Default Dose Response Relationships for Lesions caused by Chronic Arsenic Poisoning
Mallinckrodt Research Professor of Physics
Presented to the US EPA as a public comment upon the arsenic regulations proposed on Thursday June 22nd.
Modified from a poster presentation at the 4th International Arsenic Conference, San Diego, June 2000 and modified from a presentation to the Science Advisory Board of EPA on June 9th 2000.
In considering a possible dose-response relationship it is important to distinguish two possible cases. Firstly a situation where the lesion has a unique cause, and second where the lesion is indistinguishable from a lesion that occurs naturally. Arsenic caused lesions are of both types. The skin lesions, dyspigmentation, keratoses and perhaps skin cancers all seem to be unique, whereas the internal cancers, lung, bladder, kidney and liver are indistinguishable from other cancers.
That chronic arsenic poisoning causes skin lesions was reported by Hutchinson (1887,1888) but was widely ignored until Tseng et al. (1965) found skin lesions, "blackfoot disease" and skin cancers, in Taiwan which he attributed to arsenic exposure. Tseng's work has been analyzed and reanalyzed many times. The result of one such reanalysis (Byrd et al. 1996) is shown in figure 1. The skin cancer rate is plotted against the average concentration in the water. It clearly shows a either a non-linearity or a threshold (hockey-stick) behavior with a threshold at about 130 ppb in the water.
Tseng's data are merely an "ecological" study, because only average rates are plotted against average concentrations. It has been criticized both on this ground and because the attribution to arsenic may not be correct. However skin lesions (other than blackfoot disease) have been seen among arsenic exposed people elsewhere - Inner Mongolia for example. Recently the individual outcomes and concentrations for 3000 villagers were followed (report of IMCAP to ATSDR 2000). These, plotted in figure 2 show a threshold when the probability of a lesion is plotted against concentration, at 50 ppb for dyspigmentation or keratoses and 300 ppb for skin cancer. Taken together we suggest that these data suggest an appropriate dose response relationship for skin lesions is a hockey stick function with a threshold above 50 ppb.
Although it is desirable to check this conclusion in other cohorts, especially since skin lesions might behave somewhat differently among different races, there seems little reason to doubt that there is a threshold, or at least a gross non-linearity, for the production of skin lesions.
But the situation for the internal cancers is very different. In a couple of papers which surprised most scientists, Chen and collaborators (1985,1986) showed that there is a considerable increase in mortality from internal cancers among those who drank water from the arsenic laden wells in Taiwan. Again these data have been analyzed and reanalyzed. Byrd et al show plots of rates of various internal cancers plotted against average concentrations. The plot for bladder cancer in men is shown in figure 3 and for total cancer mortality in figure 4. Again conclusions from these data must be tentative because it was an ecological" study. But the cancer rate when plotted against concentration shows an excellent fit to a straight line with no threshold - in contrast to the situation with skin lesions. This suggests (but does not prove) that internal cancers behave very differently, and may have a different mechanism, from skin cancers.
More recently, case control studies have been performed in Chile (Ferreccio et al. 1998) and Argentine (Hopenhavyn-Rich et al., 1996,1998) which show unequivocally a large increase in bladder, kidney and lung tumors with arsenic concentrations in water of 500 ppb. Figures 5 and 6 (taken from Professor Allan Smith s report to WHO) shows the data for bladder cancer and lung cancer each with a straight line from the Taiwan data superposed. Although the line fits the data adequately, inadequate statistical accuracy in distinguishing an effect from a background prevents finding effects at much lower levels so that the dose-response has to be judged from indirect data or general principles.
Relating cases to natural background
Crowther (1924) suggested a simple single stage model for radiation induced carcinogenesis whereby radiation ionizes a cell that replicates and leads to a cancer. In its simplest form the theory leads to a linear dose response relationship. But his theory was obviously incomplete early on. Cosmic rays ionize thousands of atoms in the body each second. There must exist some mechanism, repair or excretion, that prevents all but one in billion ionized atoms from proceeding to form a tumor. Nonetheless it is widely, but erroneously, believed that only mutagenic compounds can lead to a linear dose relationship. The initial step of modifying a cell to start a tumor is probably only one step in cancer formation and any step can be modified. Indeed the idea of the 1970s that genotoxic materials are especially carcinogenic and only genotoxic materials can give a low dose linearity runs into many troubles and cannot be sustained as a general principle. It is likely that the major action of even genotoxins in the environment is to promote a cancer already initiated by natural processes.
