About 50 % of all primary brain tumours are gliomas and 25 % are meningiomas. Gliomas are of three main types (astrocytoma, ependymoma, oligodendroglioma) and are often associated with poor prognosis. Meningiomas are a usually benign type of brain tumour, but some can be ‘atypical’ and behave more aggressively1. A variety of risk factors for brain cancer have been investigated in epidemiological studies, but the evidence for environmental causation is inconsistent. Associations observed include certain medical conditions, exposures to radiation, viruses and chemicalsReference Wrensch, Minn, Chew, Bondy and Berger2, Reference Schwartzbaum, Fisher, Aldape and Wrensch3. The relationship between dietary trace elements and adult brain tumour aetiology has not yet been fully investigated, as few studies involving trace elements have been conductedReference Wrensch, Minn, Chew, Bondy and Berger2.
Zn is a trace element with antioxidant properties; such elements have been suggestedReference Chen, Ward, Tucker, Graubard, McComb, Potischman, Weisenburger and Heineman4 to be protective against brain tumour development. The main role of Zn is the maintenance of a healthy central nervous system. Zn is also important for DNA replication, protein synthesis and metabolismReference Vallee and Falchuk5 and oxidative stress protectionReference Stehbens6. The present study was prompted by an a priori hypothesis suggested by animal models. In rat models, Zn is essential for good neuronal functionReference Colvin, Davis, Nipper and Carter7–Reference Warming, Rachel, Jenkins and Copeland9. It has been shown that, in rat glioma cellsReference Ho and Ames10, increased oxidative stress occurs during Zn deficiency. Ho & AmesReference Ho and Ames10 also reported that, under low intracellular Zn status, proper DNA repair could not be achieved, and after Zn repletion DNA damage was reversed. Yousef et al. reported a significant increase in the levels of free radicals with Zn deficiency in the rat brainReference Yousef, El-Hendy, El-Demerdash and Elagamy11. In addition, some human case–control studies have yielded inverse associations between Zn consumption and various cancers, such as oesophageal squamous cell carcinomaReference Lu, Cai and Mu12 and lung cancerReference Zhou, Park, Liu, Miller, Wang, Pothier, Wain, Lynch, Giovannucci and Christiani13. Zn adjusted for Fe intake was inversely associated with upper digestive tract cancer in the follow-up Iowa Women's Health StudyReference Lee, Anderson, Folsom and Jacobs14. In contrast, a recent case–control study found that excessive Zn intake ( ≥ 15·7 mg/d) increases prostate cancer risk in humansReference Gallus, Foschi, Negri, Talamini, Franceschi, Montella, Ramazzotti, Tavani, Dal Maso and La Vecchia15.
The concentration of Zn in the brain is higher than elsewhere in the body (about 150 μmol/l)Reference Mocchegiani, Bertoni-Freddari, Marcellini and Malavolta16. Zn is most concentrated in neuron-abundant forebrain regions (for example, hippocampus) serving as an endogenous modulator in neurotransmissionReference Frederickson, Suh, Silva, Frederickson and Thompson17. Excess excitation of Zn-containing neurons causes Zn decrease and neuronal damage. Dietary Zn deprivation may influence Zn balance in the brain, resulting in brain dysfunctionReference Takeda18. Other dietary nutrients affect Zn concentrations in the brain and blood and, possibly, Zn availability for transport into the brain through the blood–brain barrierReference Takeda19. A number of nutrient elements, such as Ca, Fe, Cu and P, act as antagonists to ZnReference Sandstrom20, while other nutrients such as PUFA, fibre and protein facilitate Zn absorptionReference Greger and Snedeker21, Reference Solomons, Cousins, Solomons and Rosenberg22, and some are able to cross the blood–brain barrier via different transport systems.
The present study investigated the a priori hypothesis that higher dietary Zn levels may be associated with a decreased risk of brain tumour development in a large population-based case–control studyReference Hepworth, Schoemaker, Muir, Swerdlow, van Tongeren and McKinney23.
Subjects and methods
The UK Adult Brain Tumour Study (UKABTS) is a population-based case–control study conducted in the Trent, West Midlands, West Yorkshire and central Scotland regions of the UK. A common protocol was followed with identical methods for case ascertainment, control selection and data collectionReference Hepworth, Schoemaker, Muir, Swerdlow, van Tongeren and McKinney23.
