Introduction

Gliomas are the most common primary brain tumors accounting for about 70 % of adult cases [1, 2]. They have an incidence of 5–10 cases per 100,000 [2]. The etiology of glioma is still largely uncertain [3]. Besides a genetic predisposition, ionizing radiation is a known risk factor [1]. Other potential risk factors include exposure to occupational and viral agents and, among dietary factors, exposure to N-nitroso compounds [1, 2].

Coffee has many effects on the central nervous system, mainly mediated by caffeine, which represents about 2–4 % of coffee powder compounds and easily crosses the blood–brain barrier. Also, a few tea components like theophylline and caffeine easily cross the blood–brain barrier. However, both coffee and tea contain many other compounds, including antioxidants, minerals, vitamin precursors, and phenols with potential anticancerogenic activities [4].

Besides general anticancerogenic activity of antioxidants, caffeine inhibits calcium ion increase by inositol 1,4,5-trisphosphate receptor subtype 3, the expression of which is increased in glioblastoma cells, consequently inhibiting migration of glioblastoma cells in various in vitro assays. Moreover, caffeine increases mean survival in a mouse xenograft model of glioblastoma [5]. Another potential mechanism through which coffee and tea might exert an anticancerogenic effect is that a tea component (polyphenol (−)-epigallocatechin-3-gallate) and two coffee components (kahweol and cafestol) inhibit DNA methyltransferase and reactivate methylation-silenced genes, including the DNA repair protein O6-methylguanine-DNA methyltransferase (MGMT) [6, 7]. The higher activation state of MGMT has been proposed to protect against glioma and other cancers [1]. Moreover, the genetic polymorphism of MGMT has been related to the risk of glioma [8].

In epidemiological studies coffee has been related with reduced inflammation [9] and with lower incidence of cancer at the oropharynx [10, 11], liver [12], endometrium [13, 14], and colon in case–control [15], but not in prospective studies [16, 17]. The relation between black and green tea and cancer risk has been investigated in many case–control and cohort studies, showing inconsistent results [18].

The relation between coffee and/or tea consumption and glioma risk has been considered in three prospective studies [19–21] and one case–control study [22]. Two other studies, one prospective [23] and one case–control [24], considered only the relation between coffee consumption and brain cancer risk. The results of any single study are, however, based on a relatively low number of subjects. Three of these studies included also information on coffee plus tea consumption [19–21], including a recent, large European cohort study [21] finding an inverse relation.

Thus, to provide a quantitative overall estimate on this issue, we combined all published data on the relation of coffee or tea intake with glioma, using a meta-analytic approach.

Materials and methods

In September 2012, a bibliography search was carried out in both Medline and Embase to identify observational studies providing quantitative estimates for brain cancer risk in relation to coffee or tea consumption. The search was carried out following the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines [25] and using the strings reported in the Appendix 1. The search was limited to human studies written in English. Two authors (C.P. and S.M.) independently selected the articles that reported data on the association between coffee and tea consumption and the risk of glioma or brain cancer. All case–control and cohort studies which provided estimates of the relative risk (RR) and corresponding confidence interval (CI), or information sufficient to calculate them, were included in the analyses. The reference lists of the selected papers were checked to obtain other pertinent publications. Abstracts and unpublished results were excluded. Figure 1 gives the flowchart for the selection of the articles included in the present analysis.

Fig. 1
figure 1

Flowchart for literature search and selection of publications

Full size image

For each study, we extracted data on the study design, sex, country, number of cases and controls/cohort size, duration of follow-up (for cohort studies) or period of enrollment (for case–control studies), histological type of the brain tumor investigated, RR estimates for the different categories of coffee or tea intake and the corresponding 95 % CI, and, when available, the number of cases and noncases for each level of coffee or tea consumption. When more than one risk estimate was reported, we used the one adjusted for the largest number of potential confounding factors. Of the six selected papers, four considered only gliomas [19, 20, 23, 24], one considered separately gliomas and meningiomas [21], and one considered brain tumors [22]. However, of the 215 cases included in the latter study, 203 had a glioma histologic type, while the remaining 12 cases had an unspecified type of brain cancer. Thus, the analyses in the present manuscript were conducted to evaluate the relation between coffee consumption and the risk of gliomas, without considering the estimates on the risk of meningioma, since only one paper analyzed such tumor [21] and since gliomas and meningiomas are very different types of brain tumors from a pathological and clinical point of view.

