Mortality among Swedish chimney sweeps (1952–2006): an extended cohort study
- 1Unit of Occupational Medicine, Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden
- 2Division of Occupational and Environmental Medicine, Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden
- 3Research Department, Swedish National Institute of Public Health, Östersund, Sweden
- Correspondence to Professor Per Gustavsson, Unit of Occupational Medicine, Institute of Environmental Medicine, Karolinska Institutet, Norrbacka, Level 4, SE-171 76 Stockholm, Sweden;
- Accepted 29 April 2011
- Published Online First 24 June 2011
Objectives We extended a cohort study of Swedish chimney sweeps and prolonged follow-up in order to increase power and study those first employed after 1950 when oil began to replace wood as a main fuel for heating in Sweden.
Methods Male Swedish chimney sweeps who were members of the national trade union in 1981–2006 were identified (n=1087) and included to a previous cohort of those employed in 1918–1980 (n=5287). All employment histories were updated, and the total extended cohort (n=6374) was linked to the registers of Causes of Death and Total Population and followed for mortality from 1952 through 2006. Standardised mortality ratios (SMRs) were estimated using the Swedish male population as reference.
Results 1841 observed deaths resulted in an SMR for all causes of deaths of 1.29 (95% CI 1.24 to 1.36). Mortality was significantly increased for all malignant tumours, oesophageal cancer, bowel cancer, liver cancer, lung cancer, alcoholism, ischaemic heart disease, non-malignant respiratory diseases, liver cirrhosis, external causes and suicides. The lung cancer SMR remained increased, although attenuated, after adjustment for group-level smoking data, SMR of 1.52 (95% CI 1.26 to 1.89). Duration of employment showed no consistent evidence of dose–response associations. Alcohol-related deaths (liver cirrhosis and alcoholism) were not increased among those employed >30 years. Mortality among those employed after 1950 was similar to that of the entire cohort.
Conclusions Chimney sweeps are exposed to high levels of toxic substances in the occupation, but excess alcohol and smoking habits were also observed, and the results must be interpreted cautiously. However, group-level data on tobacco smoking indicated that the lung cancer excess only to some extent could be explained by smoking habits, and the increased mortality from oesophageal cancer and ischaemic heart disease among chimney sweeps employed >30 years is less likely to be caused by excess alcohol habits.
- circulatory diseases
- respiratory diseases
- polycyclic aromatic hydrocarbons
- standardised mortality ratio
- hygiene/occupational hygiene
- public health
- mortality studies
- polyaromatic hydrocarbons (PAHs)
What this paper adds
This large study of chimney sweeps revealed new findings of a significantly increased mortality from bowel cancer and of asthma, bronchitis and emphysema. The study confirmed previously reported excess risks of mortality due to all causes, all malignant tumours, oesophageal cancer, liver cancer, lung cancer, psychological diseases and disorders, alcoholism, diseases of the circulatory system, ischaemic heart disease, diseases of digestive organs, liver cirrhosis, external causes and suicides.
Tobacco smoking and alcohol intake may partly explain the findings, especially among those with shorter employment periods, while occupational exposure is a more likely explanation for the increased risk of ischaemic heart disease observed among chimney sweeps employed 30 years or longer.
Mortality among chimney sweeps who started their employment after 1950 was similar to that of the entire cohort, indicating no major influence of changes in fuel from wood to oil or changes in work tasks among the chimney sweeps.
Chimney sweeping has been associated with adverse health outcomes since the late 18th century, first noted in a classic report from 1775 of scrotal cancer among British chimney sweeps.1 Traditional chimney sweeping (‘black sweeping’) includes removing of soot from chimneys and connecting pipes, furnaces and boilers. In Sweden, chimney sweeps have gradually started performing additional work tasks, for example, degreasing of kitchen ventilation equipment and fire safety inspections (‘white sweeping’). This change in work tasks began in the middle of the 20th century, and today, Swedish chimney sweeps on average do black sweeping about half of their working hours.2 Wood, as well as coal and coke, were the dominating fuels in Sweden until the successive introduction of oil beginning around 1950.2 During the last decades, there has been a partial shift back to burning of wood pellets.
