Risk Assessment in California and Biological Effects of Low Level Exposures
Joseph P. Brown, Ph.D.*, David W. Morry, Ph.D., and Robert A. Howd, Ph.D.
Pesticide and Environmental Toxicology Section
Office of Environmental Health Hazard Assessment
California Environmental Protection Agency
Oakland, CA, 94612
* To whom correspondence may be addressed
Risk assessment requires estimation of effects at low levels of toxicants, so understanding biological effects of low level exposures (BELLE) is critical. As much information as possible about mechanism(s) of action should be obtained. The usual presumption for extrapolation from frank effect levels to lower levels is that adverse effects decrease with decreasing dose in some regular fashion until they become unmeasurable. However, all chemicals can cause multiple effects, and the ratios of these effects will also vary with concentration or dose. Qualitative variations in effects over a range of doses can be due to differences in affinities among multiple receptors as well as to gradations in physiological responses to stimulation of a specific receptor. Thus it should be clear that a simple linear or log-linear dose response to toxicants is always likely to be an oversimplification. Nonlinear responses including U-shaped curves (mixed inhibition and stimulation) can also occur. The question for regulatory or risk assessment agencies is whether the assumption of low-level stimulatory or hormetic effects is usefulfor public health protection. If so, how would it be used, and what level of evidence would be required to introduce the hormetic concept into a specific assessment?
Stimulatory or hormetic effects occurring at doses below those usually considered to be toxic have recently been reviewed by Calabrese and Baldwin (1998) for chemical hormesis and by Pollycove (1998) for radiation hormesis. Theoretical aspects of hormesis including mathematical models were reviewed by Boxenbaum et al. (1988), while some monotonic and U-shaped dose-response curves for environmental agents were described by Anderson and Barton (1998). It is important to consider that some low level effects with U-shaped dose-response curves may represent replacement of one toxic effect with another as well as possible beneficial effects. This may be particularly true in the case of receptor-mediated endocrine effects (e.g., Cassidy et al., 1994).
In this article we attempt to address three questions related to how a public health agency uses mechanistic information, which might include evidence of low-level, possibly beneficial effects in the risk assessment process. Specifically, how can we incorporate into public health decisions information on (1) mechanisms of toxicity, (2) mechanisms of adaptation, and (3) beneficial low-level effects? Additionally we describe two examples of risk assessments, that involved consideration of beneficial low-level effects. Our own agency, the California Environmental Protection Agency (Cal/EPA) and its risk assessment branch, the Office of Environmental Health Hazard Assessment (OEHHA), presently have no specific policy regarding BELLE, hormesis, or related effects. Each chemical is assessed for the health risk resulting from human exposure according to the particular legal mandate under which it falls. OEHHA performs risk assessments or develops risk assessment guidelines for toxic substances in air, drinking water, food, and sport fish as well as providing hazard identification of carcinogens and for developmental and reproductive toxicants (DARTs) under the Safe Drinking Water and Toxic Enforcement Act of 1986, "Proposition 65."
For OEHHA, the important role played by mechanistic data on the toxicity of chemicals is most evident for cancer risk assessment. Under the proposed 1996 Guidelines for Carcinogen Risk Assessment (U.S. EPA, 1996), the U.S. Environmental Protection Agency (U.S. EPA) describes three options or "science policy defaults" for dose-response approaches based on understanding the carcinogenic mode of action (MOA). These are: low-dose linear extrapolation; nonlinear (which may incorporate various mechanistic models); and both procedures combined. The initial OEHHA implementation of these guidelines views these options as proceeding along a continuum of weight of evidence. When little or no confidence can be placed in a specific MOA(s) then the default approach is linear using the essentially model-free extrapolation recommended by U.S. EPA. When data provide support for both linear and nonlinear approaches for a specific cancer endpoint both approaches should be applied. For the nonlinear approach there should be convincing evidence that one and only one mode of action is operative in producing the tumors. Positive evidence including nonlinear dose-response data is more persuasive than negative data which might support another MOA, e.g., lack of genotoxicity. U.S. EPA provides a number of questions that should be addressed before choosing a dose-response approach. Most of these address the degree of confidence the risk assessor has in the evidence supporting one or another proposed MOA.
