Knowledge of mortality patterns following exposure to asbestos have been determined mostly from cohort studies of men who were exposed to asbestos in their workplace, such as those involved in asbestos mining and milling,1 2 manufacturing asbestos insulation and textiles3 and asbestos cement,4 and shipyard workers5 and insulators.6 When women worked in these industries they were a small subset of the cohort and include female east London factory workers,7 World War II gas mask assemblers,8 9 Danish asbestos cement workers,4 German women exposed in various workplaces10 and Italian women compensated for asbestosis.11 More than 1000 women worked in asbestos textile factories in Italy12 and the United States.13 However, women are more likely to have obtained their asbestos exposure domestically: washing the clothes of an asbestos worker14 15 or from their general environment, living close to an asbestos mine, mill or manufacturing industry,16^–19 or living in some other environment contaminated with asbestos.20 21 Few studies examining women with this type of exposure have looked at all causes of mortality, and have instead focused on malignant mesothelioma.16 18 21 All of the earlier studies assessing women and asbestos exposure had limitations, such as the small numbers of women examined or an absence of quantitative exposure measures, and so were not able to examine exposure–response relationships.

The township of Wittenoom, located in a remote part of Western Australia, was formed to support the crocidolite mine and mill which was operated from 1943 to 1966 by the Australian Blue Asbestos Company (ABA). The original settlement was located within Wittenoom Gorge, 1.6 km from the first mine site, but as the workforce grew, the town of Wittenoom was built at the entrance to the Gorge, about 12 km away from the mine. Tailings from the mine, rich in crocidolite fibres, were used throughout the town to pave roads and footpaths, parking areas and the school playground, and for reducing the red dust in backyards. Tailings continued to be used in the town until the mid 1960s. Vehicle movement over the roads raised a great deal of dust and airline pilots claimed they homed in on Wittenoom from considerable distances by identifying the blue haze on the horizon. The state government began decommissioning the town in the 1980s, but it was not until 1992 that all government-owned buildings were demolished and new residents to the town were discouraged.22

The aims of this paper are (i) to examine the causes of death of the women of Wittenoom who were exposed to asbestos in their environment or home and (ii) to examine the exposure–response relationship between mortality outcomes and quantitative measures of asbestos exposure.

METHODS

Wittenoom residents’ cohort

Establishment of the Wittenoom residents’ cohort has been described before.19 The main sources used to identify the cohort and the percentage of people they identified are state primary school records (22%), hospital records (20%), the electoral roll for the Pilbara district (12%), participants in a cancer prevention program23 and associated publicity (18%), questionnaires sent to ABA workers (14%) and Wittenoom birth records (4%). Other sources included records from the Asbestos Diseases Society of Western Australia, the Catholic Church, Wittenoom burial records, and employment lists from the local school, hotel, police, hospital and banks (10%).24 A total of 5097 people were identified as being former residents of Wittenoom and not employed in asbestos mining or milling from 18 553 records collected.19

A questionnaire was sent to all persons traced to an address in Australia between 1991 and 1993 (n = 3244, 64%) excepting those participating in a cancer prevention program from whom the information had already been obtained (n = 641, 13%). Demographic information, date, length and place of residence at Wittenoom, any work they did at Wittenoom, whether they lived with an asbestos worker, whether they washed the clothes of an asbestos worker, smoking and past medical history were collected.24

Consideration of questionnaire responses resulted in 438 people being deleted from the cohort for various reasons including: denied being at Wittenoom, duplicate records, no details of birth or period of residence or lived at Wittenoom for less than 1 month. Therefore, the follow-up status of the cohort to the end of 1993 was as follows: 2173 (47%) had returned a questionnaire, 641 (14%) were participating in a vitamin A program, 460 (10%) were dead, 55 (1%) had permanently departed Australia, 785 (17%) had not returned a questionnaire and whereabouts were unknown since leaving Wittenoom for 549 (11%).20 Among those who did not return a questionnaire, remained untraced or were dead, if they were related to an ABA worker dates and place of residence were assumed to be the same as those of the worker. For those unrelated to an ABA worker, dates of residence were assumed to be the same as other family members provided that at least one family member had known exposure. Dates of residence were taken as those found on the various sources used to establish the cohort for all other residents. If the untraced person was the wife of a worker, it was assumed that she washed his clothes.20 25 The residents’ cohort was considered complete when comparisons between it and the population of Wittenoom recorded at various census dates showed a close correspondence.19

