Use Of Wood Stoves And Risk Of Cancers Of The Upper Aero

Transcript

O International Epidemiological Association 1998 Printed in Great Britain International Journal of Epidemiology 1998;27:936-940 Use of wood stoves and risk of cancers of the upper aero-digestive tract: a case-control study Javier Pintos, a Eduardo L Franco,a Luiz P Kowalski, b Benedito V Oliveirac and Maria P Curado d Background Incidence rates for cancers of the upper aero-digestive tract in Southern Brazil are among the highest in the world. A case-control study was designed to identify the main risk factors for carcinomas of mouth, pharynx, and larynx in the region. We tested the hypothesis of whether use of wood stoves is associated with these cancers. Methods Information on known and potential risk factors was obtained from interviews with 784 cases and 1568 non-cancer controls. We estimated the effect of use of wood stove by conditional logistic regression, with adjustment for smoking, alcohol consumption and for other sociodemographic and dietary variables chosen as empirical confounders based on a change-in-estimate criterion. Results After extensive adjustment for all the empirical confounders the odds ratio (OR) for all upper aero-digestive tract cancers was 2.68 (95% confidence interval [CI] : 2.2-3.3). Increased risks were also seen in site-specific analyses for mouth (OR = 2.73; 95% CI: 1.8-4.2), pharyngeal (OR = 3.82; 95% CI: 2.0-7.4), and laryngeal carcinomas (OR = 2.34; 95% CI: 1.2^i.7). Significant risk elevations remained for each of the three anatomic sites and for all sites combined even after we purposefully biased the analyses towards the null hypothesis by adjusting the effect of wood stove use only for positive empirical confounders. Conclusions The association of use of wood stoves with cancers of the upper aero-digestive tract is genuine and unlikely to result from insufficient control of confounding. Due to its high prevalence, use of wood stoves may be linked to as many as 30% of all cancers occurring in the region. Keywords Oral neoplasms, larynx neoplasms, case-control studies, environmental exposures, logistic regression, confounding Accepted 9 April 1998 Cancers of the upper aero-digestive tract (UADT) are among the most common neoplasms worldwide. Incidence rates of cancer of the oral cavity and pharynx are rising in most areas.1 The UADT cancers rank as the third most frequent group of neoplasms among males and the fourth most frequent among females in developing countries.2 Incidence rates in Southern Brazil are among the highest in the world; the combined annual age-standardized rates for oral, pharyngeal, and laryngeal cancer are 49.7 and 36.1 per 100 000 males in Sao Paulo and Porto Alegre, respectively.2'3 Several epidemiological studies have identified tobacco and alcohol consumption, and their joint effect, as the most ' Departments of Oncology and Epidemiology, McGill University, Montreal Canada. b AC Camargo Hospital SJo Paulo, Brazil. c Erasto Gaertner Hospital Curitiba, Brazil. d Araujo Jorge Hospital Gdynia. Brazil. Reprint requests to: Dr E Franco, Department of Oncology, McGill University. 546 Pine Avenue West, Montreal. QC. Canada R2W 1S6. important risk factors for the development of UADT cancers. 4 " 6 Diets low in carotenes and vitamin A have also been identified as determinants of these neoplasms. 7 " 9 Other regionally specific dietary factors, such as consumption of mati or chimarrSo, have been found consistently associated with increased risks for UADT cancers in temperate South America. 