Dietary and tobacco risk factors for adult onset glioma in the San Francisco Bay Area (California, USA) Cancer Causes and Control. Vol 8. 1997

Marion Lee, Margaret Wrensch, and Rei Miike

(Received 7 March 1996; accepted in revised form 8 May 1996)

The roles of diet and tobacco in the etiology of primary brain cancer are controversial. In this report, we comparedietary and cigarette smoking histories among 434 adults newly diagnosed with glioma in the San Francisco BayArea (California, USA) between 1991 and 1994 with frequency age, gender, and ethnicity-matched population-basedcontrols. Data were obtained on use of vitamin supplements and mean weekly consumption of each of 24 foodgroups. Adjusted for age, family income, and education, for both men and women, cases had higher mean weeklyconsumption of cured meats and other cured foods, lower consumption of high vitamin A and C fruits and vegetables,and higher average intakes of beer and other alcohol than controls. Men with brain cancer were twice as likely ascontrol men to report high consumption of cured foods and low consumption of foods rich in vitamin C (oddsratio [OR] = 2.0, 95 percent confidence interval [CI] = 1.2-3.5). This association was less pronounced and notstatistically significant in women (OR = 1.5, CI = 0.8-2.7). Similarly, men with brain cancer were twice as likely ascontrols to have high nitrite and low vitamin C consumption. Among men only, cases were 1.8 times more likelythan controls to report smoking unfiltered cigarettes (CI = 0.9-3.4). Moreover, among smokers, cases smoked un-filtered cigarettes almost twice as long as controls (P = 0.04) and cases’ average total pack-years also significantlyexceeded controls. Although these findings support the hypothesis that N-nitroso compounds might be a factor in adultglioma, they are compatible with other dietary hypotheses. In particular, these results also favor the hypothesis thattotal body burden of oxidants may play a role in brain cancer causation. Cancer Causes and Control 1997, 8, 13-24

Key words: Diet, glioma, tobacco, United States.

IntroductionThe role of diet in the etiology of primary brain canceris controversial.1,2 Although animal studies have indicatedthat dietary nitroso compounds are capable of inducingbrain tumors transplacentally, human studies have beenplagued by a variety of methodologic problems includingsmall sample sizes and heterogeneous tumor classifica-tions. Despite these and the usual dietary assessmentdifficulties, several studies3-8 have suggested that brain

cancer cases have had increased consumption of curedmeats and decreased consumption of vitamin C or vitaminsupplements compared with controls. Most studies havenot found significant associations of cigarette smokingand brain cancer,4,5,8-16 but one study7 showed a dose-response relationship of smoking unfiltered cigarettes andbrain cancer risk.

In this report, we compare dietary and cigarette

Cancer Causes and Control, 1997, 8, pp. 13-24

Authors are affiliated with the School of Medicine, Department of Epidemiology and Biostatistics, University of California, San Francisco,CA, USA. Address correspondence to Dr Wrensch, Box 0560, University of California, San Francisco, CA 94143-0560, USA. This workwas supported by grant RO1CA52689 from the US National Institutes of Health.

© 1997 Rapid Science Publishers Cancer Causes and Control. Vol 8. 1997 13

histories of 247 men and 187 women with glioma andfrequency age, gender and ethnicity-matched population-based controls in California (United States).

Materials and methods

Cases

All histologically confirmed incident cases of glioma(ICD-O-217 morphology codes 9380-9481) in adults aged20 years and older diagnosed in six San Francisco BayArea counties (Alameda, Contra Costa, Marin, SanMateo, San Francisco, and Santa Clara) between 1 August1991 and 31 March 1994 were eligible for inclusion. Caseswere ascertained within two to eight weeks of diagnosisusing the Northern California Cancer Center’s rapid case-ascertainment system which searched hospital pathology,radiotherapy, inpatient, and outpatient records in the sixBay Area counties to identify cases. Cases, or next-of-kinof deceased cases, were first sent a letter describing thestudy, then telephoned to arrange an in-person interview.

Controls

We obtained age (five-year age groups), gender, andrace/ethnicity (White, Black, Hispanic, Asian, and other)frequency-matched controls through random-digit dial-ing using methods described by Waksberg18 and refinedby Harlow and Davis.19 Telephone area code, three-digitprefix, and the next two digits of cases’ telephone numberswere the initial sampling units. For each sampling unit,two-digit suffixes were generated from random numbertables and called until the necessary eligible matches werefound. We sent a letter to eligible controls, thentelephoned to arrange an in-person interview.

Interview, dietary questionnaire, and analyses

Structured interviews were conducted with consentingcases (or their proxies) and controls in their homes oranother location of their choosing. Prior to the interview,subjects were sent a packet of materials including a self-administered dietary questionnaire. The questionnairewas collected or completed at the interview.

