Importance of Dose And Dose-Response Relationships:

Dose is a key component of quantitative human health risk assessment and reflects a central premise of toxicology; namely, that "the dose makes the poison."  This phrase was used by Paracelsus in the 16th century when he wanted to use mercury to treat syphilis.  The dose is the amount of chemical or other substance entering the body.  The risk is the probability of a specified chemical agent causing a disease or other response of concern and is a function of the dose of that chemical received.  That function is called the dose-response relationship or dose-response model.

In order to estimate the probability of the occurrence of an adverse health effect, the dose of the suspect chemical must be determined.  In other words, the dose to a specific individual must be determined before the risk (if any) to that individual can be determined.  While the dose depends on the actual media (water, air, soil, etc.) concentrations, it also depends on several other factors.  Even if the media concentrations themselves are unbiased estimates, the use of default, hypothetical, or worst case values for the other factors in the dose calculation makes the calculated doses upper bounds on the true doses and not accurate estimates of the doses for specific individuals.

Almost all chemical substances have the potential to be poisonous (toxic) if the dose is sufficiently high.  However, at lower doses many chemicals are harmless and indeed, like drugs and vitamins, may be beneficial to health.  Dose-response models reflect the importance of the size of the dose in determining the probability of a response of concern such as a disease or death.  Thus, dose-response modeling is a key component of quantitative human health risk assessment.

Upper-bound characterizations of the dose-response relationship do not provide (nor are they intended to provide) accurate estimates of the dose-response relationship and the probabilities of adverse health effects at individual doses.  These upper-bound characterizations, however, are consistent with the declared purpose of many regulatory risk assessments which utilize easily administered procedures that consistently provide upper bounds on those risks which can be used for risk comparisons and regulatory decision making.  For example, in the EPA's 1986 guidelines, these regulatory upper bounds are described as "a plausible upper limit to the risk that is consistent with some proposed mechanisms."  EPA cautions that "[s]uch an estimate, however, does not necessarily give a realistic prediction of risk.  The true value of risk is unknown, and may be as low as zero."

The regulatory upper-bound characterizations of the dose-response relationship fail to provide (nor were they intended to provide) accurate estimates of the probability of an adverse health effect at the low doses. 

Figure 0 helps to illustrate some aspects of this failure.  The figure shows a hypothetical curve representing the dose-response model fit to the animal experimental data at high doses on a specific response (usually an adverse health effect) in a specific gender, strain, and species of experimental animal in a specific study. (The health effect is usually a specific type of tumor in a cancer analysis; however, neither the animal tumor nor the mechanism of action in the specific gender, strain, and species are necessarily relevant to humans.)

The curve is intended to represent the dose-response relationship (i.e., how the probability, P(d), of a response at a dose, d, decreases as the dose decreases from high to low doses for a specified response in a specified gender, strain, and species of experimental animal in a specific study).  The probability P(d) refers to a population and not to a specific individual.  That is, P(d) does not refer to a specific experimental animal and certainly does not refer to a specific human. 

Furthermore, regulatory agencies do not use the best estimates of the increase, P(d) - P(0), in the probability of a response at dose d above the probability of a response at dose zero.  Instead, the regulatory agencies calculate a straight line starting at the estimated value of P(0) and increasing with the dose d faster than P(d) does.  The slope of that straight line is called the cancer slope factor (CSF) or cancer potency. 

The regulatory agencies intend that the straight line with slope CSF overstate the increased probability of a response at dose d (that is, overstate the risk at dose d).  That is, the regulatory agencies intend that the increase, CSF d, along the straight line from dose zero to dose d overstate the increase in the probability of response from dose zero to dose d.  (Specifically, regulatory agencies intend that CSF d exceed not only the added risk, P(d) - P(0), but also the extra risk, [ P(d) - P(0) ] / [ 1 - P(0) ] which is even larger than the added risk.)  Thus, the regulatory agencies intend that CSF d overstate the actual risk. 

Furthermore, the regulatory agencies make various policy-driven choices designed to increase the cancer slope factor, CSF, so that, CSF d exceeds the actual risk not only in the most sensitive experimental animals but also in humans. 

Risk characterizations in toxic torts and similar settings frequently use this increased CSF which was designed for regulatory purposes.  For example, the risk characterization for a specific location with an potential dose d is CSF d even though CSF is designed by the regulatory agencies to overstate the actual risk and even though the regulatory agencies state that the actual risk could be as low as zero.

Figure 0.  A hypothetical example of a regulatory upper-bound characterization of the
  dose-response relationship in the low-dose region and the cancer slope factor (CSF)