Lung Cancer (2008) 61, 275—282

avai lab le at www.sc iencedi rec t .com

journa l homepage: www.e lsev ier .com/ locate / lungcan

Dietary factors, food contamination and lungcancer risk in Xuanwei, China

Min Shena,∗, Robert S. Chapmanb, Xingzhou Hec, Larry Z. Liud,Hong Laie, Wei Chenf, Qing Lana

a Division of Cancer Epidemiology and Genetics, NCI, NIH, DHHS, Bethesda, MD, United Statesb College of Public Health, Chulalongkorn University, Bangkok, Thailandc Institute of Environmental Health and Engineering, Chinese Center for Disease Control and Prevention,Beijing, Chinad Department of Public Health, Weill Medical College of Cornell University, New York, NY, United Statese The Wilmer Eye Institute at Johns Hopkins University, Baltimore, MD, United Statesf Forest Laboratories, Inc., Harborside Financial Center, Jersey City, NJ, United States

Received 14 October 2007; received in revised form 17 December 2007; accepted 30 December 2007

KEYWORDSRisk factor;Lung cancer;Indoor air pollution;Case—control study;Food contamination;Polycyclic aromatichydrocarbon

SummaryBackground: In rural Xuanwei County, China, the high incidence of lung cancer is attributablelargely to burning smoky coal indoors for heating and cooking without adequate ventilation. Suchburning generates very high levels of indoor air pollutants, including carcinogenic polycyclicaromatic hydrocarbons, which could contaminate foodstuffs in the home. Thus, residents couldbe exposed to carcinogenic coal emissions not only via inhalation but also via ingestion of thesefoodstuffs.Methods: A population-based case—control study of 498 lung cancer patients and 498 controlswas conducted from 1985 through 1990 in Xuanwei. The interviewer-administered study ques-tionnaire queried the frequency of food items commonly consumed in this region. Overall andsex-specific multiple logistic regression models were constructed to estimate Odds ratios (OR)and 95% confidence intervals (CI) for consumption of these foods.Results: Intake of rice, green vegetables, mushrooms and fresh meat was associated with anincreased risk of lung cancer. In contrast, intake of corn, buckwheat, radishes, peppers, melons,pickled vegetables, and salt-preserved meats was associated with reduced risk. The detrimental

effect of ingesting green vegetables (OR, 2.39; 95% CI, 1.28—4.48) is consistent with previousreports.Conclusions: These findings suggest that in Xuanwei, food contamination by environmentalpolycyclic aromatic hydrocarbons may be an important risk factor for lung cancer, and that

∗ Corresponding author at: Occupational & Environmental Epidemiology Branch, Division of Cancer Epidemiology and Genetics,National Cancer Institute, 6120 Executive Blvd, EPS 8122, MSC 7240, Rockville, MD 20892-7240, USA. Tel.: +1 301 451 8791;fax: +1 301 402 1819.

E-mail address: shenmi@mail.nih.gov (M. Shen).

0169-5002/$ — see front matter © 2008 Elsevier Ireland Ltd. All rights reserved.doi:10.1016/j.lungcan.2007.12.024

276 M. Shen et al.

differential contamination of foods by polycyclic aromatic hydrocarbons possibly explained theng cancer risk.

All ri

1

IifiaacpeIcbw

XEX[sbif

iceacasiahorlAtacPic

lGbiaeafan

rttp

niPewfr

srsR

2

2

AcstwiClAcwfiTcwsf

awwe9AiDeclaration of Helsinki for human study subject protection.Oral informed consent was obtained from all subjects.

Trained interviewers administered a standardized ques-

different associations with lu© 2008 Elsevier Ireland Ltd.

. Introduction

n rural Xuanwei County, Yunnan Province, the lung cancerncidence rate is among the highest in China, and is aboutve times the Chinese national average [1,2]. This has beenttributed mainly to burning smoky coal indoors for heatingnd cooking without adequate ventilation [1]. When smokyoal is burned unvented, the indoor air concentration ofarticulate matter and extractable organic matter can bextremely high, up to 24.4 and 17.6 mg/m3 [1], respectively.ndoor levels of benzo(a)pyrene (BaP), an indicator of car-inogenic polycyclic aromatic hydrocarbons (PAH), can alsoe very high, comparable to those experienced by coke ovenorkers [1].