The 79-item food-frequency questionnaire, involvedrecollection of usual food consumption habits during theyear before diagnosis for cases and for the previous yearfor controls. It was modified from US National CancerInstitute’s (Block’s) Health Habits and History Question-naire20 and from the Los Angeles glioma and meningiomastudy.3 The questionnaire emphasized foods containingnitrite and nitrate and antioxidant foods. The food listwas divided into the following categories: fruits and juices;vegetables; breakfast foods; lunch items; meat, fish, poultry,and mixed dishes; breads/salty snacks/spreads; sweets;dairy products; and beverages. The choices of frequency

of consumption of each of the 79 food items were ‘never,’‘less than 1 per month,’ ‘1 per month,’ ‘2-3 per month,’‘1 per week,’ ‘2 per week,’ ‘3-4 per week,’ ‘5-6 per week,’‘1 per day,’ and ‘2+ per day.’ In addition, questions wereasked about how often the respondent added lemon juiceto fish, ate fruit juices or tea with cured meat or bacon,and drank beer with meals. For each choice of frequency,only one value was assigned by converting the unit ofyear, month, or day to number of times per week. If rangeof times was given, then a mid-value was assigned. Bythis conversion, each individual’s average weekly fre-quency of consumption of each food item was calculated.

The questionnaire also asked about use of vitaminsupplements (vitamins A, C, E, D, and multiple vitamins),height, weight, and weight five years previously. Respon-dents were asked how many times a day, week, month.or year they usually took each type of vitamin or whetherthey never took them in the previous year for controls,and the year before diagnosis for cases.

The food items were combined into 24 food groups(Table 1). Mean weekly consumption of each food groupfor cases and controls was computed separately for menand women; values were compared within gender usinggeneralized linear models21 to adjust for age, familyincome (less than cf greater than or equal to $70,000 peryear), and education (less than cf greater than or equal to16 years). Comparisons were made for all cases cf controlsand self-reporting cases compared with controls.

In addition to mean comparisons, the reported gender-specific frequency intakes of each food group were rankedby quartiles using the distribution from the controls. Age,income, and education-adjusted odds ratios (OR) by quar-tiles of consumption between all and self-reporting casescompared with controls were then calculated using logis-tic regression22 separately for men and women. Likelihoodratio chi-square tests were used for linear trend analysis.

Intakes of N-nitroso compounds (NOC) and theirprecursors were estimated from the food questionnaire.The database for food contents of nitrites, nitrates, andvitamin C were gathered from various references; 23-26

major food sources include types of processed meats,cured fish, and vegetables and fruits. The level of vitaminC in the diet is an important factor in affecting nitriteconcentration in the body since vitamin C can react withand inhibit nitrite formation of nitrosamines. Therefore,the intake of vitamin C also was calculated. A sum scoreof nitrate, nitrite, and vitamin C intake for each subjectthen was computed assuming average portion sizes of thefood items.27 T-tests and generalized linear models toadjust for age, income, and education were used to testfor differences between cases and controls for averagenitrate, nitrite, and vitamin C intakes.

We also grouped our subjects into four categoriesaccording to their combined intakes of cured foods and

M. Lee et al

14 Cancer Causes and Control. Vol 8. 1997

Table 1 . List of food groups used in analyses and correspondingitems asked in food frequency questionnaire, San FranciscoBay Area Adult Glioma Study, 1991-95

Food groups Food items

Bacon Bacon

Cured meats Sausage

Hot dogs

Ham, lunch meats

Other cured meat

Red meat Hamburgers, cheeseburgers, meat loaf

Beef – steaks, roasts (raw or rare)

Beef stew or pot pie with carrots, other vegetables

Pork, including chops, roasts

Poultry Fried chicken

Chicken or turkey, roasted, stewed or broiled

Fish Fried fish or fish sandwich

Other fish broiled or baked

Cured fish Smoked fish

Pickled fish

Salted fish

Cured foods Cured meats

Cured fish

Bacon

Eggs Eggs

Dairy Cheeses and cheese spreads

Whole milk and beverages with whole milk

2% milk and beverages with 2% milk

Skim milk, 1% milk, or buttermilk

Vegetables/fruits high invitamin A

Cantaloupe (in season)

Carrots, or mixed vegetables with carrots

Sweet potatoes, yams

Broccoli

Spinach (raw and cooked)

Mustard greens, turnip greens, collards

Vegetables/fruits high invitamin C

Cantaloupe (in season)

Other melons

Oranges

Orange juice or grapefruit juice

Grapefruit

Other fruit juices, fortified fruit drinks

Green salad (e.g., lettuce, celery, bell pepper)

Cole slaw, cabbage, sauerkraut

Tomatoes, tomato juice

Broccoli

Beans andlegumes

String beans, green beans

Other beans – baked beans/pintos/kidney/limas

Other fruit Apples, applesauce, pears

Any other fruit, e.g., berries, fruit cocktail

Table 1, continued .

Food groups Food items

Othervegetables

French fries and fried potatoes

Other potatoes – boiled/baked/potatosalad, etc.

Any other vegetable – cooked onions,summer squash, etc.

Rice, breads,and cereals

Rice

White bread (sandwiches), bagels,crackers, etc.

Highly fortified cereals – Product19/Total/Most

Other cold cereals – corn flakes, RiceKrispies

Cooked cereals

Corn bread, corn muffins, corn tortillas

Salty snacks (chips, popcorn, etc.)