The interpretation of mechanistic data is not entirely objective, and different agencies may come to different conclusions from the same body of data. Since the conclusion about MOA will determine if low-dose linear extrapolation is applied (which often leads to higher hazard estimates), it is an issue of key importance. For example, recently the gasoline additive and fuel oxygenate methyl t-butyl ether (MTBE) was assessed for risk via drinking water exposure by U.S. EPA (1997) and OEHHA (1998). MTBE was found to have caused cancer in three studies in mice or rats by both inhalation and oral (gavage) administration. In rats the tumor sites were renal, testicular, and leukemia/lymphoma. In mice hepatic tumors were observed. MTBE appears to be non-genotoxic but little mechanistic data is available supporting any specific non-genotoxic MOA. U.S. EPA assumed a nonlinear MOA for tumor induction based, in part, on lack of genetic toxicity. U.S. EPA's drinking water advisory levels for longer-term health and consumer acceptability (taste and odor) were 20-40 ppb. This range was compared with the margins of exposure (MOE) calculated for three cancer and three noncancer toxic endpoints. OEHHA could not identify a plausible MOA for MTBE although each tumor site appeared potentially relevant to human toxicity. OEHHA chose the linear approach for the cancer endpoints. This approach yielded a health protective water concentration of 13 ppb, and a higher value of 47 ppb for the most sensitive noncancer toxic endpoint (increased kidney weight/body weight). Despite the differing approaches chosen by these agencies, the final numerical advisories differ by less than a factor of two. This may be an unusual example, but it illustrates the practical fact that even when different mechanisms are assumed, the final values for risk assessment can be similar.
Some hormetic effects may result from adaptation to environmental stressors such as oxidative or caloric stress (Hart and Frame, 1996). A number of experimental and epidemiologic studies support adaptive or dose rate dependent responses occurring on the cell or organism levels. Cells in vitro or organisms exposed to low levels of physical or chemical agents had less biologic damage following subsequent higher level exposure than controls that had the latter dose applied acutely. Various levels of action can be postulated, and any MOA explanations will need to cover these different levels of action. As Calabrese and Baldwin (1998) note, the wide range of hormetic effects suggests action at a relatively high organizational level probably involving different gene linkages. Most of the mechanistic data on adaptation come from the radiation literature. As discussed by Trosko (1998), low-level exposure could activate DNA repair enzymes, glutathione synthetase, and other protective or repair mechanisms to allow a less injurious response to subsequent toxic exposures. Whether the adaptive response on the cell level impacts the next higher level would depend on the cell in which the adaptive response occurred, whether it prevented any stable mutagenic or epigenetic event in a surviving cell, and whether the adapted cell could be clonally multiplied. An abnormal cell will have a chance to cause a pathologic or physiologic change only if it multiplies. At the cellular level there is evidence that the adaptive response is mediated by gap junction intercellular communication (GJIC) (Ishii and Watanabe, 1996). At low doses of radiation (<0.5 Gy) gap junction structure appears unaffected but functionality has not yet been determined. Gap junctions may play a role in intercellular movement of protective antioxidants or cell cycle signals inducing apoptosis (Trosko and Goodman, 1994). These factors and observations would need to be extended from the cellular to the organ and organism levels to be useful in risk assessment.
In another biochemical context, Anderson and Barton (1998) have evaluated dose-response relationships of TCDD on the liver. At low doses (1 to 5 ng/kg-day), TCDD induces the rat hepatic microsomal cytochrome CYP1A1 and inhibits cellular proliferation induced by mitogenic stimuli. With slightly higher doses (10 to 30 ng/kg-day) TCDD causes moderate cytochrome induction, associated with mild cell proliferation and moderate cellular toxicity. These effects may occur in different populations of hepatocytes. At higher doses (35 to 125 ng/kg-day), TCDD strongly induces CYP1A1 and CYP1A2 and stimulates cellular proliferation associated with severe hepatocellular toxicity (which would lead to hepatocellular carcinoma in a chronic study). Together the inhibitory and stimulatory effects of TCDD on cell proliferation can produce a U-shaped dose-response curve as measured by hepatic cell labeling index. The authors interpret the suppression of response to mitogenic stimuli as an enhancement of a normal adaptive response. This mechanism would normally help maintain the correct size of the liver. At higher doses of TCDD the mitosuppression effect is antagonized by cell toxicity-associated proliferative responses. However, it is not clear whether it would be appropriate to call the mitosuppression a "low-dose protection" effect of TCDD (Anderson and Barton, 1998). If applied continuously, this TCDD effect might help select for precursor lesions with mutations in growth regulatory genes. This region of the dose response is merely where mitoinhibition by TCDD in liver is the predominant toxic effect.
