What are the four steps that describe what the body does with an environmental hazard?

5.1 Specific modes of action of pharmaceuticals

When classical ERA is conducted on the basis of the conceptual framework here adopted [i.e., based on available information on concentrations measured in coastal waters and PNEC values obtained from animal testing with model marine species], very low risk is frequently predicted. This is a consequence of the several orders of magnitude of difference between MEC and toxicity threshold values, particularly for nonantibiotic PPCPs [e.g., [2,16,63,192]] even in freshwater ecosystems [76]. However, it is necessary to bear in mind that this risk characterization procedure is based on standard toxicity testing, targeting universal mechanisms of toxicity, mainly nonpolar narcosis, ignoring the specific modes of action displayed by pharmaceuticals and biocides. While this drawback is shared by many environmental pollutants, it is particularly relevant for the case of pharmaceuticals, which have been designed on purpose to act selectively on certain metabolic routes by targeting very specific enzymes, receptors, or cellular compartments.

European guidelines establish that when PEC in surface water is lower than 10 ng L− 1, low environmental risk is expected, otherwise further assessment on ecotoxicity at different environmental compartments should be conducted [19]. As reflected in Table 2, toxicity thresholds below 10 ng L− 1 [i.e., 0.01 μg L− 1] were never recorded using conventional endpoints. This apparently supports the threshold for further assessment. However, as discussed earlier, pharmaceuticals may act at lower concentrations on very specific targets not studied in conventional toxicity testing. According to Christen et al. [193], the potential risk of one pharmaceutically active compound to induce adverse effects on aquatic organisms increases with the degree of similarity between the mode of action in humans and aquatic fauna. Pharmaceuticals are synthetic molecules intended to act very selectively on a given key process of the metabolism, frequently at plasma concentrations within the ng L− 1 range. Just to mention a few examples: [a] in mammals EE2 is a selective agonist of the estrogen receptor, [b] propranolol is a selective antagonist of adrenergic β-receptors, [c] fluoxetine is a selective inhibitor of serotonin reuptake, and [d] diazepam specifically binds to GABA receptors. It seems reasonable to expect higher environmental risk for chemicals with conserved molecular targets, compared to those whose molecular targets highly differ between systematic groups. In the same way, organisms whose molecular targets are more similar to those of mammals are more prone to suffer selective effects of pharmaceuticals. Homologies between human receptors targeted by pharmaceutical compounds and those from fish range between 50% and 93% [193], and thus selective effects are also expected in aquatic fauna at concentrations below the 10 ng L− 1 threshold. Even invertebrates may share metabolic pathways targeted by pharmaceuticals. Bossus et al. [194] showed that the serotonin reuptake inhibitors fluoxetine and sertraline significantly enhance the amphipod velocity upon light exposure at concentrations as low as 0.01 μg L− 1 in the short term, although the ecological relevance of these findings is unclear since effects disappear after 8 days of exposure.

In other instances, even when the specific targets of the pharmaceuticals were common for mammals and aquatic animals, exposure to the drug may imply different effects due to different downstream pathways. This is the case of simvastatin, a hypocholesterolaemic pharmaceutical of broad consumption that acts by inhibiting the 3-hydroxy-3-methylglutaryl coenzyme A reductase [MGCR], an enzyme responsible for a rate-limiting step in the synthesis of cholesterol in mammals. In particular, MGCR catalyzes the synthesis of mavelonate from an oxidized precursor obtained by condensation of acetoacetyl-CoA. Mavelonate is a precursor of isoprenoids, the structural base of cholesterol, but also of hormones related to reproduction and development in some insects and crustaceans. Neuparth et al. [122] found that chronic exposure to simvastatin severely impacted the reproduction of amphipods at ng L− 1 levels in water, i.e., 3–4 orders of magnitude below those causing lethal effects in standard bioassays with crustacean adults or even larvae [121]. The deleterious effect of simvastatin on amphipod reproduction may be due to the fact that the specific target of simvastatin, the MGCR enzyme, is also key for the synthesis of methyl-farnesoate, an isoprenoid equivalent to the insect juvenile hormone implicated in crustaceans’ vitellogenin and egg production [195].