Dark breadsand high fibercereals

Dark bread – including wholewheat/rye/pumpernickel

High fiber bran/granola cereals, shreddedwheat

Nuts Peanuts, peanut butter

Nuts

Spreads anddressings

Butter on bread or rolls

Margarine on bread or rolls

Salad dressing, mayonnaise – onsandwiches, etc.

Sweets Ice cream

Doughnuts, cookies, cakes, pastry

Pumpkin pie, sweet potato pie

Other pies

Chocolate candy

Sugar added to cereal

Soup Vegetable soup, beef, minestrone, tomato soup

Other soups

Beer Beer

Alcohol Wine

Liquor

Other drinks Regular soft drinks (not diet)

Diet soft drinks

Decaffeinated coffee

Coffee, not decaffeinated

Tea (hot or iced)

Lemon in tea

Milk or cream in coffee or tea

Sugar in coffee or tea

Artificial sugar Artificial sweetener in coffee or tea

Artificial sweetener on other foods

Diet, tobacco, and adult brain cancer

Cancer Causes and Control. Vol 8. 1997 15

vegetable and fruits rich in vitamin C, using the mediansas the cutoff. ORs were computed comparing the otherthree categories to the assumed ‘low risk’ category; lowconsumption of cured foods combined with high con-sumption of vitamin C rich fruits and vegetables. The‘high risk’ category was hypothesized to be characterizedby high consumption of cured foods combined with lowconsumption of vitamin C rich fruits and vegetables. Thishypothesis makes no assumptions about the additivity ofeffects of consumption of cured foods and vitamin C richfruits and vegetables. Similarly, we categorized subjectsinto four categories of combined intakes of nitrite andvitamin C, using the medians as cutoff.

The in-person interview asked whether or not the sub-ject had ever smoked cigarettes. If so, the numbers ofyears the subject smoked less than one-half pack a day,one-half to less than one pack, one pack to two packs,and more than two packs a day were asked separately forfiltered and unfiltered cigarettes. Pack-years were calcu-lated for each type of cigarette by assigning the mid-pointto each quantity and multiplying by the numbers of years.Total pack-years was taken as the sum of filtered andunfiltered pack-years.

To assess potential sources of participation bias amongcontrols, part-way through the study, a five-minutetelephone interview was conducted with controls whodeclined full participation. Questions pertinent to thesecomparisons included whether or not the subject hadtaken any vitamin supplements during the previous year,whether they had ever smoked cigarettes, and their heightand weight.

Results

Subjects

The Northern California Cancer Center identified 604eligible cases. We completed interviews for 494 cases (82percent); 12 percent declined to participate and physiciansrefused contact with two percent; we were unable to locateor contact 25 cases (four percent) or a suitable proxy. Onaverage, we interviewed self-reporting cases within fourmonths and proxies within eight months of case’s diag-nosis. Proxy interviews were necessary for 46 percent ofthe cases due to cases’ death or disability. Of 229 proxyrespondents, 46 percent were spouses, 35 percent werechildren, five percent were siblings, four percent wereparents, and 10 percent were more distant relatives orfriends of cases.

Seven hundred and fifty-four apparently eligible con-trols were obtained through random-digit dialing 6,612numbers. Of 5,858 numbers that did not yield eligiblecontrols, 49 percent were either businesses (1,024), fax ormodems (347), or not working (1,495); nine percent (547)

gave no response after 10 tries; 26 percent (1,502) wereenumerated but had no eligible household member; fivepercent (283) refused information; 0.6 percent (37) wereeligible but too ill; four percent (231) were not Englishspeaking; and seven percent (329) were eligible but thequota for their age/gender/ethnic group had already beenfilled. Of the 754 apparently eligible controls, contactindicated that two were related to cases, 11 lived out ofthe area, and nine did not have sufficient English forinterview. Interviews were completed with 63 percent(462/732) of eligibles, 32 percent (236) declined toparticipate, and five percent could not be reached forinterview.

Usable dietary data were obtained for 94 percent(434/462) of cases and 99 percent (439/443) of controls.Usable tobacco-smoking data including cigarette typeswere obtained for 434 cases and 430 controls.

A more complete description of cases and controls ispresented elsewhere.28 Briefly, Whites comprised 88 per-cent of subjects. Average ages were 54.2 years for controls,53.9 years for all cases, and 46.0 years for self-reportingcases. Among controls, 55 percent were men, while 57percent of all cases and 59 percent of self-reporting caseswere men. Controls were more likely to have completed16 or more years of education than cases (46 percent cf39 percent, respectively) and were somewhat less likelyto report household incomes greater than or equal to$70,000 per year (30 percent cf 33 percent respectively).Among self-reporting cases, 48 percent completed highereducation and 42 percent had high incomes. Furthercomparisons in this report are adjusted for age, education,and family income to control for these differences.

Diet, vitamin supplements, and tobacco

Comparisons of mean weekly consumption of foodgroups between cases and controls are presented in Table2. For both men and women, cases had higher consump-tion of cured meats and other cured foods, lowerconsumption of high vitamin A and C fruits and vegeta-bles, and higher intakes of beer and other alcohol thancontrols. Regardless of proxy status, male cases reportedsignificantly less average consumption of poultry andfiber, and female cases reported significantly less con-sumption of dairy products than controls. Otherwise,average differences were not consistently significant forall cases or self-reporting cases cf controls.