The multistage models, although originating in the 1930s were developed by Doll and Armitage (1954, 1957) to describe the distribution of cancer as a function of age. To do this they found that it was necessary to consider that cancers develop in 4 or 5 stages and that each stage may be influenced by a different biological mechanism. In applying the model to cancers caused by anthropogenic activity they suggested that one (or at most two) stages be influenced by pollutants. Inherent in this description was the assumption that this influence would act in the same way as whatever in the background influenced this stage.
It was evident in 1954 that, given this assumption, at low doses of pollutants that the effect would be linear with dose and effects of different pollutants would add. But at high doses, a multiplicative synergism between two pollutants would naturally occur as the probability of a stage exceeds the background probability in two separate stages. According to these ideas, therefore, non genotoxic substances are as likely to lead to a linear dose response as genotoxic substances.
Crump et al. (1976) and Guess et al. (1977) pointed out that the argument for low dose linearity is far more general than the Doll Armitage theory and depends solely on the fact that cancers caused by the pollutant and natural (background) processes are indistinguishable and therefore it is likely that the pollutant and the background act in a similar way at some stage in the cancer induction process. Crawford and Wilson (1996) showed that the argument is even more general and applies to a wide variety of non-cancer outcomes.
These analyses were used by US EPA 25 years ago as a justification for assuming low dose linearity as a general default. But the issue arises how to derive the low dose slope from high dose data. Unfortunately EPA's attempts confused the issue. Their use of the words "Linearized MultiStage (LMS) model" implied more biological and mathematical justification than existed. Zeise et al (1987) objected in vain and proposed that they more honestly say "truncated polynomial model". More recently Cox (1997) and Chiu et al. (1999) have produced a most welcome precise mathematical formulation.
These general ideas should therefore be used to suggest a dose-response relationship for the internal cancers produced by arsenic. The default will then be linear at low doses. When combined with a desire of the US EPA to regulate any (lifetime) risk larger than one in a million there are difficulties. The dose for a one in a million risk is between 1 and 5 parts per trillion when lung, kidney and bladder cancers are all included. Background levels exceed this by a factor of 1000! This then is the core of the problem regulators have faced for the last 14 years in considering the standard for arsenic in drinking water.
There is general agreement that one should use "scientifically motivated risk assessment" whenever possible, although there is far less agreement about what that means. I contend that the general (default) arguments above are very scientific. What data, direct or indirect, might be obtained to move away from the default? In this problem I find that most of the discussions of toxicologists are not helpful since they fail to discuss the natural processes at the same time as they discuss their ideas about arsenic related processes. Indeed the whole world was misled by a misunderstanding of animal toxicology. Rats and mice cannot (easily) be persuaded to get cancer from arsenic. Ergo, men cannot get cancer either and for a century (1888 to 1986) data that suggested otherwise (albeit with small statistical samples) were discounted and thought to be in error.
The task for a toxicologist who wishes to depart from the linear default is a daunting one. It is insufficient for him/her to have a theory that describes how arsenic produces a cancer, unless that theory also describes how the natural cancers occur and whether there is a difference. I am unaware of any such complete description.
But an important purpose of the mathematical models must be to point out where scientific (usually biological) data will be most useful in elucidating the low dose behavior. Statistical sampling errors would prevent any direct demonstration of a threshold in internal cancers if such a threshold were at 50 ppb or below. But we can ask an indirect (but leading) question. Are internal cancers always preceded by, or accompanied by, a skin lesion? As stated the answer must be no. For with no arsenic exposure at all (natural background) there exist internal cancers. But I note that at arsenic levels of 500 ppb the rate of skin lesions is only about 20%. Then we can modify the question: "at 500 ppb is the increase in internal cancers solely among those with skin tumors, or is it also among the larger group of persons without skin tumors?" If the former, then I would argue that the dose response for the internal cancers might well follow that for the skin lesions and show a threshold. For example, if the ideas of Dr Menzel presented at the 4th International arsenic conference at San Diego are correct I would expect just such a difference in the epidemiological studies. It is of course important to select in a blind fashion, persons with the 500 ppb exposure BEFORE asking whether or not they have skin lesions or internal lesions.
This question is very similar to that asked by Dr Mereweather, Chief Inspector of Factories in UK in 1938. "Is it asbestos, or the asbestosis caused by asbestos, which is the cause of the lung cancers? For if the former, a linear dose response relationship is likely from the Crump et al. arguments, if the latter then a threshold is probable. It is also similar to the question asked about benzene: "are the leukemias caused by benzene always preceded by pancytopenia or not?" For if the former, a threshold is probable. I note that neither in the asbestos case nor in the benzene case has a definitive answer yet been forthcoming, and EPA assume the linear default.