Statistical analysis

Information on coffee and tea intake was collected as cups/day, except in an article reporting coffee drinking in quintiles [21]. The reference category of coffee consumption was nondrinkers in four studies [19, 20, 22, 24], occasional drinkers (<1 cup/day) in one study [23], and the lowest quintile of intake in another study [21]. The reference category of tea consumption was nondrinkers in three studies [19, 20, 22] and the lowest quintile of intake in another study [21].

The analysis comparing drinkers versus non/occasional drinkers was based on five studies for coffee intake [19, 20, 22–24] and three studies for tea intake [19, 20, 22]. One article [21] was excluded from this analysis because the risk estimates were based on quintiles of consumption, whose cutpoints were not available. To obtain the RR for “drinkers” in each study, we first pooled the RRs of each category of consumption of regular drinkers of coffee or tea within each study using the Hamling method [26], which allows us to take into account the correlation between different estimates of the same study, and then we pooled these study-specific estimates using random-effects models, which consider both within- and between-study variation.

In the analyses evaluating the highest level of coffee or tea consumption versus the lowest one, we pooled the RRs for the highest versus the lowest exposure categories of coffee or tea consumption from each study, calculating the summary estimate of the RR using random-effects models. Further, three studies analyzed total coffee plus tea consumption [19–21]. We evaluated the risk of glioma in relation to this combined exposure, as it can be considered as a proxy of total caffeine intake (approximately 90 % of all caffeine) [20]. We pooled the RRs for the highest versus the lowest exposure categories of coffee plus tea consumption from each study to calculate the summary RR using random-effects models.

Finally, we estimated the RR for the increment of coffee of one cup/day using the method proposed by Greenland and Longnecker [27], including studies with at least three categories of coffee consumption [19, 20, 23, 24]. A dose–response analysis was not conducted for tea intake, because only two studies satisfied this criterion. Before conducting a linear dose–response analysis, we checked for a nonlinear relationship between coffee consumption and the risk of glioma, by including a square term in the model. However, all the p-values for such tests were higher than 0.40.

We presented combined estimates using forest plots. In these graphs, a square was plotted for each study, whose center projection on the underlying scale corresponds to the study-specific RR and whose area is proportional to the inverse of the variance of the natural logarithm of the RR. A diamond was used to plot the summary RR, whose center represents the RR and the extremes the 95 % CIs. Statistical heterogeneity among studies was assessed using the X 2 tests (results were defined as heterogeneous for p < 0.10) [28] and was quantified using the I 2 statistic, which describes the percentage of total variation across studies that is due to heterogeneity rather than chance [29]. Usually, values of the I2 statistic <25 % are indicative of low heterogeneity, those ranging between 25  and 75 % of moderate heterogeneity, and those >75 % of high heterogeneity [29]. To evaluate publication bias, we used funnel plots [30] and Begg’s and Egger’s tests [31], where p values of the test less than 0.05 indicate the presence of publication bias.

All the analyses were performed using STATA (version 11; StataCorp, College Station, TX, USA).

Results

Figure 1 shows the flowchart for literature search and selection of articles. Review team members identified 341 articles with PubMed and 1,030 with Embase. Most of them were not focused on the relation between coffee or tea consumption and glioma/brain cancer and were therefore not considered. Of the 28 publications selected for full-text examination, 20 did not report original results from cohort or case–control studies, investigated parental coffee consumption in relation to childhood brain cancer, analyzed patients’ survival rather than cancer incidence, or analyzed patients with other specific diseases or conditions, and were therefore excluded. Eight publications were retained for the meta-analysis. Two publications [32, 33] were further excluded since they reported information on total caffeinated drinks which included also cola intake. The review of the reference lists of these publications did not identify any additional reports. Thus, the present analyses were based overall on six studies. All six studies analyzed coffee consumption (four cohort [19–21, 23] and two case–control [22, 24] studies), and four studies analyzed tea consumption (three cohort [19–21] and one case–control study [22]); three prospective studies analyzed coffee plus tea consumption [19–21] (Table 1).

Table 1 Epidemiological studies on coffee or tea drinking and adult brain cancer included in the meta-analysis
Full size table

Table 1 reports the main characteristics of the six studies included in the meta-analysis. Four studies were conducted in USA, one in Europe, and one in Canada. Four studies considered gliomas only, one provided separate RRs for gliomas and meningiomas, and one considered brain cancer in general. Overall, there were 2,332 cases of adult brain cancer, of which 2,075 were gliomas, 245 meningiomas (excluded from the analyses), and 12 brain cancer of unspecified histological type. Of the 1,239 cases of gliomas for which sex was specified, 768 were men and 471 women.