Examination of ORs of quartiles of weekly consump-tion yielded similar findings for foods and non-alcoholicbeverages (results not shown). Distributions of beer andalcohol consumption were skewed in both men andwomen; the mean levels of beer and alcohol consumptionwere higher for cases than controls, but median levels ofconsumption were lower for cases compared withcontrols. No significant trends were found among men

M. Lee et al

16 Cancer Causes and Control. Vol 8. 1997

Diet, tobacco, and adult brain cancer

Cancer Causes and Control. Vol 8. 1997 17

or women of increased or decreased consumption of beeror other alcohol for cases cf controls.

No meaningful or significant findings emerged fromanalyses of food combinations, such as how often subjectsate fish with lemon, juices with bacon, or beer with meals.

Regardless of proxy status, men with brain cancer weretwice as likely as control men to report high consumptionof cured foods and low consumption of foods rich invitamin C (Table 3; tests of interaction between consump-tion of cured foods or foods rich in vitamin C were notsignificant). This association was less pronounced and notsignificant in women.

Average weekly nitrate, nitrite, and vitamin Cconsumption are shown in Table 4. On average, casesconsumed less nitrate and vitamin C and more nitrite thancontrols; the results are significant only for nitrate con-sumption among women. Combinations of nitrite andvitamin C show similar findings to combinations of curedfoods and vitamin C rich fruits and vegetables (Table 5;tests of interaction between consumption of nitrite and

vitamin C were not significant).Heights of cases and controls were nearly identical

(64.0 inches ± 0.2 for both female cases and controls, and70.2 ± 0.2 for male controls and 70.1 ± 0.2 for male cases).Among women, cases were somewhat, but not signifi-cantly, lighter than controls (141.6 lbs ± 2.4 cf 147.8 lbs± 2.5); however, five years prior to diagnosis or interview,cases were slightly heavier, on average, than controls(142.2 lbs ± 2.4 cf 140.8 ± 2.2). Among men, cases’ currentweights were only slightly less than controls (180.2 ± 2.2cf 182.7 ± 1.8 lbs), while cases were slightly heavier thancontrols five years previously (179.5 ± 2.0 cf 178.4 ± 2.2).

Cases were less likely than controls to report usingvitamin supplements including vitamins C and E, andmultiple vitamins (Table 6). The association was morepronounced in women than in men, but none of thefindings was significant for self-reporting or all cases foreither gender.

Controls who participated in the study were less likelythan controls who did not participate to report taking

Table 3 . Case-control comparisons of combinations of cured foods and vitamin C rich fruits and vegetables, San FranciscoBay Area Adult Glioma Study, 1991-95

Controls All cases Self-reported cases

No. No. OR a

(Cl)bNo. ORa

(Cl)b

Men

Lowc consumption of cured foods andhighc consumption of vegetables andfruits rich in vitamin C

61 44 1.0—

25 1.0—

Low consumption of cured foods andlow consumption of vegetables andfruits rich in vitamin C

61 56 1.3(0.7-2.2)

32 1.1(0.6-2.2)

High consumption of cured foods andhigh consumption of vegetables andfruits rich in vitamin C

57 59 1.5(0.8-2.6)

27 1.0(0.5-2.0)

High consumption of cured foods andlow consumption of vegetables andfruits rich in vitamin C

60 85 2.0(1.2-3.5)

58 2.0(1.1-3.8)

Women

Low consumption of cured foods andhigh consumption of vegetables andfruits rich in vitamin C

51 35 1.0—

18 1.0—

Low consumption of cured foods andlow consumption of vegetables andfruits rich in vitamin C

52 41 1.2(0.6-2.2)

21 1.2(0.6-2.6)

High consumption of cured foods andhigh consumption of vegetables andfruits rich in vitamin C

47 58 1.7(0.9-3.1)

32 1.6(0.8-3.4)

High consumption of cured foods andlow consumption of vegetables andfruits rich in vitamin C

49 50 1.5(0.8-2.7)

24 1.1(0.5-2.4)

a OR = Odds ratio adjusted for age, education and income. 16 controls and 26 cases did not have income information soare not included in the adjusted OR.

b CI = 95% confidence interval.c High and low indicate above median and below median, respectively.

M. Lee et al

18 Cancer Causes and Control. Vol 8. 1997

any vitamins in the previous year. To adjust for thisparticipation bias, we computed an expected prevalenceof vitamin use for controls by weighting the observedprevalences of use in participants and nonparticipants bythe participation rate. For females controls, the adjustedprevalence was 0.64 (0.63 participation rate × 0.70 preva-lence of vitamin use among participants) + (0.37nonparticipation rate × 0.55 prevalence of vitamin useamong nonparticipants). For male controls, the adjustedprevalence was 0.57 (0.63 × 0.64 + 0.37 × 0.46). Partici-pation did not vary much by gender, thus the sameparticipation rates were used for both calculations.