How should society cope with default linearity?
The above merely notes what a default risk assessment might be. It should not by itself be used to argue for any particular level for an arsenic standard. I note that if the default linear dose response applies to radiation induced cancers also, as is often believed, in spite of Dr Cohen'. s studies that show unequivocally that lung cancer rates are LOWER in counties with high radon levels than in counties with low radon levels, then the lifetime risk at the level of the natural background radiation is about 0.2%. This is about the risk of arsenic at a 5 ppb concentration of arsenic in the drinking water - including a water equivalent of arsenic in the foodstuffs. Neither radiation nor arsenic can be regulated at a one in a million level. As I have stated publicly many times in the last 21 years EPA cannot do so consistently. Their attempts to do so are arbitrary, capricious and possibly illegal. Society, Congress representing the society, and EPA executing the will of Congress, must come to grips with this issue.
Since society has coped moderately well with radiation, I suggest that regulation of arsenic should follow similar rules. Many years ago the International Commission on Radiation Protection (ICRP) proposed that average anthropogenic radiation doses to the public be kept below 170 mrem/year, although individuals might reach 500 mrem/yr. 500 mrem/yr adds a risk, assuming a linear no threshold theory with the usual slope of nearly 1%. In order to achieve this society has set a few rules, such as the NRC rule that radiation levels at the site boundary of a nuclear power plant should be kept less that 10 mrem/yr. But the main rule is As Low As Reasonably Achievable (ALARA) which could apply to arsenic (and many other pollutants) also. This was interpreted (although not yet applied very often) by NRC (1976) as meaning that society should spend $1000 per Man Rem on reducing exposure, (updated in 1992 to account for inflation and political correctness) to $200,000 per person Sievert and the implication (not always followed) is that ONE SHOULD NOT SPEND MORE. Using a slope of an assumed linear dose response of one fatal cancer per 30 Sv a linear dose response and the usual slope, this corresponds to about $6,000,000 per calculated cancer averted. I note that this is about the same as the $4,000,000 per statistical life that EPA proposed in summer 1998 for cost benefit calculations but a little less than EPA proposed in discussion of the proposed arsenic standard.
Although it has been said that consistency is a refuge of small minds, it is worth enquiring what consistency between regulation of radiation exposure and of arsenic exposure would entail. For arsenic, a literal following of ICRP would lead to an acceptance of 50 ppb as a principle, with rules to keep individual exposure from single large facilities below 1 ppb. An ALARA principle could be that one should spend a sum of $1,000,000 per person-ppb to reduce arsenic exposure. (At a concentration of 1 ppb the risk is about 25% and 25% of $4,000,000 is $1,000,000). The 1996 Amendments to the Safe Drinking Water Act (SDWA) for the first time explicitly granted EPA discretionary authority, if it determines that the technically feasible level does not justify the costs, to adjust the standard to a level that maximizes health risk reduction benefits at a cost that is justified by the benefits. Now that this discretionary authority exists, there seems to me no good reason why the EPA should not use cost explicitly in the discussion of alternatives and come into line with the thinking of risk analysts world-wide and recent thinking of other regulatory bodies.
It is clear that the main cost burden of nation-wide compliance with a reduced arsenic standard will fall upon some small towns in the western states. At the 4th International arsenic conference a representative from a small California town complained about this. But if he fails to meet the new standard, those affected are only his voters and not any do-gooders on the eastern seaboard. Following this line of reasoning I suggest an alternative to the compulsion that the EPA proposed on May 24th. There should be an absolute limit of 50 ppb as now, compulsory for all. Above this level health effects have been definitively observed and below it they have not. Each water district would be at liberty to vote on whether to adopt a lower standard of 3, 5, 10 or 20 ppb and could do so if they could justify to EPA that this meets the ALARA principle. This justification would be based on the exposure averaged over time and averaged over people to arsenic laden water not the peak exposure. A standard of 50 ppb probably leads to average exposures in a community of 20 ppb. Then this economic rule suggests that a community of 1000 people should be willing to spend $20 million to reduce their exposure below the 450 ppb standard, but not more. Since most of the water supplied to a household is used for functions such as flushing a toilet, or bathing, and since dermal absorption and evaporation of arsenic is small, a community should be permitted the option of switching to bottled water for drinking and leaving the standard alone.