As work has continued on the development of the cohort since 1993, this accounts for any difference in the number of subjects from the earlier reports.19 20 25 As at the end of 2000 there were 2608 women and 2160 men in the residents’ cohort.26

Follow-up of Wittenoom women

In 2005 and 2006 in order to update our follow-up of the women from Wittenoom, all females without a death record who were not attending a cancer prevention program23 and were not located on the electoral roll for Western Australia were searched for in the marriage register of Western Australia to determine if a change of surname had occurred. The search commenced from the year they were last known to be alive. Death certificates of any spouse or birth, death and marriage certificates of children were also sought in an attempt to obtain the wife or mother’s maiden name and date of birth. A total of 235 women were identified as having changed their surname. Fifty six women were excluded from the cohort because they had insufficient identifying information (missing date of birth, first name, etc) or because they were residents of Wittenoom for less than 1 month. This left 2552 women in the cohort.

Asbestos exposure assessment

Between 1948 and 1958, dust concentrations were periodically measured in the mine and the mill by the Mines Department of Western Australia using a koniometer.1 The first fibre count survey of the Wittenoom environment as well as the mine and mill was conducted in 1966 for the Mines Department of Western Australia using a Casella long running thermal precipitator.1 Monitoring next occurred in 1973 when dust samples were obtained from personal and fixed positional monitors in and around the township. Further samples were taken in 1977, 1978, 1980, 1984, 1986 and 1992.25 Based on these surveys, former residents of the township of Wittenoom not working directly with asbestos were assigned an intensity of exposure of 1.0 fibre per millilitre of air (f/ml) from 1943 to 1957, when a new mill was commissioned, and then 0.5 f/ml between 1958 and 1966, when mining operations ceased.20 25 Interpolation between those surveys which used personal monitors assigned exposures from 0.5 f/ml in 1966 to 0.010 f/ml in 1992.

Cumulative exposure was calculated by summing over all years of residence the product of fibre concentration for each year and length of time spent in Wittenoom during that year. This figure was adjusted by a factor of 4.2 ((24×7)/(8×5)) to allow for continuous 24 h exposure rather than the 8 h a day 5 days a week measurement method used to determine the fibre levels.20 25 The average intensity of exposure was calculated by dividing cumulative exposure by duration of residence at Wittenoom.

Case ascertainment

The cohort was linked to the Western Australian Registrar General’s Mortality Database to the end of 2004. National deaths to the end of September 2000 (for those women not resident in Western Australia) were obtained from the National Death Index and linked by the Australian Institute of Health and Welfare. Cause of death as recorded on the death certificate was defined using the International Classification of Disease (ICD), Revisions 7–10 as appropriate for the time period. Expected numbers of deaths were estimated using age, period and cause specific death rates calculated for the Western Australian female population in 5-year periods from 1970 to 2004. Death rates for the period 1970–74 were used to calculate expected rates from the person-years accumulated for the first period of 1950–1969 as period specific rates were not derived separately for those years. Standardised mortality ratios (SMRs) were calculated as the ratio of the observed deaths to expected deaths. Confidence intervals were assessed by treating the observed number as a Poisson count with expectation equal to the particular expected numbers. Given that many of the women (21%) (table 1) were lost to follow-up and given the almost complete ascertainment of deaths in Western Australia, the usual method for calculating expected deaths would lead to a probable overestimate of risk.1 Therefore, two methods were used to derive expected deaths, to show a minimum and maximum estimate of effect, based on differing censoring dates. The first method assumed that all women not known to be dead and not known to have migrated were alive at the end of 2004 (or if they were residents of other Australian states alive until September 2000). This assumes that all those who were lost to follow-up remained alive to the end of follow-up and is therefore likely to overestimate person-years at risk and so underestimate the SMR. The second method censors women at their date last known to be alive if they were not known to be dead or to have migrated and is therefore likely to underestimate person-years at risk and so overestimate the SMR. Both methods censored women at age 85 years if they were not known to have died before that age.

Ethics approval was obtained from the Human Research Ethics Committee of the University of Western Australia.