Chimney sweeps are, in addition to soot, exposed to asbestos from pipe and furnace insulation, to organic solvents used for degreasing and to some extent also to combustion gases (eg, sulphur dioxide and carbon monoxide).2 3 Chimney soot contains several toxic agents such as carbon particles with adsorbed polycyclic aromatic hydrocarbons (PAHs) and metals (eg, arsenic, chromium, cadmium, nickel and lead).3 An extensive occupational hygiene survey performed in Sweden in 1985–1986 showed that organic dust levels were high during all work operations involving contact with soot. Average levels of total dust during sweeping were 3, 9, 11 and 19 mg/m3 for the four most common sweeping operations, thus clearly exceeding the Swedish threshold limit value of 3 mg/m3 in most work operations. Exposure to PAH and benso(a)pyrene (BaP) was highly variable; BaP varied from less than detectable (∼0.2 mg/m3) to 9.1 μg/m3 in 15 samples. Exposure to metals and asbestos was mostly below the respective Swedish threshold limit values.2
Experimental animal studies show that several PAHs, including BaP, cause cancer.4 Increased risks of cancer, predominantly lung and bladder cancers,5 as well as ischaemic heart disease6 7 have been observed in epidemiological studies of workers exposed to PAHs. However, it is unknown how much of these excesses that may be attributed to PAHs and how much exposure to particles or other substances in the work environment. A meta-analysis of lung cancer risk per unit of PAH exposure in various occupations showed a much higher unit RR estimate for chimney sweeps than for coke oven, gas or aluminium smelter workers, indicating that exposures other than PAH play an important role for the chimney sweeps at risk for lung cancer.8
There are several previous observations of an increased risk of various cancer types,9–11 ischaemic heart disease10 as well as asthma12 among chimney sweeps. The hitherto largest epidemiological study of chimney sweeps is a Swedish cohort study including >5000 chimney sweeps showing increased risks of several cancer types, ischaemic heart disease, diseases of the respiratory system and accidents, injuries or other external causes.13–16 The most recent update of this cohort included follow-up for mortality until 1990 and for cancer incidence until 1987.16 In the present study, we have extended this cohort with >1000 additional chimney sweeps first employed after 1980 (when the earlier cohort was closed), updated the employment histories and included 16 years of follow-up, with the aim of studying mortality among chimney sweeps exposed only since oil begun to replace wood as a main fuel and enhancing the precision to study long-term effects among chimney sweeps starting their employment early.
The design of the original cohort study has been described in detail previously.16 The original cohort comprised all registered male members of the Swedish Chimney Sweeps' Trade Union. This union was established in 1918 and organised about 95% of all active chimney sweeps in Sweden. In 1980, the Chimney Sweeps' Trade Union was merged into the Swedish Municipal Workers' Union. Chimney sweeps joining the union in 1997–2006 were identified from the union's nationwide computerised membership register. However, occupational titles were not registered by this union between 1981 and 1997. Chimney sweeps who both started and ended their employment within this period were instead identified through name lists provided by the local chimney sweeps' sections of the Municipal Workers' Union in the large regions of Stockholm, Scania and Gothenburg. In total, 118 male chimney sweeps were identified this way and verified in the union's membership register. However, chimney sweeps who worked in rural areas and both entered and left the occupation between 1980 and 1997 could not be included. Nine hundred and sixty-nine male chimney sweeps active in 1997–2006 were identified from the computerised register of the Municipal Workers' Union, and the additional cohort thus includes 1087 chimney sweeps in total.
The trade unions' membership registers include information on personal identity numbers (a unique 10-digit number assigned to all Swedish residents since 1947) and employment histories (ie, start and end dates for each membership/employment period). Each study participant's employment history was reviewed and updated manually, also for the original cohort, and these histories are thus complete for all chimney sweeps included in the study. We restricted the study to men due to the very small number of women working as a chimney sweep.
The cohort was linked to the nationwide Causes of Death Register to identify underlying causes of deaths.17 The International Classification of Diseases (ICD) codes, including ICD revisions, used for this study are listed in the online appendix. The cohort was also matched to the register of the Total Population to identify men who had emigrated. At the end of follow-up, all cohort members were verified as either living, dead or emigrated.
One thousand and forty-two chimney sweeps attended a health examination directed to all union-organised Swedish chimney sweeps that was performed in 1972. Information on current smoking and alcohol habits was obtained by a questionnaire.18
The cohort members were followed from the date of their first membership/employment or from 1 January 1952, whichever came last, through 31 December 2006, date of death or date of emigration, whichever came first. Standardised mortality ratios (SMRs) were estimated using 5-year stratification of age and calendar time and using the Swedish male population as reference.19 We estimated exact 95% CIs for the SMRs assuming that the observed events followed a Poisson distribution.20
We further analysed mortality by duration of union membership as a proxy for cumulative exposure. Person-years were allocated dynamically over duration categories, treating duration of employment as a time-dependent covariate. We also analysed mortality by latency, that is, time since first employment.19
Chimney sweep masters were considered occupationally exposed, and employment as sweep and master sweep were combined in analyses subdivided by duration. Finally, due to the shifts in fuels and work tasks described above, we analysed two subcohorts, one restricted to those starting their employment after 31 December 1950 and the other restricted to those first employed after 31 December 1986. All statistical analyses were performed with SAS, release 9.1 (SAS Institute Inc.).