The major exposure route for combustion emissions inuanwei is thought to be inhalation of smoky coal emissions.ven though few women smoked and most men smoked inuanwei, women had comparable lung cancer rates to men1,2]. Women almost certainly had higher long-term andhort-term exposures to indoor fuel emissions than men,ecause they stayed indoors and cooked food for the fam-ly, whereas the men spent considerable time outdoors awayrom the fire.

Given the high air pollution levels and rural lifestylen Xuanwei, it is highly plausible that some foodstuffs areontaminated by PAHs. Thus, beyond inhalation, dietaryxposure may be an important route of exposure to PAHs,nd may thereby play a role in the development of lung can-er. PAHs can adhere to the surfaces of small particles or dustnd travel long distances. The particles can then settle onoil, crops and food or can pollute them via rainfall. Contam-nation with PAHs is more serious if crops or sources of foodre close to human living areas where the level of PAHs isigh. It was found that BaP content was higher in soil in areasf high lung cancer rates than in areas of low lung cancerates [2]. Crops and vegetables, especially those with broadeaves, can uptake PAHs directly from the atmosphere [3].lso, food stored indoors can absorb airborne PAHs. Addi-ionally, livestock and poultry ingesting such polluted foodnd water can accumulate PAHs. Finally, the processing andooking of foodstuffs may increase or chemically alter theirAH content [4]. The extent of food contamination would benfluenced primarily by the exposure levels and absorptionapabilities of particular foodstuffs.

Following oral administration, PAHs and/or their metabo-ites were detected in the lung [5]. It was reported byodschalk et al. that BaP-DNA adducts formed in rat whitelood cells could reflect their presence in lung and it wasndependent of the exposure via either intratracheal or oraldministration [6]. Another animal study reported that lev-

ls of anti-BaP-DNA adducts were highest in liver and lungfter gavage of transgenic mice with BaP [7]. Weyand et al.,ound increases in the number of lung tumors in A/J micefter exposure to food for 260 days that had been contami-ated with BaP (16 and 98 ppm) and manufactured gas plant

thvir

ghts reserved.

esidue (1000 or 2500 ppm) [8]. Likewise, with PAH-rich coalar mixtures in food, DNA adduct levels were increased andhe incidence of lung tumors increased in a dose-dependentattern in female B6C3F1 mice [9,10].

The relationship of dietary PAH with lung cancer risk hasot yet been explored in humans, reflecting the difficultyn assessing dietary exposure to PAHs [11]. Nevertheless,avanello et al. found that anti-BaP-DNA adducts in periph-ral mononuclear lymphocytes in the general populationere influenced significantly by dietary intake of PAH-rich

ood [12] and such high DNA adduct levels may predict higherisk of lung cancer [13].

We conducted a dietary survey as part of a case—controltudy of lung cancer in Xuanwei, and hypothesized that theisk of lung cancer was associated with the ingestion of food-tuffs potentially contaminated with indoor air pollutants.esults are presented below.

. Methods

.1. Study population

population-based case—control study of lung cancer wasonducted from November 1985 through February 1990. Thetudy was confined to all farmers who had been living inhe county for most of their lives. All lung cancer patientsho had been diagnosed in that period were eligible for

nclusion. Cases were collected from all 4 hospitals in theounty that were able to make chest X-rays. A total of 500

ung cancer patients were enrolled from the 4 hospitals.fter exclusion of 2 patients due to incorrect address, 498ases were used in the analysis. Among these, 195 (39%)ere diagnosed based on sputum cytology or pathologicalndings from needle biopsy, bronchofiberscopy or surgery.he remaining lung cancer diagnoses were based on X-rays,linical findings, and clinical history. Among 51 subjectsith cytologic or histologic classification, there were 39%

quamous cell carcinomas, 43% adenocarcinomas, 10% undif-erentiated carcinomas, and 8% small cell carcinomas.