Exposure–response relationships

A nested case–control design was used to examine exposure–response relationships. Cases were those women who died of an event of interest during the study period. Controls were those not known to have died by the year of death of the case and who were the same age as the case, in 5-year age bands. Controls could be controls for more than one case and cases could be controls in years prior to their death. All eligible controls were used in the analysis rather than a fixed number of controls per case in order to maximise information. Conditional logistic regression was used to relate asbestos exposure to mortality outcome. Skewed asbestos exposure measurements were transformed to their natural log. All analysis was undertaken using Stata 9.0.

RESULTS

Follow-up

Of the former female residents of Wittenoom, 1586 were known to be alive at the end of 2004 (table 1), 425 were dead and 541 (21%) were lost to follow-up, most having not been traced beyond the date that they left Wittenoom.

Most women arrived at Wittenoom in the 1950s and 1960s (table 2). Forty four per cent were aged between 15 and 40 years when they first came to Wittenoom, while 46% arrived as children aged less than 15 years or were born there. Forty per cent stayed for less than 1 year, with the median length of stay being 484 days (range 183–1096 days), while 59% stayed longer than 1 year. Ninety nine per cent of residents had asbestos intensity exposures estimates of less than 5 f/ml and 81% had estimated cumulative asbestos exposure measurements of less than 10 f/ml years. The median estimated cumulative exposure for all resident women was 3.15 f/ml years (range 1.23–7.35 f/ml years). Sixty four per cent of former residents lived with an ABA asbestos worker while they were at Wittenoom and 22% reported washing the work clothes of an ABA asbestos worker (table 2).

Mortality

There were 425 deaths in the cohort between 1950 and the end of 2004. Six of these deaths were excluded from further analysis because of missing data.

Mortality from all causes was around 1.0 for SMR1 (women lost to follow-up censored at December 2004) and was increased by 13% for SMR2 (women lost to follow-up censored at date last known to be alive) (table 3). Mortality from neoplasms and symptoms, signs and ill defined conditions was significantly greater than that of the Western Australian female population irrespective of which censoring method was used. Mortality from circulatory diseases was significantly lower in this cohort (SMR1) than in the Western Australian female population.

Of the 19 women whose cause of death was recorded as symptoms, signs and ill defined conditions, two were coded as dehydration while the remainder were unknown.

More specific causes of death showed that mortality from lung cancer and pneumoconiosis (asbestosis) was significantly raised irrespective of which censoring method was used (table 4). Mortality from upper aerodigestive cancers, bronchitis and emphysema, cancer of the ovary and tuberculosis was increased irrespective of which censoring method was used, but not statistically significantly. Mortality from cancer of the small intestine and cancer of the colon and rectum was less in the Wittenoom women (SMR1) than in the Western Australian female population.

We were not able to fully assess the impact of tobacco smoking on lung cancer mortality because we do not have smoking information for all of the women. Of the 60% of women residents who completed a smoking questionnaire in the early 1990s, 26% reported currently smoking, compared with 20% in the Australian female population.27 Using these figures and applying Axelson’s adjustment,28 we estimate that the SMR for lung cancer was increased by a factor of 1.42 due to confounding by smoking. So SMR1 might be adjusted down to 1.23 and SMR2 to 1.51.

Malignant mesothelioma

There were 30 deaths from malignant mesothelioma of the pleura. The first death from mesothelioma, in a woman aged 74 years, occurred in 1975, 23 years after being first exposed to asbestos. Cases had a greater intensity of asbestos exposure, a longer duration of residence at Wittenoom and therefore a greater cumulative asbestos exposure than controls. Twenty six (90%) lived with an ABA asbestos worker and 16 (53%) reported washing the clothes of an ABA asbestos miner or miller (table 5).

Of the 30 known cases of malignant mesothelioma of the pleura in this cohort, 14 were recorded on the death certificate as having died from malignant neoplasm of the pleura, six had malignant mesothelioma of the pleura, one had mesothelioma of other sites and one had mesothelioma site unspecified. Three were recorded as having died from lung cancer, one from renal disease, one from a malignancy with ill defined or unspecified site, one from a benign neoplasm of the pleura, one from asbestosis and one had no cause of death coded. There were no deaths from peritoneal mesothelioma.

Exposure–response relationships

Mortality from mesothelioma was significantly associated with cumulative asbestos exposure (table 6). The risk of mesothelioma tended to increase for those residents who were known to have washed the clothes of or lived with an ABA asbestos worker (HR = 2.61, 95% CI 0.85 to 7.99 and HR = 2.67, 95% CI 0.77 to 9.21, respectively). Mortality from lung cancer was not significantly associated with cumulative asbestos exposure in this cohort of women. Repeating this model with an untransformed cumulative exposure (eg, unlogged) gave an OR of 1.00 (95% CI 0.95 to 1.04) for cumulative exposure.

