Tobacco Documents Online

, American Jorxnaf of Epidetniology Vol. 13y;,trlo. 2' Copyright ® 1993'by The Johns Hopkins tJniver'sity Schoot lof HY*ne and IPublic Health Allirights reserved Indoor Air Pollution and Lunq Cancerr im Guangzhou, People's Republic of' China Qing Liu,'-2 Annie'J. Sasco;23 'Elio Riboli ~ and Meng Xuan Hiu' Pdnteaf'in tY.S.A. A case-aontrol study comprising 224 male and 92 female incident lung cancer cases andl the same number of i individually matched hospital controls was conducted from June 1983 to June 1984 in Guangzhou, People's Republic of China, to evaluate the association between indoor air pollint'ton and lung cancer risk. Guangzhou residents were exposed to several sources of pollhtion in i their homes, most importantly to cooking fumes: Increased risks were found among subjects living in a house without a separate kitchen (the exposure odds ratio was 2.4 (95% confidence interval (Cl) 1.4= 4.2)~ for men and 5.9 (95% Cl 2.1-16.0) for women). Similarly, living in a house with poor air circulation i was associated with an exposure odds ratio of' 2.11 (95% Cl 1.2- 3! 8) for men and 3_6 {J5"k Cl 1.4-9.3) for women. A trend in i the association between lung cancer risk and' factors pertaining to house and kitchen ventilation was observed, and a decreasing risk of lung cancer was observed for several variables indicating better ventilation, even after adjustment for potential confounders such as edtucation, occu- patiorn„living area, smoking, and history of'¢hroni+c respiratory diseases. No statistically significant differences were found between cases and controls for frequency of cooking at~ home, presence of a chirnney in the kitchen, or type of cooking fuel. Smoking was clearly related to risk of lung i cancer in both men and women; and among I nonsmoking women, exposure to tobacco smoke from their spouses was also associated with an increased risk. These results suggest that, in' addition to smoking, indoor air pollution may be a risk factor for lung,cancer. Arrt J Epidertlio/'1I993;137:14'5'-54. air pollutants, environmental;, lung neoplasms; smoking; tobacco smoke pollution; ventilation Lung cancer is the major cause of cancer dc: h for both men and women in Guang- zhou, the main city inithe Guangdong prov- ince of the People's Republic of China, rep- resenting about one fourth of all cancer deaths in that city. During the period' 19'$1- 19'84'y the crude mortality rates for lung,can- cer in the'_ Guangzhou metropolitan are~a Received for publication March 18, 19971, and in final form August 17„ 1992. Abbreviation: CI, confidence interval. "Departmentlof'Medical Statistics, Sun YatTSenUniver- sity of' Medical Sciences, Guangzhou 510089, People's Republic of~China. z'Unit of Anatytical! Epidemiology, International Agency for Research on Cancer, 150 cours Albert ~ Thomas, 69372' Lyon iCedex 08, France. 3'Institut National de la Sante et de la Recherche Medi- cate (INSERM), Lyon;,France. were, respectively, 45' male deaths and 24 female deaths per 100,000 person-years (1). Although smoking is the most important cause of lung cancer (2-4), smoking,behav- ior cannot fully explain the epidem2ology of lung cancer in Chinese women, in whom there is a rather high incidence of lung cancer, predominantly adenocarcinom'a, dt- Reprint requests to Dr. Annie J. Saseo„Unit of Analytical Epidemiolbgy; International'Agency for Research on Can- cer, 150 cours Albert Thomas, 69372' Lyon Cedex 08, France. Dr. Qing Liu was supported by a Research Training Fellowship awarded by the International Agency for Re- search on Cancer:. The authors are grateful'to Dr. Jing Yuan, Dr. Hao~Ran Guo, and Jiu-Lan Zhang for conducting the interviews and to the staffs of the Guangzhou hospitals who gave invalu- able access to their patients and records. 145

~ 146 Liu et~ al. spite relartively low smoking rates (5-12). The male:female ratio of lung cancer inci- dence is about 2 in China (1, 13); even though there is a much bigger difference than that in smoking rates between, men and women. In contrast, this ratio varies fromi 4 to 10 in Western, Europe andl is about 2 in the United States (13), but' withi a muchh higher prevalence of current and former smokers among American, women. During the ! last decade, increasing atten- tion has been, paid to the health effects of the indoor microenvironment. Some studies have shown a positive relation of passive smoking,to lung cancer risk (14-18), as well as ani association with exposure to radon and its decay products (19-25). However, re- search! results for other sources of indoora'ir pollution, suchi as cooking,and heating, have been rather elusive and inconclusive (26- 29). The goal of this study was to evaluate whether there is any relation between indoor air polbut2on resulting from