Controls were selected using stratified random samplingt 3 levels, commune, large team and small team. Controlsere matched to cases on age (±2 years) and sex. A controlas enrolled within four weeks after the matching case wasnrolled. The participation rates were 100% for cases and7% for controls. This study was approved by the Chinesecademy of Preventive Medicine and was conducted accord-

ng to the recommendations of the World Medical Association

ionnaire at hospitals (81% cases and 4% controls) or atome (19% for cases and 96% for controls). The inter-iew took an average of about 40 min. The questionnairencluded questions for demographic information, detailedesidence history and fuel use, dwelling type, cooking

Diet and lung cancer risk in Xuanwei, China 277

Table 1 Characteristics of lung cancer patients and controls, Xuanwei, China, 1985—1990

Cases N (%) Controls N (%) P-valuea

Ageb 53 ± 11 53 ± 11 0.46

SexFemale 238 (48) 238 (48)Male 260 (52) 260 (52) 1

LiteracyNot literate 330 (66) 363 (73)Literate 167 (34) 135 (27) 0.03

Ever smoke (males)No 26 (10) 34 (13)Yes 234 (90) 226 (87) 0.24

Duration of smoking (males)No 26 (10) 34 (13)<20 years 34 (13) 40 (15)≥20 to <40 years 116 (45) 117 (45)≥40 years 81 (31) 68 (26) 0.44

Tobacco smoked per day (grams or cigarettes in males)No 26 (10) 34 (13)<15 44 (17) 44 (17)≥15 to <30 121 (46) 114 (44)≥30 66 (25) 67 (26) 0.74

Ever a coal miner (males)No 212 (82) 236 (91)Yes 48 (18) 24 (9) 0.002

Lung cancer in first-degree relativesNo 460 (92) 495 (99)Yes 38 (8) 3 (1) <0.0001c

Non-malignant lung disease historyd

No 393 (79) 461 (93)Yes 105 (21) 37 (7) <0.0001

Food sufficiencyNot enough 42 (8) 82 (16)Just enough 127 (26) 224 (45)Enough with surplus 328 (66) 192 (39) <0.0001

Logistic regression of fuel type and lung cancere

Fuel typeSmokeless coal or wood 23 (5) 146 (30) 1Smoky coal 475 (95) 351 (70) 8.38 (5.18—13.56)

a Pearson Chi-square test.b Mean and standard deviation.c Fisher’s exact test.

losisstedine w

qi

d Including chronic bronchitis, emphysema, asthma, and tubercue Odds ratios and 95% confidence intervals were calculated adju

spent at home per day, non-malignant lung disease history, coal m

practice, lifestyle, tobacco smoking, alcohol drinking, occu-pational history, medical history, family medical history, anddietary intake. The questionnaire contained a number ofquestions on the frequency (‘‘eat frequently’’, ‘‘eat some-

times’’, or ‘‘eat rarely’’) of the intake of foodstuffs thatsubjects commonly ate, including rice, wheat, corn, potato,buckwheat, green vegetables, peppers, melons, carrots,radishes, pickled vegetables, mushrooms, soybean products,salt-preserved meats, fresh meat and eggs. Based on these

2

Tvw

.for age, sex, literacy, lung cancer in first-degree relatives, hoursork history, ever smoking, and passive smoke exposure history.

uestions, we compared the intake frequency of these foodtems between lung cancer cases and controls.

.2. Statistical analysis

he variables for foodstuffs were transformed into dummyariables to examine the effect of each category separately,ith ‘‘eat rarely’’ as referent. For some food items, the cat-

278 M. Shen et al.

Table 2 Odds ratios and 95% confidence intervals of consumption of foodstuffs for lung cancer, Xuanwei, China, 1985—1990

Food itema Cases Controls Unadjusted Adjusted

N (%) N (%) ORb 95% CI ORc 95% CI

RiceRarely + Sometimes 106 (21) 245 (49) 1.00 1.00Frequently 392 (79) 253 (51) 1.53 1.08—2.18 1.53 1.08—2.18

WheatRarely 32 (6) 38 (8) 1.00 1.00Sometimes 402 (81) 413 (83) 1.16 0.71—1.89 0.73 0.40—1.33Frequently 61 (12) 44 (9) 1.65 0.90—3.03 1.10 0.50—2.40P for trend 0.08 0.66

CornRarely + Sometimes 175 (35) 78 (16) 1.00 1.00Frequently 322 (65) 420 (84) 0.34 0.25—0.46 0.52 0.35—0.77

PotatoesRarely + Sometimes 198 (40) 156 (31) 1.00 1.00Frequently 300 (60) 342 (69) 0.69 0.53—0.90 0.75 0.53—1.05