DISCUSSION

This is one of only a few studies that have examined mortality in women exposed to asbestos in their environment or home.29 30 It builds upon this work having a larger number of exposed women and a longer follow-up. All the women were exposed to crocidolite and quantitative estimates of exposure have been calculated and exposure–response relationships examined. Excess mortality was found for all cause mortality (SMR2), neoplasms, malignant mesothelioma, lung cancer, and cancers of the upper aerodigestive tract, pneumoconiosis and symptoms, signs and ill defined conditions. Circulatory disease mortality and mortality from cancer of the intestine and cancer of the colon and rectum were lower than in the general population. There was a statistically significant exposure–response relationship between cumulative asbestos exposure and pleural mesothelioma.

Twenty one per cent of the women were lost to follow-up, most from the time they left Wittenoom. The Western Australian Marriage Registry was searched to ascertain any changes of name, but this was not possible for other states of Australia. The high loss to follow-up reported in this cohort may have led to either an underestimation or an overestimation of mortality from asbestos-related disease.

Selection bias may have occurred because of the methods used to establish the residents’ cohort. Together, the records from the Wittenoom doctor, hospital and school provided information on 41% of residents. Those who were resident at Wittenoom for only short periods, did not have or were not children at the local school and did not attend the doctor or hospital had a lower chance of being included in the cohort. Therefore, there may have been some exclusion of healthy adults in favour of the less healthy. This bias may have overestimated our results with respect to some diseases and all cause mortality. However, it is unlikely to have influenced the estimate of risk associated with asbestos exposure as good health during residence at Wittenoom was unlikely to be related to exposure.

Period specific mortality rates for the period 1950–1969 were not available for Western Australia, so the rates for 1970–1974 were used to calculate expected rates for those years. If the mortality rate for 1970–1974 was very different to actual mortality in 1950–1969, then any differences in mortality among the Wittenoom women might not be seen. However, we do not think this is a big problem for this cohort as there were only 41 deaths between 1950 and 1969 and so any effect would be small.

Our estimates of mortality are comparable to those few studies that examined causes of mortality among women environmentally exposed to asbestos. Ferrante et al report on a cohort of wives of asbestos cement factory workers. They show an SMR of 1.02 (95% CI 0.92 to 1.12) for all cause mortality, 1.42 (95% CI 0.71 to 2.54) for ovarian cancer, 2.69 (95% CI 1.85 to 3.77) for cancers of the respiratory system (including pleural neoplasms) and 4.86 (95% CI 2.66 to 8.16) for poorly defined causes. Their estimates of lung cancer of 1.17 (95% CI 0.60 to 2.04) and neoplasms of 1.07 (95% CI 0.91 to 1.26) were lower than those we reported.29 Luo et al report an SMR of 0.59 for all cause mortality and 1.34 for lung cancer among 5603 adults born between 1915 and 1955 in Da-yao, China exposed to crocidolite in surface soil.30

Few studies reporting on non-occupational exposure are able to examine exposure–response relationships. We reported a statistically significant exposure–response relationship for cumulative asbestos exposure and mesothelioma. Similarly, Ferrante et al report an (irregular) increase in pleural mesothelioma SMRs by duration of exposure and latency period.29 An earlier report on the same cohort of women showed increased SMRs for mesothelioma and lung cancer by duration of exposure.15 Estimated cumulative domestic exposure of 32.9 f/ml years (SD 18.5) and residential exposure of 102 f/ml years (SD 10.1) was higher among 10 women with mesothelioma from the Thetford Mines area of Quebec than among controls at 32.5 f/ml years (SD 15.9) for domestic and 63.1 f/ml years (SD 44.8) for residential exposure.31 Metintas et al report 12 cases of pleural mesothelioma among women in a cohort study in Turkey giving an SMR of 143.9 (95% CI 81.7 to 253.4), but the exposure–response relationship was not examined.21 Case studies also report mesothelioma following domestic exposure.32 33

