Arguments presented by third parties

Asbestosis

1. 
   Please see my answer to Question 3(b).
   

2. Lung Cancer

1. 
   Please see discussion in Section C.1.(i). Please see also the statement from EHC 203 below on the question of a threshold for carcinogenesis by asbestos. Some authorities favour a linear no-threshold model, whereas others argue that a threshold probably exists; nonetheless, there is no general agreement on a numerical threshold for asbestos-induced lung cancer. Dement et al. [171] observed odds ratios of > 2.5 at 2.7-6.8 fibre-years of exposure among South Carolina asbestos textile workers.
   

1. Mesothelioma

   1. 
      As indicated in previous discussion in this report, a linear dose-response model has been identified for mesothelioma induction by the amphiboles, and the dose-response relationship is maintained at low occupational levels of exposure that overlap with environmental exposures: e.g. definite dose-response relationship reported by Rödelsperger et al. [25, 137] at asbestos fibre concentrations in lung tissue of 100,000-200,000 fibres per gram dry lung tissue, with an indication that this relationship is maintained at lower levels of 50,000-100,000 fibres per gram dry lung (the level of 100,000-200,000 fibres corresponds to a cumulative dose of 1-2 fibre-years). Iwatsubo et al. [136] identified an increase in the relative risk at 0.5-0.99 fibre-year. No threshold has been identified for amphibole-related mesothelioma. For chrysotile exposures, a dose-response relationship has also been identified at high exposures, but to the best of my knowledge, there are no dose-response data for low-level exposures to chrysotile.
      
   2. 
      On this point, EHC 203 states that "No threshold has been identified for carcinogenic risks" [for chrysotile; p 144]. At the same time, no increase in risk for mesothelioma has been identified at very low-levels of exposure, of the type associated with well-maintained asbestos in place, in public buildings. However, it is impossible to ascertain whether or not there is an increase in risk at this order of exposure, because no control or reference group can be assembled where there is no asbestos content in lung tissue. If a threshold exists, it must lie somewhere in this area, between no exposure, low-level environmental exposure, and low-level occupational exposure.
      
   3. 
      De Klerk [115] has also commented on the difficulty or impossibility of distinguishing between background versus environmental mesotheliomas:
      

A final consideration in the use of national trend data for estimating environmental effects is the comparison of the likely extrapolated risk from occupational data with the background risk estimated here or from Peto's Los Angeles data. From the Peto paper ... , the incidence of mesothelioma is related to age in the following way:

Incidence = 1.7 • 10^-12 • (age)^3.5

which translates to a lifetime risk to age 80 of just over 100 per million people, which is much greater than any of the estimated environmental risks described earlier. An equivalent risk from the adjusted Western Australian data is about 160 per million lifetimes. The question remains as to how these background risks and environmental risks interact. Is the postulated environmental incidence already included in the background incidence, or should the risks be added or even multiplied together? This question is almost certainly unanswerable using epidemiological methods. ...

4. 
   Clearly, this issue is one major focus of dispute between experts; from the preceding discussions, it is my perception that — with the exception of asbestosis — no threshold has been delineated, and that even those who claim that a practical threshold must exist cannot delineate such a threshold in precise numerical terms (in this respect, I do not know what the expression "practical threshold" really means).
   

5. 
   Thresholds have not been demonstrated for any substance known to cause cancer and there is no theoretical basis to assume a threshold for the diseases related to chrysotile, or other forms of asbestos, particularly when the mechanisms involved in the pathology are not fully understood. Furthermore, it is not possible to determine thresholds from epidemiological studies because of lack of statistical power to distinguish that the risk is virtually zero. [Note: At times, some investigators state that a single point estimate from the lowest dose evaluated in an epidemiological study that does not demonstrate a significant elevation in cancer risk constitutes a threshold level for the carcinogen. Such a conclusion is scientifically invalid. When estimating dose-response, one has more confidence in the risk related to a particular dose level by using all of the data available in the study. Using only a single point estimate results in more instability in the estimate of risk for that data point in contrast to using all of the available data in the study.]
   
6. 
   Dose response analyses and modeling specifically for chrysotile asbestos exposure and lung cancer and asbestosis have been conducted recently by Stayner et al. (1997) using data from the Dement et al. (1994) study. Alternative exposure-response models were evaluated as part of the study. A model designed to evaluate evidence of a threshold also was fitted for asbestos exposure in relation to lung cancer and asbestosis. There was no significant evidence for a threshold in models pertaining to either lung cancer or asbestosis.
   
