2.1.1. Lung cancer

(a) Case–control studies in China (organized from North to South) (see Table 2.1)

The text below summarizes the individual studies from China that assess indoor air pollution resulting from the burning of coal for cooking or heating. Selected results are highlighted in the text below but more detailed information can be found in Table 2.1.

Table 2.1

Case–control studies of lung cancer and use of coal in China.

(i) Northern China

Xu et al. (1989) conducted a case–control study in Shenyang that included 1249 lung cancer cases (729 men, 520 women) and 1345 population-based controls (788 men, 557 women); 86% of male and 55% of female cases and 70% of male and 35% of female controls were tobacco smokers. Pathological or cytological confirmation was obtained for 85.1% and 75.0% of lung cancers in men and women, respectively; 31% of these were adenocarcinoma of the lung. The risk for lung cancer was generally positively associated with exposure metrics reflecting coal use for heating, and, to a more limited degree, with coal use for cooking. After adjustment for age, education and active smoking, lung cancer risk increased in a dose–response fashion with increasing duration of using coal-heated ‘burning kangs’ (beds heated by stoves). Risk also increased with increasing duration of use of a coal stove with pipes to other rooms. Risk was higher when cooking took place in the bedroom or entry corridor to the bedroom than in a separate kitchen or elsewhere in the house. In men, the adjusted odds ratios were 1.0, 1.2 and 2.1 in relation to cooking in the bedroom for 0, 1–29 and ≥30 years, respectively (p for trend <0.05); the corresponding adjusted odds ratios in women were 1.0, 1.5 and 1.8 (p for trend <0.05). [This study overlaps with Sun et al. (1991).]

Wu-Williams et al. (1990) conducted a case–control study of 965 female lung cancer cases in northern China (445 in Harbin, 520 in Shenyang) and 959 female controls (404 in Harbin, 555 in Shenyang); 417 cases and 602 controls were nonsmokers. Seventy-four per cent (714) of the lung cancers were histologically/cytologically confirmed, of which 44% were adenocarcinoma. Cases and controls were compared with respect to indoor coal use (largely use of coal-heated kangs, cooking practices) and other risk factors. In multivariable logistic regression models, risk for lung cancer was generally positively and statistically significantly associated with duration of use of the following: kangs (especially directly-heated ones), coal stoves or floor/wall heating by pipes from the cooking stove.

[The Working Group noted that female lung cancer cases and controls included in the study by Xu et al. (1989) were also included in the study by Wu-Williams et al. (1990). Assessment of cooking practices was relatively limited in these two studies. It was also noted that both studies were large and well-conducted, but assessment of coal-related exposure may have been hampered by lack of exposure contrasts, i.e. coal exposures were mainly compared with exposure to biomass fuels possibly leading to an underestimation of the effect. This comment also applies to Wang et al. (1996) and Zhou et al. (2000).]

Dai et al. (1996) conducted a study of 120 nonsmoking women with adenocarcinoma of the lung and an equal number of nonsmoking population controls; all were long-term residents of Harbin. After adjustment, risk for adenocarcinoma was positively and significantly associated with several metrics of domestic coal use, including having a coal stove in the bedroom, having coal heating and long-term residential ‘exposure to coal dust’. [The Working Group noted that confidence intervals in this report were wide, reflecting the relatively small number of subjects.]

In Shenyang, Wang et al. (1996) compared the experiences of 135 female lifetime nonsmokers diagnosed with primary lung cancer with those of an equal number of nonsmoking population control women. Of the lung cancers included, 57.2% were diagnosed pathologically or cytologically, 54.5% of which were adenocarcinoma. In bivariate analysis, but not multivariable analysis, exposure to coal smoke during cooking was positively and ‘significantly’ associated with lung cancer risk. Other metrics of coal use were not associated with risk. [The Working Group noted that this study was small and the exposure was limited to dichotomized (no/yes) assessment. The specific variables included in the multivariable analysis were not described. The validity of a diagnosis of adenocarcinoma is questionable because the authors stated that determining the histological cell type was based on relevant medical records, chest X-rays, CT films, and cytological and histological slides.]

Two studies in the Kaohsiung area, Taiwan, were reported by Ko et al. (1997) and Le et al. (2001) (described in detail in the monograph on high-temperature frying). Compared with those who did not cook or cooked with gas, the odds ratio for lung cancer was near unity for those who cooked with coal (odds ratio, 1.1; 95% confidence interval [CI], 0.4–3.6) (Ko et al., 1997).

In a subsequent report (Le et al., 2001), the relationship between cooking fuel and risk for lung cancer was examined separately by lung cancer cell type. In the analysis which included 82 squamous-cell and small-cell lung cancers and 129 controls, the odds ratio was 1.2 for use of coal and/or anthracite (95% CI, 0.5–3.0) when compared with women who did not cook or used gas for cooking. In contrast, the risk for adenocarcinoma of the lung (158 cases, 262 controls) increased in relation to use of coal (odds ratio, 2.1; 95% CI, 1.2–3.7). Tobacco smoking, residential area, education and social class were adjusted for in the analysis. [The Working Group noted that information on duration of wood and coal use was not reported in these two studies.]

Zhou et al. (2000) published another report using a subset of women from Wang et al. (1996) in Shenyang. Specifically, 72 women (52 nonsmokers) who had been diagnosed with adenocarcinoma of the lung between 1991 and 1995 were compared with an equal number of control women (49 of whom were nonsmokers). There was no association between coal burning and risk for lung cancer. [The Working Group noted that most of the lung cancer cases and controls included in the analysis by Zhou et al. (2000) were already in the report by Wang et al. (1996). Unadjusted odds ratios were reported. This study was small and the confidence intervals were very wide.]

Kleinerman et al. (2002) conducted a case–control study of lung cancer in relation to household use of coal and biomass fuel in two rural prefectures of Gansu Province, China; about 25% of subjects used coal and most coal users reported using bituminous coal. Of the patients, 846 were deemed to have lung cancer by an expert review panel, and 1740 controls were frequency-matched to patients on gender, age and prefecture of residence. Multivariable logistic regression analyses were conducted separately for men and women. In men, the risk for lung cancer was associated positively and significantly with use of coal (versus use of biomass), the amount of coal used and the percentage of time that coal was most frequently used as fuel in the past 30 years. None of these metrics for coal versus biomass exposure were significantly associated with the risk for lung cancer in women.

(ii) Xuan Wei County, Yunnan Province

Rural Xuan Wei County, Yunnan Province, is impacted by indoor air pollution due to traditionally used fuel types: ‘smoky coal’ (bituminous coal), ‘smokeless coal’ (anthracite) and wood. The great majority of residents were farmers, and residential stability was very high. There have been few stationary or mobile sources of outdoor air pollution.

Liu et al. (1991) and He et al. (1991) reported a case–control study in Xuan Wei that included 110 incident cases of lung cancer (56 men, 54 women) identified in regional hospitals/clinics in 1985–86, and 426 population controls (224 men, 202 women). Cases and controls were matched on gender, age, occupation (all were farmers) and village of residence. [Matching on village of residence overmatched on type of indoor fuel and type of home.] ‘Smoky’ coal was used at least four times more often than wood in this population. While duration and frequency of cooking food were significantly associated with risk for lung cancer in an exposure–response manner after adjustment for other risk factors, this precluded assessment of risk for lung cancer in relation to specific fuel types. [Although the analyses were duplicated in both studies, the Working Group noted discrepancies in the results. The instability of the risk estimates is demonstrated with the different choice of category cut-points in both studies. The Working Group also noted that inferences from these two studies are limited by the relatively small sample size and the uncertainty of the significance of the reference group.]

Another case–control study conducted among female farmers in Xuan Wei (Lan et al., 1993) was based on 139 incident female lung cancers that were diagnosed between 1988 and 1990 and 139 age-matched controls. Of the lung cancer cases, 55 (39.6%) were diagnosed cytologically/ pathologically. All cases and controls were current nonsmokers but all were former smokers. Use of smoky coal from the Laibin mine was significantly associated with risk for lung cancer by frequency and duration of use. [Although all participants were former smokers, the amount of past smoking was not specified nor was it considered as a potential confounder in the analysis.]

Lan et al. (2000) carried out a population-based case–control study of 122 cases and 122 controls. This study was designed to evaluate the relationship between genetic susceptibility and lung cancer. Controls were individually matched to cases by sex, age, village and type of fuel currently used for cooking and heating. Compared with subjects whose cumulative smoky coal use was less than 130 tonnes, subjects who used more than 130 tonnes of smoky coal had a 2.4-fold increased risk for lung cancer (95% CI, 1.3–4.4; 71 exposed cases; adjusted for total smoky coal use without ventilation, pack–years of smoking, chronic obstructive pulmonary disease and family history of lung cancer). [The Working Group noted that even with matching on fuel type, the study observed a cumulative effect of smoky coal.]

(iii) Central China excluding Xuan Wei

In Chengdu, Sichuan Province, Huang et al. (1992) performed a case–control study of 135 ‘pre-invasive’ lung cancer patients drawn from three provincial hospitals and 135 healthy controls individually matched to cases on gender and age. Controls were enlisted from persons coming to the same three hospitals for routine health check-ups and matched on residential area. The primary goal of this study was to assess dietary risk factors for lung cancer. Burning coal indoors was associated with a statistically significantly higher risk for lung cancer than not burning coal indoors (Odds ratio, 1.59; 95% CI, 1.01–2.07). [It is unclear what the exposure of the reference group was, but it may have been biomass.]

Shen et al. (1998) conducted a case–control study in women in Nanjing that included 70 never-smoking lung cancer patients and 70 healthy community controls, matched 1:1 with cases on gender, age, neighbourhood and occupation. Subjects appear to have been a subset of those in Shen et al. (1996). Use of solid fuel (versus non-solid fuel) and of coal stoves was assessed, together with cooking-related metrics and other covariates. Use of a coal stove for heating was marginally significantly associated with lung cancer risk (odds ratio, 1.78; 95% CI, 0.79–4.02, p=0.08). [The Working Group noted several limitations in this study. The report lacked details regarding the study design (e.g. response rate), characteristics of the study population (e.g. gender distribution, active smoking history) and covariates included in the statistical models.]