Cases were ascertained from hospital departments (for example, neurosurgery, neuro-oncology, neuropathology). Study subjects were aged 18–69 years, resident in the study areas and first diagnosed between 1 December 2000 and 30 June 2003 with a glioma (International Classification of Diseases (ICD)-O-3, topography: C71, morphology: 9380–9411, 9420–9451, 9480, 9505) or meningioma (ICD-O-3, topography: C70, morphology: 9530–9539). Controls were randomly sampled from general practitioner lists and individually matched to cases on age and sex. Non-participating controls were replaced. Eligible subjects were approached by their treating consultant or general practitioner either personally or by letter.
Participants were interviewed using a computer-assisted personal interview system, and then given a FFQ to complete and return by post. Information was collected on dietary intake and use of vitamin, mineral and other dietary supplements. The FFQ includes questions on the average consumption frequency of a medium portion of 132 food items (the most commonly consumed in this population). The subjects were asked about their usual diet during the 2 years preceding diagnosis, to reduce the possibility of reverse causation. Consumption frequency categories varied from ‘six or more per day to ‘never or less than once per month’.
Data analysis
Average daily nutrient intake was calculated by multiplying the daily consumption frequency of each food item by the content of the examined nutrient in the respective food item obtained from food composition tablesReference Holland, Welch, Unwin, Buss, Paul and Southgate24. Data were then processed by the nutritional software based on the program used for the European Prospective Investigation into Cancer (EPIC) study.
Dietary Zn intake was adjusted for energy intake using the residual methodReference Willett, Howe and Kushi25, Reference Willett26 and intake levels were defined by quartiles of the control distribution (lowest category used as the reference group).
Standard unconditional logistic regression was used to estimate OR and 95 % CI in univariate and multivariate analyses, for gliomas and meningiomas separately. All controls were used in the analyses, as in a previous analysis from the UKABTS on the association between the use of mobile phones and risk of developing brain tumoursReference Hepworth, Schoemaker, Muir, Swerdlow, van Tongeren and McKinney23. In addition to sex, age (in 5-year groups) and region, the multivariate standard logistic regression adjusted simultaneously for the following variables: deprivation category (Townsend score reflecting social class)Reference Townsend, Phillimore and Beattie27, season of dietary questionnaire return, multivitamin supplement use and energy intakeReference Willett28 (pp. 288–291). Because energy intake may be an important disease predictor, it was included in the regression model together with the nutrient energy-adjusted termReference Willett28 (pp. 288–291).
Subjects' intake of other nutrients besides Zn was also assessed and included as terms in the regression analysis. The literature suggests that nutrients having a biological relevance to Zn are the following: Ca, Fe, Cu, P (the main Zn antagonists), PUFA, protein and dietary fibre (the last three are thought to affect Zn absorption and amounts in the body, for example, protein promotes Zn absorption). These were tested for interaction with Zn, by including interaction terms in the model. Nutrients were also assessed for confounding. In the regression analysis, non-significant nutrient terms were taken out of the model, also provided that excluding them did not largely inflate the standard error while changing very little the corresponding effect size of the examined variable (Zn intake). Presented results are those obtained with inclusion of only those nutrients that remained significant.
Data analysis was carried out using the SPSS statistical package (version 11.5; SPSS, Inc., Chicago, IL, USA). All presented P values are two-sided.
Ethical approval
Approval has been obtained from multi-centre research ethics committees (MREC/99/0/77) and all relevant local research ethics committees.
Results
Of those who returned an FFQ, 637 cases (436 gliomas, 201 meningiomas) and 876 controls were included in the analyses, after fifteen subjects (eleven cases, four controls) were excluded as their energy intake and BMI were incompatible.
Table 1 gives the response rates – for the dietary FFQ – of cases and controls grouped by tumour type. Table 2 presents the demographic and social characteristics of subjects who returned the dietary FFQ. Table 3 shows the results of analysis by brain tumour subtype. For glioma, no association was seen with Zn before or after adjustment for confounders. A statistically significant risk reduction for meningioma was observed only in the 3rd quartile of dietary Zn intake (adjusted OR 0·62 (95 %CI: 0·39, 0·99); P = 0·048). The crude results were not significant.
Table 1 Interview and questionnaire response rate in the study
(Frequencies and percentages)
* Includes no permission by consultant or general practitioner, non-English speaking, mental impairment or institutionalised.
Table 2 Characteristics of cases and controls in the study
(Frequencies and percentages)
Table 3 Association between dietary zinc intake and gliomas and meningiomas
(Odds ratios and 95 % confidence intervals)
Q, quartile.
* Significant at P < 0·05 level; two-tailed P value.
† Adjusted for age (in 5-year groups), sex, study region, deprivation category (Townsend score), season of FFQ return and multivitamin supplementation.