7. 
   With regard to mesothelioma, and other asbestos-related diseases, I am not aware of any evidence of a threshold pertaining to either chrysotile, or other forms of asbestos. Furthermore, from a practical standpoint, even if there were a threshold for the chrysotile related diseases, the exposures that workers will routinely encounter in the future through continued use of chrysotile in commerce, will expose them to concentrations of asbestos that have already been related to pathology in humans. In other words, continued use of asbestos will continue to expose individuals to levels and exposure circumstances that have already been related to disease. The threshold question, therefore, seems moot.
   

8. 
   It is my understanding that a threshold for disease has not been scientifically established.
   

9. 
   The linear relationship model is generally used as a so-called "conservative" estimate, that is, if it is incorrect, it is more likely to err on the side of safety. In some ways, how one extrapolates risk assessment outside the range of available data is more of a societal decision than a scientific one. Biological plausibility could probably be given to any model.
   

10. 
    In the absence of alternatives because of unavailability of data on exposure-response relationships at low levels of exposure to chrysotile, the linear relationship model is widely employed. Under these circumstances, this may be an appropriate method for risk assessment at low levels of exposure. Whether or not it is a valid method is unknown.
    
11. 
    NICNAS 99 (p. 72) observes that:
    

confounding factors in estimating risks from epidemiological data are possible contamination by other fibre types and inaccurate estimates of historical exposures."

12. 
    Clearly, this too is one major focus of dispute between experts.
    

13. 
    A linear relationship model is appropriate for determining dose response for chrysotile exposure and lung cancer, and perhaps asbestosis and mesothelioma as well, but the most reasonable model for the latter two diseases are less clear than for lung cancer. With regard to this issue, Stayner et al. (1997) evaluated exposure-response relationships for chrysotile asbestos, and lung cancer and asbestosis by applying several alternate models. The exposure-response relation for asbestos and lung cancer gave the best fit when using a linear model. This observation is consistent with the conclusions of other investigators, who have evaluated dose response for chrysotile asbestos textile workers, or other asbestos workers and mortality from lung cancer (McDonald et al., 1983; Peto et al., 1985; Enterline, Hartley & Henderson, 1987). Furthermore, there appears to be a linear relationship between asbestos exposure and lung cancer over a wide range of exposures where such data are available. Therefore, it seems reasonable to accept a linear relationship for lung cancer when extrapolating risks to exposures below the ranges that have been evaluated in epidemiological studies. Moreover, I am not aware of any literature that convincingly proves that the dose response for asbestos and lung cancer is non-linear. Thus, in my opinion, the linear model is the most appropriate model for estimating dose-response for chrysotile exposure and lung cancer.
    
14. 
    A linear relationship might also be used for chrysotile asbestos exposure and asbestosis although one might make the argument that a non-linear model is also appropriate for asbestosis. Stayner et al. (1997) evaluated this issue using data from the Dement et al. (1994) study and concluded that the association between chrysotile exposure and asbestosis appeared to be non-linear. Stayner et al. (1997) used a non-threshold, non-linear model and the estimates of asbestosis predicted from the model seem to fit very closely with point estimates for asbestosis from other studies of chrysotile exposed populations as mentioned in my responses to Questions 3(c) and 4(a).
    
15. 
    An analysis by Peto et al. (1985) of chrysotile asbestos textile workers shows that a linear model fits the data for mesothelioma with the cube of time since first exposure. In this non-threshold model, the response is linear with dose of asbestos, but exponential with time since initial exposure. The predicted number of mesotheliomas by dose and time since first exposure was in reasonable agreement with the observed number. According to the authors, however, there were too few cases to test the model stringently and they did not attempt to fit other models to their data. Nevertheless, given the consistent observations of the long latency period between initial exposure to various forms of asbestos and the clinical manifestation of mesothelioma, it seems reasonable to use a model that is linear with exposure and exponential with time from initial exposure for chrysotile asbestos and mesothelioma.
    

16. 
    It is my opinion that the linear relationship model is the most appropriate one.
    

17. 
    While a threshold model suggests a lack of risk below a fixed level, it is unlikely that this risk would be completely zero, so that if applied to a much larger population, such a risk could lead to cases of disease.
    

18. 
    I am not aware of any other methods that have met with broad scientific acceptance or a consensus. It has been suggested that an S-shaped curve might be more appropriate, but I have not seen any data on what the form of the S-curve might be; in other words, the S-shaped model appears to presuppose the existence of a threshold, but no such threshold has been established to the best of my knowledge.
    
