007) 165:477-85.
Okada H, Villa L, Attanucci J, Erff M, Fellows WK, = Lotze MT,=20 et al. Cytokine gene therapy of gliomas: effective = induction of=20 therapeutic immunity to intracranial tumors by peripheral=20 immunization with interleukin-4 transduced glioma cells. = Gene Ther=20 (2001) 8:1157-66.
Benedetti S, Di Meco F, Cirenei N, Bruzzone MG, Pollo = B,=20 Florio N, et al. IL-4 gene transfer for the treatment of=20 experimental gliomas. Adv Exp Med Biol (1998) 451:315-21.
Schwartzbaum J, Ahlbom A, Malmer B, Lonn S, Brookes = AJ, Doss=20 H, et al. Polymorphisms associated with asthma are = inversely=20 related to glioblastoma multiforme. Cancer Res (2005) = 65:6459-65.
Wiemels JL, Wiencke JK, Patoka J, Moghadassi M, Chew = T,=20 McMillan A, et al. Reduced immunoglobulin E and allergy = among=20 adults with glioma compared with controls. Cancer Res = (2004)=20 64:8468-73.
DerSimonian R, Laird N. Meta-analysis in clinical = trials.=20 Control Clin Trials (1986) 7:177-88.
Egger M, Davey Smith G, Schneider M, Minder C. Bias in = meta-analysis detected by a simple, graphical test. BMJ = (1997)=20 315:629-34.
Egger M, Smith GD, Altman DG, eds. Systematic reviews = in=20 healthcare: meta-analysis in context. BMJ (2001) 2nd ed.
Rosenberg MS. The file-drawer problem revisited: a = general=20 weighted method for calculating fail-safe numbers in=20 meta-analysis. Evolution Int J Org Evolution (2005) = 59:464-8.
Sutton AJ, Song F, Gilbody SM, Abrams KR. Modelling=20 publication bias in meta-analysis: a review. Stat Methods = Med Res=20 (2000) 9:421-45.
StataCorp. Stata Statistical Software: release 9 = (2005)=20 College Station (TX): StatCorp LP.
Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson = GD,=20 Rennie D, et al. Meta-analysis of observational studies in = epidemiology: a proposal for reporting. Meta-analysis of=20 Observational Studies in Epidemiology (MOOSE) group. JAMA = (2000)=20 283:2008-12.
Howard TD, Koppelman GH, Xu J, Zheng SL, Postma DS, = Meyers DA,=20 et al. Gene-gene interaction in asthma: IL4RA and IL13 in = a Dutch=20 population with asthma. Am J Hum Genet (2002) 70:230-6.
Heinzmann A, Mao XQ, Akaiwa M, Kreomer RT, Gao PS, = Ohshima K,=20 et al. Genetic variants of IL-13 signalling and human = asthma and=20 atopy. Hum Mol Genet (2000) 9:549-59.
Howard TD, Whittaker PA, Zaiman AL, Koppelman GH, Xu = J, Hanley=20 MT, et al. Identification and association of polymorphisms = in the=20 interleukin-13 gene with asthma and atopy in a Dutch = population.=20 Am J Respir Cell Mol Biol (2001) 25:377-84.
Barker CF, Billingham RE. Immunologically privileged = sites.=20 Adv Immunol (1977) 25:1-54.
Mahaley MS Jr, Brooks WH, Roszman TL, Bigner DD, Dudka = L,=20 Richardson S. Immunobiology of primary intracranial = tumors. Part=20 1: studies of the cellular and humoral general immune = competence=20 of brain-tumor patients. J Neurosurg (1977) 46:467-76.
Palma L, Di Lorenzo N, Guidetti B. Lymphocytic = infiltrates in=20 primary glioblastomas and recidivous gliomas. Incidence, = fate,=20 relevance to prognosis in 228 operated cases. J Neurosurg = (1978)=20 49:854-61.
Nano R, Ceroni M. The immunobiology of malignant = gliomas.=20 Funct Neurol (2005) 20:39-42.
Benedetti S, Pirola B, Pollo B, Magrassi L, Bruzzone = MG,=20 Rigamonti D, et al. Gene therapy of experimental brain = tumors=20 using neural progenitor cells. Nat Med (2000) 6:447-50.
Kuppner MC, Hamou MF, Sawamura Y, Bodmer S, de = Tribolet N.=20 Inhibition of lymphocyte function by glioblastoma-derived=20 transforming growth factor beta 2. J Neurosurg (1989) = 71:211-7.
Lingnau K, Hoehn P, Kerdine S, Koelsch S, Neudoerfl C, = Palm N,=20 et al. IL-4 in combination with TGF-beta favors an = alternative=20 pathway of Th1 development independent of IL-12. J Immunol = (1998)=20 161:4709-18.
Giezeman-Smits KM, Okada H, Brissette-Storkus CS, = Villa LA,=20 Attanucci J, Lotze MT, et al. Cytokine gene therapy of = gliomas:=20 induction of reactive CD4+ T cells by = interleukin-4-transfected 9L=20 gliosarcoma is essential for protective immunity. Cancer = Res=20 (2000) 60:2449-57.
Volpert OV, Fong T, Koch AE, Peterson JD, Waltenbaugh = C,=20 Tepper RI, et al. Inhibition of angiogenesis by = interleukin 4. J=20 Exp Med (1998) 188:1039-46.
Barna BP, Barnett GH, Jacobs BS, Estes ML. Divergent = responses=20 of human astrocytoma and non-neoplastic astrocytes to = tumor=20 necrosis factor alpha involve the 55 kDa tumor necrosis = factor=20 receptor. J Neuroimmunol (1993) 43:185-90.
Liu H, Jacobs BS, Liu J, Prayson RA, Estes ML, Barnett = GH, et=20 al. Interleukin-13 sensitivity and receptor phenotypes of = human=20 glial cell lines: non-neoplastic glia and low-grade = astrocytoma=20 differ from malignant glioma. Cancer Immunol Immunother = (2000)=20 49:319-24.
Toren K, Brisman J, Jarvholm B. Asthma and asthma-like = symptoms in adults assessed by questionnaires. A = literature=20 review. Chest (1993) 104:600-8.
Walker PR, Calzascia T, Dietrich PY. All in the head:=20 obstacles for immune rejection of brain tumours. = Immunology (2002)=20 107:28-38.

