|
|
| |||||||||||||||||||||||||||||||||||||||
Background It has been hypothesized that a high dietary intake of nitrosamines and their precursors, nitrites and nitrates, is a risk factor for brain tumors. Vitamins C and E inhibit the formation of nitrosamines and thus may be protective.
Methods We conducted a case-control study of maternal diet and the risk of primitive neuroectodermal tumors of the brain in children. The case patients were under the age of six years at diagnosis in 1986 to 1989. The controls were selected by random-digit telephone dialing and were matched for age and race to 166 case patients. Telephone interviews with the mothers included questions on the frequency of consumption of alcohol, vitamin and mineral supplements, and 53 foods during pregnancy.
Results Significant protective trends were observed for vegetables (odds ratio for the highest quartile group for intake relative to the lowest, 0.37; P for trend = 0.005), fruits and fruit juices (odds ratio, 0.28; P = 0.003), vitamin A (odds ratio, 0.59; P = 0.03), vitamin C (odds ratio, 0.42; P = 0.009), nitrate (odds ratio, 0.44; P = 0.002), and folate (odds ratio, 0.38; P = 0.005). A nonsignificant trend of increasing risk was observed for nitrosamine (odds ratio, 1.65; P = 0.15). The use of iron (odds ratio, 0.43; P = 0.004), calcium (odds ratio, 0.42; P = 0.05), and vitamin C (odds ratio, 0.35; P = 0.04) supplements at any time during the pregnancy and the use of multivitamins during the first six weeks (odds ratio, 0.56; P = 0.02) were associated with decreased risk. In multivariate analyses, folate, early multivitamin use, and iron supplements generally remained protective.
Conclusions These results do not support the hypothesis that nitrosamines have a role in the development of primitive neuroectodermal tumors in young children, but they do suggest that certain other aspects of maternal diet can influence the risk.
We conducted simultaneous studies to investigate the effect of maternal diet during pregnancy on the risk of the two most common brain tumors in children, astrocytoma and primitive neuroectodermal tumor. The two types differ in their age-incidence curves and sex ratios,9,10,11 and their causes may therefore be different as well. We restricted our study to young children, since gestational factors would be expected to act early in life. In this paper, we present the results for primitive neuroectodermal tumor, with some reference to the results for astrocytoma.
Methods
Eligible patients had a primitive neuroectodermal tumor12 in the brain diagnosed before the age of six years between 1986 and 1989 and were registered with the Children's Cancer Group, one of the two pediatric oncology cooperative groups in North America (see the Appendix for a list of the principal investigators). We attempted to have all the potentially eligible cases reviewed by one pathologist. If the reviewing pathologist did not classify the tumor as a primitive neuroectodermal tumor, the patient was excluded from the study. When material was not available for pathological review, we used the diagnosis of the original pathologist.
Of the 281 potentially eligible patients, 28 were excluded after the pathologist's review. The physician's consent was denied for another 15 patients, the mothers of 10 patients did not speak English, the biologic mothers of 9 were unavailable, 16 had no telephone, 2 resided outside the United States or Canada, and for 1 no response was obtained from the hospital. Of the remaining 200 eligible patients, 6 could not be located, the parents of 23 refused to participate, and 5 could not be contacted and interviewed before the study ended. A total of 166 case mothers were interviewed.
Nearly 90 percent of the primitive neuroectodermal tumors in the study would previously have been described as medulloblastomas. The remaining tumors included 14 neuroblastomas in the brain (8 percent), 3 ependymoblastomas (2 percent), and 1 pineoblastoma (1 percent). Of the 166 tumors under study, 147 (89 percent) were reviewed by the pathologist and confirmed as primitive neuroectodermal tumors. The reviewing pathologist classified 15 percent of the submitted materials as from something other than a primitive neuroectodermal tumor. Thus, about 3 of the 19 tumors without pathological review were probably not primitive neuroectodermal tumors.