Zhong et al. (1999) conducted a case–control study in Shanghai that included a total of 649 women who had been diagnosed with incident lung cancer during 1992–94 and 675 population controls. Subjects who had smoked at least one cigarette a day for at least 6 months (145 cases, 74 controls) were excluded from the analyses. Thus, results were based on 504 cases and 601 controls who were lifetime nonsmokers. Seventy-seven per cent of the lung cancers were diagnosed histologically or cytologically and 76.5% (n=296) of these were adenocarcinoma. The analysis explored cooking-related associations in more detail than those related to fuel. In multivariable logistic regression models, coal and gas use versus coal only was not associated with risk for lung cancer (odds ratio, 0.92; 95% CI, 0.63–1.35) but kitchen smokiness during cooking was positively and significantly associated with risk for lung cancer in a dose–response manner. [The Working Group noted that the contrast was between coal and gas versus coal only and therefore does not address the risk associated with exposure to coal.]

(iv) Southern China

Koo et al. (1983) reported a case–control study in Hong Kong Special Administrative Region that included 200 women hospitalized with lung cancer and 200 community controls matched on gender, age (±5 years), residential district and type of housing. Data were obtained by in-person interview using a semi-structured questionnaire and taking a life-history approach. The investigators assessed use and duration of use of biomass fuels, coal, kerosene, LPG and gas. Only 12 of 400 subjects (3%) used coal as cooking fuel—this proportion was 1.5% in cases and 4.5% in controls. The few subjects who had ever used coal had used it only when they had lived in mainland China but discontinued its use when they moved to Hong Kong Special Administrative Region. Unadjusted relative risks were calculated for matched and unmatched data. Use of coal was inversely and non-significantly associated with the risk for lung cancer (odds ratio, 0.32; p=0.15). [The Working Group noted that, in view of the few subjects using coal (three cases, nine controls), this study probably did not allow a very informative test of lung cancer risk in relation to coal use.]

Liu et al. (1993) conducted a case–control study in Guangzhou during 1983–1984 that included 316 incident lung cancer cases (224 men, 92 women) and 316 hospital controls individually matched by gender, age (±2 years), residential district and date of diagnosis. Controls were not chosen from the Tumor Hospital or Chest Hospital and those with respiratory and coronary heart disease were further excluded. Analyses using multivariable conditional logistic regression models adjusted for education, occupation, occupational exposure, history of tuberculosis, chronic bronchitis, family history of cancer, smoking, living area and passive smoke (in women only) were stratified by gender. Use of coal for cooking was the reference category (odds ratio, 1.0) for examining the risk for lung cancer associated with other cooking fuels (gas, wood); 89% (200/224) of men and 88% (81/92) of women had ever used coal for cooking.

Du et al. (1996) reported another case–control study in Ghangzhou. Cases were 849 deceased lung cancer patients (566 men, 283 women) who died in 1985 and controls were persons who died of causes unrelated to lung cancer, matched with cases on gender, age (±2 years) and residence. A standardized questionnaire was used to interview next of kin of the lung cancer patients [and presumably the controls]. Lung cancer risk in women was positively and significantly associated with smoking and ‘exposure to coal fumes’. Risk in men was significantly associated only with smoking.

Luo et al. (1996) reported a case–control study in Fuzhou that included 102 lung cancer cases (78 men, 24 women) and 306 community controls, matched with cases on gender, age and ethnicity. Data were analysed using conditional logistic regression, and separate analyses were conducted for squamous-cell carcinoma and adenocarcinoma. The presence of smoke in the living room during cooking with coal was positively and statistically significantly associated with risk for lung cancer (odds ratio, 7.6; 95% CI, 3.7–15.7 for all lung cancers; odds ratio, 14.1 [95% CI, 11.8–16.4]; p=0.026 for squamous-cell carcinoma; odds ratio, 6.0 [95% CI, 4.9–7.1]; p=0.002 for adenocarcinoma). [The Working Group noted that smoking was significantly associated with the risk for squamous-cell carcinoma, but only moderately for adenocarcinoma (not statistically significant).]

(v) Taiwan

Ger et al. (1993) conducted a case–control study in Taipei that included 131 primary lung cancers (92 men, 39 women) identified between 1990 and 1991. All were histologically confirmed (59 squamous-cell/small-cell carcinoma, 72 adenocarcinoma). Two control groups were interviewed: 262 hospital controls were matched to cases on sex, date of birth (±5 years), date of interview and insurance status whereas 262 neighbourhood controls were matched to cases on age, sex and residence of the case at the time of diagnosis. When lung cancer cases were compared with neighbourhood controls, use of coal for cooking was unrelated to risk for adenocarcinoma (odds ratio, 0.56; 95% CI, 0.20–1.54; adjusted for matching factors; seven exposed cases), but was strongly associated with risk for squamous-cell/small-cell carcinoma (odds ratio, 10.00; 95% CI, 2.19–45.61; adjusted for matching factors; 10 exposed cases). In multivariable analysis, use of coal as a cooking fuel remained a significant risk factor for squamous-cell/small-cell carcinomas. The magnitude of effect was smaller but also statistically significant when cases were compared with hospital controls. [The Working Group noted that this study included few female nonsmoking lung cancer patients: 48 cases compared with 229 controls (111 hospital controls, 118 neighbourhood controls) were nonsmokers. It is not clear whether the results for coal burning related to usual, past or current practices. The prevalence of coal burning differed substantially between the neighbourhood controls who were matched to the adenocarcinoma (14.6%) or the squamous-cell lung cancer cases (1.7%). The corresponding figures for the hospital controls selected for adenocarcinoma and squamous-cell lung cancer patients were 7.6% and 5.1%, respectively. Thus, the significantly increased risk for squamous-cell/small-cell cancers associated with coal burning may be related to differences among the control subjects. Furthermore, the Working Group also noted that a higher percentage of proxy interviews was conducted for cases (21%) than controls from the neighbourhood (17%) or hospitals (12%), and that matching on location of residence may have influenced effect estimates for fuel type due to overmatching.]

The study by Le et al. (2001) included lung cancer patients who had been diagnosed between 1993 and 1999. Women diagnosed with squamous-cell or small-cell carcinoma (n=84) or adenocarcinoma of the lung (n=162) and corresponding controls (n=407) were included in the analysis. Women with other lung cancer cell types (n=45 cases) were excluded. Risk for lung cancer was associated with type of cooking fuel: women who used coal or anthracite as a cooking fuel versus those who used gas or no fuel had a significantly increased risk for adenocarcinoma (odds ratio, 2.1; 95% CI, 1.2–3.7; 49 exposed cases) but not for squamous- or small-cell carcinoma (odds ratio, 1.2; 95% CI, 0.5–3.0; 14 exposed cases).

A third study by this research group (Ko et al., 2000) addressed lung cancer and cooking in some detail, but did not specifically address fuel or fuel smoke. This study is described in the monograph on high-temperature frying.

(b) Case–control studies outside China (see Table 2.2)

Table 2.2

Case–control studies of lung cancer and use of coal outside China.

Wu et al. (1985) conducted a case–control study among white women in Los Angeles County, CA, USA. One hundred and forty-nine cases of adenocarcinoma and 71 cases of squamous-cell carcinoma of the lung identified from population-based tumour registry and a group of age-matched neighbourhood controls were interviewed by telephone. No information on the number of years that coal was used was available. Exposure to burning coal (used for heating or cooking) during the majority of childhood and teenage years increased the risk for lung cancer: the odds ratio for adenocarcinoma was 2.3 (95% CI, 1.0–5.5) and that for squamous-cell carcinoma was 1.9 (95% CI, 0.5–6.5) after adjusting for tobacco smoking. An increased risk for lung adenocarcinoma was seen when stratified by smoking status (odds ratio for nonsmokers, 3.2; 95% CI, 0.9–11.8; odds ratio for former smokers, 4.3; 95% CI, 1.0–17.8; odds ratio for current smokers, 9.5; 95% CI, 2.1–41.9). Multivariable logistic regression that adjusted for personal smoking, childhood pneumonia and β-carotene intake produced similar results.

Sharpe et al. (1989) conducted a case–control study of renal-cell carcinoma in Montreal, Canada. One hundred and sixty-four histologically confirmed cases of renal-cell carcinoma, diagnosed between 1982 and 1987 in four hospitals in the Montreal area, who responded to a mailed questionnaire were included in this analysis. One hundred and sixty-one controls without urinary tract tumours who were identified from urology files and who were matched to cases on sex, date of birth and urologist were also included in the analysis. Those who had been exposed indoors to burning coal had a non-statistically significantly increased risk for lung cancer compared with subjects who were not exposed. For those who lived in a house where coal was used as a fuel but did not handle it, the odds ratio was 1.07 (95% CI, 0.58–1.96; 53 exposed cases). The odds ratio for subjects who had handled coal only domestically was 1.41 (95% CI, 0.76–2.62; 56 exposed cases). [The covariates included in the multivariable logistic regression models were not specified. Four hundred and three cases were identified but only 168 cases and matched controls agreed by telephone to receive a mailed questionnaire. Of those who received a mailed questionnaire, 164 cases and 161 controls responded.]

Malats et al. (2000) conducted a multicentre case–control study in eight countries (Brazil, France, Germany, Italy, Poland, Romania, the Russian Federation, Sweden) that primarily investigated the interaction between glutathione S-transferase (GST) M1 and T1 genotypes and environmental risk factors in 122 lung cancer cases (14% male) and 121 controls (58 population-based, 63 hospital-based). All cases were confirmed by histology or cytology. Information on exposure was obtained through personal interview using a standardized questionnaire. Indoor air pollution from the use of wood or coal for cooking or heating was dichotomized according to the median number of years of exposure to both sources of combustion among controls. Using coal for more than 17 years (versus less than 17 years) was not associated with an increased risk for lung cancer (odds ratio, 0.4; 95% CI, 0.1–1.1). [The Working Group noted that there could be possible confounding by smoking because nonsmokers were defined as occasional smokers (up to 400 cigarettes in a lifetime) and never smokers. The Working Group also noted the heterogeneous background of the study subjects and the fact that controls were not matched by age and sex may present difficulties in the interpretation of the results. Furthermore, the crude exposure indices used (i.e. <20 years of fuel use) did not enable a good assessment of the exposure–response relationship.]