In other tissues TCDD may cause other toxic effects. The dose that causes a small response in one tissue may give a greater response in another tissue. The pathogen-icities of most toxic responses are combinations of multiple effects of the toxicant on the organ system(s) and complex dose-response curves may result from differing dose responses.
Based on the foregoing discussion the risk assessor would need to ask a number of questions of a proposed MOA for an adaptive response in order to interpret the effect in risk assessment. How universal is the response with respect to tissue and species specificity? If it is a cellular or organ response what does it mean at the whole organism level? What are the temporal effects? Does the adaptive effect persist? How does the adaptive effect differ in individuals of different ages, e.g., children and the elderly, or in well-defined disease states? Based on the earlier discussion of the MOAs applied to carcinogenic dose response approaches, positive data supporting a specific MOA for an adaptive low-dose response is superior to a lack of data which would tend to support an alternative MOA.
It follows from the discussion above that considerable caution would be needed before any low-dose beneficial effects could be incorporated into a risk assessment. Exposures to toxic substances are already sanctioned in order to achieve beneficial effects in well-known cases such as use of chlorine for disinfecting drinking water, and fluoride for prevention of dental caries. In these cases, the environmental and public health risk assessment has much in common with a medical therapeutic use of a potentially toxic substance. However, there is no accepted beneficial purpose for addition of most toxicants to drinking water. Thus even if benefits may be predicted at some low exposure level, public health regulators are not obligated to accept the exposure as desirable. Addressing preexisting and continuing exposures to low-level environmental contaminants would more likely take the form of altered uncertainty factors and margins of exposure.
In the case of convincing low-level beneficial effects of an agent carcinogenic at higher levels, a less conservative
dose-response approach might be chosen, e.g., nonlinear instead of linear. Ideally, specific dose-
response models that fit observed data and incorporate hormetic potential might be employed (e.g., Bogen 1997; Bogen and Layton, 1998). As with the establishment of an adaptive MOA above, questions must be asked about the claimed beneficial effects at low dose. How universal with regard to cell type, organ, sex or species is the beneficial effect? Is it an animal effect being extrapolated to humans? Has the effect been observed over the entire life cycle? What is the magnitude of the effect? Is the effect likely to be independent of interaction with other environmental contaminants? Does epidemiologic data support the claim for hormesis? Finally, even if the beneficial effect appears real, would the exposed population favor inclusion of this chemical in their drinking water? The following examples illustrate responses to hormesis-like situations in risk assessments conducted by OEHHA.
Uranium is being evaluated for cancer risk due to ionizing radiation. There has long been a controversy as to whether such radiation has a linear dose response at low environmental doses of a few pCi/day.
Uranium, a radioactive element, occurs as a trace element in many types of rocks. Because its abundance in geological formations varies from place to place, uranium is a highly variable source of contamination in drinking water. The kidneys and bones are the principal sites of accumulation and toxic action of uranium (Yuile, 1973; Stevens et al., 1980; Morrow et al., 1982). Following uranium administration, 80% is excreted in urine and feces, 10% is deposited in the kidneys and the remaining 10% is deposited in the skeleton with negligible concentrations appearing in other tissues (NRC, 1983). The skeleton is the major site of long-term storage of uranium (Wrenn and Singh, 1982).
The studies of Gilman et al. (1998a-c) provide a basis for a dose-response assessment of noncarcinogenic effects caused by chronic uranium intake via drinking water. In particular a 91-day study in Sprague-Dawley rats identifying a subchronic oral LOAEL of 0.06 mg/kg-d for renal and liver lesions in males appears to be the most reliable basis. The findings of effects on human kidney function resulting from quite low exposure levels of uranium in drinking water (0.004 to 9 mg/kg-d) tend to support the animal data and suggest a very broad dose-response range (Zamora et al., 1998).