Control children were selected by random-digit telephone dialing13 and were matched to the case patients according to area code and the next five digits of the telephone number, date of birth (±one year), and race (black or nonblack). Enough information to determine eligibility was obtained from 89 percent of the residences called. Of the 166 controls, 123 (74 percent) were the first eligible controls identified. In 43 instances (26 percent), the first eligible control refused; the second control identified was enrolled in 32 instances (19 percent), and the third, fourth, fifth, or sixth control identified was enrolled in 11 instances (7 percent). In 10 instances, one of the matching criteria had to be relaxed in order to find a control.
Trained interviewers conducted telephone interviews with the mothers of the case patients and controls an average of 4.5 to 4.7 years, respectively, after the index child's birth. The structured interview, with a mean length of 62 minutes, included questions about household, personal, and medical exposures during the pregnancy; occupation; and family medical history. It also included a food-frequency questionnaire with 53 food items, similar in design to that of Block et al14. The food items were chosen to permit the efficient assessment of dietary components relevant to the nitrosamine hypothesis, as suggested by Howe et al15. Additional fruits and vegetables were added to cover geographic and ethnic diversity in diet. Other items, such as coffee, tea, diet soda, and raw meat, were added for reasons unrelated to the nitrosamine hypothesis. Mothers provided data on their food consumption during pregnancy in terms of frequency per day, week, or month. Data on portion size were not collected. The frequency of consumption of alcoholic beverages, which contain nitrosamines, was determined in another part of the questionnaire.
We calculated the intake of nine dietary components: vitamin A, beta carotene, retinol, vitamin C, vitamin E, nitrate, nitrite, dimethylnitrosamine, and folate. Although not relevant to the nitrosamine hypothesis, vitamin A or beta carotene may be protective against some cancers, and folate is protective against neural-tube defects. We generally used data on portion size and content from the foods data base of Block et al14. For the five foods not in that data base, we used data from Howe et al15. The portion sizes used in our analyses were medium portions for women of reproductive age. Information on the nitrate, nitrite, and dimethylnitrosamine content was obtained from two sources15,16.
Cola drinks, coffee, and tea were assigned weights based on their content of caffeine,17 and these weights were combined with the frequency of consumption to produce an index of caffeine consumption. Data on sources of drinking water, the occurrence of nausea, and the use of vitamin supplements during pregnancy were collected. Each mother was also asked about her child's frequency of consumption of four fruits and fruit juices, six cured meats, and multivitamins and other vitamin supplements in the first year of life.
The effects of individual food items were investigated by comparing consumption at least once per week with consumption less than once per week. For the analysis of food groups and food components, we classified the mothers according to quartile of intake, using the combined distribution of intake among case and control mothers.
Matched-pairs odds ratios were used to measure the effects of individual food items and other dichotomous variables, and the corresponding two-sided P values were calculated with McNemar's test. The effect of quartile group for food-group and food-component intake was investigated with conditional multiple logistic regression18. This method was also used to investigate the independent effects of food components, to adjust for possible confounding by factors such as the child's diet, the family income, and maternal smoking, and to explore possible interactions among dietary, demographic, and other factors. Statistical significance was indicated by a P value of less than 0.05.
The parallel study of astrocytoma included 155 case-control pairs and was conducted by the same investigators and interviewers using the same questionnaire. The study groups were not separated until the analysis phase. The results of the astrocytoma study will be reported elsewhere.
Results
The median age of the case patients at diagnosis was 33 months. As expected with a matched characteristic, the racial distribution of the case patients and controls was nearly identical; 92 percent of the study subjects were white. Sixty percent of the case patients and 58 percent of the controls were boys. The case and control mothers were similar in demographic characteristics, including household income, educational level, type of housing, and number of residences since the index pregnancy (Table 1). The mothers were also similar in their sources of drinking water and in the month of their first visit to a physician during the index pregnancy (Table 1).
|
|
|
The highest quartile groups for vitamin C, nitrate, and folate intake were associated with significantly decreased odds ratios, and the trends were significant (Table 3). Although the odds ratio for the highest quartile group for vitamin A consumption was not significant, the trend of decreasing risk was. The results for beta carotene were virtually identical to those for vitamin A, whereas retinol had no association with risk. A slight, but not statistically significant, trend of increasing risk was observed for dimethylnitrosamine intake. There was a significantly decreased odds ratio for the third quartile group for vitamin E intake, but no trend was evident. No effect of nitrite was observed.