Two hundred and sixty-five histologically confirmed, incident cases of lung cancer (235 men, 30 women) who were seen at the Department of Pulmonary Medicine, Chandigarh, India, and 525 age- and sex-matched controls (435 men, 90 women) who were selected among visitors and attendants of the patients were recruited in a case–control study conducted between January 1995 and June 1997 (Gupta et al., 2001). Trained interviewers collected information on demographic factors, lifetime tobacco smoking history, detailed occupational history and residence. Exposure to indoor air pollution was assessed on the basis of the type of fuel used for cooking or heating and the number of years spent in that household. Unconditional logistic regression models stratified by gender were adjusted for age, cumulative tobacco consumption and education. The odds ratio for exposure to indoor air pollution as measured by 45 or more years of exposure to coal for heating was 1.20 (95% CI, 0.75–1.91; 42 exposed cases) for men and 1.12 (95% CI, 0.26–4.84; 5 exposed cases) for women. The odds ratio for the use of coal for cooking for 45 or more years was 0.88 (95% CI, 0.49–1.57; 23 exposed cases) for men and 1.52 (95% CI, 0.33–6.98; six exposed cases) for women.

Lissowska et al. (2005) conducted a multicentre case–control study during 1998–2002 in six eastern and central European countries (Czech Republic, Hungary, Poland, Romania, the Russian Federation and Slovakia) and the United Kingdom to examine the association between burning coal and unprocessed biomass and the incidence of lung cancer in men and women. Cases included 2861 histologically or cytologically confirmed incident lung cancer patients (2205 men, 656 women) who were identified through the main hospitals in the 15 participating centres. A total of 3118 (2305 men, 813 women) hospital-based (13 centres) and population-based controls (two centres), who were frequency-matched to cases by geographic area, 5-year age group and gender, were studied. A common structured questionnaire was used to collect information on risk factors for lung cancer such as active and passive tobacco smoking, occupational history, lifetime residential history and fuel use at every residence of at least 1 year. The study examined risk patterns in relation to modern non-solid fuels (gas, kerosene and electricity) versus traditional solid fuels (coal and biomass, mainly wood) used for cooking and heating after adjusting for centre, age, gender, education and tobacco pack–years. The odds ratios for exposure to coal were 1.13 (95% CI, 0.94–1.38; 872 exposed cases) for cooking and 1.08 (95% CI, 0.89–1.31; 772 exposed cases) for heating. [The Working Group noted many strengths in this large multicentre case–control study which used a common, standardized study protocol and questionnaire and collected information on lifetime fuel use and relevant covariates from in-person interviews with the study participants. The response rate was high in both cases and controls (>90%). Although the analysis was very thorough, exposure–response analyses by type of solid fuels (separately for coal and wood) were not provided.]

(c) Cohort studies (see Table 2.3)

Table 2.3

Cohort studies of use of coal and cancer in China.

Lan et al. (2002) followed a cohort of 21 232 farmers (11 168 men, 10 064 women) in China retrospectively from 1976 to 1992. The farmers were born between 1917 and 1951 into homes in which smoky coal and unvented stoves were used; however, 17 184 of these subjects (80.9%) later changed permanently to the use of vented stoves with chimneys. (The Chinese Government offered a small subsidy for the purchase of stoves with chimneys in 1976). Nearly all subjects were born in Xuan Wei (about 1% of men and 13% of women were born outside the immediate study area). In multivariable Cox proportional hazard models stratified by gender, in which stove improvement, duration of cooking (for women only) and duration of smoking (for men only; only 0.9% of women had ever smoked) were treated as time-dependent covariates, stove improvement was associated with a statistically significant reduction in the incidence rate of lung cancer (hazard ratio for men, 0.59; 95% CI, 0.49–0.71; hazard ratio for women, 0.54; 95% CI, 0.44–0.65). In both men and women, duration of cooking food was also positively and significantly associated with an increased risk for lung cancer, as was the daily average number of hours spent indoors through to the age of 20 years. [The Working Group noted that although fewer men than women cooked food, the significantly protective effect of stove improvement was of similar magnitude in both genders, which suggests that the protective effect of stove improvement related largely to the reduction of cooking fuel smoke.]

2.1.2. Cancer of the salivary glands

All residents of urban Shanghai aged 20–75 years who were newly diagnosed with cancer of the salivary glands (International Classification of Diseases [ICD]-9, 142) during the period from 1 January 1988 to 28 February 1990 were eligible to participate in a case–control study (Zheng et al., 1996). A total of 44 eligible cases (19 men, 25 women) were identified from the Shanghai population-based cancer registry during the period in question. Of all identified cases, 41 (93.2%) were interviewed and three other cases could not be located or were too ill to be interviewed. Adenocarcinoma and adenoid cystic carcinoma were the two major cancers diagnosed and accounted for 46.3% and 24.4% of total cases, respectively. Controls were randomly selected from the general population of urban Shanghai by use of a frequency-matching method in accordance with the sex–age distribution of cases of all head and neck cancers reported to the Shanghai cancer registry during 1985–86. A total of 462 controls were selected, among whom 414 (89.6%) were interviewed. Information on demographic factors, tobacco and alcohol consumption, dietary habits, lifetime job history, occupational and household exposures and previous disease history was collected from each study subject. After adjusting for gender, age and income, the odds ratio was 1.6 for the use of coal for cooking (95% CI, 0.5–5.6; 38 exposed cases). [The Working Group noted that the sample size was relatively small.]

2.1.3. Cancer of the oesophagus

A nested case–control study within a cohort of workers of an iron–steel complex was carried out in Anshan, China, to evaluate the relationship of oesophageal cancer with occupational exposure to silica and other dusts, taking lifestyle exposure factors into consideration (Pan et al., 1999). A total of 141 men who were confirmed as having died from oesophageal cancer during 1980–88 were selected as cases. Two male controls were randomly selected and matched on age (within 5-year age groups) for each case from the death registry file of the company over the same period. The first control group consisted of workers who had died of diseases other than cancer or respiratory or digestive diseases (non-cancer controls) and the second control group consisted of workers who had died of cancers other than of the stomach or respiratory system (cancer controls). The number of cases whose relatives could be interviewed was 125 (88.7%). Each of the two groups of controls consisted of the same number of subjects (125) as the cases. Either the wife or a first-degree relative was interviewed for 95.2% of cases and 96.0% of controls. Information was obtained for smoking, drinking, diet, method of cooking and heating in the household and lifetime occupational history. A job-exposure matrix was applied to lifetime job histories and dust exposures were categorized into no exposure, refractory silica dust, other silica dust, iron dust, founding dust, coal dust, wood dust, welding dust and other dusts. The results were presented for both controls combined since the results of the analysis using two control groups were very similar. In a univariate analysis, occupational exposure to silica dust, “other silica dust” and “any dust”, domestic exposure to coal heating, cooking with coal, heavy smoking, alcohol drinking and consumption of salted vegetables were shown to be risk factors. Central heating, cooking with gas and consumption of fish, meat, eggs and fruit were found to be protective factors. In multivariate analysis, the odds ratio was 2.01 for cooking with coal (95% CI, 1.09–3.70). Exposure to silica dust for 25 years and more and cooking with coal for 20 years and more gave the highest risk: the odds ratio for the former was 8.87 (95% CI, 1.67–47.08) and that for the latter was 2.48 (95% CI, 1.29–4.78). [The Working Group noted that the findings were among long-term male steelworkers and not the general population, thus making extrapolations to other groups difficult. Comparison of deceased cases with deceased controls and the use of information obtained from relatives, not from subjects themselves, are the weakness of the study.]

2.1.4. Cancer of the nasal cavity, paranasal sinuses and middle ear

Cases of cancer of the nasal cavity, paranasal sinuses and middle ear (ICD-9, 160), aged 20–75 years, who had been newly diagnosed during the period from 1 January 1988 to 28 February 1990 were eligible to participate in a case–control study (Zheng et al., 1992). A total of 63 cases were identified from the Shanghai population-based cancer registry during the study period. Of these, 60 cases (39 men and 21 women) were interviewed and three cases (4.8%) could not be located; 51 cases (85%) were pathologically diagnosed. Controls were randomly selected from the general population of the Shanghai urban area by use of frequency-matching in accordance with the sex–age distribution of incident cases of oral, pharyngeal, laryngeal and nasal cancers reported to the Shanghai cancer registry in 1985–86. A total of 462 controls were identified and 414 (89.6%) were interviewed. Information on demographic factors, tobacco and alcohol consumption, dietary habits, occupational and household exposures and previous disease history was collected. Unconditional logistic regression was used to adjust for confounders and to calculate adjusted odds ratios. After adjusting for age, the odds ratio for coal used as cooking fuel was 1.1 (95% CI, 0.5–2.6; 53 exposed cases).

2.1.5. Cancer of the cervix (see Table 2.3)

Wu et al. (2004) conducted a nested case–control study of 100 women with cervical cancer and 197 population controls selected from a cohort of 32 466 women who underwent Papanicolaou (Pap) smear screening in Chai-Yi City, Taiwan, China. Use of coal compared to gas was positively, although not significantly, associated with the risk for cervical cancer (odds ratio, 2.09; 95% CI, 0.86–5.10; 21 exposed cases).

2.1.6. Cancer of the kidney (renal-cell carcinoma) (see Table 2.4)

Table 2.4

Case–control study of renal cancer and coal use.