For nutrients biologically relevant to Zn as suggested from the literature (Ca, Fe, Cu, P, PUFA, protein and fibre), interaction terms were included in the analysis. However, all interaction terms were found to have non-significant overall P values. The above nutrients were also assessed for confounding, and those significant were entered in the regression analysis.
Confounders remaining significant when examining the Zn–disease association were Fe for gliomas and Cu for meningiomas (overall significance P = 0·05 and P = 0·02 respectively). Results appear in Table 4. Zn intake was significantly correlated with both Cu intake and Fe intake at the P < 0·01 level (the Pearson correlation coefficient between Zn and Fe intake for gliomas was 0·25, and between Zn and Cu intake for meningiomas was 0·27). However, strong collinearity was not observed in the data, as collinearity tests conducted were not significant (gliomas R 2 0·05; meningiomas R 2 0·13). Note that, after taking account of Cu intake in the analysis for meningiomas, the result for the 3rd quartile was no longer statistically significant (Tables 3 and 4).
Table 4 Association between dietary zinc intake and gliomas and meningiomas adjusted for iron and copper intake respectively
(Odds ratios and 95 % confidence intervals)
Q, quartile.
* Two-tailed P values; P < 0·05 significance level.
† Energy-adjusted mean intake of Fe and Cu was 28·82 and 2·42 mg respectively for controls.
‡ For glioma, adjusted for age (in 5-year groups), sex, study region, deprivation category (Townsend score), season of FFQ return, multivitamin supplementation and Fe intake.
§ For meningioma, adjusted for age (in 5-year groups), sex, study region, deprivation category (Townsend score), season of FFQ return, multivitamin supplementation and Cu intake.
‖ Energy-adjusted mean Fe intake for glioma cases was 27·19 mg.
¶ Energy-adjusted mean Cu intake for meningioma cases was 2·46 mg.
We also obtained results for groupings of tertiles according to the RDA recommendations for Zn (8–11 mg/d). Results were similar to those already obtained (before adjustment, P gliomas = 0·561, P meningiomas = 0·125; after adjustment for Fe, Cu respectively, P gliomas = 0·577, P meningiomas = 0·224) and no significant associations were observed.
Additional adjustment of Zn intake quartiles for dietary intake of vitamins A (carotene), B12, B6, C, D, E, biotin, niacin, retinol, riboflavin and thiamin did not alter any of the results.
Discussion
Zn is involved in cell division and differentiation, in tumour cell metabolismReference Vallee and Falchuk5, and in the normal development of natural killer cellsReference Shankar and Prasad29. Normal Zn levels work against superoxide free radicalsReference Shankar and Prasad29, Reference Williams, Spencer, Goni and Rice-Evans30, and it is often suggested that free radical reduction may help to lower cancer riskReference Chan, Gerson and Subramaniam31, Reference Prasad, Bao, Beck, Kucuk and Sarkar32. Zn deficiency is prevalent in some cancers, and low Zn levels may reduce the number of helper T-cells and thymic hormone levelsReference Hadden33, thereby weakening immune functionReference Shankar and Prasad29. Cancer, in general, arises more frequently against a background of immunodeficiencyReference Hadden33.
On the other hand, in animal models with existing tumours, depletion of dietary Zn has been proven to suppress tumour growthReference Takeda, Tamano and Oku34, Reference Takeda, Goto and Okada35. Excess Zn intake has been linked to disease and toxicityReference Chan, Gerson and Subramaniam31. Reduced immune function can result from both excessive Zn intakeReference Santillo and Lowe36 and low Zn intake, as mentioned earlier. The above contradicting evidence shows that the mechanisms behind the Zn intake–brain tumour relationship are not yet fully comprehended; thus, a balanced intake is recommendedReference Naganska and Matyja37.
Blood–brain barrier dysfunction has been linked to neurological conditions and brain tumour development, i.e. the barrier is usually non-existent in brain tumoursReference Rapoport38. Intake of normal Zn levels is required for a healthy blood–brain barrierReference Takeda19, as enhanced dietary Zn consumption does not affect Zn concentration in the brain except for the case where Zn deprivation already existsReference Takeda18.
Results of a recent study show that Zn depletion damages non-brain endothelial cells; however, the brain endothelial cells respond by enhancing barrier propertyReference Di Cello, Siddharthan, Paul-Satyaseela and Kim39. Levels of other elements in the brain play an important role, as interaction with elements transported across the blood–brain barrier affects Zn absorption and its concentrations in the brain and, subsequently, its contribution to normal brain function. Fe and Cu are both elements that can pass through the blood–brain barrier and affect Zn levelsReference Sandstrom20. Fe is transported through the barrier by p97Reference Moroo, Ujiie, Walker, Tiong, Vitalis, Karkan, Gabathuler, Moise and Jefferies40 and Cu via a Cu-transporting ATPase mechanismReference Qian, Tiffany-Castiglioni, Welsh and Harris41.