U.S. EPA has classified natural uranium as a Group A carcinogen ("human carcinogen based on sufficient evidence from epidemiological studies") because it is an emitter of ionizing radiation. U.S. EPA classifies all emitters of ionizing radiation as Group A carcinogens. U.S. EPA acknowledges that "studies using natural uranium do not provide direct evidence of carcinogenic potential." However, studies with radium and certain isotopes of uranium provide evidence for the carcinogenicity of ionizing radiation in humans (U.S. EPA, 1991a, b). The two most common uranium isotopes in drinking water are 234U and 238U with lifetime risk coefficients for exposure via tap water of 4.59 x 10-11 (pCi)-1 and 4.18 x 10-11 (pCi)-1, respectively (U.S. EPA, 1998). Since the isotopic ratio of 234U/238U in California groundwater is 1.32 ± 0.3 (DHS, 1997) the combined risk coefficient for uranium in California water is 4.20 x 10-11 (pCi)-1.
In this calculation it was assumed that ionizing radiation (particularly alpha particles) emitted by natural uranium would be as carcinogenic as ionizing radiation emitted by more highly radioactive substances including man-made isotopes of uranium. This assumption and linear extrapolation is uncertain because of the studies indicating that low level exposure to radiation may have beneficial effects, that radiogenic dose response may be nonlinear or linear-quadratic, or that tumors develop more slowly at low doses (e.g., Raabe et al.,1980; Billen, 1990; Makinodan and James, 1990; Cohen, 1995; and Pollycove, 1998). For example, Bogen (1997) successfully fit a new cytodynamic 2-stage (CD2) cancer model jointly to data compiled by Cohen (1995) on U.S. county lung cancer residential radon exposure data and to lung cancer mortality radon exposure data for uranium workers (NRC, 1988). The CD2 model fit to the data was very good (C2 goodness-of-fit, p = 0.25). This model fit supports the idea proposed by Cohen (1995) that the dose response of lung cancer mortality and residential radon air concentrations are consistent with either a threshold or a hormesis model. An earlier cell kinetic 2-stage cancer model which does not accommodate a negative slope gave a poor fit to the data (p < 10-6). The linear relative-risk-model based on the same data set gives higher risk estimates between ca. 1 pCi/L and 19.5 pCi/L relative to the CD2 model. It is difficult to extrapolate directly from inhaled radon to ingested uranium but notwithstanding that problem, extension of the bimodal dose response in Bogen's CD2 model to radionuclide risk assessment is also problematic. The implication that we would be able to convince the public that a little dose of ionizing radiation is good for them strains credulity. For the sake of simplicity a linear dose response would still appear to be the most viable solution for regulatory purposes.
U.S. EPA has characterized some of the effects noted above in terms of a dose and dose rate effectiveness factor (DDREF):
e.g., a DDREF of three means the risk per unit dose observed at high acute doses should be divided by three before being applied
to low dose (dose rate) conditions (U.S. EPA, 1998). With the possible exception of lung cancer, current scientific data
generally indicate DDREFs between 1 and 3 for human cancer induction. For uranium and radionuclides emitting
high-linear energy transfer (LET) alpha radiation, the radiobiological results generally support a linear non-threshold dose response, except for a possible fall-off in effectiveness at high doses (U.S. EPA, 1998). At this point it appears to us that the low-dose-linear paradigm of radiogenic risk dose response best satisfies the requirements of the California mandate with respect to assessing risks of waterborne natural uranium at low levels. It must be emphasized that risk estimates for consumption of uranium in drinking water represent theoretical risks.
Fluoride occurs naturally in some drinking water sources but is added to water in other communities to achieve beneficial effects on dental health. Fluoride does not fit the typical hormetic model because its low-level effects are probably not mediated by adaptive responses, but rather by chemical and structural alterations to tooth enamel. Nonetheless, it presents similar issues to the risk assessor. Fluoride's beneficial effects appear to be optimal at drinking water levels of 0.7 to 1.0 ppm. A number of increasingly adverse effects occur with increasing doses corresponding to water concentrations of about 3 ppm and above. In children mild dental fluorosis occurs at this level and becomes more severe at somewhat higher levels. Fluoride is a well-known metabolic inhibitor. It has shown equivocal evidence of cancer causation in animal bioassays, and has exhibited some epidemiological association with increased hip fracture in the elderly.