BuckwheatRarely 149 (30) 104 (21) 1.00 1.00Sometimes 340 (69) 367 (74) 0.65 0.48—0.86 0.58 0.40—0.85Frequently 7 (1) 26 (5) 0.19 0.08—0.45 0.21 0.08—0.57P for trend <0.0001 0.0003

Green vegetablesRarely + Sometimes 21 (4) 65 (13) 1.00 1.00Frequently 477 (96) 433 (87) 3.41 2.05—5.67 2.39 1.28—4.48

PeppersRarely + Sometimes 164 (33) 80 (16) 1.00 1.00Frequently 334 (67) 417 (84) 0.39 0.29—0.53 0.36 0.25—0.53

MelonsRarely + Sometimes 313 (63) 274 (55) 1.00 1.00Frequently 185 (37) 223 (45) 0.73 0.56—0.94 0.66 0.48—0.92

Pickled vegetablesRarely + Sometimes 322 (65) 244 (49) 1.00 1.00Frequently 176 (35) 249 (51) 0.54 0.42—0.69 0.46 0.33—0.64

Mushroomsd

Rarely 38 (8) 85 (17) 1.00 1.00Sometimes + Frequently 454 (92) 409 (83) 2.48 1.66—3.72 2.03 1.22—3.37

CarrotsRarely 110 (22) 126 (25) 1.00 1.00Sometimes 361 (72) 342 (69) 1.21 0.90—1.62 1.07 0.73—1.56Frequently 27 (5) 29 (6) 1.07 0.60—1.91 0.78 0.37—1.65P for trend 0.38 0.83

RadishesRarely 63 (13) 40 (8) 1.00 1.00Sometimes 387 (78) 395 (79) 0.62 0.41—0.95 0.53 0.30—0.92Frequently 47 (9) 62 (12) 0.48 0.28—0.83 0.35 0.17—0.72P for trend 0.009 0.004

Soybean productsRarely + Sometimes 336 (68) 392 (79) 1.00 1.00Frequently 161 (32) 106 (21) 1.77 1.33—2.36 1.39 0.96—2.02

Fresh meatRarely + Sometimes 406 (82) 475 (95) 1.00 1.00Frequently 92 (18) 23 (5) 4.68 2.91—7.53 4.41 2.40—8.12

Diet and lung cancer risk in Xuanwei, China 279

Table 2 ( continued )

Food itema Cases Controls Unadjusted Adjusted

N (%) N (%) ORb 95% CI ORc 95% CI

Preserved meatsRarely + Sometimes 218 (44) 139 (28) 1.00 1.00Frequently 280 (56) 359 (72) 0.50 0.38—0.65 0.50 0.35—0.71

EggsRarely 16 (3) 13 (3) 1.00 1.00Sometimes 439 (88) 464 (93) 0.77 0.37—1.62 0.49 0.20—1.20Frequently 42 (8) 20 (4) 1.71 0.69—4.22 0.88 0.29—2.66P for trend 0.05 0.66

a ‘Rarely’ and ‘Sometimes’ were combined as reference if the number of ‘Rarely’ was smaller than 5.b Crude unadjusted OR.c ives,

at bif ‘Fr

cst1taodcvi(tm

idqWsqaaqrC1b(00b(ilC

Adjusted for age, sex, literacy, lung cancer in first-degree relatcoal mine work history, ever smoking, passive smoking, coal type

d ‘Sometimes’ and ‘Frequently’ were combined as the number o

egories ‘‘eat rarely’’ and ‘‘eat sometimes’’ were combineddue to small sample sizes.

Some subjects had missing information for some vari-ables, and therefore cases and controls were not completelymatched in data analysis. Thus, unconditional logisticregression models were constructed to calculate odds ratios(OR) and 95% confidence intervals (CI), with two-sided P-value (conditional logistic models yielded similar results,not reported here). The independent variables includedage, sex, literacy (adjusting for socioeconomic situation),having enough food (adjusting for socioeconomic situationand possibly amount of ingested food), having been a coalminer (adjusting for occupational exposure), ever smok-ing, family members’ smoking, lifetime hours staying athome (adjusting for indoor fuel smoke exposure), historyof non-malignant lung diseases (associated in Xuanwei andelsewhere with lung cancer risk), having a first-degree rela-tive who had lung cancer, and the specific mine from whichthe family bought coal (adjusting for exposure to coalswith potentially different carcinogenic potencies, referencegroup = using smokeless coal or wood). Different variablesof tobacco smoking were modeled (pack-years, duration,intensity, etc.) and similar results were obtained. So inthe final model, smoking was adjusted as ever smoking.Unadjusted ORs were also presented for overall analysis.Interaction with sex was tested in the model combiningmen and women by adding a multiplicative term for intakefrequencies of specific foods by sex. Sex-specific logisticregression models were also constructed. Very few womenhad smoked or been coal miners, so their models were notadjusted for these variables.