Using this adjusted data, cases were still less likely than

controls to report vitamin use, but the association wasattenuated towards the null. Comparing any reportedvitamin use among cases cf controls, the Mantel-HaenszelOR (to adjust for gender) using participant data only was0.66 (CI = 0.5-0.9, P = 0.004). Using the estimated preva-lence for controls incorporating nonparticipant data, theOR was 0.86 (CI = 0.6-1.2, P = 0.33). Although we donot have other dietary data for nonparticipants, averageweights of participating and non-participating controlsdid not differ very much (166 lbs ± 1.7 cf 170 lbs ± 3.7,respectively).

Among men only, cases were 1.8 times more likely thancontrols to report smoking unfiltered cigarettes (Table 7).

Table 4 . Average weekly nitrate, nitrite, and vitamin C levels consumption by cases and controls, San Francisco Bay Area AdultGlioma Study, 1991-95

A. Men Controls All cases Self-reported cases

(n = 235) Mean ± SEa

(n = 240) Mean ± SEa

(n = 140) Mean ± SEa

Nitrates 775.8 ± 26.3 726.3 ± 26.8 681.9 ± 34.7

Nitrites 6.0 ± 0.2 6.5 ± 0.3 6.4 ± 0.4

Vitamin C 943.4 ± 34.6 880.1 ± 39.6 864.7 ± 57.3

B. Women Controls All cases Self-reported cases

(n = 197) Mean ± SEa

(n = 181) Mean ± SEa

(n = 93) Mean ± SEa

Nitrates 660.9 ± 41.2 552.1 ± 22.8* 571.5 ± 29.9

Nitrites 4.6 ± 0.2 4.9 ± 0.2 4.9 ± 0.3

Vitamin C 885.6 ± 50.7 800.0 ± 38.1 769.6 ± 42.9

a SE = standard error.* P ≤ 0.05; adjusted for age, education, and income.

Table 5 . Case-control comparisons of combinations of nitrites and vitamin C consumption, San Francisco Bay Area Adult GliomaStudy, 1991-95

Controls All cases Self-reported cases

No. No. OR a (Cl)b No. ORa (Cl)b

MenLowc nitrite + highc vitamin C 45 36 1.0 — 19 1.0 —Low nitrite + low vitamin C 73 75 1.3 (0.7-2.3) 51 1.5 (0.7-3.0)High nitrite + high vitamin C 72 58 1.0 (0.6-1.8) 29 0.9 (0.4-1.9)High nitrite + low vitamin C 45 71 2.1 (1.1-3.8) 41 2.1 (1.0-4.3)

WomenLow nitrite + high vitamin C 35 24 1.0 — 13 1.0 —Low nitrite + low vitamin C 64 52 1.2 (0.6-2.4) 26 1.2 (0.5-2.8)High nitrite + high vitamin C 63 69 1.6 (0.9-3.0) 36 1.6 (0.7-3.7)High nitrite + low vitamin C 35 36 1.5 (0.7-3.1) 18 1.2 (0.5-3.0)

a OR = Odds ratio adjusted for age, education, and income. 16 controls and 26 cases did not have income information soare not included in the logistic analyses.

b CI = 95% confidence interval.c High and low indicate above median and below median, respectively.

Diet, tobacco, and adult brain cancer

Cancer Causes and Control. Vol 8. 1997 19

Table 6 . Odds ratios (OR) for use of vitamin supplements among cases and controls, San Francisco Bay Area Adult GliomaStudy, 1991-95

A. Men Controls ( n = 239) All cases ( n = 243) Self-reported cases ( n = 139)

% % ORa (Cl)b % ORa (Cl)b

Percent consuming any within previous yearc

Vitamin E 21.3% 17.3% 0.9 (0.5-1.4) 16.6% 0.8 (0.5-1.5)Vitamin C 41.4% 35.8% 0.9 (0.6-1.3) 38.1% 0.8 (0.5-1.3)Multiple vitamin 43.5% 34.2% 0.7 (0.5-1.0) 35.2% 0.7 (0.4-1.1)Other vitamin 22.2% 18.9% 0.9 (0.6-1.4) 18.7% 0.9 (0.5-1.6)

B. Women Controls ( n = 199) All cases ( n = 176) Self-reported cases ( n = 94)

% % ORa (Cl)b % ORa (Cl)b

Percent consuming any within previous yearc

Vitamin E 25.6% 19.9% 0.7 (0.4-1.1) 13.8% 0.8 (0.5-1.5)Vitamin C 44.2% 32.8% 0.6 (0.4-0.97) 28.7% 0.8 (0.5-1.3)Multiple vitamin 50.8% 42.6% 0.7 (0.5-1.1) 36.2% 0.7 (0.4-1.1)Other vitamin 32.5% 21.0% 0.6 (0.3-0.9) 18.1% 0.9 (0.5-1.6)

a OR = Odds ratios compare any cf no use of vitamins adjusted for age, education and income.b CI = 95% confidence interval.c Questions are asked for year previous to interview for controls and year before diagnosis for cases.