domestic cook- ing practices and lung cancer occurrence: MATERIALS AND METHODS New1y diagnosed cases of primary lung cancer (Int'ernational Classif ration of Dis- eases, Ninth Revision,, code 162) were se- lected from eight major hospitals covering most of Guangzhou from June 1'983 to June 1984. All' cases occurred in permanent resi- dents of t'he city of Guangzhou. A totall of 327 lung cancer cases were identified from the medical records of these~ eight hospitals during that year. It was possible to complete an interview for 316 cases (224 men and 92' women; 96.6 percent). Eleven cases were excluded because either they were too illl to answer the interview, they couil'd not be traced, or they had already died. Fifty-five percent of cases were diagnosed by means of repeated chest radiographic examinations and clinical examination;, 13 percent were diagnosed by bronchoscopy alone; and 322 percent of the cases had cytologic or histo- logic confirmation. Controls who were also permanent resi- dents of the city of Guangzhou were individ- uall'y matched'. to cases on age (to within 2 years), sex; residential district (Liwan, Yuexiu, Haizhu, or Dongshan), and date of' diagnosis or hospital'" admissioni. Controls were selected according to the matching cri- teria among, i'npatients of the surgery de- partments am six of the same eight major hospitals. No controls were chosen from either the Tumor Hospitall or the Chest Hos- pital. Patients who had been adinittedl for chronic obstructive diseases of the lung, pul- monary t'uberculosis, malignant tumors, andd coronary heart disease were excluded. Each control was to be interviewed during the 2 mont'hs following the interview of his/her matched case. In most instances, the first selected patient control was interviewed; only for 26 subjects were second-choice con- trols chosen because of an inability to trace the subject or because an individual was found to, be inadequate regarding one or severallof the matching variables. No subject refused the interview. The interview was carried out at the sub- ject's home by trained epidemaologic work- ers using a structured questionnaire. All cases and controls were interviewed~ in per- son. The interviewers obtained extensive in- formation about the subject's generaf de+ mographic characteristics, occupationall his- tory, history of respiratory diseases, family cancer history; smoking habits, spouse's smoking habits; cooking practices (including, domestic fuel use), and residence history. After completing the, questionnaire,, the in- terviewer measured the size of the windows and doors that opened onto the outside of the building, thereby providing, an estima- tion of ventilation capacity. The ventilation capacity of the kitchen was analyzed sepa- rately from tlhat' of the rest of the dwelling; hereafter designated the living area. If the subject had lived in his or her present home for less than 20 years, the interviewer asked similar questions regarding the preceding residences and~ their vent'ilation, conditions_ In the latter case, given that no measure- ments were~ available, the ventilation: condi- tions of previous residences were simply ranked as poor, average, or good based on the questionnaire data. Data on up to three residences were collected; however„since the

Indoor Air Pollution and Lung Cancer 147 population of the People's Republic of China is relatively stable, very few subjects had moved more than twice. Exposure odds ratios, their 95 percent confidence intervals, and significance levels were computed using a matched-pair analy= sis method (30). A conditionaf logistic regression model was used to estimate ex- posure odds ratios for multiple class vari- ables and toiadjust fior confounding,factors. The SAS (31) 1 and EGRET (32) software packages were used for the analysis. RESULTS Table l presents the! demographic char- acteristics of lung,cancer cases and controls. The ages of cases ranged~ from 28 years to 83' years, but the majority' of cases were in their fifties or sixties. The median ages of male and female cases were ! 62 years and! 60':5 years, respectively. The median ages of mate:and fentale controls were 62 and 61.5. Almost all subjects were of Han nationality, and most' subjects had been born in, the province of Guangdong. IViaritall status; ed- ucational level, dialect, occupation, and'.liv- ing area did not differ significantly between cases andl controls, although cases had a lower educational level and a smaller aver- age living area than controls. Thus, we con- trolled fbr education, occupation, and living areai wheni we analyzed other variables. TABLE 1. Demographic characteristics of lung cancer cases and controls, Guangzhou, People's Republic of China,1983-1J84 Men Women Characteristic. No. of cases. ~ No. of comtrols % No. of' cases % fifo. of controls % Age (years) <4'0 2 0;9 3 1.3' 1' 1.1 1 1.11 40-49 13 5:8 12 5.4 9 9.8 110 10:9 50-59 76 33;9 80 35.7' 33 35.9 33 35.9 60-69 90 402 84 37.5 24' 2'611 23 25.0 >70, 43' 19.2 45 20.1 25 27:2 25 27.2 Marital ,status Married 193 86.2 194 86.6 56I 60.9 53 57.6': W idowed 27 12.1 24 10.7 36'. 