3. Results

As expected, sex and age were very similar in cases and

controls. However, more cases could read (P = 0.03) thancontrols (Table 1). Most men (88%) smoked. Overall, tobaccosmoking was not associated with increased risk of lung can-cer. However, cigarette smoking was marginally significantlyassociated with lung cancer risk at the highest exposure

r

cIc

hours spent at home per day, non-malignant lung disease history,rth, and if having enough food.equently’ was smaller than 5.

ategory (pack-years lower than the median value vs. never-moking: OR, 0.98; 95% CI, 0.48—2.01; pack-years abovehe median value vs. never-smoking: OR, 1.95; 95% CI,.00—3.82) (data not shown in Table 1), which was similaro the results obtained from other studies conducted in thisrea [1,2]. Ever being a coal miner, positive family historyf lung cancer, and personal history of non-malignant lungiseases were associated with an increased risk of lung can-er. More cases than controls had enough food to eat (92%s. 84%, P < 0.0001). Overall, lung cancer risk was far highern smoky coal users than in smokeless coal and wood usersOR, 8.38; 95% CI, 5.18—13.56) (Table 1). There was substan-ial heterogeneity in risk estimates for coals from differentines.Table 2 summarizes the analytical results for foodstuffs

n men and women combined. Rice was a staple in theiets of many subjects, and cases ate rice more fre-uently than did controls (OR, 1.53; 95% CI, 1.08—2.18).heat, corn, potatoes, or buckwheat were staples for

ome subjects. Cases ate corn and buckwheat less fre-uently than did controls (OR, 0.52; 95% CI, 0.35—0.77,nd OR, 0.21; 95% CI, 0.08—0.57, respectively). Subjectste vegetables more frequently than meats. Vegetables fre-uently consumed were green vegetables, peppers, melons,adishes, wild mushrooms and pickled leafy vegetables.ases ingested more green vegetables (OR, 2.39; 95% CI,.28—4.48) and mushrooms (OR, 2.03; 95% CI, 1.22—3.37)ut less peppers (OR, 0.36; 95% CI, 0.25—0.53), melonsOR, 0.66; 95% CI, 0.48—0.92), pickled vegetables (OR,.46; 95% CI, 0.33—0.64) and radishes (OR, 0.35; 95% CI,.17—0.72) than controls. Major sources of protein were soy-ean products (Tofu, soybean milk), salt-preserved meatse.g., ham), fresh meat (pork), and eggs. More frequentntake of fresh meat (OR, 4.41; 95% CI, 2.40—8.12) andess frequent intake of preserved meats (OR, 0.50; 95%I, 0.35—0.71) were associated with increased lung cancer

isk.

There were significant interactions between a few spe-ific foods and sex in models including interaction terms.n stratified analysis by sex, the associations of lung can-er risk with rice (OR, 2.33; 95% CI, 1.37—3.96; Pinteraction,

280 M. Shen et al.

Table 3 Sex-specific adjusted odds ratios and 95% confidence intervals of consumption of foodstuffs for lung cancer, Xuanwei,China, 1985—1990

Food itema Female Male

Cases Controls ORb 95% CI Cases Controls ORc 95% CI

N (%) N (%) N (%) N (%)

RiceRarely + Sometimes 65 (27) 117 (49) 1.00 41 (16) 128 (49) 1.00Frequently 173 (73) 121 (51) 1.01 0.61—1.65 219 (84) 132 (51) 2.33 1.37—3.96

CornRarely + Sometimes 68 (27) 40 (17) 1.00 110 (42) 38 (15) 1.00Frequently 173 (73) 198 (83) 1.14 0.63—2.06 149 (58) 222 (85) 0.26 0.15—0.45