Both beneficial (tooth decay prevention) and adverse (dental fluorosis) effects of fluoride ingestion were originally studied by Dean (1942). These effects and others were subsequently studied by many investigators, and reviewed by expert panels (National Research Council, 1993; U.S. Public Health Service, 1991). The consensus of these reviews is that the best drinking water concentration for fluoride, for preventing dental caries without incurring increased dental fluorosis, is 1 ppm. OEHHA took this as a NOAEL because some milder forms of dental fluorosis begin to increase in frequency (although not dramatically) at drinking water concentrations just above 1 ppm. In calculating a public health goal (PHG) for fluoride, OEHHA used an uncertainty factor of one to account for numerous human studies and to avoid setting the PHG below the beneficial level.
If fluoride were evaluated based only on toxicity, the typical uncertainty factors and acceptable margins of safety would require a water level close to 0.1 ppm. An uncertainty factor could be used to protect sensitive individuals or individuals who have unusually large exposure to fluoride from other sources. An additional factor might be incorporated to allow for uncertainty about the long-term effects of fluoride ingestion, such as possible effects on bone brittleness and possible carcinogenicity. However, based on the considerable knowledge of the multiple effects, we could balance the beneficial effects of added fluoride with a sufficient margin of safety from the adverse effects of dental fluorosis to choose 1.0 ppm as a health protective Public Health Goal (PHG).
In this article we have discussed some of our concerns about the application of the hormesis concept in regulatory toxicology and risk assessment. The use of mechanistic data, adaptation, and beneficial low level effects were also discussed briefly. We gave short summaries of two recent situations where hormesis-like considerations were raised. In the case of fluoride beneficial effects on dental health were balanced against the minimal adverse effects of dental mottling in children. In the case of uranium evidence of radiation hormesis was not considered sufficient to depart from the linear-low-dose dose-response paradigm. A quantitatively similar low-level toxic endpoint of uranium-induced kidney and liver toxicity was also factored into this decision.
Extension of this practice to other low-level effects must address the issue of achieving net public health benefit. It is insufficient to merely demonstrate that there can be a hormetic effect at low levels. Risk assessors must determine whether and how to incorporate this information into public health decisions in a manner consistent with regulations, public safety, and public trust.
Andersen ME, Barton HA. The use of biochemical and molecular parameters to estimate dose-response relationships at low levels of exposure. Environmental Health Perspectives 1998; 106:349-355.
Billen D. Spontaneous DNA damage and its significance for the "negligible dose" controversy in radiation protection. Radiation Research 1990;124:242-245.
Bogen KT. Do U.S. county data disprove linear no-threshold predictions of lung cancer risk for residential radon? A preliminary assessment of biological plausibility. Human and Ecological Risk Assessment 1997; 3:157-186.
Bogen KT, Layton DW. Risk management for plausibly hormetic environmental carcinogens: The case of radon. BELLE Newsletter 1998; 7:30-37.
Boxenbaum H, Neafsey PJ, Fournier DJ. Hormesis, Gompertz functions, and risk assessment. Drug Metabolism Reviews 1988; 19:195-229.
Calabrese EJ, Baldwin LA. Hormesis as a biological hypothesis. Environmental Health Perspectives 1998 106:357-362.
Cassidy RA, Vorhees CV, Minnema DJ, Hastings L. The effects of chlordane exposure during pre- and postnatal periods at environmentally relevant levels on sex steroid-mediated behaviors and functions in the rat. Toxicology and Applied Pharmacology 1994; 126:326-337.
Cohen BL. Test of the linear no-threshold theory of radiation carcinogenesis in the low dose, low dose rate region. Health Physics 1995; 68:157-174.
Dean H. The investigation of physiological effects by the epidemiological method. In: Fluoride and Dental Health, F.R. Moulton, ed. Washington, DC: American Association for Advancement of Science, 1942, p. 23.
DHS. A Study of the Isotopic Abundance Ratios for Select California Groundwater. Sanitation and Radiation Laboratories Branch-North, Division of Drinking Water and Environmental Management, California Department of Health Services, Berkeley, CA, 1997.
Gilman AP, Villeneuve DC, Secours VE, Yagminas A, Tracy BL, Quinn JM, Valli VE, Willes RJ, Moss MA. Uranyl nitrate: 28-day and 91-day toxicity studies in the Sprague-Dawley rat. Toxicological Sciences 1998a; 41:117-128.
Gilman AP, Villeneuve DC, Secours VE, Yagminas AP, Tracy BL, Quinn JM, Valli VE, Moss MA. Uranyl nitrate: 91-day toxicity studies in the New Zealand White rabbit. Toxicological Sciences 1998b; 41:129-137.