39.1 33 35.91 Divorced 3 1.3' 4 1.8 0 0.0 0 0.01 Single . 11 0.4 2' 0.9 01 0A 6 6.5 Years of'eiducation <1 22 93 10, 4'.5 40 415 41 44.6 1-6 139 62.1 134 59.8' 35 38.01 35 3'8:0 7-12 50 22:3 60 26'.8' 17 18!5 16 17.4 >_13 13 5.8 20 8'.91 0 010 0 0.0 Province of'birth Guangdong 207' 92.4 206 92.0 80 87.0 83' 90.2 Other 17' 7.6 18 8:0 12 13.0 9 1 9.8' Occupation Worker 152 67.8' 153' 68.3 59~ 64.1 55 59.8' Other 70 31.3' 67' 29.9 11 12.0 13 14! 1 None 2 0.91 4 11.8 22 23.9' 24 26.1 Living area (tm2 per person) <2.00 11 4.9 6 2'7 2 2.2 2 2.2 2.00-3;99 56' 25.0 42 18;7 27 29.3 15 16.3 4_00-7:99' 1100: 44_6 109 48;7 38 41.3 42'. 45.6 >_8:00 57 25.5 67 29:9 25' 27.2 31 35.9

. 148 Liu et a1. TABLE 2. History of occupational exposure and personal andlfamiliallhistory of selected diseases among lung cancer cases and controls, Guangzhou, People's Republic of China„ 1983-1984 Men Women Exposure No. of' easeS'. wo' of ~trOIS'. EOR,• E0R2 95% Clt fdo: of cases. ~h' of CUn[YOIS. EOR, EORx, 95% Cl History of occupational ex• posure Nonet 1511 184 1.0 1.0 69 78 1.0 1.0 Dust 20 8 2.6' 2.2' 0:78-6,2 13 4 4.3' 8.8 1_6-16.6'. Smoke 23 12 2.1 2.1 0A9-5,1 1' 2 0.74' 0.57 0.00-73.4 Other 30 20 1.7' 1.6 072-15 9 8 1.5 1.0 0.26-4.2 History of pulmonary, tuber, culbsis filot' 137 176 1.0 1.0 78 85 1.0 1.0 Yes 87 48 2.8 2:4 1.3-4.6 14 7 2.2' 1.3 0.39-4.3' History of chronic bnonehftis Not 125 185 1.0 1.0 63 79 1.0 1 L0 Yes 99 39 3.9 3;0 1.6-5;5 29 13 2.6' 0.90 : 0.36-2.4 Fartu'ly history of canoer Not 203 216 1.0 1.0 83 86 1.0 1LO Yes 21 8 2:9 0:90 0.62-1.3 9 6 1.5' 0.94 054-1_6'. * EOR,, matched exposure odds ratio; , EORz,, matched logistic exposure odds ratio, adjusted for education, occupation, living area, and smoking; CI, confidence internral. t Referent. As expected; the subjects with lung cancer showedl increased frequencies of occupa- tio',nall exposure to hazardous working envi- ronments, a history of pulmonary tubercu- losis„ and a history of chronic bronchntisiin both sex groups (table 2): When educartion, occupation, living area, and smoking were controlled for, the associations of lung can- cer with, the above risk factors were not substanRially modified among men, but they were attenuated or even disappeared among wornen; except! for exposure to dust. In con- trast, the increased risk of lung cancer asso- ciated with a family history of cancer, found in, uniivariate analysis, disappeared for both men and women when we controlled for the same variables. Smoking, was strongly re-latedl in a dose-response manner to the risk of' lung cancer in both men and women. (table 3). Niinety-f ve percent of male cases and 59 percenrtt of' female cases had ever smoked, compared with 80 percent and 25 percent of male andi female controls, respec- tively. The exposure odds ratios for lung cancer were 6.1 (95percent~ conifidencein_ terval (CI) 2.7-15.0) and 4'.91(95 pereent CI'. 2'.3-10.4) among male and female smokers, respectively. An increased risk oflungcancer was, also observed among nonsmoking women who lived', with a husband who smoked the equivalent of 20: or more ciga- rettes per dhy (table 3): Table 4 presents infprmationon the cook- ing practices of lung cancer cases and con- trols. R'esults, showed that women! borethee primary burden of cooking, in the family. The majority of families cookedl three times per day at home, and coal was the basic cooking fuel used. Questions were asked onn the type of fuel used dluring three historical periods (1194'9-195'7, 19'58-19716, and 1977 to the present). Since cooking fuel is rationed' by t!he government, there was little variation, in fuel usage among familiies in Guangzhou. Before 1958, most families used~ wood and charcoal. From 19'58 to 19'76, they gradually turned to coal, and gas was usedl oniiy after 1976. Results are presentedl for the most recent period„whiieh is the only one in which differences between cases and controls could be distinguished. The history of coal use in Guangzhou was usually longer than 20 years, and only in the past 3^5 years had a small propo,rtion of families changed'to pe- troleum gas as a: domestic fuel. Very few families used" rnaiinlh, electriciRy or kerosene