Therefore, toxicity experiments designed for the specific target of each pharmaceutical compound in aquatic invertebrates and fish are so necessary. Information about the effects of pharmaceuticals on the molecular targets of aquatic organisms homologous to those in mammals are essential in order to conduct more realistic risk assessment studies, which are applicable to aquatic ecosystems. In the few instances where these studies are available much higher risk quotients emerge. For example, fluoxetine LC50 for marine fish ranged from 500 to 2000 μg L− 1 [reviewed in Ref. [106]]. However, predatory performance in striped bass decreases irreversibly [within the experimental period] by exposure to 67 μg L− 1 of fluoxetine, and reversibly at a concentration of 15 μg L− 1 [196], concomitantly to a dose-dependent decrease of brain serotonin. In another study, the Arabian killifish exposed to 0.3 μg L− 1 of fluoxetine reduced swimming speed by 38% [197]. Moreover, exposure to 0.1 and 0.001 μg L− 1 of fluoxetine for 1.5 months seemed to impair predatory learning in cuttlefish, although for unexplained reasons the effects were more remarkable when the lowest dose was supplied [198].

Similarly, Olsén et al. [199] observed anxiolytic effects of citalopram on females of the freshwater fish Poecilia wingei, when concentration levels as low as 2.3 μg L− 1 were tested. These anxiolytic effects reduced the locomotion and were able to increase the risk of predation, and therefore be relevant at population level. There are thus more than three orders of magnitude between concentrations that actually affect standard toxicity testing endpoints used in ERA and those concentrations leading to specific mode of action of some psychiatric drugs. Unfortunately, the degree of standardization of those behavioral endpoints is still very low, preventing comparability among substances or species, and strongly limiting their utility in risk assessment.

5.2 Positive effects of pharmaceuticals on aquatic organisms

Due to reasons related to the pharmaceutical mode of action or not, instances of positive effects of pharmaceutical residues on aquatic organisms have frequently been reported in the scientific literature. Solé et al. [118] recorded increased feeding rates in mussels exposed to 23 and 147 μg L− 1 of paracetamol. Bossus et al. [194] showed that, in the short term, serotonin reuptake inhibitors fluoxetine and sertraline significantly enhance amphipod velocity upon light exposure at concentrations as low as 0.01 μg L− 1. This might be related to the natural function of serotonine. Serotonergic activity increases with motor activity, while firing rates of serotonergic neurons increase with intense visual stimuli. When amphipods are exposed to these drugs, serotonin reuptake is blocked and the neurotransmitter remains in the neuronal synapses longer than in control individuals.

Positive effects of toxicants at low doses are a common empirical finding termed hormesis [200] and are also described for several pharmaceuticals and their mixtures [e.g., [98,120]]. Munari et al. [201] found increased total haemocyte count [THC] in clams exposed to 1–25 μg L− 1 of fluoxetine, whereas THC dropped below control levels when higher concentrations were supplied. Fluoxetine significantly increased offspring production in D. magna exposed to concentration levels between 15 and 40 μg L− 1 [202,203], and this positive effect seemed not to be related with the specific mode of action of the pharmaceutical as it was observed also in exposures to a completely different chemical [nonyphenol]. Pomati et al. [204] observed that, for unexplained reasons, exposure to 10 μg L− 1 of ibuprofen stimulated the growth of the cyanobacterium Synechocystis sp. by more than 70%. Belhaj et al. [205] reported that concentrations as low as 10 ng L− 1 of EE2 could stimulate the growth of Dunaliella salina and increase its cellular content of photosynthetic pigments; in contrast higher concentrations significantly inhibited the growth and decreased the cellular content of photosynthetic pigments. Nonmonotonic toxicity curves are a serious challenge for ERA studies not contemplated in current standard procedures.

8.2.1 Overview of Environmental Risk Assessment

A brief introduction to the concepts of ERA is given in this section to introduce readers to the concepts and ideas behind the application of ecotoxicological models to assess an environmental risk.