Gilman AP, Moss MA, Villeneuve DC, Secours VE, Yagminas AP, Tracy BL, Quinn JM, Long G, Valli VE. Uranyl nitrate: 91-day exposure and recovery studies in the male New Zealand White rabbit. Toxicological Sciences 1998c; 41:138-151.
Hart RW, Frame LT. Toxicological defense mechanisms and how they may affect the nature of dose-response relationships. BELLE Newsletter 1996; 5:1-16.
Ishii K, Watanabe M. Participation of gap junctional cell communication on the adaptive response in human cells induced by low dose of X-rays. International Journal of Radiation Biology 1996; 69:291-299.
Makinodan T, James SJ. T cell potentiation by low dose ionizing radiation: Possible mechanisms. Health Physics 1990; 59:29-34.
Morrow P, Gelein R, Beiter H, Scott J, Picano J, and Yuile C. Inhalation and intravenous studies of UF6/UO2F2 in dogs. Health Physics 1982; 43:859-873.
NRC. Health Effects of Radon and Other Internally Deposited Alpha Emitters (BEIR IV). National Research Council Committee on Biological Effects of Ionizing Radiations (BEIR), National Academy Press, Washington, DC, 1988.
NRC. "Uranium." In: Drinking Water and Health. Volume 5. National Research Council, Washington, DC, 1983, pp. 90-97.
NRC. Health Effects of Ingested Fluoride. National Research Council, National Academy Press, Washington, DC, 1993.
OEHHA. Public Health Goal for Methyl Tertiary Butyl Ether (MTBE) in Drinking Water. Draft. Office of Environmental Health Hazard Assessment, California Environmental Protection Agency. Sacramento, CA, 1998.
Pollycove M. Nonlinearity of radiation health effects. Environmental Health Perspectives 1998; 106:363-368.
Raabe OG, Book SA, Parks NJ. Bone cancer from radium: Canine dose response explains data for mice and humans. Science 1980; 208:61-64.
Stevens W, Bruenger FW, Atherton DR, Smith JM, Taylor GN. The distribution and retention of hexavalent 233U in the beagle. Radiation Research 1980; 83:109-126.
Trosko JE. Hierarchical and cybernetic nature of biologic systems and their relevance to homeostatic adaptation to low-level exposures to oxidative stress-inducing agents. Environmental Health Perspectives 1998; 106:331-339.
Trosko JE, Goodman JA. Intercellular communication may facilitate apoptosis: implications for tumor promotion. Molecular Carcinogenesis 1994; 11:8-12.
U.S. EPA. National Primary Drinking Water Regulations; Radionuclides; Proposed Rule. Federal Register. 1991a, 56, No. 138, 33050-33127. Thursday, July 18, 1991.
U.S. EPA. Drinking Water Criteria Document for Uranium (Draft). U.S. Environmental Protection Agency, 1991b, Washington, DC.
U.S. EPA. Proposed Guidelines for Carcinogen Risk Assessment. U.S. Environmental Protection Agency, Federal Register 1996; 61:17960-18011.
U.S. EPA. Provisional Health and Consumer Acceptability Advisory for Methyl Tertiary-Butyl Ether (MtBE). Office of Water, U.S. Environmental Protection Agency, 1997, Washington, DC.
U.S. EPA. Health Risks from Low-level Environmental Exposure to Radionuclides. Federal Guidance Report No.13-Part 1 Interim Version. EPA 402-R-97-014. U.S. Environmental Protection Agency, 1998, Washington, DC.
U.S. Public Health Service. Review of fluoride, benefits and risks, report of the ad hoc Subcommittee on Fluoride of the Committee to Coordinate Environmental Health and Related Programs. February, 1991.
Wrenn MD, Singh NP. Comparative distribution of uranium, thorium and plutonium in human tissues of the general population. In: Natural Radiation Environment. Edited by Vohra KC, Wiley Eastern Ltd., New Delhi, 1982, pp. 144-154.
Yuile CL. Animal experiments. In: Handbook of Experimental Pharmacology. Edited by Hodge HC, Stannard JN, Hursh JB. Volume 36. Springer-Verlag, Berlin, 1973, pp. 165-196.
Zamora M, Tracy BL, Zielinski JM, Meyerhof DP, Moss MA. Chronic ingestion of uranium in drinking water: A study of kidney bioeffects
in humans. Toxicological Sciences 1998;