In+door Air Pollution and Lung Cancer 149 TABLE 3. Exposure to smoking among lung i cancer cases and controls, Guangzhou, People's Republic of China, 1983 -1984 Exposure No. of cases Nb: of EOR"' 95% Cl• controls Active smoking (cigarette equivalents smoked periday)among i men Plever smokert 2 4 .0 1-19 21 93 1.2 0.4".5 20-29 , 97 66 7.1 2.6'-19.5'. Z30 94' 21 21.41 7:1-G4.0 X=fortrendi= 99.6 p<0.001 Active smoking (agarette equivalents smoked per day) emons. wonoen i Mever smoker fii 8 9, .0 11-9 8 1i0 1.8 0.57-5.9 10-19 16 9 3.5 12-9:8 ;_-20 30 4 17.9 4!0-80:6 for trend = 28',0 p <0i001 Passivesmoking among nonsmmkinr,q wornen (cigarette equiva- lents smoked per day by husband14', Wot ettpbsedt 3 2 .0. 1-19 6 21 0.7 0.23-2.2 >_20 19 16 2:9 1.2-7.3 X''for'trend i= 4:5 P = 0 084' * EOR, matched logistic exposure odds ratio, adjusted for edwcation; oaaupationj and living area, cornparing never smokers to present and former smokers combined;'C/. confidence intervaL tl. Referent. t', unmatched method was used. i,.. cooking. In the kitchens of more than % percent of families„there was no chiimneyor other apparatus for extracting, fumes. Nol significant case-control differences were as- sociated withi the above variables. Information on the use of cooking,oil was not formally included in our study. In the Peoplc's Republic of China, cooking oil' is rationed by the government-controlled food and oil company, and in Guangzhou, the main cooking, oil used is peanut oil (for a short time it: was bean oil). Rapeseed oil was not widely available, and we did', not expect to find any diilff'erences in the types of oil being used, because all Guangzhou residents are dependent on the same government sup- p1y- Analysis of the effect of ventilation con- ditions on lung cancer risk was conducted for housing at the time of diagnosis and interview„ as well'as housing where subjects had lived!the longest; finally, data were sum- marized over all residences. The results of these analyses were similar (data na shown); therefore, this paper presents only the xesults of the analysis for current ventilation con- ditions. Iniaddition, the proportions of cases and controls who ~had moved during the past. 110 or 20 years did not differ significantly, and duration of residence in the current housing was similar for cases and controls. For most subjects, the current residence was also the longest residence. Table 5 shows that after adjustment for education, occupation, occupational ex- posure, pulmonary tuberculosis history, chronic bronchitis history, faRnily cancer his- tory,,arnount of smoking per day, and living area, as well as passive smoking (for women only), several variables pertaining to venti- lation conditions were strongly associated with lung cancer risk. There was increased risk associated with having a window or door opening, from the kitchen directly into the liiving,area or bedroom, and for cooking in the living area or bedroom. Fon cooking in the living area or bedroomy the exposure odds ratio was 2.4 (95 percent CI 1.4'-4'.2) , for men and 5.9 (9S percent CI 2:1-16.0), for wamen. Having windows or doors that openedl in d(ffierent directions so that indoor air could circulate! alsol signifcantly influ~ enced the risk of lung, eancer: The relative risk for lung,cancer tended' to decrease withi increasing size of ventilation openings ini living areas and kitchens. For t'he best ven- tilated living area as campared with the lcast ventilated, the exposure odds ratiolfor lung cancer was reduced to only 0:14 (95 percent', CI 0.04-0.5 111) for men and 0!02 (95 percent CI 0.00-0.21) for women. The exposure odds ratios for kitchen ventilation, were 0.15. (95 percent CI 0AS-0.44) for men and 0.06' (95 percent CI 0.01-0.32): women, re- spectively. The differences were statistically significant. A similar trend was found for the ceiiling,height throughout the apartment, but no clear trend was seen for the floor oni

150 Liu etal. TASLE 4. Cooking practices of lungicancer cases and controls, Guangzhou, People's Republic of China, 1983-1984 Men Wbmen Cooking practice No: of aases No: of oontrds EOR,'' E0112- ' 9596 C1* No. of fiW. of cases oontrols EOR, EOR.` 95% Cl Frequency of cooking at home RarelYl 107 100 1.0 1.0 4 5 1.01 Oocasionally 45 63 0 65' 0.52 025-1.11 8 8 1.2' 1'.2 0.17 -9.2 Frequently, 72 61 1.1 1.11 0.69-1.9 80 79 L.3'. 1.1 0.19-6.1 Filavig a chimney in kitchen CYMS) No ctrrrv*yt 160 147 1.0 1.01 56 54 1.0 1.0 1-9 33' 37 0.80 0.55' 0.22-1.6' 16 10 1.7 3;6' 0:72-17:5 ;--10 31 40 0.70 0.80 0.34-1.9 20 28 0.77 1.1 0:40-3:0 Cooking fuel Coa1t 200 193 1.0 1.0 81 79 1:01 1.0 Gas 14 22 0.59 0.48 0.15-1.6 8 91 0.90 0.90, 0.24-3.3 Woldd 8 9 0.79 0.57 0:11-3.0 3 4 0.67 0.67 0.04-11.7 Other 2 0 01 0 No. of meals prepared at't home per day 0-1 t 18I 26' 1l0 1.0 2 3 1.0 1.0. 