Table 7 . Tobacco use histories among cases and controls, San Francisco Bay Area Adult Glioma Study, 1991-95

A. Men Controls ( n = 229) All cases ( n = 237) Self-reported cases ( n = 139)

% % ORa (Cl)b % ORa (Cl)b

Cigarette smokingNever 36.2 38.4 1.0 — 41.7 1.0 —Filtered 33.2 27.9 0.8 (0.5-1.2) 32.4 0.9 (0.5-1.5)Unfiltered 10.5 17.3 1.8 (0.9-3.4) 11.5 1.7 (0.7-3.7)Both 20.1 16.5 0.9 (0.5-1.5) 14.4 0.9 (0.5-1.9)

Pack-yearscMean (SE)d Mean (SE)d Mean (SE)d

Filtered cigarettes 15.6 (1.5) 17.2 (1.9) 17.4 (2.5)Unfiltered cigarettes 6.6 (1.2) 11.3 (1.7)* 7.6 (1.9)Total 22.2 (1.9) 28.8 (2.2)* 25.0 (2.6)

B. Women Controls ( n = 201) All cases ( n = 197) Self-reported cases ( n = 102)

% % ORa (Cl)b % ORa (Cl)b

Cigarette smokingNever 52.7 51.3 1.0 — 54.9 1.0 —Filtered 31.8 33.0 1.1 (0.7-1.7) 34.3 0.9 (0.5-1.6)Unfiltered 5.0 4.1 1.0 (0.4-2.7) 2.9 1.1 (0.3-4.3)Both 10.5 11.7 1.3 (0.6-2.6) 7.9 1.3 (0.5-3.4)

Pack-yearscMean (SE)d Mean (SE)d Mean (SE)d

Filtered cigarettes 17.2 (1.9) 17.4 (2.3) 12.9 (3.2)Unfiltered cigarettes 4.1 (1.3) 4.3 (1.0) 1.3 (0.6)Total 21.2 (2.2) 21.7 (2.5) 14.2 (3.5)

a OR = Odds ratios compare each category of smoking with the never-smoked category adjusted for age, education, and income.b CI = 95% confidence interval.c Pack-years are computed for smokers only; means compared with general linearized model controlling for age, education,

and income.d SE = standard error.* P ≤ 0.05.

M. Lee et al

20 Cancer Causes and Control. Vol 8. 1997

Diet, tobacco, and adult brain cancer

Cancer Causes and Control. Vol 8. 1997 21

Moreover, among smokers, cases smoked unfiltered ciga-rettes almost twice as long as controls (P = 0.04) and cases’average total pack-years also significantly exceeded con-trols. Participating controls were slightly more likely thannonparticipants to report having ever smoked (57 percentcf 53 percent).

Discussion

Attempts to associate N-nitroso compounds with adultglioma derive mainly from the observation that intra-venous administration of nitroso compounds, such asN-methyl-N-nitroso urea and N-ethyl-N-nitroso ureainduce glioma in experimental animals.29 The furtherobservation that ethyl urea and nitrite fed to pregnantrats caused sufficient formation of ethyl nitroso urea tocause brain tumors in all their offspring prompted Preston-Martin et al 30 to propose that prenatal and early exposuresmight be related to childhood brain tumors in humans.

Seven previous studies addressing the role of diet inadult glioma3-8,15 are summarized in Table 8. Examinationof this table indicates some of the difficulties in comparingresults of these studies including differences in foodsasked or reported, choices of case and control groups,and rates of participation and proxy-reporting. Despitethese difficulties, the general impression from thesestudies is that cases were more likely than controls toconsume cured foods (especially cured meats) and wereless likely than controls to consume vegetables or fruits.

Perhaps one of the most serious problems in case-control studies of glioma is differential data quality dueto proxy reporting. Individual matching by proxy respon-dent was impractical in this population-based study.Frequency matching of controls considerably facilitatedboth control recruitment and statistical analyses. Also,proxy data from controls would not necessarily be com-parable to proxy data from cases; a devastating illnesscould influence a person’s relationships and the informa-tion a proxy might have about him or her. To compensatefor the likely differential data quality from proxies, ourapproach was to emphasize associations that were con-sistent among all cases and controls and self-reportingcases and controls. Ahlbohm et al, 5 Boeing et al,4 andGiles et al 6 also reported no substantial differences intheir results when proxy data were deleted.

Selection bias related to dietary habits and controls isanother serious concern in this and other studies3,4,6-8

where control participation rates ranged from 45 percentto 72 percent. We were able to assess the effect of thisbias for one dietary variable, use of vitamin supplementsthrough the use of an abbreviated questionnaire forrefusing controls. The difference between participatingand nonparticipating controls was large. Nevertheless, theadjusted OR for any use of vitamins in cases compared

with controls, although attenuated, remained less than1.0. If there was no similar bias in case participation, ourcalculations are conservative; however, if participatingcases were more likely than nonparticipating cases to usevitamins, the true OR would be smaller (farther from thenull) than that estimated adjusting only for controlselection bias. Further, the method we used might under-estimate vitamin use among nonparticipating controls,because they were only asked ‘any’ cf ‘no’ use, whereasparticipants were asked (even infrequent) use of severalparticular types of vitamins. Future studies should try toestimate bias associated with other dietary componentsof special interest such as vegetables, fruits, and curedmeats.