10 ' 11 Indoor air pollution has been postulated as a potential risk factor for head and neck neoplasms. Kodama and Dollar12 suggested that the high incidence of cancers of the UADT in Southeastern Asia could be related to chronic exposure to the smoke from indoor fires. A main difficulty in evaluating the risk of UADT cancers due to indoor fires is the appropriate control of confounding, mainly tobacco smoking, alcohol consumption, and dietary factors. Incomplete adjustment for these variables may leave residual confounding, manifested as spuriously elevated risks attributable to indoor air pollution. We analysed the data from a case-control study to assess whether use of wood stoves for cooking and heating is associated with an increased risk of UADT cancer in Southern Brazil. To circumvent the confounding problem by prominent risk 936 WOOD STOVES AND UPPER AERO-DIGESTIVE TRACT CANCER factors we used a very conservative approach to adjust the analysis of the association between use of wood stove and UADT cancer risk for an extensive list of empirical confounders. Subjects and Methods Study subjects The investigation followed a hospital-based, case-control study design. AH patients with newly diagnosed squamous cell carcinoma of the mouth (International Classification of Diseases, 9th revision rubrics [ICD-9] 140-145), pharynx (ICD-9 146149), and larynx (ICD-9 161) referred to three head-and-neck surgery services in Sao Paulo (Heliopolis Hospital), Curitiba (Erasto Gaertner Hospital), and Goiania (Araujo Jorge Hospital) between January 1987 and January 1989, were considered eligible for the study. Patients with tumours of the salivary gland (ICD-9 142) or of the nasopharynx (ICD-9 147) were excluded from the investigation. All diagnoses were confirmed histopathologically and the anatomical site was ascertained post-surgically. Control subjects were selected among inpatients from the same hospitals where cases had been identified or from another general hospital with the same catchment area as the index hospital. Two control patients were matched to each case on the basis of gender, 5-year age group, and trimester of hospital admission. Patients with neoplastic diseases (ICD-9 140-239) or mental disorders (ICD-9 290-319) were not eligible as controls. Risk factor information Nurses, who had been specially trained for this study, conducted questionnaire-based, structured interviews with all subjects to elicit information on sociodemographic variables, health conditions, environmental and occupational exposures, tobacco and alcohol consumption, diet, and oral hygiene. Interviews were carried out before treatment was initiated and were interrupted if patients experienced difficulty communicating due to pain or speech problems. Altogether, there were nine cases eliminated from the study: one refusal, seven due to physical conditions, and one due to our inability to identify suitable controls. There were no refusals to participate among controls. Methods of analysis The odds ratio (OR) was the measure of association used to estimate the relative risk of disease due to each study factor. The OR and their respective 95% confidence intervals (CI) were calculated by conditional logistic regression, which maintained the matching used in the design of the study. 13 Analyses were done with MULTLR, a public domain software for both unconditional and conditional logistic regression (available at hup:// www.epi.mcgill.ca).14 Adjustment for tobacco and alcohol consumption was based on the lifetime cumulative exposure using the pack-years equivalent of dgarette smoking and the sum over all beverage types in kg of ethanol consumption. The variable subsuming cumulative tobacco consumption included other tobacco types, besides cigarettes. For the computation of packyears of consumption we assumed the following equivalence: 20 manufactured cigarettes = 4 hand-rolled, black tobacco cigarettes = 4 cigars = 5 pipes with pipe tobacco. A pack-year was defined as the cumulative exposure equivalent to smoking one pack of cigarettes daily during one year. Likewise, the 937 lifetime alcohol consumption variable included individual beverages reported during interview. The ethanol content was assumed to be 5% in beer, 10% in wine, and 50% in hard liquor. Since residual confounding was a major concern in the interpretation of the association between use of wood stove and risk of UADT cancer, control of confounding was done by extending covariate adjustment to other variables that were considered empirical confounders. Following a change-in-estimate confounder selection criterion, 15 ' 16 we identified all potential confounders based on their ability to alter the point estimates of the OR for wood stove exposure. We defined as empirical