2 26' 33 1.2 0:83' 0.27-2.6' 7 12 0.67 0A8 0.03-8.8 3 180 165 1.7' t.3' 060-3:3 83 77 1.5 1.5 0.12-17.7 •'EOR;, matdwd expmsure odds ratio: EOR2. matched logistic exposure odds ratio, adjwsted for educationl, occupation, oecupationalexposure, history oi'tuberautosisf chronic bronchitis.,fandly'Fustory of canoer. smoking, and living area; G. confidence intereral: EORb, matched logistic exposure odds rati©, adjusted for passive smoking,in addition to all of the variables listed labove: t'i Referent. which the subject lived, with, the exception of a higher risk for people living on the ground flbor as opposed to a higher flopr: When we analyzed the effect of floor by district of residence and by sex;, we found the higher risk for persons living on the groundl floor in, three out of four districts for women and in two out four districts for men (data not shown): D'IS'CUSSI ON' Usually, h ospital control's are selected from the same hospitals as the cases, but in this study some of the controls were nott chosen fromi the same hospital as their matched cases because two of the hospitals that were ! sources of cases were del[berately, exclluded as sources of controls. We did this to avoid selectioni of subjects with, diseases that could potential'l'y' be linked t;o smoking and/or air pollution. To reduce potential bias in, the control selection, procedure, we matched cases and' controls oni district of residence, thereby controlling for the referral pattern of patients. Most importantly, this matching also: controlled, for atmospheric air pollution. Results of environrnentaU surveil. lance were on'ly availablle at the district level, as were lung cancer mortality rates. One district, Liwan, has the highest levels for all indexes of air pollution (falling dust, benzo(a)pyrene, total suspended partRcul'ate: matter„ sulfur dioxide, nitrogen oxide, andd carbon monoxide), whereas the other t'hreee districts do not differ from one another ap- preciably (33). The higmest age-standardized (world standard population) lung cancer mortality rates for women are recorded, in Liwan at 22.2' deat'hs per 100,000 woman- years, compared with the lowest rate for the: district of Dongshan at 18'.3: For men, the highest mortality rate is in Yuexiu at 51.8, versus the lowest rate in Dongshan at 40.9. All of the pairs of cases and! controls lived in the same residential district, in an, area of about 10 kmZ. They were exposed to ap- proximately similar levels of outdoor air pollution, although variations may have oc- curred within districts. No data were avail-

IndoorAirPollution and Lung Cancer 151 TABLE 5. Ventilation conditions in the homes of7ungicancer cases andlcontnols, Guangzhou, People's Republic of'China, 1983-1984 Men i Women Ventilation factor Wo: of cases No. of EOR,' EOR7 control$ 95% CI' Mo' of' Ddo' of EOR, EOR; ' cases controls 95% Ct Separate kitchen Yest 72 113' 1.0. 1.0 52 1.0 1.0 No 152 11'1 2:0. 2.4 114-4:2 40 3.1 5.9 2.1-16.0i ('aopd'air circwlation Yest 90 120 1,.0, 1.0 38 60 1.0 1.0 No 134 104 1.7' 2.1 12-3.8' 54 32 3.0 3.6 1.4-9.3 Size of ventilation openings in. iving area (m2 per'perr son) 0.0-OAt 47 48 1.0' 1.0 16' 6 1.0 1.0 0.5-0.9 86 69 0.53' 0.39 0:14-1.1 36 23 0.69, 0:36 0A9-1.5 1.0-1.9 38 53 0.33' 0.30 0:10-0.91 22 24 0.40, 0:25 0.05-1.1 I 2.0-3J9 22 30 0.33' 0.24: 0J06-0.90 9. 13' 0.24 0.14 0.02-0.89 z410 31 24 0.28' 0:14I 0J0M-0.51 9 26' 0.11 0J02 0'.00-02!t' Xz for trend 14.5 (p <'0~:001). 18.0 (p < 0.001) . 'ize of ventilation operings in kitchen (n>='per famrly) 0i0-0.4t 79 41 1.0 1 _0 22' 8 1.0 1.0 0:5-U.9 58 52 0.55 0.77' 0.36-1t7 27 25 0.29 0.11 0J02-0J60 1.0-1.4 48 59 0.38 0.23' 0.10-0.56 24 22 0.31 0:13' 0J02-0J74 1.5'-1.9 19 31 0.28 0.49 0.16'-1:5 7 13 0.16 0.09 0J0n+-0 M _2A 20 41 0.20 0.15 ~ 0.05-0.44 12' 24 0.16 0.06' 0M-0J32 X2 for trend 25.7 (p <0!001) 10.3(p<0:001) Height of room (m) Q.8t 61 45' 1.0 1~0 22 12 1.0 1.01 2:".1 72 71 0.77 0.91 0:41-2:0 30 26 0.62 0.54 0.09-1.3' >_3.2' 91 108' 0.65 0.64 0.31-1.3 40 54 0:39 0.23' 0.06-0.84 xz for, trend 3!5 (p = 0.06). 5!5'(p = 0.02). Floor of apartment Groundt 106' 92 1.0 1.0 50 32 1.0 1i0 1 77' 78 0:85 0.79 0.40-1.5' 23 34 0!30 0.12 0.03~-0.46' 2-3 27' 34 0.66 0.88 0.36-2:1 14 22 029 0.11 0.02 -0.54 _4 14 20 0.59 0.62 0.20-1.9 5 4 056 0.72 0.07-7.01 X2 for trend 3,0(p=0.08). 5:0 (p = 0.02). • EOR,, matchedl exposure odds ratio;, EOR2, matched logistic exposure odds ratio; adjusted for education, occupation, occupational exposure, history of tuberculosis, chronic brorxhitis; family history of'cancer; smoking, and tiving,area; CI, confidence interval; EOR,, matched lkogistic exposure odds ratio, adjusted for passive smoking in addition to all of the variables listed above: t Referent: able on a, scale smaller than the district. We knew that separation of the effects of out, door air pollution from those of indoor air pollution might be diff icult~ This led us to match on residence as a proxy for matching on outdoor air polllution, level. In addition, our results, demonstrating a clear reductionn in risk of llung cancer for living, in a house with large openings onto the outside, irldii- cate that outdoor air pollution is unlikely to explain or confound the association found between lung cancer risk and indoor sources of air pollutiom Our results showed't'hat comparability be- tweenicases and'controls withiregard tolbasic demographic variables was good, suggesting that, these demographic variables might not have a major eonfounding effect. Further- more, education and occupation, which were considered good representative vari- j I

