The new tapestry of risk assessment

Abstract

Neurotoxicology is entering a new phase in how it views and practices risk assessment. Perhaps more than any of the other disciplines that comprise the science of toxicology, it has been compelled to consider a daunting array of factors other than those directly coupled to chemical and dose, and the age and sex of the subject population. In epidemiological investigations, researchers are increasingly cognizant of the problems introduced by allegedly controlling for variables classified as confounders or covariates. In essence, they reason, the consequence is blurring or even concealing interactions of exposure with modifiers such as the individual's social ecology. Other researchers question the traditional practice of relying on values such as NOAELs when they are abstracted from a biological entity that in reality represents a multiplicity of intertwined systems. Although neurotoxicologists have come to recognize the complexities of assessing risk in all its dimensions, they still face the challenge of communicating this view to the health professions at large.

Introduction

Neurological diseases and disorders are rarely unidimensional or unifactorial. Even those whose etiologies seem closely linked to genetic predispositions, such as autism, tend to be the product of multiple and intertwined risk factors, of which environmental chemical exposures may serve as one component. The list of such factors can be intimidating: age, sex, dietary practices, immune status, and intercurrent disease state comprise only a small portion of a much larger list.

Despite such complexities, which are widely recognized, traditional risk assessment practices manage to elude this reality. They tend to focus instead on exposures to single chemicals in isolation from other risk factors. Animal studies all too often examine the effects of a single chemical as though it were independent of age, sex, and early environment. Epidemiological and clinical studies tend to emphasize the main effects of environmental exposures, stripping away interactions by allegedly controlling for confounders. The result is a wide gulf between current models of diseases and disorders and the actual conditions under which they emerge.

The vulnerability of an organism to disease also depends upon its confrontation with ecological stressors and modifiers. Advances in our ability to interdict threats to neurobehavioral integrity require us to take such modifiers into account in addition to xenobiotic exposures. These modifiers are embedded in the individual's social ecology, the web of factors such as class status, neighborhood and income that envelops that person in a particular social setting. We already possess substantial evidence of the health impact of single ecological factors. We lack the data required to model the interactions between these risk modifiers and chemical exposures and to translate them into risk policy.

One way to exemplify the challenges is to view the current scene in neurotoxic risk assessment against the wider social ecological background. A useful frame of reference is offered by early education programs such as Head Start. Fig. 1 presents the argument by economists such as James Heckman (Cunha, 2005) that the dividends from such an investment are highest when the investments are made during early development. Later investments in education and training, according to calculations from his group at the University of Chicago (Cunha et al., 2005), may fall below the opportunity costs.

Early interventions, according to a now extensive literature, without question provide substantial benefits. Fig. 2, based on the Abcedarian project (Barnett and Masse, 2002) testifies to their effectiveness.

Such early intervention programs have been aimed primarily at disadvantaged populations. In weighing costs and benefits, they unfortunately overlook an impediment to their success that, if accounted for, might demonstrate an even great scope of benefits. They do not take account of environmental chemical exposures, which almost always are greater in such populations. (Nor, in fact, do the bulk of the numerous studies examining the relationship between health and socioeconomic status).

Toxicology offers a complementary history from the perspective of neurotoxicology. Early developmental effects tended to be ignored until the thalidomide catastrophe (Weiss and Landrigan, 2000). Later, we recognized the cogency of the question posed by David Rall, the second director of NIEHS, at a neurotoxicology meeting in the 1980s: “Suppose that thalidomide, instead of causing the birth of children with missing limbs, had instead reduced their intellectual potential by 10%. Would we be aware, even today, of its toxic potency?” This is the kind of question that has animated developmental neurotoxicology for much of its life.

It also leads us to the presentations given in a session from the 24th International Neurotoxicology Conference held in 2007. It was convened to broadly examine the multifaceted nature of neurotoxic risk. It was structured around a series of four questions formulated to provoke a myriad of positions. Panelists were then requested to respond, on the basis of their own background, to the questions they had been assigned. Their answers represent a sample of the multiplicity of views among those who practice neurotoxicology in one of its many forms. Rather than melding into a unified but narrow point of view, the respondents offered a personal account of where they see deficiencies in current practice and the direction toward which they believe the discipline needs to move to overcome these deficiencies. They demonstrate the diversity of perspectives reflected in a meeting such as the 24th International Neurotoxicology Conference and offer guidance for future conferences.