confounders any variables that, when included in the models, changed the point estimate of the crude or of the adjusted (for tobacco and alcohol consumption) OR by 2% or more. We arbitrarily defined the cutoff at this conservative level to select all variables that could confound the association either individually or jointly, even at the expense of a probable loss in precision when examining the wood stove-risk association. This process was repeated for all sites combined and for mouth, pharynx, and larynx, independently. In addition to the tobacco and alcohol consumption variables defined above, we considered the following factors as potential empirical confounders: (i) Sociodemographic variables: ethnicity, rural residency, schooling level, household income; (ii) Diet: past frequency of consumption of chimarrao, coffee, tea, and drinking temperature of these beverages, smoked or cured meat, barbecued or broiled meat, orange, lemon, tomato, carrot, pumpkin, papaya, pequi (a fibrous fruit common in central Brazil), leafy and green vegetables, pickles, pepper, eggs, milk, pinh&o (the fruit of a coniferous tree common in Southern Brazil), manioc, corn; (iii) History of employment in specific industries: textile, wood and paper, mining, leather, metallurgy, sugar and alcohol refining, and rubber. In the interest of 'capturing' the confounding effect, variables with multiple categories were preserved as such when added to models as sets of dummy regressors; we made no attempt to use simplified dichotomous forms of any of the potential confounders. All models adjusted for smoking and alcohol contained at least four dummy factors for each of these variables to accommodate five ordinal categories of exposure. The statistical assessment of interaction (effect modification) was based on a multiplicative scale by fitting models containing main effect variables and their cross-product terms, inference was based on the statistic representing the ratio of the estimated coefficient to its standard error assuming a standard normal distribution. Results A total of 784 cases and 1568 controls were included in the study. Case accrual by city was as follows: 213 (27.2%) in Sao Paulo, 380 (48.5%) in Curitiba, and 191 (24.4%) in Goiania. There were 373 (47.6%) patients with cancer of the mouth, 217 (27.7%) with pharyngeal cancer, and 194 (24.7%) with laryngeal cancer. The underlying causes of hospitalization among control patients could be grouped into 13 diagnostic categories of the ICD-9, of which the most common were: digestive system diseases (26.0%), cardiovascular system diseases (24.9%), ill-defined diagnostic conditions (8.6%), and genito-urinary tract conditions (7.5%). 938 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Table 1 summarizes the distribution of cases and controls according to selected characteristics. There were slightly more non-white subjects among controls than among cases, and controls had more years of schooling. The proportion of individuals using wood stove for cooking and heating was higher among cases (50.7%) than among controls (32.8%). As expected, the intensity of smoking and alcohol drinking was also higher among cases than controls. The alcohol-adjusted OR contrasting ever versus never smokers was 7.6 (95% CI : 4.9-11.9). Similarly, the tobacco-adjusted OR for heavy alcohol drinking Table 1 Distribution of cases and controls according to selected sociodemographic characteristics Variable Age Gender Ethnic background Schooling level Cases n(%) <50 169 (21.6) 50-59 60-69 70 + 278 (35.5) 556 (35.5) 218 (27.8) 438 (27.9) 119 (15.2) Male Female White Mulatto Black Other Illiterate Grade school High school College 683 (87.1) 236 (15.1) 1366 (87.1) 202 (12.9) 1236 (79.2) 236 (15.1) 78 (5.0) 10 (0.6) 433 (27.6) 961 (61.3) 124 (7.9) Use of wood stove for cooking or heating Tobacco smoking (in pack-years) Alcohol consumptioni (inkgs) Controls Categories 101 (12.9) 660 (84.5) 86 (110) 31 4 252 467 51 14 (4.0) (0.5) (32.1) (59.6) (6.5) (1.8) No 386 (49.3) Yes 397 (50.7) <1 30 1-22 23-45 46-91 207 >91 202 124S 142 200 (3.8) (18.2) (26.5) (25.6) (25.9) n(%) 338 (21.6) 49 ?