Sharpe et al. (1989) conducted a hospital based case–control study in Montreal, Canada, that included 164 cases of renal-cell carcinoma (62% men) identified retrospectively from medical records and 161 age- and sex-matched controls who attended the same urologists; all were given mailed questionnaire with telephone follow-up. Cases were survivors and had less advanced disease. Occupational exposure to burning coal showed a dose–response relationship with renal- cell carcinoma by duration in months (p<0.05) Intensity of exposure, assessed by combining domestic and occupational exposures in a hierarchical ordinal manner, also showed a dose–response trend (p<0.025). Multivariable analysis adjusting for sex and age (smoking was not a statistically significant confounder and was removed from final model) showed that handling coal in a setting where it was being burned between the ages of 10 and 24 years was a risk factor. [Results on the duration of occupational exposure or the intensity of exposure (combining domestic and occupational) were not reported. In general, the results were not presented in a clear or understandable manner.]

2.2.1. Cancer of the lung (see Table 2.5)

Table 2.5

Case–control studies of lung cancer and use of biomass fuel.

Koo et al. (1983) (described in detail in Section 2.1.1) reported a case–control study in Hong Kong Special Administrative Region, China. Use of wood/grass did not have any appreciable effect on the risk for lung cancer (odds ratio, 0.74; p=0.50).

Sobue (1990) conducted a hospital-based case–control study of lung cancer in Osaka, Japan, that included 144 nonsmoking female lung cancer patients and 731 nonsmoking female controls. Newly admitted patients were asked to complete a questionnaire that asked about tobacco smoking habits, exposure to environmental tobacco smoke and other sources of indoor air pollution (i.e. use of straw or wood for cooking, use of heating appliances fueled with kerosene, gas, coal and charcoal and wood stoves without chimneys, and use of charcoal footwarmers). The odds ratio for lung cancer in nonsmoking Japanese women who used wood or straw compared with those who did not at age 30 years was 1.77 (95% CI, 1.08–2.91; 32 exposed cases; adjusted for age, exposure to passive smoke in adulthood and mother smoked during childhood). Current and previous (at ages 15 and 30 years) use of charcoal footwarmers for sleeping was not associated with risk. [The Working Group noted that the exposure variables were not well defined and were limited to dichotomized categorization. The reference category was not clearly defined and may have included the use of coal.]

In the study of Wu-Williams et al. (1990) (described in detail in Section 2.1.1) duration of use of non-coal, generally biomass-fueled stoves was not associated with risk, and there was no discernible association of risk with central heating (in which the heating fuel source was outside the home).

In a hospital-based case–control study conducted in Guangzhou, China, Liu et al. (1993) studied the risk for lung cancer associated with the use of different cooking fuels (described in Section 2.1.1). Analyses using multivariable conditional logistic regression models adjusted for education, occupation, occupational exposure, history of tuberculosis, chronic bronchitis, family history of cancer, smoking, living area and passive smoke (in women only) were stratified by gender. Compared with use of coal for cooking, use of wood was inversely associated (although not statistically significant) with the risk for lung cancer in both men (odds ratio, 0.57; 95% CI, 0.11–3.0; eight exposed cases) and women (odds ratio: 0.67; 95% CI, 0.04–11.7; three exposed cases).

Two hospital-based case–control studies were conducted in Kaohsiung, a heavily industrialized city in Taiwan, China (Ko et al., 1997; Le et al., 2001). The first study included 117 female lung cancer cases identified between 1992 and 1993 who were compared with 117 female hospital controls admitted for a health check-up (n=55) or for eye diseases (n=62) (Ko et al., 1997). Information on histological type was not provided. Active smokers (11 cases, three controls) were excluded so that the analysis was based on 105 case–control pairs who were nonsmokers. Use of wood or charcoal before the age of 40, as opposed to other fuels including coal, was associated with an increased risk for lung cancer, after adjusting for socioeconomic status, education and residential area.

The case–control study by Le et al. (2001) included women diagnosed with squamous-cell or small-cell carcinoma (n=84) or adenocarcinoma of the lung (n=162) and corresponding controls (n=407) (described in detail in Section 2.1.1). Risk for lung cancer was associated with type of cooking fuel: women who used wood or charcoal as a cooking fuel compared with those who cooked with gas or did not cook showed a 3.1-fold (95% CI, 1.0–9.2; 22 exposed cases) increased risk for squamous-cell and small-cell cancer and a 3.0-fold (95% CI, 1.4–6.4; 40 exposed cases) increased risk for adenocarcinoma. Risk was also significantly higher for those who cooked in a kitchen without a fume extractor: the odds ratio was 3.0 (95% CI, 1.3–7.1; 31 exposed cases) for squamous-/small-cell cancer and 3.9 (95% CI, 2.3–6.6; 74 exposed cases) for adenocarcinoma. Only 7% of men reported cooking for the family and thus these data were not reported.

In a hospital-based case–control study among nonsmokers (including occasional smokers of up to 400 cigarettes in a lifetime), Malats et al. (2000) (described in detail in Section 2.1.1) presented an overall odds ratio for lung cancer for >20 years of use of wood of 2.5 (95% CI, 1.0–6.2). [The Working Group noted that there could be possible confounding by smoking because nonsmokers were defined as never smokers and occasional smokers (up to 400 cigarettes in a lifetime). The Working Group also noted the heterogeneous background of the study subjects and the fact that controls were not matched by age and sex may present difficulties in the interpretation of the results. Furthermore, the crude exposure indices used (i.e. <20 years of fuel use) did not enable a good assessment of the exposure–response relationship.]

In a case–control study of lung cancer in Chandigarh, India, Gupta et al. (2001) (described in detail in Section 2.1.1) reported that the odds ratio for the use of wood for heating for 45 or more years was 0.97 (95% CI, 0.65–1.43; 67 exposed cases) for men and 2.78 (95% CI, 0.97–7.98; 13 exposed cases) for women. The odds ratio for the use of wood for cooking for 45 or more years was 0.87 (95% CI, 0.58–1.30; 105 exposed cases) for men and 1.11 (95% CI, 0.34–3.60; 12 exposed cases) for women. [The Working Group noted that the selection of controls may have resulted in overadjustment.]

In a study conducted in Mexico, Hernández-Garduño et al. (2004) determined the association between long-term exposure to wood smoke from cooking and lung cancer in nonsmoking Mexican women. Cases and controls, aged 44 years or more, were identified through a review of patient records (discharged between 1986 and 1994) at the Instituto Nacional de Enfermedades Respiratorias, a specialized hospital for respiratory diseases in Mexico City. All cases (n=113) were nonsmoking women with a histological confirmation of lung adenocarcinoma. Controls (n=273) were hospitalized at the same institute during the same period of time for pulmonary tuberculosis (n=99), interstitial lung disease (n=110) and other miscellaneous pulmonary conditions (n=64), of which 55 were pneumonia. Information on environmental exposures including wood smoke (ever used wood for cooking in their household, years of exposure) was obtained by personal interview at admission and abstracted from medical records for this study. Potential cases and controls with no information on cooking fuel exposure or socioeconomic status were excluded [numbers not given]. Of the patients, 75% were currently living in Mexico City or in the state of Mexico while the remainder lived in other states of Mexico. Cases were slightly older and more likely to come from rural areas and had lower socioeconomic status than controls, although the difference was not significant. The majority of women reported some use of wood for cooking during their lifetime (68.1% of cases; control groups: 67.7% with tuberculosis, 67.0% with interstitial lung disease, 62.5% with miscellaneous pulmonary conditions and 66.1% of the combined control group). For those who had ever used wood for cooking, the duration of exposure to wood smoke was significantly higher in the cases (median, 56 years) than in each of the control groups (median for the combined controls, 38 years). The percentage of women exposed for 1–20 and 20–50 years to wood smoke was higher in the control groups; however, the percentage of exposure over 50 years was higher in the cases. In a multivariate analysis adjusting for age, exposure to environmental tobacco smoke, education and socioeconomic status, the odds ratio for lung cancer for more than 50 years of exposure to wood smoke was 1.9 (95% CI, 1.1–3.5) compared with all control groups combined. This risk was higher when compared with the control group of miscellaneous pulmonary conditions (odds ratio, 2.6; 95% CI, 1.0–6.3). However, for the duration of exposure of 1–20 and 21–50 years, the odds ratios were less than 1 and not statistically significant. [The Working Group noted that the use of controls whose disease might be related to wood exposure could underestimate the relative risk but may also reduce interviewer bias. No information was provided on the cases and controls that were excluded due to lack of information on exposure to wood smoke. Since coal is not used in Mexico, confounding by the use of coal is not a concern in this study. The dose–response results are difficult to interpret with the use of multiple control groups.]

In a multicentre case–control study of lung cancer, Lissowska et al. (2005) (described in detail in Section 2.1.1) reported odds ratios for principal exposure to wood for cooking (odds ratio, 1.23; 95% CI, 1.00–1.52; 1065 exposed cases) and heating (odds ratio, 1.31; 95% CI, 1.06–1.61; 1105 exposed cases) after adjusting for centre, age, gender, education and tobacco pack–years. [The Working Group noted many strengths in this large multicentre case–control study which used a common, standardized study protocol and questionnaire and collected information on lifetime fuel use and relevant covariates from in-person interviews with the study participants. The response rate was high in both cases and controls (>90%). Although the analysis was very thorough, dose–response analyses by type of solid fuels (separately for coal and wood) were not provided.]