Nineteen women died from symptoms, signs and ill defined conditions. A comparable number of cases (n = 23) was reported among the male ABA workers to the end of 2000.34 Similarly, Magnani et al reported seven cases (SMR = 356.3, 95% CI 143.2 to 734.2) among 1964 wives of asbestos cement workers,15 increasing to 14 deaths (SMR = 4.86, 95% CI 2.66 to 8.16) with follow-up to April 2003.29 Most of the deaths (n = 14) among the Wittenoom women occurred in the period 1950–1974. A further three deaths, occurring between 1984 and 1990, were coded as malignant neoplasm without the site being specified (ICD-9). Traditionally mesothelioma has been difficult to diagnose and historically mesothelioma has been recorded poorly on death certificates,35 often being coded as lung cancer, carcinomatosis or site unspecified. This problem may be compounded among women. Data from the Western Australian mesothelioma register show that 41% of women, compared with 12% of men, with mesothelioma report no known asbestos exposure or that no source of asbestos exposure can be ascertained (work in progress). Given this and the (albeit earlier) difficulty of diagnosing mesothelioma, it is not unexpected that a rare industrial disease was not the obvious choice of the diagnosing clinician, although diagnostic accuracy appears to be improving with time. It may be possible that some of the ill defined deaths among the Wittenoom women were mesotheliomas, although the latency period (for those who died between 1950 and 1974) is somewhat short for mesothelioma.

The quantitative asbestos exposure measurements derived for each woman in the cohort were estimated from environmental monitoring that occurred at various time points and assigned to individuals based on their period of residence at Wittenoom. Therefore, they may be different to the actual exposure experience of each woman. We have sought information on whether the women lived with or washed the clothes of an ABA asbestos worker in order to obtain a more accurate depiction of their asbestos exposure. However, we have no extra knowledge as to behaviour at Wittenoom that may have increased asbestos exposure (eg, did the woman garden while at Wittenoom, play outdoor sport, etc). These measures have demonstrated internal validity as exposure–response relationships have previously shown.20 25 26

Mortality from lung cancer was high in this cohort, but we did not find a significant exposure–response relationship between asbestos exposure and lung cancer. There are several potential reasons for this. Firstly we had only a small number of cases (n = 30). Secondly, possible errors in estimating the women’s individual asbestos exposures may bias the results towards the null hypothesis, therefore showing no relationship. Also, the way that the exposures were derived based on environmental measures at certain time points provided a somewhat limited range of exposure values, which may have led to a reduced statistical power to detect any possible association. Other studies that have examined communities environmentally exposed to asbestos have also failed to report an excess risk of mortality from lung cancer. Among men living near an amosite asbestos factory in New Jersey, lung cancer deaths represented 2.3% of all deaths compared to 2.6% in the comparison population.36 Women from the asbestos mining areas of Quebec had an SMR of 0.99 (95% CI 0.78 to 1.25).37 Lung cancer mortality rates of 18.7 per 100 000 and 80.6 per 100 000 were reported for non-occupationally exposed female and male residents of Casale Monferrato, the location of the largest Italian asbestos cement factory. These rates were not different to those of the rest of the population in the region.38 Similarly, an excess risk in lung cancer mortality was not found among Italian women with domestic asbestos exposure (SMR = 149.8, 95% CI 55.0 to 326.1).15 The ABA workers for whom we have previously reported a significant exposure–response relationship between asbestos exposure and lung cancer mortality (RR = 1.12 per year of exposure (95% CI 1.04 to 1.20) and RR = 1.01 (95% CI 1.00 to 1.01) per fibre/ml) were probably exposed to more respirable size fibres than the women.39

We found reduced mortality from cancers of the intestine and colon and rectum in this study. Among the male Wittenoom workers we reported an SMR of 1.31 (95% CI 0.99 to 1.74) for the period 1979–2004, which was not associated with asbestos exposure.40 Other studies have suggested an association between asbestos exposure and cancer of the colon and rectum, but the relationship was not strong or consistent.41 42 Berry et al reported an exposure related excess of 1.83 (95% CI 1.20 to 2.66) among their cohort of east London asbestos factory workers but suggested that the excess may have been due to peritoneal mesotheliomas being certified as cancers of the colon.3

CONCLUSION

Women who were former residents of Wittenoom and were exposed to asbestos in their environment or in their home have excess cancer mortality, including mesothelioma, compared with the Western Australian female population. The observed exposure–response relationship between cumulative asbestos exposure and malignant mesothelioma was consistent with that seen among male asbestos miners and millers at Wittenoom.