9.8. DOSE-RESPONSE CONSIDERATIONS

The current efforts to evaluate the risks of dioxin and related compounds have focused on the understanding of the biological basis of response as well as evaluation of the weight of the empirical observations on inferences regarding hazard and risk. Previous sections have discussed the relationship of binding of this class of compounds to a specific receptor and subsequent events. It is generally accepted that all well-studied responses to dioxin appear to be mediated by receptor binding. This situation is not unlike the signal transduction pathways that have been described for hormone action, particularly exemplified by the well-studied family of steroid hormones, although the dioxin receptor does not belong to the steroid receptor family.

The fact that much of the biological activity of this class of compounds follows the rank order of binding affinity of the congeners to the Ah-receptor supports the concept that these earliest steps play a determining role in the probability that later responses will occur. This does not imply that a simple proportional relationship between receptor binding and biological response can explain the diversity of biological responses described for dioxin and related compounds. It is likely that differences in response will be due to tissue and cell-specific factors that modulate the qualitative relationship between receptor binding, or more precisely, occupancy and response. It is expected that there may be markedly different dose-response relationships for different effects of dioxin, depending on the respective roles of modulating activities. Coordinated biological responses, such as TCDD-mediated increases in cell proliferation, likely involve numerous cellular factors and hormone systems. This means that the dose-response for relatively simple sequelae of the early binding events such as cytochrome (CYP1A1) induction may not accurately predict dose-response relationships for more complex responses such as cancer. Much additional knowledge will be required before we can accurately predict these complex dose-response relationships.

Development of biologically based dose-response models for dioxin and related compounds as a part of this reassessment has led to considerable and valuable insights regarding both mechanisms of dioxin action and dose-response relationships for dioxin effects. These are described in some detail in Chapter 8. These efforts have provided additional perspectives on traditional methods such as the linearized multistage (LMS) procedure for estimating cancer potency or the uncertainty factor approach for estimating levels below which noncancer effects are not likely to occur. These methods have also provided a biologically based rationale for what had been primarily statistical approaches. The development of models like those in Chapter 8 allows for an iterative process of data development, hypothesis testing, and model development. These efforts have resulted in incorporation of more of the available biological data into models to predict human risk at low increments of exposure.

Tables 9-3 through 9-6 summarize estimated body burdens and effect levels for a variety of species, including the lowest observed effect levels (LOELs) for some of the more sensitive indicators of biological response induced by dioxin and related compounds. Important assumptions used in deriving these values are included as part of this table. It is particularly important to note that the estimated body burdens associated with several of these experimental doses are quite low relative to background body burdens in the general human population. The implications of this observation will be discussed later in this chapter.

Dose-response modeling efforts in Chapter 8 for liver cancer in female rats and for lung cancer and all cancers combined in humans have produced results that can be used to estimate risk-specific doses and risk estimates. Estimates from these efforts differ with models based on the human data, providing somewhat higher risk estimates than the animal-based estimates. The risk estimates resulting from these models have uncertainties that cause their ranges to overlap, and all models produce fits that are consistent with the linearized multistage model commonly used for cancer risk estimates. By the definitions of mechanistic modeling given in Chapter 8, both modeling efforts fall short of completely explaining conditions of biology or exposure. However, because the animal modeling establishes a better mechanistic basis for extrapolation to low doses and the animal data have greater certainty in terms of causal association for cancer (especially considering that smoking may have a potentiating effect with dioxin for lung cancer in the human studies), the animal model will be the focus of estimates of cancer risk.

Many scientists agree that the cancer modeling efforts should continue to focus on the animal studies in the absence of better quantitative human data. Others suggest that there is no compelling reason to conclude that estimates derived from the human data are any more uncertain than the estimates based on the rodent bioassay. In both cases, modeling efforts have indicated the sensitivity of certain model parameters to choice of data sets and/or assumptions. Analyses in Chapter 8 illustrate that the slope of the dose-response curve for surrogate markers of low-dose response such as enzyme induction or indirect mutagenic activity on estimates of cancer risk using animal data are highly dependent on the assumptions that go into the modeling. Dependent on assumptions, use of the obvious dose surrogates could predict very different low-dose risks, differing by orders of magnitude from the estimates described above. For gene expression of biological markers, the major factor controlling this broad range of low-dose risk estimates is the mechanism by which dioxin modifies constitutive expression. However, as expressed in Appendix D of Chapter 8, reasonable assumptions concerning constitutive expression of the biochemical markers will result in low-dose linearity and risk estimates consistent with that obtained using the linearized multistage approach.

The two-stage modeling of the Kociba et al. (1978) female rat liver tumor data in Chapter 8 incorporates data from earlier events in the carcinogenic process into the estimation of model parameters. In fact, the results using the two-stage model incorporating dioxin-altered hepatic foci data to estimate mutation and growth parameters provide nearly the same low-dose estimates as the linearized multistage model using only the tumor data. When using the default species extrapolation from animals to humans (body weight ratio to the 3/4 power), both models yield oral intake risk-specific doses of slightly less than 0.01 pg TCDD/kg/day, corresponding to unit risk estimates of 1´ 10^-4 per pg TCDD/kg/day. Chapter 8 discusses other potential models that might fit these data as well as the best-fitting model (Appendix C, Chapter 8). These analyses indicate that, unless a protective effect of TCDD on mutation rates occurs at low doses, low-dose risk will remain proportionate to exposure and consistent with the linearized multistage model. If protective effects are allowed in the model, the low-dose risks may be substantially reduced. The focal lesion data and the biochemical markers generally agree and do not suggest the protective effect discussed above. These models assume that the PGST foci are precursors to cancer. Other hepatic focal lesion markers could be used in this context and may lead to different dose-response curves for tumor response (see discussion in Chapter 8).

Uncertainty in estimates of human half-life for dioxin and related compounds represents an important factor in comparing human-based risk estimates versus animal-based risk estimates. For instance, if the dose-dependent pharmacokinetic model of Carrier (1991) is correct, exposures in the occupational studies must have been greater than the fixed half-life model would suggest, so that the estimated risk per unit of exposure may well have been lower. However, this reduction will be relatively small and is unlikely to move the risks outside the range of risk estimated by the linearized multistage model.

An additional consideration regarding the evaluation of dose response for dioxin and related compounds involves the ubiquity of background exposure to these compounds. Body burdens of these compounds have been discussed previously in several parts of this assessment. In all studies, both in laboratory animals and in humans, incremental exposures are being added onto an existing body burden that is present at birth and appears to increase with age. This background is often insignificant from the standpoint of added dose in experimental studies or for highly exposed human cohorts. On the other hand, it has real implications relative to the detectability of response at low incremental exposures and may have implications for the use of models that assume additivity to ongoing processes that may have been stimulated by background levels. Modeling estimates suggest that, if dioxin and related compounds are adding to human cancer burden, current background exposures may result in upper bound population cancer risk estimates attributable to exposure to dioxin and related compounds in the range of 1 in 10,000 (10^-4) for the average population exposures to 1 in 1,000 (10^-3) for more highly exposed members of the population (e.g., individuals consuming high levels of dioxin-containing foods). Actual risk for individuals exposed to background levels in the population is likely to be less than these upper bound estimates and, for some, may even be zero. More highly exposed populations with exposures to specific sources of dioxin and related compounds such as those exposed under occupational or accident conditions may, on average, experience proportionately higher risk.