Characteristics of the study participants
In the prior update, 5313 men were included in the mortality analyses.16 We excluded 10 men with incomplete or incorrect personal identity numbers and 16 men who died before 1952, resulting in 5287 male chimney sweeps. The additional part of the cohort comprised 1087 male chimney sweeps, after exclusion of 29 women. Thus, the extended total cohort consists of 6374 male chimney sweeps who were included in the statistical analyses, contributing 214 759 person-years. The mean employment duration in the cohort was 14.8 years.
Mortality among male Swedish chimney sweeps 1952–2006
In total, 1841 chimney sweeps had died during follow-up compared with 1422 expected (SMR 1.29, 95% CI 1.24 to 1.36) (table 1). We identified 484 deaths due to malignant tumours compared with 347 expected (SMR 1.39, 95% CI 1.27 to 1.52). Statistically significantly increased mortality risks were observed for cancer of the oesophagus, bowel, liver and lung. SMRs were increased, although not significantly, for cancer of the stomach, rectum, bladder and haematolymphatic organs (table 1).
There were 59 deaths from psychological disorders and diseases compared with 31 expected (SMR 1.91, 95% CI 1.45 to 2.46), mainly due to an increased mortality from alcoholism. We identified increased risks of death due to diseases of the circulatory system (SMR 1.18, 95% CI 1.10 to 1.27), including ischaemic heart disease (SMR 1.20, 95% CI 1.10 to 1.32), while no increased risk was observed for cerebrovascular diseases.
There were 113 observed deaths due to non-malignant diseases of the respiratory system compared with 74 expected (SMR 1.53, 95% CI 1.26 to 1.83). The SMR was strongly increased for asthma, bronchitis and emphysema. The mortality from diseases of the digestive organs was significantly increased (SMR 1.71, 95% CI 1.38 to 2.08), mainly due to a more than twofold increased risk of liver cirrhosis. We identified 263 deaths due to external causes resulting in an SMR of 1.47 (95% CI 1.30 to 1.66), partly due to an increased SMR for suicides. The point estimate for fall accidents was increased, although not significantly.
Mortality by duration of employment
All-cause mortality was significantly increased in all four categories of duration, but the risk increase was highest among the short-term employed (ie, >0–9.9 years) (table 2). All-cancer mortality was significantly increased in all four categories of duration, and there was no trend with duration of employment. No consistent dose–response associations regarding duration were observed for any of the studied cancers (table 2).
The SMR for death from alcoholism was increased in all duration categories, but most strongly increased in duration categories below 30 years. No significant excess risk was observed among those employed 30 years or longer. The mortality from ischaemic heart disease was increased in all duration categories, showing no consistent dose–response association.
Strongly increased risks of death from asthma, bronchitis and emphysema were seen among those employed >0–9.9 and 10–19.9 years, while the excess risk was moderate and non-significant for those employed 20 years or more. Significantly increased risks of death from liver cirrhosis were noted in all duration categories below 30 years of employment, while the SMR was close to unity among those employed 30 years or longer. All-cause SMRs were essentially similar in an analysis restricted to employment as sweep only (table 2).
Mortality by latency
Most chimney sweeps start employment at a young age, and consequently, a majority of observed and expected deaths were found in the group followed for >30 years since first employment (table 3). No cases were found in latency categories below 30 years for many causes of deaths, and data are shown for the most frequent causes of deaths only. All-cause mortality showed a positive association with time since first employment and was close to the expected during the first 10 years of follow-up. An almost significantly increased SMR was observed for the period 10–19 years since first employment, and the SMRs were significantly increased in the categories 20–29 and 30+ years since first employment. Mortality from all malignant tumours showed a similar pattern with SMRs below unity in the two first decades of follow-up, a non-significant excess in the 20–29 years category and a significant excess risk in the >30 years category.
SMRs for ischaemic heart disease, respiratory diseases and digestive diseases were increased in the 30+ category, while the observed and expected numbers were few in the categories with >0–29.9 years since first employment. Mortality from external causes was increased in all four latency categories (table 3).