•}) 1050 (67.2) 513 (32.8) 358 (23.0) 362 (23.2) 333 (21.4) 267 (17.1) 239 (15.3) 95 (12.1) 416 (26.7) 80 (10.2) 295 (18.9) 181 (23.1) 366 (23.5) 166 (21-2>. 261 (33.3) 233 (14.9) 249 (16.0) relative to light drinking, defined as the upper two quintiles versus the lowest quintile of cumulative consumption, was 4.1 (95% 0:2.8-6.0). Matched analysis In the matched, unadjusted analysis, patients using wood stoves for cooking or heating had 2.7 times (95% CI: 2.2-3.3) the risk of developing UADT cancers compared with non-users (Table 2). Crude OR of similar magnitude were obtained when the analysis was repeated for individual cancer sites, with values ranging from 2.6, for mouth, to 2.8, for pharynx. The excess risk was only partially due to confounding by smoking or alcohol consumption. Addition of covariates indicating lifetime consumption of tobacco and alcohol reduced only slightly the magnitude of OR for use of wood stove, for all cancers and for each individual site. All adjusted estimates were significant at the 5% level (Table 2). At least part of the apparent excess risk could have been due to additional confounding by variables other than tobacco or alcohol. To test this hypothesis we identified all the empirical confounders in our data set based on a change-in-estimate confounder selection criterion using a 2% cutoff. Following this selection process we identified 13 variables for the analysis of all sites combined, 18 for the site-specific analysis for mouth, 18 for pharynx, and 19 for larynx. The confounders yielding the highest changes (*4% in either direction) in the sitespecific adjusted OR estimates for wood stove were: (i) mouth: smoking, rural residence, ethnicity, income, and consumption of smoked meat and oranges; (ii) pharynx: smoking, rural residence, schooling, employment in the mining industry, and consumption of chimarrdo, coffee, smoked meat, pumpkins, papaya, and corn; (iii) larynx: alcohol drinking, rural residence, and consumption of chimarr&o, coffee, carrots, and pequi. After adjustment for the extensive sets of confounders, the significant elevation in risk persisted for all cancer sites combined as well as in the analyses for individual sites (Table 2). In fact, except for laryngeal cancer, controlling for the net effect of all empirical confounders tended to increase the magnitude of the OR. Addition to these models of cross-product terms to account for the potential joint effect of smoking and alcohol resulted in significant contribution to the goodness-of-fit of the models for pharynx, larynx and all three sites combined. It had little effect (<2%), however, on the magnitude of the OR for wood stove in each of the four analyses. We further decided to purposefully bias the analysis towards the null hypothesis by following an overly conservative approach. Table 2 Odds ratios8 (OR) of upper aero-digestive traa cancers due to use of wood stove Topography All sites Mouth Pharynx Larynx Crude OR (95% CI) 2.68(2.17-3.31) 2.62 (1.94-3.53) 2.83 (1.85-4.32) 2.65 (1.72-1.10) Adjusted1" OR (95% CI) 2.39(1.88-3.05) 2.34(1.67-3.29) 2.78 (1.70-4.53) 2.37 (1.40-4.02) * By conditional logistic regression (matching variables: age, gender, hospital and admission period). b Adjusted for tobacco and alcohol consumption. c Adjusted additionally for all empirical confounders (see text for explanation). d Adjusted only for positive conlounders (see text for explanation). Adjusted 0 OR (95% CI) 2.45 (1.84-3.26) 2.73 (1.76-4.24) 3.82 (1.96-7.42) 2.34(1.17-1.67) Adjusted d OR (95% CI) 2.21 (1.71-2.87) 2.21 (1.54-3.16) 2.51 (1.96-7.42) 1.83(1.02-3.29) WOOD STOVES AND UPPER AERO-DIGESTIVE TRACT CANCER 9 3 9 We estimated the OR for wood stove adjusting only for 'positive confounders', i.e. for those empirical confounders whose inclusion decreased the point estimate of the OR for use of wood stove. It is dear that this approach causes a bias in our estimation of the excess risk, a bias towards no association. Although this model is not valid for risk estimation, it serves the purpose of illustrating that even with such a conservative approach, we still found persistent and significant elevations in risk (ranging from 80% to 