Figure 2 shows the RRs and 95 % CIs of glioma for coffee (Panel A) and tea (Panel B) drinkers versus non/occasional drinkers from cohort and case–control studies. The summary RR from all combined studies was 0.96 (95 % CI: 0.81–1.13, p for heterogeneity = 0.271, I 2 = 22.6 %) for the five studies analyzing coffee consumption and 0.86 (95 % CI: 0.78–0.94, p for heterogeneity = 0.419, I 2 = 0.0 %) for the three studies analyzing tea consumption. When excluding the study whose reference category was occasional drinkers [23], the summary RR for coffee consumption became 0.90 (95 % CI: 0.78–1.04, p for heterogeneity = 0.391, I 2 = 0.1 %) for drinkers verus non drinkers.

Fig. 2
figure 2

Forest plot for study-specific and pooled relative risks (RR) and 95 % confidence intervals (CI) of adult glioma for coffee (PANEL A) and tea (PANEL B) drinkers versus non/occasional drinkers. RR: relative risk, CI, confidence interval, NR, not reported, PR, persons at risk, PY, person-years. * 12 cases were brain cancers of unspecified histological type

Full size image

Figure 3 shows the RRs and 95 % CIs of gliomas for the highest versus the lowest intake of coffee (Panel A), tea (Panel B), and coffee plus tea (Panel C) from cohort and case–control studies. The overall RR was 1.01 (95 % CI: 0.83–1.22, p for heterogeneity = 0.479, I2 = 0.0 %) for the six studies that analyzed coffee consumption and 0.88 (95 % CI: 0.69–1.12, p for heterogeneity = 0.208, I 2 = 34.1 %) for the four studies analyzing tea consumption. When we pooled the results of the three studies considering coffee and tea together [19–21], the overall RR for the highest versus the lowest consumption was 0.75 (95 % CI: 0.54-1.05, p for heterogeneity = 0.101, I 2 = 56.3 %).

Fig. 3
figure 3

Forest plot for study-specific and pooled relative risks (RR) and 95 % confidence intervals (CI) of adult glioma for the highest versus the lowest consumption of coffee (PANEL A), tea (PANEL B), and coffee plus tea (PANEL C). RR, relative risk, CI, confidence interval, NR, not reported, PR, persons at risk, PY, person-years. * 12 cases were brain cancers of unspecified histological type

Full size image

The RR for the increment of one cup/day of coffee was 1.00 (95 % CI: 0.96–1.05, p = 0.160, I 2 = 41.9 %) for the four studies reporting three or more categories of coffee intake (Fig 4).

Fig. 4
figure 4

Pooled relative risk (RR) of adult glioma for an increment of coffee of one cup/day. RR relative risk, CI confidence interval, PR persons at risk, PY, person-years

Full size image

There was no significant asymmetry in the funnel plots, suggesting the absence of publication bias, and the tests for publication bias provided p values >0.25. However, this analysis was limited by the small number of studies.

Discussion

The present meta-analysis, based on six studies, found no association between coffee consumption and the risk of adult glioma, with consistent results between cohort and case–control studies. The finding of a slight decreased risk for tea drinkers versus non/occasional drinkers was based on three studies only and was not confirmed when the highest dose of intake was compared with the lowest, leaving the issue open to discussion. Similarly, consumption of coffee plus tea showed a tendency to a lower risk of glioma.

Among the prospective studies included in this meta-analysis, the NIH-AARP Diet and Health Study, based on 904 cases of glioma, found no relation with coffee intake up to 6 or more cups/day and a 25 % lower risk of borderline significance for drinkers of 3 or more cups/day of tea [19]. A pooled analysis of three studies (the Nurses’ Health Study I and II, and the Health Professionals Follow-Up Study) [20] found no relation of coffee or tea intake with glioma in either sex, but a RR in men of 0.46 (95 % CI: 0.26–0.81) for the highest quintile of total caffeine intake compared with the lowest one. No relation in women was found with tea intake up to ≥8 cups/week [20]. The European Prospective Investigation into Cancer and nutrition (EPIC) cohort study found no relation of glioma or meningioma with coffee or tea intake, but reported a hazard ratio of 0.66 (95 % CI: 0.45–0.96) in drinkers of >100 ml/day of coffee plus tea compared with consumers of <100 ml/day of the two beverages [21]. The Kaiser Permanente Medical Care Program of Northern California study found a slight, nonsignificant, increased risk of glioma in heavy coffee drinkers, based on 125 cases [23]. Two case–control studies were available. The first, based on 215 adult brain cancer cases (203 of which being gliomas), found no relation for ever coffee or tea drinking [22], while the second one found no relation between the risk of glioblastoma and coffee drinking up to 4 or more cups/day, as compared with nondrinkers [24].

