Health & Environment

Assessing Risk Posed by Chemicals in Mixtures

Current practice in risk assessment, although changing, is more-or-less grounded in the 1970s, when pollution from industrial smokestacks and waste outlets was seen as the primary source of risk.

Pollution from such tangible sources can be assessed relatively easily and normally has a technological solution, such as filters in a chimney stack, which can reduce the degree of risk to a point where it is considered acceptable.

Since the 1980s, however, our understanding of where chemicals may pose a risk to health has changed. It is now recognised that people are exposed to a wide variety of chemicals from many sources, not just industrial pollution.

Rather than anticipating the risk to health and the environment posed by a specific waste outflow, the problem has become one of understanding and managing the risk that multiple, everyday exposures may pose to health. And as things stand, chemical risk assessment is not able to do the job.

Mixture effects are nothing new in toxicology, where it is known that the total toxicity of a mixture of chemicals can exceed the toxicity of its most toxic component. For example, asbestos and tobacco are more toxic in combination than they are independently of one another.

Xenoestrogens can act in concert to be more potent in a mixture than they are individually. In this case, the potency can be predicted with a DA calculation. (Rajapakse et al. 2002). Click to enlarge.

In spite of this being well-understood, however, chemicals are currently only assessed for toxicity on an individual basis. Even multiple exposure routes to the same chemical are barely considered; the effects chemicals may have in combination with each other are not considered at all.

The concern is that, even with safety factors built into individual chemical risk assessments, the toxicity of mixtures is therefore being underestimated.

The process for dealing with this which risk assessors will use is known as cumulative risk assessment (CRA). CRA attempts to combine calculations of risk from all sources into one overall risk measurement. In chemicals policy, it has generally been used to describe the risk to health posed by all exposure routes to single or multiple chemicals.

The EU Scientific Committees have published a preliminary opinion on the best way to do CRA (DG Health 2011), while Denmark recently submitted the first proposal for restricting the use of several phthalates because the risk they pose to children’s health when simultaneously exposed to all four exceeds what is acceptable.

There are a number of challenges in implementing CRA. The biggest problem with chemical mixtures is the sheer, overwhelming number of them. With approximately 80,000 chemicals in mass production there are more potential combinations than can possibly be assessed for toxicity.

Calculating risk from multiple sources

When the components of a mix are relatively consistent the toxicity of some mixtures can be measured, such as diesel exhaust and tobacco smoke. The task is simply to measure the toxicity of the mix as a whole, and this can be done in the same way as for individual chemicals, through epidemiology and animal tests.

The problem arrives when the components of a mix are not known – a person can have any number of chemicals in their body at varying concentrations (see e.g. PSR 2009). An actual measurement of toxicity can be accurate for a specific mix at a specific time only; since it is not possible to test every single possible combination a person is exposed to, the risk then has to be estimated.

There are two basic approaches for predicting the toxicity of mixtures: dose/concentration addition (DA) and independent action (IA). DA assumes that the toxicity of chemicals in a mix add up to produce the total toxicity of the mixture. IA assumes that chemicals act independently of each other in the mixture, so the mixture is only as toxic as its most toxic component.

A simplified representation of the difference between IA and DA in CRA. Click to enlarge.

These approaches are pragmatic, intended to provide approximations of sum total risk based on the risk posed to health by each component of the mixture. Evidence indicates these predictions, particularly for DA, tend to be accurate (Kortenkamp et al. 2009).

Amongst all the potential mixtures, there are several options for identifying mixtures of chemicals which are likely to be more toxic than their individual components. One way is to identify them by structure, the idea being that if a chemical with a particular structure causes a problem, then other chemicals with the same structure are more likely to cause a problem as well.

This method is not entirely reliable, since there is not a particularly tight correlation between a chemical’s structure and its particular effect in the body. Selection by structure could also miss chemicals which act in concert even though they are very different to each other, so the overall selection of chemicals may well be too narrow.