~ 1152' Liu etal~. ables of social class„ were also controlled for in the analysis. Since the number of cases in eachi histo- logic category was limited, a separate analy- sis could not be carried out for the associa- tion of indoor air pollution, with specif c histologic types of lung cancer: Sorne:diag- nost'rc error may also have occurred. How- ever, undifferential misclassif icationi of cases and controls usually leads only to a bias toward the null and cannot explain the ob- servedresults. It is possible t'hat some social class di#1'er- ences between cases and controls could ac- count for part of the difl'erence in ventilation conditions. The : finding of a higher risk of lung cancer among subjects living at street level could be explained by a higher concen- tration, of bothi indbor and outdoor air pol- lutants on the ground floor as opposed to higher floors, where natural venti.lation! is usually better. Social' class, which is closely linked with housing, conditiions in China, was nevertheless more strongly associated with the total: surface of the living area than was the specific measure of ventilation. Al- though cases had smaller living areas than controls, the difference was not substantial. After adjustment for education, occupation, and living area, the relation between venti- lation conditiions: and lung cancer risk re- mained. This means that any bias linked to social'status cannot be a major confounder of the observed association. Recall bias can, affect most retrospective studies; however;, in this study, the size of ventilation openings was objectively measured, so the association, between ventilation conditions and lung cancer risk cannot be an artifact of recallll bias: Smoking was the major cause of lung cancer in both sex groups, and a clear dose- response relation was observed'betiNreen the amount~ of tobacco srnokedl and lung,cancer risk. Highi smoking rates were observed in both male and female groups. These rates were higher thant'hoseobtained~ froam aicase- control study, of lung cancer carried out among Shanghai Chinese women (11) but similar to rates from other studies (6, 1©„ 18.. 34). The subjects of this study belonged to an older population and! had relatively low social'status (a low educational levell and the major occupation of worker)1 These two fac- tors are known to be associated with high smoking rates (35). Furthermore, the def- nition of a smoker in this study included both ever and!current smokers. For instance, if we had counted only current smokers, the smoking rate would have been reduced~ to 141 . percent in female controls. Passive smok - ing may also account for some : excess risk, although increased risk was only observed in the women living with husbands who smoked heavily. No effect was seen for women married to light smokers. This may be explained by the reduced sample size and by the imprecise quant2f cation of passive smoking. The results of other studies in Chinese women have suggested t'hat, passive smoking contributes to a slight increase in lung cancer risk (11, 18).. Severaf reports of environmental moni- toring showed that the concentrations of nitrogen oxide, sulfur dioxide, benzo(a)- pyrene, and! total suspended particulate mat- ter in Giuangzhoui were higher in dwellings than in the outdoor atmosphere and varied according to, time of day and season (36). Three peaks of pollutantconcentratiions dur- ing the day were correlated! with cooking activities. Results of studies also showed that there were higher concentrations of sus- pended dlust and suspended benzo(a);pyrene in the rooms and' urinary benzo(a)pyrene in housewives in households where coall was used as a cooking fuel compared with house- holds where pet'rolieum gas was used (36, 37),. Results of a study conducted! inthe; north of China showed that use of coal for heating elevated the risk of lung cancer inn comparisoni with use of gas. Cooking, in the bedroom was also related to an excess risk ('34). High miortality from female iung can- cer in Xuan.Wei County, Yunnani Province, was also associated with the combustioni of smoky coal at home (38). These results in} dicated that: home cooking, practices were a major source of indoor air pollutants and that coal produced more severe air pollution than other kinds of d'omestic fuel. Evidencealsocomesfromtheevaluat~ion