19. 
    The problem with arguing that there exists a practical threshold level for lung cancer and mesothelioma induction is that it is impossible to delineate such a threshold in numerical terms, because of a lack of observational data. (Please see also my answer to Q.5(c)).
    

20. 
    From the public health perspective, it has been the convention to use non-threshold linear models for estimating cancer risk to humans. This is particularly the case for substances known to cause cancer in humans. One might deviate from this concept if the mechanism(s) by which the substance causes the cancer were known. This is not the case with chrysotile, or any other form of asbestos. In the particular case of chrysotile asbestos, Stayner et al. (1997) selected several Poisson regression models to explore the shape of the exposure-response relationship between chrysotile asbestos exposure and risk of death from lung cancer. The models were capable of reflecting a wide range of exposure response patterns, including linear, sublinear and supralinear relationships. They also considered a threshold model to determine whether there was evidence that exposures below a certain exposure concentration were equivalent to zero, i.e., that a threshold was present. As mentioned above, for lung cancer, a linear model gave the best fit; for asbestosis, the response preferred was that based on a non-linear, non-threshold model. In both cases, the models did not provide any support for the existence of a threshold. Thus, in my opinion, these models are appropriate to assess risk for these diseases as a result of occupational exposure to chrysotile asbestos. With regard to lung cancer and asbestos, I am not aware of any public health organization, or governmental agency that has ever used a non-linear model to estimate risk. During the hearings held by the U.S. Occupational Safety and Health Administration (OSHA) as part of its rulemaking related to the standard promulgated for asbestos in 1994, numerous scientists were of the opinion that a non-threshold linear model was the preferred model to use for estimating the relationship between asbestos exposure and lung cancer. It is my opinion that a non-linear model is not an acceptable model to use in estimating dose response for asbestos exposure and risk of death from lung cancer. For mesothelioma, I would favor a non-threshold model that incorporates a linear relationship with exposure.
    

21. 
    I believe that there is no reason to discard the linear model as no threshold for any carcinogen is known to exist.
    

22. 
    A person is "at risk" of developing a chrysotile asbestos-related disease after any exposure to chrysotile asbestos, the lower the amount of exposure, the lower the risk. For example, it could be estimated that there was a 50-50 chance that exposure to 1 fibre of crocidolite could cause 1 case of mesothelioma among the whole population of the world (including all those who have ever lived), i.e. a very small probability, but still greater than zero.
    

23. 
    This question iterates the issue of a threshold exposure. The answer is essentially the same as for Questions 4(a)-4(d) in the absence of exposure-response data at low levels of exposure.
    

24. 
    The answer to this question depends upon the amount of risk that is considered unacceptable by a particular country. It is a matter of health policy. In the United States, the Environmental Protection Agency (EPA) regulates risk to a level below one extra death in a population of 100,000 people over its entire lifetime. I have already provided estimates for excess risk of death from lung cancer and asbestosis from chrysotile exposure. These risk estimates, however, are average risks to a group of individuals that are based on maximum likelihood estimates (MLEs) and they do not incorporate statistical uncertainty in terms of variability, e.g., they are not based on upper 95 per cent confidence limits as is usually the health policy when estimating adverse health effects to a group of individuals at risk from exposure to an environmental insult. In addition, the risk estimates may only be appropriate to workers' health risks. They are derived from a group of healthy adults, who were able to pass a physical in order to gain employment. They are not representative of individuals in the general population who may be exposed to asbestos and have a compromised immune system, or be exposed to other conditions which may exacerbate their risk of contracting the various diseases related to chrysotile asbestos as estimated from a healthy worker population. There will always be some risk from exposure to asbestos and the degree of that risk will depend upon the amount of asbestos exposure in relation to the susceptibility of the individuals exposed in terms of their health status and other factors that interact to produce clinical manifestation of disease.
    