Buckwheatd

Rarely 57 (24) 51 (22) 1.00 92 (36) 53 (20) 1.00Sometimes + Frequently 181 (76) 186 (79) 0.84 0.49—1.46 166 (64) 207 (80) 0.34 0.19—0.60

Green vegetablesRarely + Sometimes 11 (5) 38 (16) 1.00 10 (4) 27 (10) 1.00Frequently 227 (95) 200 (84) 3.26 1.45—7.31 250 (96) 233 (90) 1.62 0.60—4.36

PeppersRarely + Sometimes 89 (37) 45 (19) 1.00 75 (29) 35 (14) 1.00Frequently 149 (63) 193 (81) 0.38 0.22—0.65 185 (71) 224 (86) 0.31 0.17—0.57

MelonsRarely + Sometimes 151 (63) 133 (56) 1.00 162 (62) 141 (54) 1.00Frequently 87 (37) 104 (44) 0.70 0.44—1.13 98 (38) 119 (46) 0.66 0.41—1.06

Pickled vegetablesRarely + Sometimes 160 (67) 105 (45) 1.00 162 (62) 139 (54) 1.00Frequently 78 (33) 129 (55) 0.34 0.21—0.55 98 (38) 120 (46) 0.59 0.37—0.96

Mushroomsd

Rarely 20 (8) 56 (24) 1.00 18 (7) 29 (11) 1.00Sometimes + Frequently 216 (92) 179 (76) 2.69 1.39—5.20 238 (93) 230 (89) 1.50 0.65—3.46

RadishesRarely 25 (11) 19 (8) 1.00 38 (15) 21 (8) 1.00Sometimes 185 (78) 188 (79) 0.53 0.22—1.24 202 (78) 207 (80) 0.49 0.23—1.05Frequently 27 (11) 31 (13) 0.43 0.15—1.21 20 (8) 31 (12) 0.28 0.10—0.79P for trend 0.13 0.02

Fresh meatRarely + Sometimes 210 (88) 228 (96) 1.00 196 (75) 247 (95) 1.00Frequently 28 (12) 10 (4) 1.72 0.66—4.50 64 (25) 13 (5) 8.50 3.70—19.52

Preserved meatsRarely + Sometimes 103 (43) 66 (28) 1.00 115 (44) 73 (28) 1.00Frequently 135 (57) 172 (72) 0.55 0.33—0.91 145 (56) 187 (72) 0.51 0.31—0.85

a ‘Rarely’ and ‘Sometimes’ were combined as reference if the number of ‘Rarely’ was smaller than 5.b Adjusted for age, literacy, lung cancer in first-degree relatives, hours spent at home per day, lung disease history, passive smoking,

coal type at birth, and if having enough food.c Adjusted for age, literacy, lung cancer in first-degree relatives, hours spent at home per day, non-malignant lung disease history, coal

rth, af ‘Fr

0ba0or

4

mine work history, ever smoking, passive smoking, coal type at bid ‘Sometimes’ and ‘Frequently’ were combined as the number o

.04), corn (OR, 0.26; 95% CI, 0.15—0.45; Pinteraction, 0.0002),

uckwheat (OR, 0.34; 95% CI, 0.19—0.60; Pinteraction, 0.02)nd fresh meat (OR, 8.50; 95% CI, 3.70—19.52; Pinteraction,.01) were restricted to men (Table 3). Ingestion ofther food items was not associated with lung cancerisk.

Weio

nd if having enough food.equently’ was smaller than 5.

. Discussion

e observed that an increased intake of rice, green veg-tables, mushrooms and fresh meat was associated with anncreased risk of lung cancer. In contrast, frequent intakef corn, buckwheat, radishes, peppers, melons, pickled

ws

orth

qqtstcidteitpwfioar

cidsiwt

C

N

R

Diet and lung cancer risk in Xuanwei, China

vegetables, and salt-preserved meats was associated withreduced risk. This suggests that dietary intake may be animportant route of exposure, and PAH-contaminated foodsmay increase the risk of lung cancer.