2.2.2. Cancers of the oral cavity, pharynx and larynx (see Table 2.6)

Two partially overlapping hospital-based case–control studies conducted in Brazil (São Paulo, Curitiba and Loiania) examined the risk for cancers of the larynx, pharynx (excluding nasopharynx) and mouth (excluding salivary glands) in relation to the use of wood stoves for cooking and/or heating (Franco et al., 1989; Pintos et al., 1998). Franco et al. (1989) compared 232 oral cancer cases with 464 non-cancer control patients matched by age, sex, study site and trimester of admission and reported an odds ratio for ever/never use of wood stove adjusted for alcohol and tobacco use of 2.5 (95% CI, 1.6–3.9; 134 exposed cases). Pintos et al. (1998) identified 784 incident cases of cancer of the larynx, pharynx and mouth and 1568 controls from hospital inpatients (patients with cancer or mental disorders were excluded), matched to cases on age, sex, study site and trimester of hospital admission. Ever/never use of wood stoves, adjusted for tobacco and alcohol consumption, was associated with an increased risk for cancer at all sites (odds ratios: for all sites, 2.39; 95% CI, 1.88–3.05; mouth, 2.34; 95% CI, 1.67–3.29; pharynx, 2.78; 95% CI, 1.70–4.53; larynx, 2.37; 95% CI, 1.40–4.02). Presenting results of wood stove exposure by levels of tobacco and alcohol consumption resulted in statistically significant odds ratios above 2.0. [The increased risk seen in the groups that consumed the least alcohol and tobacco suggested that residual confounding was unlikely.] Stratifying by gender and subsite while also adjusting for race, income, rural residence and schooling resulted in statistically significant odds ratios above 2.0 for men while the point estimates varied considerably with wide confidence intervals for women because the data for women were sparse (adjusted odds ratio for laryngeal cancer: in men 2.03; 95% CI, 1.12–3.67; in women, 16.24; 95% CI, 2.66–99.1). [The Working Group noted that exposure assessment was crude; use of wood stoves for cooking or heating was ascertained only via a single yes/no question. No attempt was made to determine duration of use and the reference group was not described.]

2.2.3. Cancer of the salivary glands

In a case–control study of cancer of the salivary glands, Zheng et al. (1996) (described in detail in Section 2.1.2), the odds ratio was 1.6 for use of wood/straw for cooking (95% CI, 0.6–4.4; six exposed cases) after adjusting for gender, age and income.

2.2.4. Cancer of the nasopharynx (see Table 2.7)

Table 2.7

Case–control studies of nasopharyngeal cancer.

The earliest publication on cancer and indoor pollution from biomass fuel focused on nasopharyngeal cancer using an ecological comparison of rates by elevation in rural Kenya (Clifford, 1972). A small (eight-household) indoor air pollution survey during evening cooking was undertaken to indicate that concentrations of total PM, total organic matter, benzo[a]pyrene and benz[a]anthracene from wood burning varied by elevation, presumably because of lower ventilation and more fuel use due to the need for more space heating at higher elevations. The incidence of nasopharyngeal cancer varied by region.

Shanmugaratnam et al. (1978) conducted a case–control study of nasopharyngeal carcinoma (NPC) among persons of Chinese origin who were permanent residents of Singapore. A total of 379 histologically confirmed NPC patients (266 men, 113 women) were recruited from the Ear, Nose and Throat (ENT) Department of Singapore General Hospital. Two control groups were enrolled: 595 (311 men, 284 women) ENT patients without NPC and 1044 (738 men, 306 women) other hospital controls. Trained interviewers conducted in-person interviews between March 1966 and August 1968 using a standardized questionnaire that included the main type of fuel used over a period of more than 10 years. Charcoal use was not significantly associated with the risk for NPC (odds ratio, <1.0; p>0.05) using either control group as a comparison. Firewood use was significantly associated with NPC compared with ENT controls (odds ratio, 1.71; p<0.01) but not when compared with other hospital controls (odds ratio, 0.87; p>0.05). [The Working Group noted that a greater proportion of the ENT controls may have used gas instead of firewood since this group had a higher socioeconomic status.]

In a case–control study (Zheng et al., 1992) of cancer of the nasal cavity, paranasal sinuses and middle ear (ICD-9, 160) (described in detail in Section 2.1.4) after adjusting for age, ever-use of wood or straw as cooking fuels was associated with a significantly increased risk for nasal cancer (ICD-9, 160) (odds ratio, 3.3; 95% CI, 1.7–6.7). In multivariate analysis after adjusting for related variables such as age, intake of oranges/tangerines, consumption of salted fish/meat/vegetables, ever exposure to wood/silica/petroleum products and ever diagnosed with chronic nasal diseases, ever use of wood/straw as cooking fuels was associated with a risk for nasal cancer and yielded an odds ratio of 3.3 (95% CI, 1.5–7.3). Consistent associations in both men and women were found for use of wood/straw as cooking fuels (odds ratio for men, 2.2; 95% CI, 0.9–5.4; odds ratio for women, 8.5; 95% CI, 2.3–31.8). [The Working Group noted that the sample size was relatively small. The variable use of wood/straw as cooking fuels was only classified into two groups (never, ever) which resulted in a lack of information on dose–response relationship.]

Eighty-eight incident cases of NPC diagnosed after 1 January 1 1986 in Zanhwu and Wuzhou, an area in China that is endemic for NPC, were enrolled in a case–control study (Hubert et al., 1993; Zheng et al., 1994). One hundred and seventy-six controls were selected in the immediate neighbourhood, matched on sex, age (±4 years) and place of residence. The controls were interviewed within the same week as the patients and under the same conditions. Information was collected on past and present conditions including educational level, marital status, place of birth, residential history, personal or family income, housing, types of fuel used, kitchen and toilet equipment and sleeping conditions, as well as diet (including methods of preparation and preservation). A sociodemographic score was established to describe living conditions according to three variables: monthly income, lack of house windows during the preceding year and type of housing in childhood. Odds ratios and 95% CIs were calculated using conditional logistic regression. Use of wood as fuel in the year before diagnosis was positively associated with the risk for NPC after adjusting for sociodemographic score (odds ratio, 6.4; p=0.003; 80 exposed cases) and after additional adjustment for consumption of herbal tea in year before diagnosis and consumption of salted fish in porridge before the age of 2 years (odds ratio, 5.4; 95% CI, 1.5–19.8; 80 exposed cases). Exposure to wood as a fuel in early childhood was not significantly associated with NPC. The risk for NPC from the use of wood fire was also studied in conjunction with other environmental factors that may affect the level of exposure. [The Working Group noted that exposure to wood fuel was assessed crudely, in that it was classified as a dichotomous variable, and no information on a dose–response relationship was presented. The Working Group noted that almost all subjects were exposed to wood combustion in early childhood and therefore ‘current use’ may be a better surrogate for cumulative exposure.]

A case–control study was designed in Nagaland, India, to evaluate the determinants of nasopharyngeal cancer (Chelleng et al., 2000). The study included 47 histologically confirmed cases recruited at the Bhobaneswar Boruah Cancer Institute between 1996 and 1997 and 94 neighbourhood controls matched on age, sex and ethnicity. Information on risk factors including dietary, environmental and sociodemographic factors was obtained from an in-person interview using a standard questionnaire. Of the cases, 72.3% were men and a large majority (68.1%) were over 40 years of age at the time of diagnosis. Questions on biomass exposure included presence of soot in the house, cooking fuel used, location of the kitchen, and type of house and the number of windows. None of these variables was significantly associated with the risk for nasopharyngeal cancer. In the multivariate logistic regression that accounted for tobacco smoking and socioeconomic status among other factors, the presence of soot in the house gave an odds ratio of 1.4 (95% CI, 0.6–3.3; 20 exposed cases) and use of wood for cooking (versus gas) gave an odds ratio of 1.6 (95% CI, 0.4–6.6; 42 exposed cases). [The Working Group noted that the consumption of smoked meat and a previous history of nasal drop use were significantly related to cancer. The Working Group further noted that this study had a poor characterization of exposure to wood burning (dichotomous wood versus gas) and no specification on current or past exposure. In addition, the power was low due to the small number of cases.]

2.2.5. Cancer of the oesophagus and its precursors

Chronic oesophagitis is considered to be a precusor condition for oesophageal cancer. A study was carried out to collect information on the prevalence of chronic oesophagitis at early ages in a high-risk area for oesophageal cancer and to identify risk factors associated with the prevalence of this disease (Chang-Claude et al., 1990). Study subjects were young adults aged 15–25 years from all households where cases of oesophageal cancer had been diagnosed after 1981 until October 1987 and twice the number of randomly selected households where no oesophageal cancer was diagnosed (with neither a diagnosis nor a family history of oesophageal cancer or dysplasia). A total of 227 and 660 young adults from these two types of household were eligible for inclusion in the study. In May 1988, 545 (62%) subjects participated in the study. They were interviewed and information was collected on dietary habits in the early 1970s and in the past 5 years, methods of food preparation, types of oil used, alcohol consumption, tobacco smoking, use of coal and other fuels, cooking fumes, ventilation, family history of oesophageal cancer, occupation and dental hygiene. A physical examination with collection of a 10-ml blood sample and early morning urine was conducted for each subject. Endoscopic examination of the oesophagus and stomach was also performed. Variables identified in univariate analysis were then evaluated in multivariate logistic regression models. A total of 538 subjects (354 men, 184 women) underwent an oesophagoscopy with biopsy. Of these, 166 came from cancer households and 372 from non-cancer households. Since the distributions of variables of interest among subjects with very mild oesophagitis were similar to those with a normal oesophagus, very mild oesophagitis was classified as normal in the analysis of risk factors for chronic oesophagitis. In univariate analysis (the household was controlled for as a confounder), in addition to other significant risk factors, cottonseed oil used for cooking most frequently (odds ratio for men, 2.3; 95% CI, 1.2–4.5; odds ratio for women, 1.6; 95% CI, 0.5–5.9), use of wood as fuel in the early 1970s (odds ratio for men, 2.5; 95% CI, 1.2–5.2; odds ratio for women, 2.6; 95% CI, 0.7–9.2) and use of wood as fuel in the past 5 years (odds ratio for men, 1.4; 95% CI, 0.3–7.0; odds ratio for women, 9.9; 95% CI, 2.3–43.2) were statistically significant variables for the risk for chronic oesophagitis in this rural area. In multivariate analysis, there was no statistically significant association of these variables, adjusted for other risk factors including age and sex (adjusted odds ratio for use of wood as fuel in the past 5 years, 1.72; p=0.19). The authors stated that the unexpected finding of an association of the use of wood as fuel with disease occurrence could be a chance association, since the relationship virtually disappeared in multivariate analysis. [The Working Group noted that the participation rate (62%) in this study was relatively low, and selection bias might exist.]