Nearly all case and control mothers (92 percent) took multivitamins during their pregnancies, and no effect on risk was observed. However, multivitamin use in the first six weeks after the last menstrual period was associated with an odds ratio of 0.56 (95 percent confidence interval, 0.32 to 0.96; P = 0.02). A significant protective effect was observed for iron supplements (odds ratio, 0.43; 95 percent confidence interval, 0.22 to 0.80; P = 0.004), calcium supplements (odds ratio, 0.42; 95 percent confidence interval, 0.16 to 1.01; P = 0.05), and vitamin C supplements (odds ratio, 0.35; 95 percent confidence interval, 0.11 to 0.94; P = 0.04).
For comparison, Table 4 shows selected results from the parallel study of astrocytoma. No evidence of an effect of folate was observed. Although the odds ratios for fruit, vegetables, vitamin C, and nitrate generally decreased in the higher quartile groups, none of the trends were statistically significant. Iron and calcium supplements had protective effects of a magnitude similar to that for primitive neuroectodermal tumor, with odds ratios of 0.49 and 0.44, respectively. However, no protective effect of multivitamin use in the first six weeks of pregnancy was observed (odds ratio, 1.16; 95 percent confidence interval, 0.72 to 1.86; P = 0.55).
|
Supplemental daily multivitamin use during the first year of life was associated with a decreased odds ratio of 0.68 (P = 0.11). No significant trends with increasing consumption were observed for cured meat, orange juice, apple juice, and other fruit juices. Although no trend was observed for increasing intake of fruit (blended or solid), not eating any fruit in the first year of life increased the risk significantly (odds ratio, 4.3; 95 percent confidence interval, 1.23 to 15.21; P = 0.009).
Since most of the dietary variables were significantly correlated, we performed multiple conditional logistic regression with seven maternal food components and four maternal supplements. The multivariate results were substantially different from the univariate effects (Table 5). Vitamins A and C, which had been protective, now had little effect on risk; vitamin E, which had had little effect, now significantly increased the risk (odds ratio for the highest as compared with the lowest quartile group, 3.80; P for trend = 0.02); the protective effect of nitrate became weaker (odds ratio, 0.54; P for trend = 0.19); the effect of folate became stronger (odds ratio, 0.15; P for trend = 0.02); and the effects of nitrite and nitrosamine remained about the same. The effect of early multivitamin use did not change, whereas the effects of iron, calcium, and vitamin C supplements became somewhat weaker. Early multivitamin use and iron were of at least borderline statistical significance. The exclusion of folate from the model resulted in a protective effect of vitamin C (odds ratio, 0.43; P for trend 
0.10) (Table 5). When margarine consumption was included in the model, the effect of vitamin E disappeared. A model with the three food groups rather than the food components produced significant results. Fruit and vegetable intakes were correlated, however, and no individual food group was significant.
|
Discussion
We conducted a study of maternal diet during pregnancy and the risk of primitive neuroectodermal tumors in children. Before discussing the implications of our results, we should consider the limitations of the dietary assessment and the possibility of bias. The dietary questionnaire was not tested for validity or reliability, but food-frequency questionnaires have been shown to be valid and reliable for many nutrients19. We did not collect data on portion size, but used average portion sizes for young, nonpregnant women from previous studies. However, the use of a standard portion may not introduce a large error in dietary assessment, since portion sizes vary more for an individual subject than among subjects20. Even if the lack of data on portion size led to error in the assessment of food components, bias would be unlikely unless portion size differed consistently between case and control mothers.