of occupational exposure to coal burning. Steel, gas, and coke oven workers have an, elevatied' risk of lung cancer (39-45). Thee results of animal experiments have demon- strated that some components of coal, fumess are carcinogenic or mutagenic (38, 46, 47). Since coal use, frequent' home cooking, and lack of an apparatus for extracting fumes are universal in Guangztiou, it, was difficult to f nd a significant diifference between any population~ groups. However, this indicates that severe indoor air pollution exists forr most, families in Guangzhou, where people live in comparatively overcrowded condi- tions with poor ventilation. Better ventila- tion of houses could thus play a key role in improvement of the indoor m4cnoenviron- ment, and dissimilar ventilation conditions c.ld be responsible for different exposure levels of lung cancer cases and controls. During the study period, there was in, most houses no artificial ventilation such as air conditioning, so the indoor microenviron- mental conditiionsdependeld, mainly onnat- ural'ventilation. The area of ventilation (as defined by the area of openings to the out, sidi;) was a good representative measure of ventilation conditions. This is in agreement with other studies on indoor air pollurtionn sbowing, that the concentrations of polllu- tants are greatly affected by ventilation (19, 34). Iln summary;,t'he results of this study sug- gest that indoor air pollution produced dur- ir<.- home cooking is a risk factor for lung cancer in Guangzhou, especially for women, who are more likely to be exposed to coal fumes and cooking oil vapors in the kitchen. This could contribute to the high rate of lung, cancer in Chinese women. Further in- vestiigatrons are needed to clariifytheprecisenature of indoor air pollutants and their carcinogenic mechanisms. It would also be informative to conduct studies by major his- tologic type, particularly in Chinese women, among whom adonocarcinomai is unusually frequent. Finally, this study was not de- signed to evaluate the effect of outdoor air pollution; that must be left to future re re- search. Iln the meantime,, our data indicate that in Guangzhou, there are sourct;s of in- IrndoorAir Pollution and Lung Cancer 153 door air pollution which play a role in the occurrence of lung cancer independent'ly, of active smoking and outdoor air pollution. REFERENCES 1. Hu MX, Guo HIR, Huang XL„ et a11 Analysis of cancer mortality and its time trends in Guangzhou (1964-1982). (in Chinese). Chin J Health Stat 1988;5:12-16. 2. International Agency for Research on Cancer. To- bacco smoking. IARC Monogr Eval Carcinog Risk Chem Hum 1986:3'8. 3. Zaridze D, Peto R, eds. Tobacco: a major interna- tional health hazard. Lyon, France: International Agency for Research on Cancer,, 1986. (IARC Sci- entific Publication no. 74). 4. US Department' of Health and Human Services. The Health Consequences of Smoking; cancer. A Report of the Surgeon General. Rockville„ MD: Office on Smoking and Health, US Public Health Service, 1982. (DHHS publication no. (PHS) 82- 50179). 5. Fraumenii JF Jr, Mason TJ. Cancer mortality amongChinese Americans,1950-69: J'Natl Cancer Inst 11974;52:659-65. 6. MacLennan R, Da Costa J,, Day NE, et al. Risk factors for lung cancer in Singapore Chinese, a population with high l female incidence rates. Int J Cancer 1'977;20:854-60: 7. Hinds M'W„ Stemmermann GN, Yang HY, et al. Differences in lung, cancer risk from smoking among,Japanese, Chinese and Hawaiian women in Hawaii. Int J Cancer 1981;27:297-302. 8. Green JP, Brophy P. Carcinoma of theiung in non: smokiiag,Chinese women. West J Med 1'982;136` 291-4. 9. Kung IT, So KF, Lam TH. Lung cancer in Hong Kong, Chinese: mortality and histological types, 1973-1982. Br J'Cancer 1984;50:381-8! 10. Koo LC, Ho JH, Lee N. An analysis of some risk factors for lung cancer in Hong Kong. IntJ Cancer 1985;35:149-55: 11. Gao YT, Blot WJ, Zheng W, et all Lung cancer among Chinese women. lnt J Cancer 1987A01 604-9. 1'2'. Koo LC,, Ho JHL Worldwide epidemiological pat- terns of 1 ung cancer in nonsmokers. Imt J Epidemiol 1990;J!9(suppl1 1);S'd4-23: 13, Muir C, Waterhouse J, Mack T, et al, eds. Cancer incidence in f ve continents, vol 5. Lyon, France: International Agency for Research on Cancer, 1'987. (IARC Scientific Publication no. 88). 14. Hirayama H. Nonsmoking wives of heavy smokers have a higher risk of lung cancer: al study from Japan. BMJ 1981;282:I183:-5. 15. Trichopoulos D, Kalandidi A, Sparros L, et al. Lung cancer and passive smoking: Int' J Cancer 19811;27c 1-4. 16. Sandler DP, Everson RB, Wilcox AJ. Passive smok- ing,in adulthood andlcaneer risk. Ami Epidemiol 1985;121:37-4'8! 17. Akiba S,, Kato H, Blot WJ. Passive smoking amdl lung cancer among Japanese women. Cancer Res 1986;46I:4804'~ 7: I :f I J i