Misclassification of dietary data, in this and other case-control studies, also occurs to the degree that recent dietdiffers from diet during an etiologically relevant period.Recent diet was assessed both to improve recall accuracyand because the etiologically relevant period for gliomais unknown.

Computed measures of dietary nutrients are question-able for two reasons. Without data on portion size,nutrient calculations assume everyone consumes averageportion sizes. This creates nondifferential misclassifica-tion and reduces the ability to detect differences. Second,the food nitrate and nitrite tables used are very generaland may be outdated. We do not know to what degreethese problems affected other studies that computednutrients. As Preston-Martin and Mack3 concluded intheir study, ‘we are far from being able to quantify anindividual’s total exposure to N-nitroso compounds inepidemiologic interview studies.’ Given the reality of thisconclusion not only for N-nitroso compounds, but alsofor most dietary nutrients, we chose a simple method ofcategorizing high cf low risk diets, as described in themethods above.

For both men and women, cases were more likely thancontrols to be categorized as having a high risk diet (highconsumption of cured foods and low consumption offruits and vegetables rich in vitamin C), although theassociation was stronger and statistically significant onlyin men. Although this finding supports the hypothesisthat N-nitroso compounds might be a factor in adultglioma, it also is compatible with other dietary hypothe-ses. Ames et al 31 suggested that while dietary mutagensmight increase cancer rates directly, the total body burdenof oxidants also is likely to be important in cancer cau-sation. Given the general protective association observedfor diets high in fruits and vegetables with many cancersstudied,32 this association also might pertain to gliomas.

Cigarette smoking has not been linked convincingly torisk of adult glioma. In a previous review,1 we noted thatof 10 studies of adult brain cancer and smoking, fourreported nonsignificant relative risks below 1.0;10,11,14,15

M. Lee et al

22 Cancer Causes and Control. Vol 8. 1997

two9,12 reported an RR of 1.0; one study5 found a RR lessthan 1.0 comparing cases with clinical controls, and some-what greater than 1.0 for population controls 1.0; andthree7,13,16 reported nonsignificant RRs greater than 1.0.Only one study7 investigated filtered and unfiltered ciga-rettes separately and found a significant dose-response ofincreased risk for unfiltered cigarettes only. A recent studyof glioma among women in Los Angeles county,8 reportedno association with any smoking, but a nonsignificantelevated RR for smoking unfiltered cigarettes. Thus, ourfindings with regard to smoking are consistent withprevious studies and suggest that future studies shouldemphasize brands and quantities smoked of unfilteredcigarettes.

With regard to assessing the possible influence of dietand cigarettes on brain cancer, it will be important toincorporate genotypes of subjects with regard to carcino-gen-metabolizing genes and other genes that affect thebody burden of oxidants. For example, no consistenteffect of cigarette smoking with breast cancer has beenreported, but a recent study33 found that when subjectswere stratified according to genotype for N-acetyltrans-ferase 2, smoking was associated with breast cancer risk,but only among those who were slow acetylators. Inaddition to categorizing people according to likelysusceptibility, perhaps some progress can be made in dis-covering the role of diet and cigarettes in adult braincancer by improving both epidemiologic methodologyto estimate the role of control participation bias anddietary assessment and classification methods to considerthe overall carcinogenic potential of a subject’s dietaryand supplement intake.

References1. Wrensch M, Bondy ML, Wiencke J, et al. Environmental

risk factors for primary malignant brain tumors: A review.J Neurooncol 1993; 17: 47-64.

2. Berleur MP, Cordier S. The role of chemical, physical, orviral exposures and health factors in neurocarcinogenesis:Implications for epidemiologic studies of brain tumors.Cancer Causes Control 1995; 6: 240-56.

3. Preston-Martin S, Mack W. Gliomas and meningiomas inmen in Los Angeles County: Investigation of exposures toN-nitroso compounds. In: O’Neill IK, Chen J, Bartsch H,eds. Relevance to Human Cancer of N-Nitroso Compounds,Tobacco Smoke and Mycotoxins. Lyon, France: Interna-tional Agency for Research on Cancer, 1991; IARC Sci.Pub. No. 150: 197-203.

4. Boeing H, Schlehofer B, Blettner M, et al. Dietary carcino-gens and the risk for glioma and meningioma in Germany.Int J Cancer 1993; 53: 561-5.

5. Ahlbom A, Navier IL, Norell S, et al. Nonoccupational riskindicators for astrocytomas in adults. Am J Epidemiol 1986;124: 334-7.

6. Giles GG, McNeil JJ, Donnan G, et al. Dietary factors andthe risk of glioma in adults: Results of a case-control study

in Melbourne, Australia. Int J Cancer 1994; 59: 357-62.7. Burch JD, Craib KJ, Choi BCK, et al. An exploratory

case-control study of brain tumors in adults. JNCI 1987;87: 601-9.