150%) for each of the three cancer sites and for all UADT cancers combined (Table 2). Joint effects of use of wood stove with tobacco and alcohol Analysis of the combined effects of wood stove and smoking, and wood stove and alcohol, were assessed by adding the crossproduct terms to the underlying multiplicative models containing these variables as main effects. As shown in Table 3, there seemed to be an indication that relative risks due to wood stove use may tend to increase with cumulative smoking exposure. However, there was no statistical evidence of effect modification between wood stove and smoking, or between wood stove and alcohol, as judged by the likelihood ratio contribution in goodness-of-fit of the cross-product terms. Due to the small number of non-smokers among cases, the OR estimates are very imprecise in this category, and prevented us from carrying out detailed site-specific analyses. No dear trend was found in the analysis of the joint effect of wood stove and alcohol, either crude or tobacco-adjusted. Gender-specific analysis Due to the relatively small number of women in our study, gender-specific analyses lacked the necessary power to control confounding for as many variables as we had accounted for in the overall analyses. Table 4 shows the OR associated with wood stove exposure separately for males and females, and adjusted for ethnidty, income, rural residence, schooling level, smoking, and alcohol consumption. The risk of disease due to use of wood stove was consistently higher for women than for men in all three cancer sites, but more noticeably for pharyngeal and laryngeal cancers. The difference in gender-spedfic excess risks reached statistica] significance in the analysis of laryngeal cancer, where women had an OR of 16.2 (95% CI: 2.7-99.1) and men an OR of 2.0 (95% CI: 1.1-3.7). Table 3 Odds ratios3 (OR) of upper aero-digestive tract cancers due to use of wood stove, according to lifetime cumulative tobacco smoking and alcohol consumption Variable Tobacco smoking Lcvel . 2-08 (0.8«.09) ~ \ (1:57~2-.?.?> 2 - 1 . 3 .('- 53 - 2 - 96 > .^?« 2.67 (1.96-3.62), _ 2.74 (1.99-3.76) Alcohol consumption (kgs) 0-10 3.12 (1.85-5.26) 2.94 (1.66-5.19) 11-793 2.25 (1.61-3.14) 2.19 (1.55-3.10) >793 2.79(2.04-3.82) 2.44(1.76-3.37) • By conditional logistic regression (matching variables: age, gender, hospital. and admission period). b Adjusted for the other variable in the Table. .. 1 45 207 Adjusted OR (95% a ) 2 16 Table 4 Gender-specific upper aero-digestive tract cancer risks associated with use of wood stove Topography All sites Mouth Pharynx Larynx Adjusted* OR (95% CI) Males Females 2.42 (1 .84-3.19) 4.11 (1.99-8.97) 2.52(1 .69-3.76) 2.77 (1.09-7.02) 2.82(1 .63-4.86) 5.78 (0.52-64.3) 2.03(1 .12-3.67) 16.24(2.66-99.1) P-value 0.1918 0.8528 0.5511 0.0154 a Adjusted for race, income, rural residence, schooling, tobacco, and alcohol consumption. b Improvement In goodness offitfor the model specifying an interaction between gender and wood stove exposure. Discussion Wood burning is a major source of energy for indoor cooking and heating in many developing countries. Cooking and heating stoves are used in more than half the world's households and have been shown in many locations to produce high indoor concentrations of particulates, carbon monoxide and other combustion-related pollutants. 17 Wood and coal fires generate a number of combustion products which are known or suspected carcinogenic agents. 18 ' 19 Prompted by the high incidence rates of lung cancer among women, several epidemiological studies have been conducted in Southeast Asia, where the vast majority of rural inhabitants use coal or wood for heating and cooking. These investigations show strong evidence suggesting use of stoves as a risk factor for 20 22 " and in Japan. 23 l u n g cancer, both in China To our knowledge the only published epidemiological study on indoor air pollution and risk of UADT cancers is the one by Dietz et al.24 The authors carried out three individual case-control studies to determine possible risk factors for the development of carcinomas of the larynx, pharynx, and oral cavity in Germany. The risk of UADT cancer was significantly elevated for those subjects using fossil fuels, mainly gas, for heating and cooking for more than 40 years, with OR adjusted for tobacco and alcohol ranging from 2.0, for larynx, to 3.3, for pharynx cancers. 