25. 
    It is noteworthy, as mentioned in my response to Question 1(a) above, that the population surrounding the Quebec mining and milling operation (Begin et al., 1992) that was exposed to background levels of chrysotile asbestos, developed mesothelioma at an incidence of 62.5 cases per million population per year, or 0.625 per 10,000 per year. This risk level translates to 0.5 per cent of the population developing mesothelioma over an 80-year lifetime from this background exposure. This estimate may well represent an underestimate of risk since the identification of cases was based on a workman's compensation board review, and additionally, out-migration of inhabitants would also result in loss of some cases. In the same report, it was pointed out that the largest increase in the mesothelioma rate in Quebec was among individuals that had occupations where their exposure would be occasional only, and that 33 per cent of these cases were exposed for less than a 5-year period. When one adds to this information to additional cases of mesothelioma reported in the literature that are associated with standby exposure to chrysotile, it leads one to the conclusion that occasional exposure for a short period of time, or constant low level exposure to chrysotile asbestos leads to death from mesothelioma (and lung cancer and asbestosis). [Note: while an excess of lung cancer was not identified in the Camus et al. (1998) study of women in the same Quebec population, the study had limited power to detect an excess of lung cancer; it did, however, demonstrate an excess of asbestosis and mesothelioma even though out-migration may have resulted in the loss of all three of these diseases in the study.]
    

26. 
    In my opinion any level of exposure to chrysotile (or other form of asbestos) constitutes some risk and that the level of "acceptable risk" is not a scientific issue but an issue for society to debate and determine at different times according to the evidence as they perceive it.
    

27. 
    This is rather outside my areas of expertise. It does however appear theoretically feasible but practically very unlikely given the problems with "downstream" use described above.
    

28. 
    In principle, regulation and control of chrysotile and high-density chrysotile products is feasible at some points of the life-cycle (manufacture and disposal), but in reality not others (please see following discussion).
    
29. 
    The manufacture of high-density products is usually carried out under closed conditions with dust extraction. As one example, the manufacture of chrysotile friction materials in Australia involves the following processes: following transfer of other ingredients required for the product mix, unopened 50 kg plastic bags of raw chrysotile are placed in the mixer and opened under dust extraction. The empty bag is then delivered into a second plastic bag attached to the mixer. When full, this second bag is sealed and taken to a controlled disposal site. Mixing is a closed process. After mixing, the material is emptied under dust extraction before decanting into smaller buckets for weighing and use in moulding and finishing processes. The moulding is a hot process and when complete, the moulded product undergoes finishing processes that include grinding, grooving and drilling — all carried out under dust extraction. The finished disc pads and commercial vehicle brake blocks and linings are then wrapped and packed into sealed containers.
    
30. 
    The potential for exposure includes opening and emptying chrysotile bags into the mixer, the moulding and finishing processes, and handling of damaged bags containing raw chrysotile. The workforce amounts to a few hundred workers, and the maximum exposures per employee vary from minimal, to the largest group involved in the processing operations. Airborne asbestos fibre levels are assessed by personal monitoring. About 84 per cent of 461 samples between 1992 and 1997 were < 0.1 f/ml; 10 per cent ≥ 0.01–≤ 0.2 f/ml; 6 per cent were ≥ 0.02–< 0.5 f/ml, and < 1 per cent were ≥ 0.5 f/ml. The manufacture of compressed asbestos fibre sheeting — most for export and the remainder processed into finished cut gaskets for industrial applications is also a closed process carried out under similar conditions to the manufacture of friction products (please see NICNAS 99, pp 32-34). A total of 232 personal samples between 1991 and 1996 showed similar low airborne fibre
    

31. 
    Although controlled use of this type is feasible for these manufacturing processes, and for disposal of materials left over (e.g. empty polythene bags), it is my perception that, historically and in reality, it is almost impossible to extend analogous controlled and regulated use to the end-users of asbestos products such as workers involved in building construction and demolition (e.g. builders' labourers, carpenters, electricians, painters, plasterers and plumbers), or to individuals who carry out maintenance or renovation work on their own homes, or to brake mechanics (please see AMR 99). This is because these groups add up to a large population of disparate and varied workers; many such individuals work for small businesses or are self-employed, so that it is difficult or impossible to extend controlled use or training to all of them.
    
32. 
    Some of the supporting documentation submitted to the WTO refers to ILO recommendations, but it is also worth emphasizing that prohibition or regulation of asbestos-containing products varies from one nation to another, with different upper limits of airborne fibre concentrations (e.g. < 1 f/ml or < 0.1 f/ml). These are summarized in Tables 27 and 28 and Appendix 7 in NICNAS 99.
    
33. 
    From the literature cited throughout this report and the reasons discussed, it is my perception that broad agreement exists among experts that controlled use of chrysotile (or other varieties of asbestos) is not a feasible option in the real world for certain worker groups, notably those involved in construction trades (e.g. see EHC 203).