8. Blowers L, Preston-Martin S, Mack W. Dietary and otherlifestyle factors of women with brain gliomas in LosAngeles County (California, USA). Cancer Causes Control1996; 8: 5-12.

9. Brownson RC, Reif JS, Chang JC, et al. An analysis ofoccupational risks for brain cancer. Am J Pub Health 1990;80: 169-72.

10. Choi NW, Schuman LM, Gullen WH, et al. Epidemiologyof primary central nervous system neoplasms. II. Case-control study. Am J Epidemiol 1970; 91: 467-85.

11. Preston-Martin S, Mack W, Henderson BE. Risk factorsfor gliomas and meningiomas in males in Los AngelesCounty. Cancer Res 1989; 49: 6137-43.

12. Hochberg F, Toniolo P, Cole P, et al. Nonoccupational riskindicators of glioblastoma in adults [published erratumappears in J Neurooncol 1990; 9: 90] J Neurooncol 1990; 8:55-60.

13. Ryan P, Lee MW, North B, et al. Risk factors for tumorsof the brain and meninges: Results from the Adelaide AdultBrain Tumor Study. Int J Cancer 1992; 51: 20-7.

14. Schlehofer B, Kunze S, Sachsenheimer W, et al. Occupa-tional risk factors for brain tumors: Results from apopulation-based case-control study in Germany. CancerCauses Control 1990; 1: 209-15.

15. Mills PK, Preston-Martin S, Annegers JF, et al. Risk factorsfor tumors of the brain and cranial meninges in Seventh-Day Adventists. Neuroepidemiology 1989; 8: 266-75.

16. Musicco M, Filippini G, Bordo BM, et al. Gliomas andoccupational exposure to carcinogens: Case-control study.Am J Epidemiol 1982; 116: 782-90.

17. Percy C, Van Holten V, Muir C, eds. International Clas-sification of Disease for Oncology, Second Edition. Geneva,Switzerland: World Health Organization, 1990.

18. Waksberg J. Sampling methods for random digit dialing. JAm Stat Assoc 1978; 73: 40-6.

19. Harlow BL, Davis S. Two one-step methods for householdscreening and interviewing using random digit dialing. AmJ Epidemiol 1988; 127: 857-63.

20. Block G, Hartman AM, Dresser CM, et al. A data-basedapproach to diet questionnaire design and testing. Am JEpidemiol 1986; 124: 453-69.

21. SAS Institute Inc. SAS/STAT User’s Guide, Version 6,Fourth Edition. Vol 1. Cary, NC (USA): SAS Institute Inc.,1990.

22. Breslow NE, Day NE. Statistical Methods in Cancer Re-search. Vol 1. The Analysis of Case-control Studies. Lyon,France: International Agency for Research on Cancer, 1980;IARC Sci. Pub. No. 32.

23. Walker R. Nitrates, nitrites and N-nitrosocompounds: Areview of the occurrence in food and diet and the toxico-logical implications. Food Addit Contam 1990; 7: 717-68.

24. Randles LL. Nitrate, nitrite, and nitrosamine values fordatabase. (Unpublished data.) California Birth DefectsMonitoring Program, January 1989.

25. Howe GR, Harrison L, Jain M. A short diet history forassessing dietary exposure to N-nitrosamines inepidemiologic studies. Am J Epidemiol 1986; 124: 595-602.

26. United States Department of Agriculture. Human Nutri-tion Information Service. Composition of Foods; Raw,Processed, Prepared. (Revised Handbooks #8-1 to #8-21,

Diet, tobacco, and adult brain cancer

Cancer Causes and Control. Vol 8. 1997 23

1988). Washington, DC: U.S. GPO, Nutrition MonitoringDivision, 1976-88.

27. National Cancer Institute. Health Habits and HistoryQuestionnaire: Diet History and Other Risk Factors. DietsysVersion 3.4. Information Management Services, Inc., BlockDietary Data Systems, 1995.

28. Wrensch M, Lee MM, Miike R, et al. Family and personalmedical history of cancer and nervous system conditionsamong adult brain cancer cases and controls. Am JEpidemiol, (in press).

29. Maekawa A, Mitsumori K. Spontaneous occurrence andchemical induction of neurogenic tumors in rats-influenceof host factors and specificity of chemical structure. CritRev Toxicol 1990; 20: 287-310.

30. Preston-Martin S, Yu MC, Benton B, et al. N-nitroso com-pounds and childhood brain tumors: A case-control study.Cancer Res 1982; 42: 5240-5.

31. Ames BN, Shigenaga MK, Hagen T. Oxidants, antioxidants,and the degenerative diseases of aging. Proc Natl Acad Sci1993; 90: 7915-22.

32. Block G, Patterson B, Subar A. Fruit, vegetables, and cancerprevention: A review of the epidemiologic evidence. NutrCancer 1992; 18: 1-29.

33. Ambrosone CB, Freudenheim JL, Marshall JR, et al. Theassociation of polymorphic N-acetyltransferase (NAT2)with breast cancer risk. Ann NY Acad Sci 1995; 768: 250-2.

M. Lee et al

24 Cancer Causes and Control. Vol 8. 1997