Potential areas of bias
Participation levels were relatively low amongst cases. This was due mainly to disease severity; patients were very ill or died before being recruited. As reported previouslyReference Hepworth, Schoemaker, Muir, Swerdlow, van Tongeren and McKinney23, high-grade glioma cases were less likely to be interviewed than those diagnosed with a low-grade tumour. Control participation was also low, a problem for many population-based studiesReference Willett28 (pp. 9–11, 90–91). This may have introduced selection bias amongst controls, as previously reported controls taking part in the study tended to be more affluent than non-interviewed controlsReference Hepworth, Schoemaker, Muir, Swerdlow, van Tongeren and McKinney23. Controls used may have better nutritional regimens than the general population. Deprivation category was adjusted for in the analysis, although the bias cannot be fully removed.
The FFQ method is cheap, easy to administer, and provides quick intake estimatesReference Willett28 (pp. 74–91). Although extreme misclassification has been shown to be minimalReference Thompson and Byers42, another problem associated with use of FFQ in case–control studies is that questions may have been misinterpreted by some subjectsReference Willett28 (pp. 101–124, 302–304).
Energy adjustment, as carried out in the present study, minimises errors resulting from general food consumption over- or under-reportingReference Willett, Howe and Kushi25. The dietary FFQ used (based on the EPIC FFQ) containing questions on as many as 132 food items commonly consumed in the UK might also have reduced under-reporting of food consumptionReference Thompson and Byers42.
In the study, the frequency question was combined with a specific ‘medium portion’ size and this can present cognitive challenges for subjectsReference Hunter, Sampson, Stampfer, Colditz, Rosner and Willett43. However, several studies have found that consumption frequency is the main determinant of between-person variation in measured dietary intake levels and that it is positively correlated with portion sizeReference Hunter, Sampson, Stampfer, Colditz, Rosner and Willett43.
There is concern that cases will report on their diets differently to controlsReference Friedenreich, Howe and Miller44. Brain tumours are associated with impaired memory and concentrationReference Nelson, Bingham, Margetts and Nelson45 and current dietary habits also considerably affect responses regarding previous dietReference Thompson and Byers42. However, recall bias is reduced by recruiting incident casesReference Willett28 (pp. 153–155), as has been done in the present study.
Missing values
The suggestive result before Cu inclusion could potentially be stronger, as dietary assessment through FFQ inherently produces measurement error and generally modest relative risksReference Willett28. In that respect, missing values are potentially a source of bias in the present study. Some foods (for example, cooked vegetables), tend to be more frequently omitted than othersReference Willett28 (pp. 61–67) and respondents tend to selectively omit foods they never or seldom eatReference Caan, Lanza, Schatzkin, Coates, Brewer, Slattery, Marshall and Bloch46. After conducting a missing values analysis, we found that there were significantly more missing values for cases than controls. Also, responses on Zn-containing foods are different from those on food items containing Cu. Of missing values for Zn, 64 % are for foods with zero Zn content; the respective percentage for Cu is 56 %. Although Cu has 4 % of its missing values for foods containing 5·8–9·9 mg Cu (the highest composition range), the highest percentage (22 %) of missing values is accumulated in the 0·01–0·09 mg range (lowest). However, the highest percentage (31 %) of missing values for Zn is found in the 0·1–1·0 mg range. Therefore, more of the missing values for Zn are for foods with a moderate composition of Zn, while more of the missing values for Cu are for foods low in Cu, indicating that Zn intake may have been underestimated. It would be interesting to see if Zn amounts greater than used here would yield an effect.
Conclusions
In this dietary investigation of the UKABTS, no associations were found between dietary Zn intake and risk of glioma or meningioma. Overall, our findings are non-significant. The specific hypothesis on a protective effect of increased compared with low levels of dietary Zn against glioma or meningioma formation is not supported.
There is no strong multi-collinearity in the data. Therefore, controlling for a confounding effect of Fe and Cu intake is helpful, as relationships of dietary elements are complex and it is difficult to separate the effects of one element alone from the effects of others.