As in any case-control study, recall and selection bias must be considered. Selection bias would occur if health-conscious parents with good diets were more likely than other parents to participate as controls. However, we did not observe differences between case and control mothers in health-related behavior such as smoking, breast-feeding, and seeing a doctor early in pregnancy. A recent study of controls obtained by random-digit telephone dialing with a participation rate of 75 percent, similar to that in the present study, found no differences between the control group and the general population in most health-related characteristics, including smoking, exercise, weight, and the frequency of health care visits21.
Whether there is recall bias in dietary factors has not been studied extensively. The existing evidence suggests that differences between case patients and controls in recalling past diet are small and unlikely to explain the findings of this study22,23,24. The parallel study of astrocytoma allowed us to assess the possibility of recall (or other) bias to some extent. Similar results for the two types of tumor would be consistent with a common bias. Two factors, multivitamin use early in pregnancy and high folate intake, were associated with a decreased risk of primitive neuroectodermal tumor but had no effect on the risk of astrocytoma. The specificity of these findings for primitive neuroectodermal tumor makes bias an unlikely explanation. In contrast, fruits, vegetables, vitamin C, and nitrate were found to have protective effects against both astrocytoma and primitive neuroectodermal tumor, although only against primitive neuroectodermal tumor were the effects strong and statistically significant. The similar findings for both types of tumor are consistent with recall bias, but shared causes and chance could also explain the data. Although the evidence of recall bias does not seem strong, we cannot exclude the possibility. We therefore have less confidence in these findings than in those for folate and multivitamins.
Like previous studies of brain tumors and diet, ours focused on the hypothesis that high intakes of nitrosamine, nitrite, and nitrate increase the risk of brain tumors, and high intakes of vitamin C and vitamin E decrease the risk. Our findings on nitrosamine, nitrite, nitrate, and vitamin E are generally inconsistent with the nitrosamine hypothesis. We observed only a small effect of nitrosamine and no effect of nitrite or vitamin E. Nitrate appeared protective, reflecting the high nitrate content of some vegetables. We assessed all major sources of dietary nitrite and nitrosamine. However, our dietary questionnaire did not include some important sources of vitamin E25. Since incomplete assessment of intake can obscure an association, we cannot exclude an effect of vitamin E. Our measure of nitrate intake was also incomplete, since we did not assess intake from drinking water, which is a major source of nitrate for some people16. It seems unlikely, however, that including drinking water would reverse our finding of a strong protective effect of nitrate.
The nitrosamine hypothesis predicts that cured meats, which contain nitrosamines and their precursors, increase risk. We did not observe an elevated risk with frequent maternal consumption of cured meats, in contrast to two previous studies4,5. The case group in one of those studies included only astrocytomas,5 and the case group in the other included all brain tumors (i.e., about 50 percent astrocytomas)4. Thus, the differences from our study may reflect different causes of astrocytoma and primitive neuroectodermal tumor.
Vitamin C is predicted to protect against the development of brain tumors. We observed significant protective dose-response relations for vitamin C and for fruits and vegetables, which are rich in vitamin C. Vitamin C supplements were also protective. The limited data from previous studies generally corroborate our results for vitamin C. Protective effects against brain tumors were observed for the consumption of fruit juice by children,6 the regular consumption of most fruits by adults,7 and the use of vitamin C supplements by adults7. Some component of fruits and vegetables other than vitamin C may be responsible for the protective effect, however, since vitamin C had little effect in multivariate analyses that included folate.
We found folate to be strongly protective against primitive neuroectodermal tumor in univariate and most multivariate analyses. Our food-frequency questionnaire was not designed to measure folate intake, however, and assessed only about 55 percent of intake26. It is possible that a study that assessed folate intake more completely would not observe an association. On the other hand, the incompleteness of our measure of folate intake will have resulted in nondifferential misclassification, which would most likely underestimate the effect. Clearly, folate intake requires further study.