Chimney sweeps first employed after 31 December 1950 and after 31 December 1986
SMRs of similar magnitude as in the entire cohort were found for chimney sweeps first employed after 1950 (table 1). Statistically significantly increased mortality from lung cancer, alcoholism, ischaemic heart disease and suicides was observed. Only six deaths were observed in the subcohort restricted to those first employed after 1986 (including 975 chimney sweeps), and the SMR was not increased for all-cause mortality (SMR 0.83, 95% CI 0.30 to 1.80) (data not shown).
Lung cancer mortality attributed to smoking
The nationwide health examination of 1040 Swedish chimney sweeps performed in 197218 showed that the proportion of current smokers was higher among the chimney sweeps than in the general population of the same ages.16 Using a method similar to that of Axelson,21 and assuming that chimney sweeps' smoking habits in 1972 were representative for the entire cohort, showed that a lung cancer SMR of 1.28 would be expected from smoking habits alone.16 Smoking adjustment of the observed SMR of 1.94 (table 1) resulted in an attenuated, although still significantly increased, SMR for lung cancer of 1.52 (95% CI 1.26 to 1.89).
This large cohort study of Swedish chimney sweeps showed, consistent with our prior findings,16 a significantly increased mortality from all causes of death combined, all malignant tumours, oesophageal cancer, liver cancer, lung cancer, diseases of the circulatory system, ischaemic heart disease, diseases of the respiratory system, liver cirrhosis, external causes and suicides. New findings included a significantly increased mortality from bowel cancer, psychological disorders and diseases, alcoholism, asthma, bronchitis and emphysema. The findings were essentially similar among those with first employment starting after 1950, indicating no major influence of changes in work tasks or fuels since the 1950s.
Strengths of our study include the large number of participants and the long-term and complete follow-up, resulting in a high statistical power. Selection bias is minimised due to the trade unions' high coverage and the nationwide registers used. The follow-up and the detection of deaths are virtually complete due to the high quality of the nationwide coverage of the Swedish population registers used.
An important limitation of our study is the lack of individual data on potential confounders such as tobacco smoking or alcohol intake, although group-level data from 1972 were available. Furthermore, variations in exposure within the occupation as chimney sweep could not be taken into account since information on work tasks was not available. Another limitation is the possibility of a ‘healthy worker effect’, that is, the cohort members may be on average healthier compared with the male background population of the same age and calendar period categories.22 This bias may contribute to an underestimation of observed risks.
The chimney sweeps were exposed to high levels of soot,2 3 rich in PAH, which are known to induce cancer.5 Exposure to soot and combustion gases in urban air are suspected to cause respiratory and cardiovascular disease.23 All these diseases were in excess among the chimney sweeps. The latency analysis showed a higher risk of all-cause mortality and all cancer after longer time since first employment, giving some evidence for a causative role of the occupational exposures. In contrast, no consistent dose–response associations were observed between employment duration and mortality. However, employment duration may be a poor proxy for exposure, and the absence of association with duration does not indicate absence of effect of the occupational exposure.
Excess alcohol and smoking habits were observed in the cross-sectional survey of chimney sweeps in 1972.16 18 Alcohol increases the risk of oesophageal squamous cell carcinoma and liver cancer, while the dose–response relation between alcohol intake and ischaemic heart disease is U-shaped.24 This gives potential for positive confounding from the rather high intake of alcohol in this cohort regarding these diseases. However, an excess of liver cirrhosis and alcoholism was only observed among chimney sweeps with employment duration of <30 years. The lack of excess of deaths from alcoholism and liver cirrhosis in the group with >30 years of employment indicates that the excess mortality from oesophageal cancer, liver cancer and ischaemic heart disease found in this category is not likely to be caused by excess alcohol habits. Thus, occupational exposure is a more likely explanation.
Tobacco smoking increases the risk of lung cancer, non-malignant respiratory disease, as well as ischaemic heart disease. The proportion of current smokers was higher among the chimney sweeps than in the general population of the same ages. The smoking-adjusted lung cancer SMR was significantly elevated, although the adjustment was based on group-level data from a cross-sectional study. This provides some evidence of an occupational origin of the lung cancer excess, but residual confounding from tobacco smoking cannot be ruled out in the absence of individual smoking data.