24 Compared with gas stoves, wood stoves typically release 50 times more pollutants during cooking of meals of equivalent size. 25 In the present investigation, we analysed data from a hospital-based case-control study. An important limitation of our study was the fact that it was planned to identify the main ^ k f actO rs assodated with these diseases in the region; the , assess ^door air pollution, d was ad tet0 °^ The exposure assessment was based only on current use of wood stoves for cooking or heating; the interview did not collect information on lifetime or average daily exposure, house ventila^on, o r (jj e presence of a kitchen separated from the rest of the h o u s e T h e p OO r measurement of exposure to smoke from wood stoves probably biased our results towards no assodation between exposure and cancer. An important concern that we had when we postulated our hypothesis was the possibility of confounding. The main determinants for upper aero-digestive tract cancers in the region are tobacco and alcohol.6 Relative risks due to the latter variables frequently exceed 10, when analysed independently, and 100, when effects are assessed jointly. 6 ' 26 At such relative risk levels, 940 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY confounding by tobacco and alcohol becomes a major concern when probing other relations in the data, for example, that of use of wood stove. In the present investigation, we performed an exhaustive control of confounding. Following the change-in-estimate confounder selection criterion, conservative approaches were used to control for the confounding effects of tobacco and alcohol, and their joint exposures, and for other potential confounders. We consistently found significantly increased risks for all UADT cancers and for each specific cancer site studied, despite the conservative strategy to reveal net confounding effects. The similarity of results bolsters the conclusion that the effea for use of wood stove in increasing UADT cancer risk may be genuine and independent from those of other risk factors. In a separate set of analyses, we purposefully biased the association towards the null hypothesis by only adjusting for positive confounders. Even at such an extreme level of conservatism relative risk estimates were consistently and significantly gTeater than one, further supporting our conclusion. The analysis of interaction effects between use of wood stove with smoking or with alcohol did not show conclusive evidence supporting that joint exposures with the latter variables may increase the risk beyond expected levels assuming independence of effects. In regard to gender-specific analysis, our results suggest that women may be at greater risk than men, specially for laryngeal cancer, a site more closely linked to exposures to airborne carcinogens than the oral cavity or the pharynx (because of the cleansing effects of deglutition). This finding is probably related to the fact that women are more exposed to emissions from wood stoves. Analogous results were found in China, where women exposed to emissions from cooking stoves were at higher risks of developing lung cancer than men. 17 Future studies in the region should include a better exposure assessment, in order to examine dose-response relationships. If a causal relation between use of wood stove and risk of developing UADT cancers is eventually established in future studies, we estimate that approximately 32% of all UADT cancers in the region may be attributable to use of wood stoves, given an estimated OR = 2.45 and a prevalence of wood stove use of 33%. Such levels of risk attribution may partially explain the increased incidence of UADT cancers in southern South America. Acknowledgements 3 Muir C, Waterhouse J, Mack T, Powell J, Whelan S Cancer Incidence in Five Continents, Vol V. IARC Scientific Publications 88. Lyon: International Agency for Research on Cancer (IARC), 1987. 