Treatment of industrial wastewater, solid waste, and smoke is very expensive. Consequently, the industries attempt to change their products and production methods in a more environmentally friendly direction to reduce the treatment costs. Therefore, industries need to know how much the different chemicals, components, and processes are polluting our environment. In other words: What is the environmental risk of using a specific material or chemical compared with other alternatives? If industries can reduce their pollution just by switching to another chemical or process, then they will reduce their environmental costs and improve their green image. An assessment of the environmental risk associated with the use of a specific chemical and a specific process enables industries to make the right selection of materials, chemicals, and processes to benefit the economy of the enterprise and the quality of the environment.

Similarly, society needs to know the environmental risks of all chemicals to phase out the most environmentally threatening chemicals and set standards for the use of all other chemicals. The standards should ensure there is no serious risk in using the chemicals, provided that the standards are followed carefully. Modern abatement of pollution includes ERA, which is defined as the process of assigning magnitudes and probabilities to the adverse effects of human activities. This process involves identification of hazards such as the release of toxic chemicals to the environment by quantifying the relationship between an activity associated with an emission to the environment and its effects. The entire ecological hierarchy is considered in this context including the effects on the cellular [biochemical] level, the organism level, the population level, the ecosystem level, and the entire ecosphere.

The application of ERA is rooted in the recognition that:

The elimination cost of all environmental effects is impossibly high.

Practical environmental management decisions must always be made on the basis of incomplete information.

We use about 100,000 chemicals in amounts that might threaten the environment, but we know only about 1% of what is necessary to make a proper and complete ERA of these chemicals. Section 8.5 is a short introduction to available estimation methods to apply if information about properties of chemical compounds is unavailable in the literature. A list of relevant properties and how they impact the environment is also given.

ERA is in the same family as environmental impact assessment [EIA], which attempts to assess the impact of a human activity. EIA is predictive, comparative, and concerned with all possible effects on the environment, including secondary and tertiary [indirect] effects, whereas ERA attempts to assess the probability of a given [defined] adverse effect as a result of human activity.

Both ERA and EIA use models to find the expected environmental concentration [EEC], which is translated into impacts for EIA and to risks of specific effects for ERA. Development of ecotoxicological models for assessing environmental risks is detailed in the following section. An overview of ecotoxicological models is given in Jørgensen et al. [1995].

Legislation and regulation of domestic and industrial chemicals for the protection of the environment have been implemented in Europe and North America for decades. Both regions distinguish between existing chemicals and introduction of new substances. For existing chemicals, the EU requires a risk assessment to humans and the environment according to a priority setting. An informal priority setting [IPS] is used for selecting chemicals among the 100,000 listed in “The European Inventory of Existing Commercial Chemical Substances.” The purpose of the IPS is to select chemicals for detailed risk assessment from among the EEC high production volume compounds, that is, >1000 t/y [about 2500 chemicals]. Data necessary for the IPS and an initial hazard assessment are called Hedset and cover issues such as environmental exposure, environmental effects, exposure to humans, and human health effects.

At the UNCED meeting on the Environment and Sustainable Development in Rio de Janeiro in 1992, it was decided to create an Intergovernmental Forum on Chemical Safety [IGFCS, Chapter 19 of Agenda 21]. Its primary task is to stimulate and coordinate global harmonization in the field of chemical safety and covers the following principal themes: assessment of chemical risks, global harmonization of classification and labeling, information exchange, risk reduction programs, and capacity building in chemical management.

Uncertainty plays an important role in risk assessment [Suter, 1993]. Risk is the probability that a specified harmful effect will occur or, in the case of a graded effect, the relationship between the magnitude of the effect and its probability of occurrence.