Sidebar: Context and Caveats

Prior knowledge

An association between atopic = disease and=20 a reduced risk of glioma has been observed in epidemiologic = studies. No=20 clear association with meningioma has been = observed.

Study=20 Design

Meta-analysis of case=96control and cohort = studies of the=20 association between a history of asthma, eczema, or hay fever or = allergy=20 and a diagnosis of glioma and=20 meningioma.

Contributions

Eight observational = studies=20 involving 3450 patients diagnosed with glioma and 1070 with = meningioma=20 were included. Individuals with a history of an atopic condition = had a=20 30%=9640% lower relative risk of glioma than those with no such = history. No=20 association with meningioma was=20 observed.

Implications

If the association is = causal,=20 it could reflect an effect of heightened immune surveillance on = brain=20 tumor development.

Limitations

Most of the = data came=20 from case=96control studies, which may not provide as clear = information on=20 the association between exposure and disease as cohort studies. = Most=20 studies used self-reported measures of atopy, which are subject to = error.=20 Complete information on the relative timing of the atopic = condition and=20 the diagnosis of brain tumors was not available.

Acknowledgements

The authors would like to thank Dr Miguel Hernan for critical = review of=20 the manuscript.

Funding Information

National Institutes of = Health=20 (R25CA098566 to E. L., R01CA114205 to D. M.).

Disclaimer

The funder was not involved in the design of the study, the = analysis=20 and interpretation of the data, the decision to publish, or the = writing of=20 the manuscript.

Reprint Address

Eleni Linos, BM, BCh, MPH, Department of Epidemiology, Harvard = School=20 of Public Health, 677 Huntington Ave, Boston, MA 02115 e-mail: elinos@hsph.harvard.edu

Eleni Linos, Tim Raine, = Alvaro=20 Alonso, Dominique Michaud, Department of Epidemiology, = Harvard=20 School of Public Health, Boston, MA (EL, AA, DM); Department of = Medicine,=20 Guy=92s and St. Thomas=92 Hospitals, London, UK (TR)

Disclosure: None of the authors of this = article declare=20 any conflict of interest.