4 Rothman KJ, Cann CL Flanders W, Fried MP. Epidemiology of laryngeal cancer. Epidemiol Rev 1980:2:195-209. 5 Blot WJ, McLaughlin JK, Wlnn DM rt al. Smoking and drinking in relation to oral and pharyngeal cancer. Cancer Res 1989:48:3282-87 6 Franco EL, Kowalski LP, Oliveira BV et al Risk factors for oral cancer in Brazil: a case-control study. Int J Cancer 1989:43:992-1000. 7 McLaughlin JK, Gridley G, Block G rt al. Dietary factors in oral and pharyngeal cancer. J Natl Cancer Inst 1988:80:1237^13. 8 Franceschi S, Bidoli E, Bar6n AE rt al Nutrition and cancer of the oral cavity and pharynx in north-east Italy. Int J Cancer 1991:47:20-25. 9 Graham S, Mettlin C, Marshall J, Priore R, Rzepka T, Shedd D. Dietary factors in the epidemiology of cancer of the larynx. Am J Epidemiol 1981:113:675-80. 10 De Stefani E, Correa P, Oreggia F rt al. Black tobacco, wine and mate in oropharyngeal cancer. Rev Epidemiol Sante Pubhque 1988;36: 389-94. 11 Pintos J, Franco EL, Oliveira B, Kowalski LP, Curado MP, Dewar R. Mate, coffee, and tea consumption and risk of cancers of the upper aero-digestive tract in Southern Brazil. Epidemiology 1994,5:583-90. 12 Kodama AM, Dollar AM. Indoor fires as a possible cause of cancers of the upper respiratory and digestive systems in certain underdeveloped countries. J Environ Health 1983:46:88-90. l3 Breslow NE, Day NE. Statistical Methods in Cancer Research, Vol I, The Analysis of Case-Control Studies. IARC Scientific Publication 32. Lyon: IARC, 1980, pp.248-79. 14 Campos-Filho N, Franco EL. A microcomputer program for multiple logistic regression by unconditional and conditional maximum likelihood methods. Am J Epidemiol 1989:129:439-44. 15 Mickey RM, Greenland S. The impact of confounder selection criteria on effect estimation. Am J Epidemiol 1989:129:125-37. 16 Maldonado G, Greenland S. Simulation study of confounder-selection strategies. Am J Epidemiol 1993:138:923-36. 17 Chen BH, Hong CJ, Pandey MR, Smith KR. Indoor air pollution in developing countries. World Health Stat Q 1990:43:127-38. 18 Cooper JA. Environmental impact of residential wood combustion emissions and its implications. J Air Pollut Contr Assoc 1980:30:855-61. l9 Matanoski G, Fishbein L, Redmond C, Rosenkranz H, Wallace L. Contribution of organic particulates to respiratory cancer. Environ Health Prespect 1986:70:37-49. 20 Xu ZY, Blot WJ, Xiao HP et al. Smoking, air pollution, and the high rates of lung cancer in Shenyang, China. J Natl Cancer Inst 1989: 81:1800-06. 21 Liu Z, He X. Chapman RS. Smoking and other risk factors for lung cancer in Xuanwei, China. Int J Epidemiol 1991:20:26-31. The authors are indebted to all the participants in the Ludwig Institute for Cancer Research's Upper Respiratory and Digestive System Cancer Study Group: ClinicaJ Committee: Drs MB Carvalho, A Rapoport, J Andrade-Sobrinho, G Ramos, JL Kanda, JF Gois, JS Chagas, and GA Teixeira; Pathology Committee: Drs H Torloni. WT Vieira, LA Sampaio, and VM Cardoso; Data acquisition and management: ME Silva, RN Pereira, N CamposFilho, L Fanes, VN Souza, and MS Morais. 22 Liu Q, Sasco AJ, Riboli E, Hu MX. Indoor air pollution and lung cancer in Guangzhou, People's Republic of China. Am J Epidemiol 1993:137:145-54. 23 Sobue T. Association of Indoor air pollution and lifestyle with lung cancer cancer in Osaka, Japan. Int J Epidemiol 1990;19(Suppl.l): S62-S66. 24 References Dietz A, Senneweld E, Maier H. Indoor air pollution by emisions of fossil fuels single stoves: possibly a hitherto underrated risk factor in the development of carcinomas in the head and neck. Otolaryngol Head NeckSurg 1995:112:308-15. 25 Smith KR. Indoor air pollution and the pollution transition. In: Kasuga M (ed.). Indoor Air Quality. Berlin: Springer Verlag, 1990, supplement to the Internationa] Archives of Environmental Health. 26 De Stefani E, Correa P, Oreggia F et al. Risk factors for laryngeal cancer. Cancer 1987:60:3087-91. 'Blot WJ, Devesa SS, McLaughlin JK, Fraumeni JF Jr. Oral and pharyngeal cancers. Cancer Surveys 1994,19-20:23—42. 2 Parkin DM, Pisani P, Ferlay J. Estimates of the worldwide incidence of eighteen majors cancers in 1985. Int J Cancer 1993:54:594-606.