A study was conducted to determine whether functional polymorphisms in xenobiotic metabolizing genes could affect the risk for oesophageal cancer in different population groups (Dandara et al., 2006, in Table 2.6). A total of 245 patients with histologically confirmed squamous-cell carcincinoma and admitted for dysphagia were recruited from a hospital in Cape Town, South Africa. A total of 288 age-matched, healthy population controls were recruited from the same geographical location as the patients; 145 cases and 194 controls were black and 100 cases and 94 controls were of mixed ancestry. As part of the questionnaire, information on cooking and heating fuels used during the past 20 years was recorded and smokers were classified as individuals who had smoked at least one cigarette per day for at least 1 year. Subjects were also classified as alcohol consumers if they consumed alcohol regularly (at least once at week). Among black subjects, the burning of wood or charcoal for cooking and heating (compared with electricity) was significantly associated with an increased risk for oesophageal cancer (odds ratio, 15.2; 95% CI, 8.15–28.2; p=0.001; 63 exposed cases), as were smoking pipes and consumption of home brewed beer. In the subjects of mixed ancestry, wood or charcoal use was not associated with oesophageal cancer (odds ratio, 1.19; 95% CI, 0.60–2.34; p=0.62; 28 exposed cases); however, alcohol consumption and tobacco smoking were strong risk factors for oesophageal cancer. [The Working Group noted that the exposure variable used was dichotomous and that no dose–response information was available. There was concern for residual confounding because the risk estimates were adjusted for overall consumption of alcohol and tobacco (not associated in the data set with oesophageal cancer in blacks) but not for pipe smoking or home brewed beer consumption (which were strong risk factors oesophageal cancer in the black population).]

Table 2.6

Case–control studies of cancer of the upper aerodigestive tract.

2.2.6. Cancer of the cervix (see Table 2.8)

Table 2.8

Case–control study of cervical cancer.

A case–control study was conducted in Honduras to determine whether exposure to wood smoke increases the risk for invasive cervical cancer (Velema et al., 2002). One hundred and twenty-five women aged 20–64 years who had different grades of cervical intraepithelial neoplasia (CIN) (44 CIN I, 36 CIN II, 45 CIN III) were recruited from a screening programme. Each case was matched by age, clinic and calendar time to two controls (241 controls in total) without cervical abnormalities. All women were from low socioeconomic backgrounds. Cervical scrapes were tested for the presence of human papilloma virus (HPV) and HPV genotyping was performed. An interview was conducted in the clinic for case and control women to determine whether they had ever cooked with wood and, if so, the duration of use and the number of years since stopping use of wood for cooking. Exposure to wood during childhood was also determined. HPV DNA was detected in 48% of women with CIN I, 67% with CIN II and 89% with CIN III. Exposure to wood smoke for 35 or more years increased the risk for CIN III (odds ratio, 4.89; 95% CI, 0.51–47.1; p=0.017; nine exposed cases) compared with women with no exposure. Restriction of the analysis to women who reported exposure yielded a positive association with development of CIN III for women exposed for more than 35 years versus those exposed for 1–14 years (odds ratio, 9.5; 95% CI, 1.16–77.4; p=0.017; nine exposed cases). There was no significant association with CIN I or II. [The Working Group noted that this significant association was observed only when unexposed women were excluded from the analysis and unstable point estimates resulted after stratification by CIN stages and duration of exposure.] Among HPV-positive women, more than 35 years of exposure to wood smoke increased the risk for CIN compared with 1–14 years of exposure to wood smoke (odds ratio, 5.69; 95% CI, 1.00–32.70). [The Working Group also noted that these results are difficult to interpret because women who reported not having used wood in the kitchen had a risk higher than those with low or intermediate exposure and the analytical approach was questionable.]

2.3.1. Cancer of the lung (see Table 2.9)

Table 2.9

Case–control studies of lung cancer and use of mixed coal/biomass fuel.

Chen et al. (1990) conducted a case–control study in Taipei, Taiwan, China that included 323 serial lung cancer cases from four teaching hospitals (133 epidermoid, 47 small-cell and 134 adenocarcinomas) and 617 healthy controls who were ophthalmic patients in the study hospitals that were frequency-matched to cases on hospital, gender and age. Logistic regression models adjusted for sex and age showed no significant effect of burning coal or wood compared to other fuels (charcoal, gas) and electricity.

Mzileni et al. (1999) conducted a hospital-based case–control study that included 288 men and 60 women who had been diagnosed with incident lung cancer between 1993 and 1995 in the main tertiary referral hospital in the northern Province of South Africa. Controls were 183 men and 197 women who had been diagnosed with other incident cancers (predominantly of the prostate, liver and breast, colorectal and haematological cancers) in the same hospital as cases during the same study period. Cases and controls were interviewed regarding their tobacco smoking habits, residence, main occupation and fuel use (wood and coal) at home. The risk for lung cancer was increased in relation to the use of wood or coal in the house in men (odds ratio, 1.9; 95% CI, 0.9–3.3; 260 exposed cases) and women (odds ratio, 1.4; 95% CI, 0.6–3.2; 51 exposed cases) after adjusting for smoking, dusty occupation and residential exposure to asbestos. The positive association between the risk for lung cancer and the use of wood or coal in the house was statistically significant in men and women combined (adjusted odds ratio, 2.0; 95% CI, 1.1–3.6). Active tobacco smoking and living in asbestos-polluted areas were also significantly associated with risk for lung cancer in men and women combined after adjusting for age, sex, dust and use of wood for fuel. [The Working Group noted that results were presented for wood and coal use combined and as a dichotomous variable (no/yes). The independent effects of coal and wood could not be examined because the information on these two types of fuel were not presented separately. Because of the strong significantly positive associations between risk for lung cancer, smoking and residence in asbestos-polluted areas, it is difficult to rule out the role potential confounding effects of these exposures in the association between wood/coal use and lung cancer.]

Two studies in the Kaohsiung area, Taiwan, were reported by Ko et al. (1997) and Le et al. (2001) (described in detail in the monograph on high-temperature frying). Risk patterns associated with cooking fuels were reported. The first report included 106 nonsmoking lung cancer cases and an equal number of hospital controls (Ko et al., 1997). Compared with women who did not cook or cooked with gas at age 20 years or younger, those who cooked with coal experienced no increased risk for lung cancer (odds ratio, 0.5; 95% CI, 0.2–1.6) whereas those who cooked with wood and/or charcoal had more than a twofold increased risk (odds ratio, 2.5; 95% CI, 1.3–5.1). Results were similar when they examined cooking practices between the ages of 20–40 years. Compared with those who did not cook or cooked with gas, the risk for lung cancer was increased among those who cooked with wood/charcoal (odds ratio, 2.5; 95% CI, 1.1–5.7). Few subjects cooked with wood or coal after 40 years of age and thus meaningful analyses could not be conducted.

In a subsequent report (Le et al., 2001) (described in detail in Section 2.1.1), the relationship between cooking fuel and risk for lung cancer was examined separately by lung cancer cell type. The use of wood was associated with a threefold increased risk for lung cancer (odds ratio, 3.1; 95% CI, 1.0–9.2) compared with women who did not cook or used gas for cooking. In contrast, the risk for adenocarcinoma of the lung (158 cases, 262 controls) increased in relation to the use of wood (odds ratio, 3.0; 95% CI, 1.4–6.4). Tobacco smoking, residential area, education and social class were adjusted for in the analysis. [The Working Group noted that information on duration of wood and coal use was not reported in these two studies.]

Sixty-seven women who had histologically or cytologically confirmed lung cancer seen at the Department of Pulmonary Medicine, Chandigarh, India, and 46 controls with non-malignant respiratory disease were recruited in a case–control study between January 1999 and December 2002 (Behera & Balamugesh, 2005). A questionnaire was used to collect information on demographic factors, lifetime exposure to smoking, detailed occupational history, residence and exposure to indoor air pollution due to burning of organic fuels. [The Working Group noted that ‘organic fuels’ were not clearly defined but probably included coal, wood, cow dung cake and agricultural waste.] Unconditional logistic regression models were used for analyses. Among the lung cancer cases, 50 (74.6%) were nonsmokers among whom adenocarcinoma was the predominant histology (50%). In women who smoked, squamous- and small-cell carcinoma were the most common histological types. When adjusted for active and passive smoking [not stated if adjusted for age] and compared with the use LPG as the reference category, the odds ratio for use of biomass fuel was 3.6 (95% CI, 1.1–12.0). Among nonsmokers, the corresponding odds ratio was 5.3 (95% CI, 1.7–16.7). [These results were presented in a table and in the text and it was unclear if the analyses were stratified or unstratified univariate. The Working Group also noted that the selection of controls with respiratory diseases could underestimate the relative risk.]

Pisani et al. (2006) carried out a case–control study of lung cancer with 211 hospital cases (66% men; including smokers and nonsmokers), 202 population controls and 211 hospital controls without tobacco-related diseases matched by age and sex (and also residence for the hospital controls). This study primarily investigated the interactions between smoking and genetic polymorphisms in northern rural Thailand. Information was obtained through in-person interviews. The exposure index of solid fuel use was calculated by combining years of cooking with coal/wood indoors (weight=1) and outdoors (weight=0.5). The majority of cases (99%) were microscopically verified; 45% were squamous-cell carcinomas and 21% were adenocarcinomas. Only 7% of men and 33% of women in the overall study group were never smokers. No significant effects or trends were found for the cumulative index of exposure to domestic fumes for the whole group after adjusting for age and sex (and also cumulative smoking in smokers). [The Working Group noted that the coal and wood exposures were not separated and therefore made the study results uninformative for either type of fuel. The use of fuel for cooking and heating was also not separated.]