Increased risks of cancer mortality among chimney sweeps have previously been observed in two small Danish studies,9 10 and an increased risk of malignant melanoma of the upper limbs was observed among Swedish male chimney sweeps employed in 1970 and followed from 1970 to 1989.11 Moreover, a recent large Nordic record-linkage study from Denmark, Finland, Iceland, Norway and Sweden showed increased risks of cancer of the lung, oesophagus, pharynx, bladder, pancreas and colon in chimney sweeps.25
Increased risks of lung cancer have been associated with occupational exposure to PAH and soot in a large number of studies.5 As discussed above, our finding of an increased risk of lung cancer mortality may partly be explained by tobacco smoking and partly by exposure to chemical carcinogens such as PAHs, arsenic, nickel, chromium or asbestos. The increased risk of oesophageal cancer mortality observed may partly be explained by tobacco smoking and alcohol use among the chimney sweeps. However, PAH exposure from other sources26 and occupational exposure to combustion products27 28 have been suggested as risk factors for oesophageal squamous cell carcinoma. PAHs or dust could be deposited in the airway region, then swallowed, and thereby directly act on the oesophageal mucosa,27 29 thus potentially explaining the finding of an increased risk of oesophageal cancer, especially among those employed for >30 years where alcohol is a less likely cause.
The increased risk of liver cancer among those employed <30 years may be explained by a higher intake of alcohol in this subgroup of chimney sweeps compared with the general population, whereas the excess of liver cancer mortality among those employed >30 years is less likely to be caused by alcohol use.
The reason for the increased risk of bowel cancer in the cohort is not clear. Established risk factors for colon cancer include physical inactivity. Some studies suggest an association with asbestos exposure, while the effect of diet, alcohol and tobacco is weak and under debate.30 Occupational asbestos exposure seems as the most likely explanation, although the chimney sweeps exposure to asbestos is only moderate.2
Increased mortality due to diseases of the circulatory system and ischaemic heart disease was observed in a small study of Danish chimney sweeps10 and in prior updates of this cohort.13 14 16 Furthermore, a population-based Swedish case–control study with adjustment for individual data on tobacco smoking and alcohol intake showed an increased risk of myocardial infarction among workers exposed to combustion products, as well as a dose–response pattern.7 Moreover, in a large cohort study including 248 087 Swedish construction workers, an increased smoking-adjusted risk of ischaemic heart disease mortality was observed among workers exposed to airborne particles (especially diesel exhaust),31 and in a multicenter cohort study including 12 367 asphalt workers, positive dose–response associations between occupational PAH exposure and risk of fatal ischaemic heart disease were found.6 Pathogenetic mechanisms that may explain the increased risks of circulatory diseases is inhalation of small particles inducing an inflammatory reaction in the airways leading to systemic inflammation and coagulation disturbances, or disturbances of the autonomic nervous regulation of the heart.23 31 Our finding of increased risks of diseases of the circulatory system and ischaemic heart disease may partly be explained by tobacco smoking, although occupational exposure to dust and/or PAH is an equally likely explanation.
The increased power of this study allowed us to identify a significantly increased risk of death from asthma, bronchitis and emphysema. In a Swedish study based on census data linked to the Swedish National Patient Register, chimney sweeps had an increased risk of hospitalisation due to asthma.12 Tobacco smoking is a strong risk factor for respiratory diseases, and the excessive tobacco habits among the chimney sweeps may explain our findings, but chemical exposures in the occupation such as soot and combustion gases are other possible causes.
Finally, the findings of strongly increased mortality due to injuries, intoxications and other external causes including suicides were also observed in the prior updates of this cohort.16 The excess seems not to be related to occupational accidents as there was no significant excess of fall accidents. Excess alcohol intake is a likely cause.
In conclusion, the present study provides strong evidence of increased mortality among chimney sweeps from cancer including cancer of the oesophagus, bowel, liver and lung as well as from alcoholism, ischaemic heart disease, respiratory diseases, liver cirrhosis, external causes and suicides. Exposure to toxic substances in the occupation, as well as tobacco smoking and alcohol intake, may to a varying degree have contributed to the findings. The lung cancer SMR remained significantly elevated after approximate adjustment for tobacco smoking based on group-level data. In addition, all SMRs may have been underestimated due to a healthy worker effect, especially among the long-term employed.
We thank the Swedish Municipal Workers' Union and the Swedish Association of Chimney Sweep Masters for valuable assistance and discussions during this study. We specifically thank Peter Hammarin and Anette Karlsson at the Swedish Municipal Workers' Union and Peter Stenholm, EDB Business Partner for valuable information and administration of the computerised membership register. We also thank Dr Kazem Zendehdel for valuable statistical input.
Funding This study was funded by grants from the Swedish Council for Working Life and Social Research, Stockholm, Sweden, and from the Swedish Municipal Workers' Union, Stockholm, Sweden.
Competing interests None.
Ethics approval This study was conducted with the approval of the Regional Ethics Committee, Stockholm (Dnr 2007-306-31).
Provenance and peer review Not commissioned; externally peer reviewed.