Rice, green vegetables, mushrooms and fresh meat maywell be exposed to unusually high levels of PAHs in Xuan-wei and in turn absorb and concentrate PAHs. The intakeof such concentrated amounts of PAHs could well increasethe risk of lung cancer. Traditionally, rice was prepared andsun-dried in the yards or roads in the neighborhood for afew days after it was harvested. Local people were fondof eating leafy vegetables which were often grown locallyin the neighborhood, or even in the backyards of houses. Inaddition, residents ate wild mushrooms which grew near thevillage and in turn could absorb substantial PAHs from boththe atmosphere and the soil [14]. Meats (pork) came frompigs that were raised in residents’ yards and therefore nearthe household stoves. Pigs also frequently ate cooked foodin Xuanwei. Local residents liked to eat fatty meat insteadof lean meat. These food items may all have high absorp-tion capacities for PAHs, due partly to their relatively largesurface areas and/or high lipid content.

Unlike rice and wheat, the other two staples, corn andbuckwheat, were found to be inversely associated withlung cancer risk in Xuanwei. Usually, buckwheat was grownin less fertile land and in higher altitude areas far awayfrom the village, making it exposed to less PAHs fromthe burning of fuels. Corn was a common food in Xuan-wei and was also often dried at home, making it exposedto high levels of PAHs. However, it was often threshed,washed with water, and ground into powder before cook-ing. Such processing may decrease the content of PAHs incorn.

Unlike fresh green vegetables and fresh meat, pickledvegetables and preserved meats (not smoked) were associ-ated with reduced lung cancer risk. Vegetables were oftenwashed first, processed with a large amount of salt, and thenput into a sealed container for at least one month beforeeating. Usually, a family prepared pickled vegetables onceeach year. Pork was often washed, salt-processed, put in asealed container for about 2 weeks, and then dried in a darkroom for a few months before it was eaten. PAHs may havedegraded in vegetables and meat when they were processedwith salt and water, and might also have been degraded bybacteria [15].

Peppers, radishes, and melons were associated withreduced lung cancer risk in Xuanwei. These vegetables haveboth small and smooth surfaces, which in turn enable themto be washed easily, and perhaps to resist PAH absorption.This also suggests that the major source of pollution for veg-etables comes from the atmosphere instead of from the soil[16].

Effects of rice, corn, buckwheat and fresh meat wereobserved only in men. This may reflect different ingestedamounts between men and women. In general, men con-sumed more staples and meats than did the women even ifthey ate them at the same frequencies. Therefore, expo-

sure levels could well have been different between thesexes, but these differences would not have been capturedby the questionnaire. In summary, it is plausible that PAHcontamination was relatively high in rice, green vegetables,mushrooms and fresh meat, and relatively low in corn, buck-

281

heat, radishes, peppers, melons, pickled vegetables, andalt-preserved meats in Xuanwei.

Opposite to our results, inverse associations werebserved between vegetable/fruit intake and lung cancerisk in most other studies in China [17,18]. This may reflecthat most food in these studies was not contaminated withigh environmental PAHs.

There are several limitations of this study. First, theuestions concerning dietary intake did not ascertain intakeuantitatively, and did not query intake of all foodstuffshat residents might have eaten. Second, some food waseasonal and intake frequencies would have changed overime, which were not considered. Third, we did not measureoncentrations of PAHs or other coal combustion emissionsn foodstuffs and the diverging contamination by PAHs forifferent food was largely based on assumption with lit-le analytic data. However, interim results from an ongoingxposure study in Xuanwei suggest that some foodstuffs arendeed contaminated with PAHs, and that such contamina-ion is correlated with indoor air PAH concentrations (FWei,ersonal communications). Fourth, histological informationas not available for most cases and this restricted strati-ed analysis by histology. In addition, we did not collect datan cooking oil fumes, and confounding by cooking oil fumesnd residual confounding by tobacco smoking may play aole, even though its impact should be minimal.

In summary, we observed considerable variation in lungancer risk associated with ingestion of different foodstuffsn Xuanwei. This variation could plausibly be explained byifferential rates of PAH absorption by the different food-tuffs considered. Thus, ingestion of PAHs may constitute anmportant risk factor for lung cancer in Xuanwei and else-here. Before these results can be considered conclusive,

hey should be replicated and extended in further research.

onflict of interest

one declared.

eferences

[1] Mumford JL, He XZ, Chapman RS, Cao SR, Harris DB, Li XM, etal. Lung cancer and indoor air pollution in Xuan Wei, China.Science 1987;235:217—20.