The questions and responsed are listed below:

1. We are supposed to be part of a science whose framework was constructed by the mantra that The Dose Makes the Poison. Is it now time to bid farewell to the NOAEL as it is currently applied?

Deborah C. Rice. There are multiple issues embedded in the question of whether the NOAEL approach should be discarded. As is generally acknowledged, the NOAEL/LOAEL approach is dependent on the choice of doses rather than any real biological threshold. It is also dependent on the number of subjects and the variability of responses, which determines statistical power. This may be particularly relevant in animal studies, in which power is very low and yet only the dose group(s) statistically different from the control group are considered to be affected, even in the presence of an orderly dose–effect function that includes lower doses. The doses identified as a NOAEL or LOAEL are obviously also dependent on the endpoint(s) under consideration. The NOAEL associated with effects of developmental lead exposure may differ by orders of magnitude between studies and for different endpoints, for example (Rice et al., 1996). The issue of relevant endpoint is particularly relevant for endocrine disruptors such as bisphenol A, for which different biochemical mechanisms may be responsible for low-dose effects versus effects on more standard tests of endocrine activity that are observed at much higher doses (vom Saal and Hughes, 2005).

One response to the recognition of the limitation of the NOAEL approach is the use of benchmark dose (BMD) analysis in conjunction with modeling of the dose–effect relationship. This allows the use of all the information on the relationship between dose and effect in animal studies, and the relationship between individual body burden and effect in human studies. BMD analyses require explicitly defining an acceptable risk level as well as defining an adverse effect level if using continuous data. In addition, it is typically assumed that the exposure–effect relationship is linear, which of course may not be the case. Rather, the actual shape of the relationship should be determined, as has been performed for methylmercury (Budtz-Jørgensen et al., 2000, Axtell et al., 2000).

Another fallacy embedded in the statement that “[t]he dose makes the poison” is failure to take into account potentially sensitive populations. This perhaps most obvious with respect to lifestage, for which there is overwhelming evidence for multiple chemicals that the fetus may be severely and permanently affected at doses or body burdens that have little or no effect on the mother. It is being increasingly recognized, however, that aging, disease states, genetic polymorphisms, and the greater social environment may also result in differential toxicity. The assertion that “[t]he dose makes the poison” is therefore an incomplete characterization of the relationship between exposure and effect, and may be seriously misleading in some circumstances.

What kinds of data are risk assessors willing to use, what are the rules of evidence, and what is “the real world”?

Ted Schettler. We know of many different variables that can influence risk. Some are intrinsic to an individual or community and some are external. Ecologists have been dealing with complexity in natural systems for a long time. When trying to understand system dynamics, establishing and defining system conditions is extremely important.

What happens when we think of the brain and intimately related immune, endocrine, and gastrointestinal systems as an ecosystem—and at times, like a more or less vulnerable ecosystem?

The “state space” of a system is defined by the state variables that constitute the system. It is a three-dimensional space of all the possible combinations of these variables. A “basin of attraction” is a region in state space in which the system tends to remain. For our purposes, we might think of an individual living within a community as a complex system nested within its own complex system—living within a kind of “basin of attraction”.

The trajectory of early development and ongoing functioning will be influenced by both intrinsic and extrinsic variables. An extrinsic factor of particular interest to a risk assessor must be considered within the particulars of the state space of the system and the nature of the basin of attraction.

The preponderance of neurodevelopmental and neurodegenerative disease is a result of complex interactions among numerous variables. To be sure, there are occasions when some single factor is overwhelmingly responsible, but this is generally an exception.

Within individuals there are multiple determinants of risk, and within populations the causes of neurodevelopmental and neurodegenerative conditions are heterogeneous. They include genetic makeup and a wide range of both current and historical variables. Attributes of populations are the context for the distribution of risk factors in individuals within those populations.

Multiple mechanisms and multiple pathogenic pathways are woven together in various combinations to result in similar appearing phenotypes and neuropathology. For example, the pathogenic cascade leading to cell death in Parkinson's disease can include mitochondrial dysfunction, oxidative stress, excitotoxicity, inflammation, and abnormal degradation of abnormal cellular proteins in the ubiquitin–proteasome system. Combinations of these mechanisms undoubtedly vary from person to person.

The end result of this complexity is marked system-condition heterogeneity—varying from one individual to another AND one community to another, depending also on time and place.