Acknowledgements
The dietary analysis of the study was funded by Cancer Research UK grant number C18182/A5769. The authors wish to thank P. A. McKinney and M. Grainge for their expert advice and comments. We acknowledge the support of the study steering group chaired by David Coggon. The nutritional database version based on EPIC was designed by Dr A. Lophatananon. The UKABTS received funding from the Mobile Telecommunications, Health and Research programme and as part of the Interphone study from the European Union, the Mobile Manufacturers Forum, and the Global System for Mobile Communications (GSM) Association through the scientifically independent Union Internationale Contre le Cancer, the Health and Safety Executive, the Department of Health, the UK network operators (O2, Orange, T-Mobile, Vodafone, 3) and the Scottish Executive.
We wish to thank the following neuropathologists, neuroradiologists, neurosurgeons, neuro-oncologists, clinical oncologists, neurologists, specialist nurses, administrators and secretaries: P. Barlow, I. Bone, J. Brown, J. Crowther, R. Dolan, L. Dunn, M. O. Fitzpatrick, M. Fraser, R. Grant, A. Gregor, J. Ironside, R. Johnstone, K. W. Lyndsay, S. Macnamara, J. Mair, R. Mills, L. Myles, B. O'Reilly, V. Papanastassiou, R. Rampling, M. Russell, D. Sim, P. Statham, J. Steers, W. A. Taylor, G. Teasdale and I. Whittle (Scotland); J. M. Anderson, P. Barber, C. R. Barraclough, P. Bennett, H. G. Boddie, A. Brind, P. Carey, M. Choksey, M. Christie, R. N. Corston, G. S. Cruickshank, A. Detta, P. Dias, S. J. Ellis, G. Flint, D. A. Francis, A. H. Grubneac, S. P. Harland, C. Hawkins, T. Heafield, R. C. Hughes, D. G. Jamieson, A. Logan, C. H. A. Meyer, R. Mitchell, K. Morrison, P. Newman, D. Nicholl, S. Nightingale, H. S. Pall, J. R. Ponsford, A. Shehu, J. Singh, J. A. Spillane, P. Stanworth, B. Summers, A. R. Walsh, J. Wasserberg, A. C. Williams, J. Winer and S. Zygmunt (West Midlands); R. J. Abbott, S. Adams, R. D. Ashpole, R. D. E. Battersby, L. Blumhardt, P. Byrne, M. Cartmill, S. C. Coley, P. Critchley, B. B. Faraj, A. Gibson, P. Griffiths, R. Grunwald, T. J. Hodgson, D. T. Hope, S. Howell, D. Jefferson, D. Jelinek, N. Jordan, A. Kemeny, M. C. Lawden, J. Lowe, N. Messios, K. Pardoe, S. Price, I. F. Pye, M. Radatz, I. Robertson, K. Robson, C. Romanowski, G. Sawle, B. Sharrock, P. Shaw, C. Smith, W. Temperley, G. Venables, B. White, A. M. Whiteley and A. J. Wills (Trent); A. S. N. Al-Din, D. Ash, J. Bamford, M. Bond, G. Bonsor, L. Bridges, B. Carey, A. Chakrabarty, P. Chumas, D. Dafalla, H. Ford, G. E. Gerrard, P. J. Goulding, J. Howe, S. Jamieson, M. H. Johnson, L. A. Louizou, P. Marks, M. Nelson, S. Omer, N. Phillips, S. Ross, I. Rothwell, H. Spokes, J. Straiton, G. Towns, A. Tyagi, P. Vanhille and M. Busby (West Yorkshire).
K. R. M. and M. vT. were responsible for the design and implementation of the study. P. D. and J. F. L. conducted the dietary and statistical analyses. P. D. and S. N. wrote the first draft of the paper, PD wrote all subsequent drafts. L. D. conducted data entry and checking. K. R. M., J. F. L., S. J. H. and M. vT. provided comments on the draft. The authors declare that there are no conflicts of interest.
Electronic Publications Rights
As the author(s) and beneficial owner(s) of the entire copyright and all other rights in the research paper entitled
DIETARY ZINC INTAKE AND BRAIN CANCER IN ADULTS: A CASE-CONTROL STUDY
we hereby, reserve, with the consent of our employers, for the benefit of Cancer Research UK, in recognition of the funding for the work described, the non-exclusive right to access and to distribute this paper, a summary of this paper and any extracts from this paper electronically and in paper form, unencumbered and free of any charges provided that this use is solely for the purposes of promotion and fundraising and in pursuit of the charitable aims of Cancer Research UK. These rights are reserved in accordance with the duty that Cancer Research UK has as a UK Registered Charity to make available information about the work that it has funded.
Dimitropoulou Polyxeni
Nayee Suneet
Liu Jo-Fen
Demetriou Lia
van Tongeren Martie
Hepworth Sarah
Muir Kenneth
17/01/07