The approach preferred by the EU Scientific Committees is to group chemicals which are toxic by the same mode of action in the body. This is not ideal either, since our understanding of modes of action is continually evolving and it is not clear, in a complex biological system, just how independent each mode of action might be.

One other possible solution is to assess chemicals according to total effect on a disease end-point (see e.g. H&E #34). This has precedent in ecotoxicology, where simply affecting, for example, reproductive potential of a species is sufficient to trigger a DA assessment of a mixture, regardless of whether or not there is a common mode of action behind the effect.

There is, however, as yet no overall consensus on how best to identify and evaluate the troublesome mixtures.

Synergistic effects and other knowledge gaps

Another challenge in CRA is the potential for synergistic and antagonistic interactions between the components of a mixture. Synergy is when one component of a mixture enhances the toxicity of another (as might happen if a mixture consists of a chemical which inhibits DNA repair mechanisms and another chemical which is a mutagen). Antagonism is where one chemical inhibits the toxicity of another.

In the event of synergy, DA necessarily underestimates risk, while overestimating it in the case of antagonism. Kortenkamp et al.’s 2009 report on the state-of-the-art of mixture toxicity found these interactions to be relatively rare, meaning that in most cases DA provides an accurate assessment of risk.

Is this good enough? The problem is, because of the unpredictable nature of synergistic effects, we do not know in which cases DA is reliable. Biological systems are highly interlinked so anticipating these interactions is difficult, while information on the mixtures to which people are exposed is almost non-existent.

As yet, the only proposal for anticipating these interactions is to rely on expert opinion as to where they might be. This, however, only allows regulators to respond to an identified risk, where the objective of modern chemicals regulation is to be able to prevent harm where it can be expected but its precise nature is not known.

An evaluation system which generates recommendations to act only in the small number of instances where we have enough knowledge to anticipate harm cripples our ability to prevent harm in the overwhelming majority of cases, namely the ones where we do not know enough to be able to anticipate whether harm is happening or not.

The problem of knowledge is not restricted to synergistic effects. CRA is only of use when the components of a mix, and their potency, are known (whether their potency within a particular mode of action or relative to a disease end-point). Here there is a chronic lack of data, where the toxicity of even the components themselves are largely unknown.

Risk assessment only works when risk can be accurately anticipated. Even if interactions are relatively rare, since we do not have the data we need to anticipate them (or even the data we need for doing DA assessments accurately) or the risk they pose, it is not clear how much of a step forward CRA actually represents in chemicals policy.

Is it possible that CRA is the wrong way to approach the complex problem of mixture toxicity? Yannick Vicaire, chemical processes coordinator for French environmental health group Réseau Environnement Santé, believes the key question is not about how degree of risk is assessed, it is about how exposure to potential risks is managed.

He argues that CRA is only a tool. Like risk assessment as a whole, if it does not produce accurate results, then chemicals which pass a risk assessment end up being used legally without necessarily being safe. Legal use then becomes decoupled from safe use, defeating the point of the legislation in the first place.

He says: “It is more efficient to base risk management on a precautionary approach rather than rely on a tool like risk assessment which cannot handle the complexity of the problem with which it is faced.

“There are too many chemicals in the environment, food chain and the body. The goal of chemicals policy should be to reduce flow of chemicals into the environment and reduce hazard because at the moment the problem is too big and complicated for the tools we have available for dealing with it.”

This amounts to reducing the toxicity of chemicals to which people are exposed, and reducing the number of chemicals which end up in the environment and in people by using fewer chemicals in consumer products and employing closed-loop manufacturing and disposal processes.

“…the biologic effects of xenoestrogens cannot be dismissed as insignificant solely on the basis of their low potency compared with steroidal estrogens. Considered in isolation, the contribution of individual xenoestrogens at the concentrations found in wildlife and human tissues will always be small. However, such reasoning cannot be used to support claims of negligible health risks from weak xenoestrogens, because the number of xenoestrogens present in wildlife and humans is unknown but likely to be very large.” (Rajapakse et al. 2002)