Long-term disposal of arsenic
More important, however, are the long term implications of bringing arsenic from secure storage below the ground to the environment above the ground. Of course we have been doing that in mining activities for 3000 years. In Bangladesh, for example, much irrigation is by water from arsenic-laden tube wells. The arsenic can build up and cause the arsenic level in foodstuffs to steadily increase. In considering this EPA should be guided by a similar concern that has been expressed for high level nuclear waste. Here there is considerable concern that the waste is long lived, and cannot be broken down as one hopes that organic chemicals are broken down with time. Half lives of thousands of years cause concern. But I note that the half life of arsenic is infinite. Fortunately we no longer spray 40,000 tons a year (20,000 tons imported) on our crops and forget about it. But, in line with our concerns about materials which are carcinogenic solely because of their radioactivity (where we insist on accurate tracking of radioactive sources), we might insist that ANY quantity of arsenic greater than 1 gram be tracked. The arsenic, being carcinogenic for ever, should obviously be placed in a landfill at least as secure as planned (at Yucca Mountain for example) for long lived nuclear waste. In line with the EPA requirement that no one s radiation exposure be increased by more than 2 mrem per year if there is an accident at a nuclear repository, EPA should demand that no one s exposure to arsenic be increased by more than 1/4 ppb as a result of an accident at an arsenic repository. For arsenic of course this must be satisfied for ever, in contrast to the nuclear requirement of a few thousand years.
If these suggested rules for arsenic seem unreasonably stringent to you, then I suggest that you recommend to those at EPA and NRC that are considering the matter that rules for the comparable hazards of radioactive materials be modified to match whatever rules you finally adopt for arsenic.
Armitage P. and Doll R., (1954) Brit J. Cancer 8:1-12
Armitage P. and Doll R., (1957) Brit J. Cancer 11:161-169
Byrd D.M. , M Luann Roegner, James C. Griffiths, Steven H. Lamm, Karen S. Grumski, Richard Wilson, Shenghan Lai, Carcinogenic risks of inorganic arsenic in perspective. International Arch Occupational Environmental Health, 68:484Ä494, 1996.
Chen, C.J., Y.C. Chuang, S.L. You, and H.Y. Lin. 1986. A Retrospective Study on Malignant Neoplasms of Bladder, Lung and Liver in Blackfoot Disease Endemic Area in Taiwan. Br J Cancer, 53:399Ä405.
Chen, C.J., Y.C. Chuang, T.M. Lin, and H.Y. Wu. 1985. Malignant Neoplasms Among Residents of a Blackfoot DiseaseÄEndemic Area in Taiwan: HighÄArsenic Artesian Well Water and Cancers. Cancer Res, 45:5895Ä5899.
Chiu, W.A. Hassenzahl D.M., and Kammen, DM. (1999) Risk Analysis 19:15
Cox, L.A. (1995) Risk Analysis 15: 359
Crawford, M. and Wilson, R. (1997) "Low Dose Linearity the Rule or the Exception" Human and Ecological Risk Assessment 2:305-330
Crowther J, (1924) "Some Considerations Relative to the Action of X-rays on Tissue Cells" Proc. Roy. Soc. Lond. B Biolog Sci. 96:207-211
Crump, K.S., Hoel, D.G., Langley, C.H., and Peto, R. (1976) "Fundamental Carcinogenic Processes and their Implications for Low Dose Risk Assessment" Cancer Research 36:2973
Ferreccio C, et al Lung cancer and arsenic exposure in drinking water: a caseÄcontrol study in northern Chile. Cad Saude Publica 1998;14 Suppl 3:193Ä8
Guess, H., Crump, K. and Peto R. (1977) "Uncertainty Estimates for Low Dose Rate Extrapolation of Animal Carcinogenicity Data" Cancer Research 37:3475-3483
HopenhavynÄRich C, et al Lung and kidney cancer mortality associated with arsenic in drinking water in Cordoba, Argentina. Int. J Epidemiol 1998 Aug;27(4):561Ä9
HopenhaynÄRich C, Biggs ML, Smith AH, Fuchs A, Bergoglio R, Tello E, Nicolli H. Bladder cancer mortality associated with arsenic in drinking water and in Cordoba, Argentina. Epidemiology, 7:117Ä124, 1996.
Hutchinson, J. (1888) Diseases of the skin: On some examples of Arsenic_keratosis of the skin and of ArsenicÄCancer. Transactions of the Pathological Society of London 39:352Ä363, 1888.
Hutchinson, J. 1887. Arsenic cancer. Br Med J, 2:1280Ä1281
NRC (1975) Rule making RM-30-2
NRC (1992) Letter of Tom Kress, Chairman, ACRS, updating RM-30-2 to account for inflation and political correctness.
Zeise, L., Crouch, E.A.C. and Wilson R., (1986) "Dose response for Carcinogens: A Review" Environmental Health Perspectives 73:259