Ramanakumar et al. (2007) conducted a hospital-based case–control study in Montreal, Canada, which was originally designed to examine occupational risk factors. This analysis included 1205 (739 men, 466 women) histologically confirmed lung cancer patients who were diagnosed between 1996 and 1997 at the 18 largest hospitals in the study area. A total of 1541 (925 men, 616 women) population controls were selected from electoral lists and were interviewed. Structured interviews were conducted to collect information on smoking history, occupational history, sources of traditional heating (defined as wood or coal stove in living space) and cooking (defined as a wood or gas stove) and other risk factors. To assess exposure to traditional cooking sources, subjects were asked if they had ever lived full-time in a house/apartment where the cooking was carried out on a gas or wood stove. Similarly, to assess exposure to traditional heating sources, subjects were asked if they had ever lived in a house/apartment that was mainly heated by a stove or fireplace located in the living quarters. No significant associations with exposure to traditional heating or cooking were found for men, with most of the odds ratios below 1.0. For women, however, most of the odds ratios associated with traditional heating or cooking were above 1.0 and were statistically significant. In women, elevated risks associated with traditional heating were found in subjects who were older (≥60 years) at the age of onset, self-respondent, and in smokers. An increased risk for lung cancer was also observed in women classified as nonsmokers and light smokers. [The Working Group noted that the assessment of ‘traditional cooking’ combined cooking with gas and wood whereas the assessment of ‘traditional heating’ may have included exposure to wood/coal. In addition, there was incomplete specification of fuel use; for example, no mention was made of the use of electricity for cooking or oil and electricity for heating which further complicated the interpretation of this study.]

2.3.2. Cancer of the oral cavity, pharynx and larynx (see Table 2.6)

Three concurrent case–control studies were conducted in Heidelberg, Germany, to examine the association between fossil fuel stoves and the risk for laryngeal, pharyngeal and oral cavity cancer (Dietz et al., 1995; Maier & Tisch, 1997). A total of 164, 100 and 105 cases of laryngeal, oral cavity and pharyngeal cancer, respectively, were ascertained between 1989 and 1992 from all patients seeking treatment at the Otorhinolaryngology Department within 3 years from first diagnosis. Almost all cases were current or former smokers. Controls were recruited from the same medical centre and general outpatient department at the University of Heidelberg from among non-cancer patients and matched to cases on sex, age and size of the place of residence. Fossil fuel emissions from stoves and cooking and the type of burning materials used (coal, briquette, coke, peat, gas and oil) were ascertained. Use of fossil fuel stoves or cookers were associated with all three types of cancer: adjusting for tobacco and alcohol use, the odds ratio for >40 versus 0–20 years was 2.0 for fossil fuel heating (95% CI, 1.10–3.46) and 1.4 for cooking (95% CI, 0.76–2.41) for laryngeal cancer, 3.3 for fossil fuel heating (95% CI, 1.43–7.55) and 2.5 for cooking (95% CI, 1.03–6.30) for pharyngeal cancer and 2.4 for fossil fuel heating (95% CI, 1.26–4.40) and 1.6 for cooking (95% CI, 0.90–2.97) for oral cavity cancer. [The Working Group noted that no dose–response trend with duration of use and no association for the 20–40-year exposure category for any of the cancer sites were observed; also, the specific type of fuel responsible for the increase in risk at the highest duration of stove use could not be determined.]

2.3.3. Cancer of the nasopharynx (see Table 2.7)

A case–control study of 115 cases of NPC (86 men, 29 women) newly diagnosed pathologically in the hospitals of Fujian Province during the period from March to May 1991, who had lived in Fujian at least for 10 years, was conducted (Cai & Ye, 1996). Controls (115 cancer controls and 115 non-cancer controls) were randomly selected from the patients and matched on sex, age (within 5-year age group) and date of hospitalization (same month) to the cases. The controls had also lived in Fujian at least for 10 years. The cancer controls were patients with cancers other than of the respiratory system, while the non-cancer controls were patients without cancer or respiratory diseases. Information in 1968 and 1978 on demographic factors, residential, dietary and occupational history, smoking and alcohol drinking, family history of cancer and chronic diseases of ear, nose, pharynx and larynx was collected. In multivariate analysis using non-cancer controls, the odds ratio for use of straw as domestic fuel was 6.52 in 1978 (95% CI, 1.90–22.32). Use of coal and firewood/straw in 1968 was positively associated with NPC for cancer controls (odds ratio, 3.14; 95% CI, 1.18–8.36) and for the combined control groups (odds ratio, 3.35; 95% CI, 1.36–8.26). [The Working Group noted that the odds ratio for straw was estimated using coal fuel as a reference group. Consumption of salted fish was only assessed in 1978, but not in childhood, and because both salted fish consumption and straw use may be associated with socioeconomic status, the odds ratio for straw may be confounded.]

A hospital-based case–control study of NPC (Huang et al., 2002) was carried out in Guangxi to search for risk factors of NPC other than Epstein-Barr virus (EBV) infection; 175 cases of NPC (132 men, 43 women) pathologically diagnosed and treated at First Hospital and Cancer Hospital affiliated to Guangxi Medical University during the period from March 2000 to May 2001 were involved in the study. The cases had lived in Guangxi for at least 10 years and had originated from Guangxi. A total of 350 controls (264 men, 86 women) were selected from patients without cancer or respiratory diseases treated at the same hospital and same period, and matched on sex, age (±4 years), occupation, at least 10 years of living in Guangxi and same place of origin as the cases. Information on demographic factors, occupational history, residential environment, life and dietary habits, previous diseases, family history of NPC and psychological factors was collected by use of a structured questionnaire. Use of coal or firewood as fuel was associated with a significantly increased risk for NPC (odds ratio, 3.68; 95% CI, 2.15–6.29).

2.4.1. Cancer of the lung (see Table 2.10)

The studies that are included in this section of mixed exposures did not examine specific exposures such as type of cooking oil or frequency of high-temperature cooking but rather examined measures related to general cooking practices (e.g. age started or years of cooking) or ventilation.

The first study that included data on cooking practices and lung cancer was a hospital-based case–control study conducted in Singapore (MacLennan et al., 1977) which included 233 lung cancer cases (147 men, 39 Cantonese women, 47 non-Cantonese women) and 300 hospital control subjects (134 men, 80 Cantonese women, 86 non-Cantonese women) who were identified from three main hospitals. In total, 46 lung cancer cases (20%) and 124 controls (41%) were nonsmokers. Thirty-six per cent (84/233) of the lung cancer cases were histologically confirmed. The risk for lung cancer in relation to domestic cooking was 1.55 [95% CI, 0.75–3.22] in men, 1.74 [95% CI, 0.67–4.47] in Cantonese-speaking women and 0.40 [95% CI, 0.21–0.92] in non-Cantonese-speaking women. [The Working Group noted that no adjustment for active smoking or use of gas/kerosene was made in the analysis. No description of the questions asked on cooking was given: the frequency, duration and intensity of cooking were not included.]

In a case–control study in Shenyang, China, Xu et al. (1989) (described in Section 2.1.1) developed a continuous index of average long-term indoor air pollution intensity, based on duration of living at each residence, presence of coal heating, cooking fuel and location of cooking place (e.g. separate or not). This index was positively associated with lung cancer risk in a dose–response fashion in both men and women. [This study overlaps with Sun et al. (1991).]

One study in Xuan Wei County, China (Liu et al., 1991), was based on 110 incident lung cases (56 men, 54 women) identified in regional hospitals and clinics between 1985 and 1986 and 426 population controls (224 men, 202 women) matched on age, sex, occupation and village of residence. Almost all of the men (52 cases, 205 controls) but few of the women (no cases, one control) were smokers. Only 17% of the lung cancers were pathologically or cytologically confirmed. Men who reported that they often cooked food had a significantly increased risk for lung cancer (odds ratio, 3.36; 95% CI, 1.27–8.88; 12 exposed cases) after adjusting for smoking and other unspecified covariates. Women who started cooking at a young age showed an increased risk but there was no significant trend of increasing risk with decreasing age when starting to cook. Compared with women who started cooking at age 15 years or older, the adjusted odds ratio for lung cancer was 2.37 (95% CI, 1.09–5.15; 30 exposed cases) for starting cooking at age 11–15 years and 1.25 (95% CI, 0.45–3.49; 11 exposed cases) for starting cooking at age 10 years or younger (p for trend >0.05). However, the risk increased with increasing years of cooking (adjusted odds ratios, 1.00, 9.18 and 14.70 for ≤30, 31–44 and ≥45 years of cooking, respectively; p for trend >0.05), but the confidence intervals were very wide. [The Working Group noted that this small study had only indirect measures of cooking practices, based on age when started cooking and duration of cooking. The study was limited because only 83% of the lung cancers were diagnosed clinically or radiologically. Potential confounding was not adequately addressed in the statistical analysis.]

Liu et al. (1993) presented the results of a hospital-based case–control study of indoor air pollution and lung cancer in Guangzhou, China (described in detail in Section 2.1.1). After the in-person interview, the interviewer measured the size of the windows and doors that opened onto the outside of the building, thereby providing an estimation of ventilation capacity. 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 ventilation conditions. Data on up to three residences were collected. Not having a separate kitchen and poor air circulation were significantly associated with the risk for lung cancer in men and women. There was a significant dose–response relationship (p for trend <0.05) and a significant inverse association with risk for lung cancer seen with size of ventilation openings in the kitchen and living area and room height. Increased risks for lung cancer were found for men (adjusted odds ratio, 2.4; 95% CI, 1.4–4.2) and women (adjusted odds ratio, 5.9; 95% CI, 2.1–16.0) who lived in homes that did not have a separate kitchen. Similarly, living in a house with poor air circulation was associated with an increased risk in men (adjusted odds ratio, 2.1; 95% CI, 1.2–3.6) and women (adjusted odds ratio, 3.6; 95% CI, 1.4–9.3). In contrast, significant trends of decreasing risk were observed in association with better ventilation, based on variables that measured size of ventilation openings in living areas and in kitchens (see Table 2.10). However, no differences were observed between cases and controls in the number of meals they cooked per day or the presence of chimneys in their homes. [The Working Group noted that this study did not include direct measures of cooking fumes/practices. However, this was one of the few studies that attempted to obtain a more objective measure of ventilation. As part of the study, the interviewers measured the size of the windows and doors that opened onto the outside of the building in each participant’s home. In addition, information on ventilation in up to three previous residences was obtained. A limitation is that a high percentage of the lung cancers were not cytologically/histologically confirmed].

Table 2.10

Case-control studies of lung cancer and proxies of indoor air pollution.