154 Liu et al. 118! Lam THi Kung, ITIvI, Wong, CMi et al. Smoking, passive smoking and histological types in lung can- ceri in Hong Kong, Chinese women. Br J Cancer. 1987;56`673-8. 19! Axelson O, Edling C, Kling H: Lung, cancer and residency: a case-referent study on the possible impact of expo.sure to radon and its daughters in di,veilings. Scand J' Work Environ Health 1979;5: 10-15. 20: Edling C,, Kling H; Axelson O. Radon in homes- a possible cause of lung cancer. Scand J' Work Environ Health 1!984;,10?2'5-34. 21. Damber LA, Larsson LG. Lung cancer in males and' type of dwelling: an epidemiological pilot study. Acta Onco1 1987;216:2111-I5: 2'2: Svensson C, Eklund' G, Pershagen G. Indoor ex- posure to radbn from the ground and bronchial cancer in women. InU Arch Occup Environ Health 11987;59`. I23-3'1. 2'31 Axelson 0, Andersson K, Desai G, et al. Indbor radbn~exposure and active and passive smoking in relation to.the occurrence of lung cancer. Scand4 Work Environ Health 19'88;14:286-92. 24. Svensson C, Pershagen G, Klbminek J! Lung can- cer in i women and type of diwelling, in relation i to radbniexposure. Cancer Res 1989;49`.1I861-5'. 25: Schoenberg,JiB, K1ou J'B, Wilcox HB, et al. Case- control study of residential radon and lung cancer among New Jersey women. Cancer Res 1990;50` 6520-4. 26: Leung JSM. Cigarette smoking, the kerosene stove and lung cancer in Hong Kong. Br J' Dis Chest 1'977;71:273-6. 27. Koo LC, Lee N, Ho J. Do cooking fuels pose a risk for lung cancer?'A case-control study of women in Hong Kong. Ecol Dis 1983;2:255-65. 2& Wu A, Henderson B'E; PikeMIC, et al. Simoking, and other risk factors for lung cancer in women, J Nartl'Cancer Inst 198'5;74:747-5I. 29! Sobue T. Association of indoor air pollution and lifestyle with lung cancer in Osaka, Japan. lint J Epiderniol 1990; l9(suppl I ):S62'-6, 30; Breslow NE; Day NE, eds. Statistical methods in cancer research. Vol I. The analysis of case-control studies. Lyon, France: Internationali Agency, for Research on Cancer, 1982. (IARC Scientific Pub- lication no. 32). 31. SAS Institute, lina. SAS user's guide: basics. Ist ed. Cary. NC: SAS';Institute, linc, 1986. 32. Statistics and Epidemiology Research Corporation. EGRET reference manual. l,st, ed. Seatrtle. WA: Statistics and Epidemiology Research Corporation, 1990. 33. Department of Environmental Protection. People's Government of Guangzhou. Report oni environ- mental quality' in Guangzhou during, 1981. (In Chinese). Guangzhou, China: People's Govern- ment of Guangghou„ 1'982'. 34: Xul ZY,, Blot WJ, Xiao HP, et al. Smoking, air, ppollution, and the high rates of lung cancer in Shenyang, China: J Natil Cancer Inst 1989;81: 1800-6. 35. Nationwide Investigation Group on Smoking, Re- port of nation-wide sampling investigation on smoking.. (IIn Chinese). Nati Med J China 1987;. 67:229-32.. 36. Huang LF: A report of environmental monitoring for benzo[a]tDyrene (B[a]P)in the suspended pantic- ulite matter in Guangzhou. (Working,report; in Chinese). Guangzhou, China: Health and! Anti- Epidemic Station of Guangzhou, 1987. 37. Zhang ZF, Yu SZ, Zhou GD. Indoor air pollution, of coal fumes as a risk factor of stroke,, Shanghai. Am J Public Health 1988;78:975-7. 38. Mumford JL, He XZ, Chapmen RS, et al. Lung cancer and indoor air pollution in Xuan Wei, China. Science 1987;23'5:217-20: 39. Redmond CK, Cioao A, Lloyd! JW: Long-term mortality study of steel workers. VI. Mortality from malignant neoplasms among coke oven workers. J Occup Med 1972;14:621-9: 40. Blot WJ; Brown LM, Pottern LM, et al. Lung cancer among long-terrnti steel workers. Am J Epi- demiol 1983;,1'.17:706-16. 41. Redtnond CK„ Strobino BR, Cypess RH. Cancer experience among coke by-product workers. Ann N Y AcadiSci 1976;21'7:ll02-1!5: 42. Redmond CK. Cancer mortalityarnongaoke oven workers. Environ Health Perspect I'983;52:67-73: 43. Lloyd JW. Long-term mortality study of steel work- ers. V. Respiratory cancer in coke plant workers. J Occup Med 197'Ih,13:53-68! 44. Doll R, Fisher REW, Gammon EJ, er al. Mortality of gasworkers, with special reference to cancer of the lung and bladd'er,,chronic bronchitis and pneu- mmnoconiosis, Br J Ind Medl l'965;22' 1'-12. 45. Doll R, Vessey MIP, Beasley RWR, et'a1. Mortality of gasworkers: final report of a prospective study. Br J' Ind Med I'972;29:394-4'06. 46. Internartional; Agency for Research on Cancer: In- dustnai exposures in alluminium production, coal gasificartion, coke producUion; and iron and steell founding; IARC Monogr Eval Carcinog Risks Hum 1'984;34:65 -19a: 47. Liang CK„Quan NY, Cao SR, eti all Vatural inha- lation exposure to coal smoke and wood! smoke induces lung cancer in mice and rats, Biomedi Environ Sci 1988;1:42-50!