Risk assessment has traditionally emphasized risks to human health over the concerns of ecological effects. However, some chemicals such as chlorine, ammonia, and certain pesticides — which have no risk or only a small amount of risk to human health — cause severe effects on ecosystems such as aquatic organisms. An up-to-date risk assessment is comprised of considerations of the entire ecological hierarchy, which is the ecologist's worldview in terms of levels of organization. Organisms interact directly with the environment, so they can be exposed to toxic chemicals. The species-sensitivity distribution is therefore more ecologically credible [Calow, 1998]. A reproducing population is the smallest meaningful level ecologically. However, populations do not exist in a vacuum; they require a community of other organisms of which the population is a part. The community occupies a physical environment with which it forms an ecosystem.

Moreover, both the various adverse effects and the ecological hierarchy have different scales in time and space, which must be included in a proper ERA [Figure 8.1]. For example, oil spills occur at a spatial scale similar to those of populations, but they are briefer than population processes. Therefore, a risk assessment of an oil spill requires the consideration of reproduction and recolonization on a longer time scale to determine the magnitude of the population response and its significance to natural population variance.

8.2.2 Uncertainties in Risk Assessment

Uncertainties in risk assessment are taken into account by application of safety factors. Uncertainties have three basic causes:

Inherent randomness of the world [stochasticity]

Errors in execution of assessment

Imperfect or incomplete knowledge

Inherent randomness refers to uncertainty that can be described and estimated but not reduced because it is characteristic of the system. Meteorological factors such as rainfall, temperature, and wind are effectively stochastic at levels of interest for risk assessment. Many biological processes such as colonization, reproduction, and mortality also need to be described stochastically.

Human errors are inevitably attributes of all human activities. This type of uncertainty includes incorrect measurements, data recording errors, computational errors, and so on.

Uncertainty is addressed using an assessment [safety] factor from 10 to 1000. The choice of assessment factor depends on the quantity and quality of toxicity data [Table 8.1]. The assessment or safety factor is used in step 3 of the ERA procedure presented in the following section. Relationships other than the uncertainties originating from randomness, errors, and lack of knowledge may be considered when the assessment factors are selected [e.g., cost-benefit]. This implies that the assessment factors for drugs and pesticides may be given a lower value due to their possible benefits.

Table 8.1. Selection of Assessment Factors to Derive Predicted No Effect Concentration

PNEC, Predicted No Effect Concentration. Note: See also step 3 of the procedure presented below.

Lack of knowledge results in an undefined uncertainty that cannot be described or quantified. It is a result of practical constraints on our ability to describe, count, measure, or quantify accurately everything that pertains to a risk estimate. Clear examples are the inability to test all toxicological responses of all species exposed to a pollutant and the simplifications needed in the model used to predict the EEC.

The most important feature distinguishing risk assessment from impact assessment is the emphasis in risk assessment on characterizing and quantifying uncertainty. Therefore, it is of particular interest in risk assessment to estimate the analyzable uncertainties, such as natural stochasticity, parameter errors, and model errors. Statistical methods may provide direct estimates of uncertainties, and they are widely used in model development.

The use of statistics to quantify uncertainty is complicated in practice by the need to consider errors in both the dependent and independent variables and to combine errors when multiple extrapolations should be made. Monte Carlo analysis is often used to overcome these difficulties [Bartell et al. 1992].

Model errors include inappropriate selection or aggregation of variables, incorrect functional forms, and incorrect boundaries. The uncertainty associated with model errors is usually assessed by field measurements utilized for calibration and validation of the model [see Chapter 2]. The modelling uncertainty for ecotoxicological models is no different from what was previously discussed in Chapter 2.

8.2.3 Step-by-Step Guide for Ecological Risk Assessment

Chemical risk assessment is divided into nine steps shown in Figure 8.2. The nine steps correspond to questions that the risk assessment attempts to answer when quantifying the risk associated with the use of a chemical.

Which hazards are associated with the application of the chemical? This involves gathering data on the types of hazards such as possible environmental damage and human health effects. The health effects include congenital, neurological, mutagenic, endocrine disruption [e.g., estrogen], and carcinogenic effects. It may also include characterization of the behavior of the chemical within the body [interactions with organs, cells, or genetic material]. The possible environmental damage including lethal effects and sub-lethal effects on growth and reproduction of various populations is considered in this step.