Although based on a small number of studies, the null relation of glioma risk with consumption of coffee is likely real, as most single studies show a lack of relation, results were fairly homogeneous, the exclusion of each study in turn did not materially modify the overall risk estimate, and no significant asymmetry emerged from the funnel plot, indicating that publication bias should not have appreciably changed the results. No heterogeneity was found between studies, possibly because five of the six studies included were conducted in similar geographic areas (four in the USA and one in Canada), with similar type of coffee beverage and similar pattern of consumption. Only a multicentric study [21] was conducted in several European regions, most of which have coffee habits similar to those of northern America. However, no study reported coffee brewing methods and preparation. In the analysis to evaluate the effect of high coffee intake, the highest category included consumption higher than 4 cups/day in all studies, thus confirming the null results also for relatively high amount of coffee consumption. Moreover, the analysis for an increment of one cup/day confirmed a lack of a dose–risk relation.

A limitation inherent in meta-analyses of coffee intake is that the reference category includes some occasional drinkers. However, when we excluded the only study including occasional drinkers in the reference category, there was a tendency toward an inverse relation.

Three prospective studies [19–21] and one of the case–control studies [22] considered the relation between tea and brain cancer risk. The results of any single study were based on a relatively low number of subjects and found no relation except for a slight decreased risk in the NIH-AARP Diet and Health study [19]. However, when we pooled together the results of these studies, a significant inverse association emerged for tea consumption. Few studies were included also in the meta-analysis of tea intake, and the discrepancy in results between the analyses comparing drinkers versus non/occasional drinkers and the highest versus the lowest intake might be due to a lack of dose–risk relation, but also to the inclusion of occasional drinkers in the reference category. This might be particularly important since the highest doses of tea intake considered in two studies were relatively low, equal to ≥8 cups/week [20] and >3 cups/day [19]. Also in the EPIC study, a moderate to heavy amount of tea was consumed in four out of nine countries (the United Kingdom, followed by Denmark, the Netherland, and Germany, corresponding to about half of the subjects included in the study). In the case–control study, the determination of the absolute amount of tea intake was reported in terms of lifetime consumption [22]. We did not consider the relation between tea intake and glioma at various levels of tea intake, as only two papers considered such issue [19, 20]. They both showed no relation between tea intake at doses below 3 cups/day and a protection for higher doses, suggesting a threshold effect rather than a linear relationship. The issue is, however, still open to discussion and needs more evidence. None of the study reported the type of tea consumed (black or green). Thus, the results for tea intake are less clear, and the lack of studies conducted in Asia, where generally tea intake is much higher, leaves open the discussion on a possible inverse relation of tea intake with brain cancer risk.

Coffee plus tea intake, considered as a proxy for total caffeine intake, has been suggested to have a favorable effect on glioma risk. We found a pooled RR below unity supporting such hypothesis. However, this result was based on three studies, with only one of them showing a clear protection and thus needs to be confirmed by further research.

Two studies, excluded from the meta-analysis because they considered caffeinated beverages in general [32, 33], found a slight tendency, not statistically significant, to a decreased risk when considering the highest intake of caffeinated beverages versus the lowest, with a RR of 0.3 (95 % CI: 0.1–1.2) in the prospective study [32] and an odds ratio of 0.55 (95 % CI: 0.30–1.02) in women and 1.10 (95 % CI: 0.71–1.71) in men in the case–control study [33]. Considering that coffee and tea contribute to approximately 90 % of the total amount of caffeine consumed [20], these findings confirm the results of the present meta-analysis.

There is no explanation for the potential different effect of coffee and tea intake on adult glioma risk. A hypothesis is that substances different from methylxanthines contained mostly in tea are responsible for the favorable effect, if any, of tea intake. However, the results are still too scanty and unclear.

In conclusion, the results of this meta-analysis, although based on a small number of studies, suggest that coffee consumption is not related with the risk of adult gliomas. The analyses of the relation of glioma risk with tea or with coffee plus tea intake suggest a null relation or a moderate inverse association. However, further examination in studies conducted in populations with high tea intake is needed.