[2] He X, Yang R. Lung cancer and indoor air pollution from coalburning. Yunnan Science and Technology Press; 1994.

[3] Mumtaz M, George J. Toxicological profile for polycyclic aro-matic hydrocarbons (PAHs). Atlanta, GA, U.S. Department ofHealth and Human Services, Agency for Toxic Substances andDisease Registry; 1995.

[4] Guillen MD, Sopelana P, Partearroyo MA. Food as a sourceof polycyclic aromatic carcinogens. Rev Environ Health1997;12:133—46.

[5] Ramesh A, Inyang F, Hood DB, Archibong AE, Knuckles ME,Nyanda AM. Metabolism, bioavailability, and toxicokinetics ofbenzo(alpha)pyrene in F-344 rats following oral administration.Exp Toxicol Pathol 2001;53:275—90.

[6] Godschalk RW, Moonen EJ, Schilderman PA, Broekmans WM,

Kleinjans JC, Van Schooten FJ. Exposure-route-dependent DNAadduct formation by polycyclic aromatic hydrocarbons. Car-cinogenesis 2000;21:87—92.

[7] Helleberg H, Xu H, Ehrenberg L, Hemminki K, Rannug U, Torn-qvist M. Studies of dose distribution, premutagenic events and

2

[

[

[

[

[

[

[

[

82

mutation frequencies for benzo[a]pyrene aiming at low dosecancer risk estimation. Mutagenesis 2001;16:333—7.

[8] Weyand EH, Chen YC, Wu Y, Koganti A, Dunsford HA,Rodriguez LV. Differences in the tumorigenic activity of apure hydrocarbon and a complex mixture following ingestion:benzo[a]pyrene vs manufactured gas plant residue. Chem ResToxicol 1995;8:949—54.

[9] Culp SJ, Gaylor DW, Sheldon WG, Goldstein LS, BelandFA. A comparison of the tumors induced by coal tarand benzo[a]pyrene in a 2-year bioassay. Carcinogenesis1998;19:117—24.

10] Culp SJ, Warbritton AR, Smith BA, Li EE, Beland FA. DNA adductmeasurements, cell proliferation and tumor mutation induc-tion in relation to tumor formation in B6C3F1 mice fed coal taror benzo[a]pyrene. Carcinogenesis 2000;21:1433—40.

11] Kazerouni N, Sinha R, Hsu CH, Greenberg A, Rothman N. Anal-ysis of 200 food items for benzo[a]pyrene and estimationof its intake in an epidemiologic study. Food Chem Toxicol

2001;39:423—36.

12] Pavanello S, Pulliero A, Saia BO, Clonfero E. Determinantsof anti-benzo[a]pyrene diol epoxide-DNA adduct formationin lymphomonocytes of the general population. Mutat Res2006;611:54—63.

[

M. Shen et al.

13] Tang D, Phillips DH, Stampfer M, Mooney LA, Hsu Y, Cho S, etal. Association between carcinogen-DNA adducts in white bloodcells and lung cancer risk in the physicians health study. CancerRes 2001;61:6708—12.

14] Cerniglia CE. Fungal metabolism of polycyclic aromatichydrocarbons: past, present and future applications inbioremediation. J Ind Microbiol Biotechnol 1997;19:324—33.

15] Bamforth SM, Singleton I. Bioremediation of polycyclic aro-matic hydrocarbons: current knowledge and future directions.J Chem Technol Biotechnol 2005;80:723—36.

16] Tao S, Cui YH, Xu FL, Li BG, Cao J, Liu WX, et al. Polycyclic aro-matic hydrocarbons (PAHs) in agricultural soil and vegetablesfrom Tianjin. Sci Total Environ 2004;320:11—24.

17] Galeone C, Negri E, Pelucchi C, La Vecchia C, Bosetti C, HuJ. Dietary intake of fruit and vegetable and lung cancer risk:a case—control study in Harbin, northeast China. Ann Oncol2007;18:388—92.

18] Lu Q, Yao S, Huang C, Lan Y, Cang S, Qiao Y, et al. Acohort study on the relationship between vegetable intakeand risks of lung cancer in the Tin Corporation (YTC) minersin Yunnan. Zhonghua Liu Xing Bing Xue Za Zhi 2000;21:205—7.