As an attempt to quantify the potential danger posed by chemicals, a variety of toxicity tests have been devised. Some of the recommended tests involve experiments with subsets of natural systems, such as microcosms, or with entire ecosystems. The majority of testing new chemicals for possible effects has, however, been confined to studies in the laboratory on a limited number of test species. Results from these laboratory assays provide useful information for quantification of the relative toxicity of different chemicals. They are used to forecast effects in natural systems, although their justification has been seriously questioned [Cairns et al. 1987].

What is the relation between dose and responses of the type defined in step 1? It implies knowledge of NEC and LDx values [dose that is lethal to x% of the organisms considered], LCy values [concentration lethal to y% of the organisms considered], and ECz values [concentration giving the indicated effect to z% of the considered organisms] where x, y, and z express a probability of harm. The answer can be found by laboratory examination or we may use estimation methods. Based upon these answers, a most probable level of no effect [NEL] is assessed. Data needed for steps 1 and 2 are obtained directly from scientific libraries, but are increasingly found via online data searches in bibliographic and factual databases. Data gaps should be filled with estimated data. It is very difficult to completely know about a chemical's effect on all levels from cells to ecosystem as some effects are associated with very small concentrations [the estrogen effect]. Therefore it is far from sufficient to know NEC, LDx-, LCy-, and ECz-values.

Which uncertainty [safety] factors reflect the amount of uncertainty that must be taken into account when experimental laboratory data or empirical estimation methods are extrapolated to real situations? Usually, safety factors of 10–1000 are used. The choice was discussed earlier and is usually in accordance with Table 8.1. If good knowledge about the chemical is available, then a safety factor of 10 may be applied. If, on the other hand, it is estimated that the available information has a very high uncertainty, then a safety factor of 10,000 may be recommended. Most frequently, safety factors of 50–100 are applied. NEL times the safety factor is the predicted noneffect level [PNEL]. The complexity of ERA is often simplified by deriving the predicted no-effect concentration [PNEC] for different environmental components [water, soil, air, biotas, and sediment].

What are the sources and quantities of emissions? The answer requires thorough knowledge of the production and use of the considered chemical compounds, including an assessment of how much of the chemical is wasted in the environment by production and use. The chemical may also be a waste product, which makes it very difficult to determine the amounts involved; for instance, the very toxic dioxins are waste products from incineration of organic waste.

What is [are] the actual exposure concentration[s]? The answer to this question is the PEC. Exposure can be assessed by measuring environmental concentrations. It may also be predicted by a model when the emissions are known. The use of models is necessary in most cases either because we are considering a new chemical, or because the assessment of environmental concentrations requires a very large number of measurements to determine the variations in concentrations. Furthermore, it provides an additional certainty to compare model results with measurements, which implies that it is always recommended both to develop a model and to make at least a few measurements of concentrations in the ecosystem components when and where it is expected that the highest concentration will occur. Most models demand an input of parameters, describing the properties of the chemicals and the organisms, which also requires extensive application of handbooks and a wide range of estimation methods. The development of an environmental, ecotoxicological model requires extensive knowledge of the physical-chemical-biological properties of the chemical compound[s] considered. The selection of a proper model is discussed in this chapter and in Chapter 2.

What is the ratio PEC/PNEC? This ratio is often called the risk quotient. It should not be considered an absolute assessment of risk but rather a relative ranking of risks. The ratio is usually found for a wide range of ecosystems such as aquatic and terrestrials well as ground water. Steps 1–6 shown in Figure 8.3 agree with Figure 8.2 and the information given in the previous six steps.

How will you classify the risk? Risk valuation decides on risk reductions [step 9]. Two risk levels are defined: [1] the upper limit, that is, the maximum permissible level [MPL]; and [2] the lower limit, that is, the negligible level, NL. It may also be defined as a percentage of MPL, for instance, 1% or 10% of MPL. The two risk limits create three zones: a black, unacceptable, high risk zone >MPL; a gray, medium risk level; and a white, low risk level