Du et al. (1996) also reported a case–control study in which cases were 120 nonsmokers drawn from 849 decedent cases. In conditional logistic regression models, risk for lung cancer in women was significantly associated with increasing indoor air pollution and decreasing kitchen size. These factors were not associated with risk in men. [The Working Group noted that these studies are limited by use of proxy respondents and overall exposure assessment.]

Lei et al. (1996) conducted a case–control study in Guangzhou, China, among 792 cases who had died from primary lung cancer (563 men, 229 women), identified from 1986 death certificates, and 792 controls matched on gender, age, street of residence and year of death. Controls had no history of respiratory diseases or tumours. Women with a fair or poor living conditions index (living area per person/room ventilation) had a significantly increased risk for lung cancer while this association was much weaker and only of borderline statistical significance in men.

Koo and Ho (1996) examined the role of cooking fumes and lung cancer risk in a case–control study that included 200 female lung cancer cases and 200 neighbourhood controls. Histology was determined in 90.5% of the cases of which 34% were adenocarcinoma. Forty-four per cent of cases (88/200) and 68.5% of controls (137/200) were never smokers (Koo et al., 1983). A significant inverse association was observed between the duration of cooking and risk for lung cancer among never smokers. After adjusting for age, number of live births and education, the risk for lung cancer declined with increasing years of cooking; the adjusted odds ratios were 1.0, 0.38 (95% CI, 0.17–0.88) and 0.37 (95% CI, 0.14–0.96) for 0–25, 26–40 and ≥41 years of cooking (p for trend <0.001). [The Working Group noted that this study only presented results on indirect measures of cooking and that results in smokers were not reported. From an earlier report by Koo et al. (1983), kerosene was the main fuel that was used and thus confounding by coal use is unlikely to be a main issue in this analysis.]

In Shen et al. (1996, 1998) (described in detail in Section 2.1.1), multivariable analysis showed that exposure to fumes or pollution from cooking was significantly associated with risk for lung cancer (odds ratio, 2.45; 95% CI, 1.06–5.66; p=0.02). [The Working Group noted several limitations in this study. The report lacked details regarding the study design (e.g. response rate), characteristics of the study population (e.g. gender distribution, active smoking history) and covariates included in the statistical models.]

In two hospital-based case–control studies conducted in Kaohsiung, a heavily industrialized city in Taiwan (China) (Ko et al., 1997; Le et al., 2001) (in detail in Section 2.1.1), the use of a fume extractor in the kitchen was also significantly associated with a reduced risk for lung cancer.

Zhou et al. (2000) (described in detail in Section 2.1.1) showed that, compared with women who had a separate kitchen for cooking, the risk for lung cancer was not increased for cooking in the living room (crude odds ratio, 1.40; 95% CI, 0.41–4.88) or bedroom (crude odds ratio, 1.00, 95% CI, 0.11–8.93). However, in multivariable regression analysis, frequent eye irritation from smoke had an independent impact on risk. Compared with women who reported no eye irritation from smoke, those who reported slight, medium and heavy eye irritation had elevated risks; the respective adjusted odds ratios were 1.58, 11.45 and 3.41 (p for trend=0.002). [The Working Group noted that most of the lung cancer cases and controls included in the analysis by Zhou et al. (2000) were already in the report by Wang et al. (1996). Unadjusted odds ratios were reported. This study was small and the confidence intervals were very wide.]

In the case–control study of lung cancer by Le et al. (2001) (described in detail in Section 2.1.1), risk was significantly higher for those who cooked in a kitchen without a fume extractor: the odds ratio was 3.0 (95% CI, 1.3–7.1; 31 exposed cases) for squamous-/small-cell cancer and 3.9 (95% CI, 2.3–6.6; 74 exposed cases) for adenocarcinoma. Only 7% of men reported cooking for the family and thus these data were not reported.

2.4.2. Cancer of the nasopharynx

A hospital-based case–control study of NPC was conducted in Minan Prefecture, Fujian Province, China (Ye et al., 1995), on 135 cases of NPC that were pathologically diagnosed and treated at Second Hospital affiliated to Fujian Medical University in Quanzhou. Controls were patients of surgical and osteological departments of the same hospital without cancer or respiratory diseases, were matched on sex, age (within 5-year group) and date of hospitalization and had lived in Minan Prefecture for at least for 10 years with the same place of origin. Information on demographic factors, residence, family cancer history, dietary history, smoking, alcohol drinking, tea drinking, occupational history and chronic diseases of the ear, nose, pharynx and larynx was collected. In a multivariable conditional logistic regression analysis, smokiness during cooking was associated with an elevated risk for NPC with an odds ratio of 2.30 (P=0.012). The other significant factors associated with risk were intake of green melon (protective factor), index of passive smoking during adulthood, consumption of salted preserved vegetables and cooking (risk factors).

In a case–control study (Hubert et al., 1993; Zheng et al., 1994) (described in detail in Section 2.2.4), absence of windows, poor ventilation, cooking outside the house in a shack and the presence of a fireplace in the kitchen during childhood significantly increased the excess of risk for NPC associated with using wood as a fuel.

Two hospital-based case–control studies on NPC that used a similar study design and questionnaire were conducted in Guangzhou and Heilungjiang Province (Huang et al., 1997). A total of 104 cases of NPC who were pathologically diagnosed at each of two hospitals (cancer hospitals affiliated to Sun Yat-Sen Medical University in Guangzhou and Harbin Medical University in Harbin) during the period from 1 October 1992 to 1 March 1994 and who had lived in Guangzhou or Heilungjiang for more than 80% of their life were involved in the study. The cases were recruited sequentially in accordance with the order of entrance to each hospital until 104 cases had been obtained. One control per case was matched on sex, age (±5 years) and some place of origin as the case and was selected from residents without cancer in the next residential community to the cases. Information on demographic factors, residence, occupation, dietary history, previous diseases and family cancer history was collected from cases and controls by use of a unified questionnaire. In multivariate analysis, cooking inside the house and use of a stove without a chimney were two significant variables associated with the risk for NPC. Taking the odds ratio for cooking inside the house for 35 or less than 35 years as 1.0, the odds ratio for more than 35 years was 1.96 (95% CI, 1.23–3.71). Taking the odds ratio for use of a stove without a chimney for 10 or less than 10 years as 1.0, the odds ratio for more than 10 years was 2.69 (95% CI, 1.54–4.68). In Harbin, in comparison with the use of coal or wood, the use of gas as fuel decreased the risk for NPC, with an odds ratio of 0.93 (95% CI, 0.90–0.97; p=0.027). [The Working Group noted the discrepancy between the relatively small number of subjects and the narrow confidence intervals.]

Sihui City, Guangdong Province, is one of the endemic areas of NPC in China with an incidence rate of about 20/10⁵. A case–control study on NPC was conducted that included 57 cases who were alive and pathologically diagnosed in Sihui City between January 1998 and June 1999 (Cao et al., 2000). The control group consisted of spouses and relatives of spouses of these 57 cases. In multivariate analysis, family cancer history, connection of the bedroom with the kitchen during childhood and tobacco smoking were significant risk factors for NPC. The odds ratio for separation of the bedroom from the kitchen was 0.48 (95% CI, 0.30–0.78).

2.4.3. Cancer of the cervix

A nested case–control study to investigate the association of exposure to cooking oil fumes with the risk for cervical neoplasm was carried out between October 1999 and December 2000 on 32 466 women aged over 19 years who underwent Pap smear screening in Chi-Yi City of Taiwan, China (Wu et al., 2004). Among 420 women newly diagnosed as having CIN lesions ≥CIN1, 349 were followed-up by biopsy. Among the 349 subjects with biopsy follow-up, 116 women had lesions ≥CIN2 confirmed by biopsy. These 116 women were eligible as cases for the study. The controls were randomly selected from women whose Pap smear results were negative in the first screening of the study period. The case–control ratio was 1:2 with matches for age (±2 years), residence and the time that the Pap smear was performed (within 6 months of the cases). The findings of previous Pap smears taken before this study were found to be normal in both cases and controls. The subjects were interviewed in their homes between October 2000 and March 2001. Information on demographical characteristics, education, smoking, exposure to environmental tobacco smoke, exposure to X-ray examinations or hair dye, occupation (especially professional chef), sexual and reproductive history and times of prior cervical smears, as well as cooking and kitchen ventilation status, was collected. A multivariable logistic regression model was used to assess the association between case–control status and different cooking and ventilation conditions, after adjusting for age, education, smoking, age at first intercourse, number of prior Pap smears and profession as a chef. Of the 116 cases with ≥CIN2, 16 women who had no questionnaire information were excluded from the study. Among the 100 cases with a completed questionnaire, 39, 12, 46 and three cases had CIN2, CIN3, carcinoma in situ and invasive cancer, respectively. The results of the age at which one started cooking, years of cooking and hours spent on cooking were insignificant. Subjects who cooked in a kitchen without a fume extractor at 20–40 years of age had a 2.29-fold higher risk (95% CI, 1.08–4.87) for developing CIN than those who used a fume extractor, after adjusting for corresponding factors. The odds ratio was 3.16 (95% CI, 1.19–8.43) for women who cooked in kitchens without fume extractors at >40 years of age. The use of coal as cooking fuel had a higher risk for CIN than that of gas for the women in the group who cooked at 20–40 years of age (adjusted odds ratio, 2.09; 95% CI, 0.86–5.10) and at >40 years of age (adjusted odds ratio, 1.53; 95% CI, 0.17–14.19), but the difference in risk was insignificant. Among the women over 40 years of age, those who did not have kitchen fume extractors at home at both 20–40 years and >40 years of age had a 3.46-fold greater risk of developing CIN than those who did have fume extractors during the same two periods, after adjusting for corresponding factors. The odds ratio for having a fume extractor only during one period but not the other was 2.05 (95% CI, 0.86–4.86). It was also found that women who had been professional cooks had a 3.97-fold greater risk (95% CI, 1.02–15.41) than those who had not. [The Working Group noted that the results were not stratified by or adjusted for HPV.]