The apparent protective effects of high levels of folate consumption and multivitamin use in the first six weeks of pregnancy are intriguing, because both have also been associated with a reduced risk of neural-tube defects27,28. Furthermore, periconceptional folate supplementation has been shown in clinical trials to protect against neural-tube defects29,30. The neural tube forms and closes in the first four weeks of gestation (six weeks after the last menstrual period) and is lined with neuroepithelial cells, which are the precursor cells of primitive neuroectodermal tumors31. One could thus imagine that insults early in pregnancy, such as low nutrient levels, might both interfere with neural-tube closure and increase the risk of primitive neuroectodermal tumor.
The protective effect of iron supplements is puzzling, since all prenatal multivitamin supplements contain iron32. Possible explanations include some effect of very large amounts of iron or an effect early in gestation.
In this detailed study of maternal diet and the risk of primitive neuroectodermal tumor before the age of six years, strong protective dose-response relations were observed for maternal consumption of fruit, vegetables, vitamin C, nitrate, and folate. Protective effects of early multivitamin use and iron supplements were also observed. The limitations of the dietary assessment, the correlations among dietary factors, and the possibility of bias make specific conclusions unwise. Our results nonetheless suggest that some aspects of maternal diet influence the risk of primitive neuroectodermal tumor in children. The observed protective effects of folate and multivitamin use early in pregnancy are the most intriguing, because of their specificity for primitive neuroectodermal tumor and the biologic plausibility of a causal similarity between primitive neuroectodermal tumor and neural-tube defects.
Supported in part by grants from the National Institutes of Health (CA 29275 and CA 13539); a Yamagiwa-Yoshida Memorial International Cancer Study Grant from the Japan National Committee of the International Union against Cancer and the Olympus Optical Company; and a research training fellowship from the International Agency for Research on Cancer, World Health Organization (to Dr. Kuijten).
We are indebted to our interviewers, Lorraine Bramble, Willie Mae Carroll, Jean Rodwell, Ann Murphy, Lucille D'Emilio, and Deborah Domino, for their diligent efforts, and to our other staff members, Patricia Jarrett, Xiaoyun Li, and Tanya Johnson; to Drs. Geoffrey Howe and Meera Jain for data on portion sizes and food content; to Drs. John Potter and Susan Preston-Martin for advice and guidance; and to the Children's Cancer Group E-12 and Epidemiology committees for their assistance.
Source Information
From the Division of Oncology, Children's Hospital of Philadelphia, and the University of Pennsylvania School of Medicine, both in Philadelphia (G.R.B., R.R.K., L.B.R., A.T.M.); and the Department of Preventive Medicine, University of Southern California School of Medicine, Los Angeles (J.D.B.). Presented in part at the 25th Meeting of the Society for Epidemiologic Research, Minneapolis, June 9-12, 1992.
Address reprint requests to Dr. Bunin at the Children's Cancer Group, P.O. Box 60012, Arcadia, CA 91066-6012.
References
The participating principal investigators of the Children's Cancer Group are as follows: D. Hammond, H. Sather, M. Krailo, J. Buckley, M. Bauer, D. Stram, J.W. Lee, R. Hutchinson, K. Matthay, P. Gaynon, R. Chard, S. Shurin, G. Reaman, E. Baum, J. Ortega, F. Ruymann, S. Piomelli, J. Mirro, J. Lukens, R. Neerhout, W. Woods, T. Williams, A. Meadows, P. Steinherz, P. Breitfeld, M. Greenberg, R. O'Brien, H. Cohen, C. Fryer, R. Wells, J. Finklestein, S. Feig, R. Tannous, L. Odom, G. Gilchrist, A. Pyesmany, H. Cooper, M. Donaldson, A. Freeman, P. Coccia, D. Norris, F.L. Johnson, W.A. Bleyer, and M. Willoughby.
| |||||||||||||||||||||||||||||||||||||||
Related Letters:
Maternal Diet and Primitive Neuroectodermal Brain Tumors in Children
Foreman N. K., Pearson A. D., Bunin G. R., Witman P. A., Meadows A. T.
Extract |
Full Text
N Engl J Med 1993;
329:1963, Dec 23, 1993.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
