Safety Factors

The term safety factor is synonymous with the term uncertainty factor and grew out of what was known as application factors. For the purposes of this discussion, the term safety factor will be used to encompass all these terms. In human health and ecological risk assessments, safety factor is defined as a number by which an observed or estimated no observed adverse effect concentration (NOAEC) or dose is divided to arrive at a criterion or standard that is considered safe (or has an acceptable level of risk). Safety factors are, of course, used in engineering as well, and depend on the structure that is being supported. Most codes require a safety factor of 2, although bridges and large structures generally apply safety factors of 5-10. In all cases, the safety factor will vary, depending on such things as the circumstances under which it is to be used and the amount of data available for making the decision.

Safety factors were first proposed in 1945 as a way to calculate presumed harmless concentrations in water of toxic substances based on dose-response relationships of standard aquatic test organisms such as fish, algae, or daphnia (water fleas). The initial approach was to divide the lowest acute toxicity value by 3. This eventually was increased to dividing by multiples of 10, to take into account more of the uncertainty in extrapolating data from a few species tested for a short exposure period to all species that may be chronically exposed in different types of aquatic systems. About 10 years later, in 1954, human health risk assessors proposed the use of uncertainty factors when making regulatory decisions regarding chemical use. A 100-fold uncertainty factor was proposed, and used for many years to account for uncertainties in extrapolating animal data for the protection of humans. This has since been expanded to include margins of safety for sensitive subgroups (e.g., children), different endpoints, and potential effects. However, as the discussion of safety factors in this article will be limited to ecological assessments, their use in human health protection will not be discussed further.

The use of safety factors is now an accepted part of the hazard analysis in ecological risk assessments. They are used to account for the unknowns when predicting the threshold for effects ofpotentially hazardous substances to fish, wildlife, plants, and invertebrates. Toxicity tests can be conducted on only a few selected species held in the artificial environment of the laboratory, yet the results must be used to make predictions about potential risks to all species in any natural environments. Some margin of safety is needed to ensure protection for species or environments that are more sensitive than those that are tested. It is important to emphasize that safety factors are used only for effects assessments. Exposure estimates do not include safety factors but rather are based on either worst-case scenarios or use various statistical methods to

Table 1 Safety factors used in ecological risk assessments

Type of extrapolation

Assumptions

Safety factor

Intraspecific

Includes individual variation, test-to-test variation, and differences between laboratories

2-10

Interspecific

Wildlife species, acute data Wildlife species, chronic data Aquatic species Plants

2-4 2-1S 10-1000

Acute to chronic

Primarily for aquatic species; based on some measured values. Size of safety factor depends upon number of species tested

1-40

LOAEC to NOAEC

Depends upon the number of studies of a particular chemical and the range of species tested

10-1000

Laboratory to field

Accounts for differences between laboratory test conditions and variable real world environments

10-100

Aquatic to terrestrial

Water column to sediment organisms

Individuals to populations

10000 10 10 1-10

See the European TGD listed in 'Further reading' for detailed tables for each type of environmental media (soil, water, sediments).

See the European TGD listed in 'Further reading' for detailed tables for each type of environmental media (soil, water, sediments).

be inclusive of the potential range(s) of exposures that might occur. Safety factors (Table 1) are used to account for:

• intraspecies heterogeneity: accounts for differences between individuals of the same species, due to genetics, testing differences, or laboratory factors;

• interspecies extrapolation: accounts for differences between species due to variable physiology and response mechanisms;

• acute-to-chronic comparisons: estimates effects from long-term exposures that may span the full life cycle of organisms, from data generated from short experimental exposure times;

• LOAEC-to-NOAEC extrapolations: estimates the no effects threshold from test data that predict only the lowest effect level where responses are first observed;

• laboratory-to-field extrapolations: accounts for differences in field conditions (e.g., additional stresses such as predators, climate, or poor nutrition) that may alter the response of organisms to toxicants when compared to studies conducted under controlled conditions in laboratory settings.

In general, the more the data that are available, the smaller the safety factor that is required. For example, if extrapolations across species are made on the basis of a single species, then a safety factor of 100 or 1000 may be used. Adding additional species that represent different taxonomic groups (e.g., an algae, a fish, and an invertebrate) will reduce the safety factor to 10. If more than 10 species have been tested, then a statistical distribution can be developed to predict the sensitivity of the most affected 5% of the species (see below for more detailed discussion of this approach), in which case a safety factor of 5 is applied or, in some cases, no safety factor is used at all. Occasionally, safety factors as high as 10 000 have been used, for example, when setting toxicity thresholds for marine organisms under European risk assessment guidelines. The following sections provide additional details on the use of safety factors for each of the extrapolation categories listed above.

Intraspecies Heterogeneity

Because no two organisms are exactly alike, variation in response between individuals within the same species is a common occurrence. If all organisms were exactly the same, there would simply be a threshold dose (or concentration) where all of the organisms responded and below which none responded to the chemical exposure. This variation among individuals is described by the dose-response curve (Figure 1), which shows the proportion of the animals that is affected by the chemical at a given concentration. As with most statistically generated functions, there is uncertainty around this curve which can be quantified by the 95% confidence intervals on either side of the mean (the dotted curves in Figure 1). Conservatism in the risk estimates is then included through the decision of which point on the curve to use as an effect metric. In acute (short-term) studies, this often is the LC50 which is the concentration (or dose) that kills 50% of the test animals. Using the upper confidence interval from the dose-response curve at this value adds further conservatism to the estimate (i.e., results in a benchmark dose which is the lower confidence interval around the LC50 concentration; see Figure 1). For chronic (long-term)

Benchmark dose: Lower CI on the 50th percentile

log [dose]

Figure 1 Dose-response curve showing the percent of test organisms that respond at each chemical concentration. The dotted lines on either side of the solid curve represent the 95% confidence interval. The EC50 (or LC50, if the response measured is mortality), is determined by reading across from where 50% of the organisms respond, and then down to the corresponding chemical concentration that caused this level of response.

studies, a nonlethal endpoint generally is used, often selecting the EC10 or EC2o (effects threshold for 10% or 20% of the tested organisms, respectively). Again, the lower confidence interval of the EC10 will provide additional protection and is used as the benchmark dose. An alternative to using confidence intervals about the LC(EC)X value is to apply a safety factor to the value from the calculated curve (i.e., the solid line in Figure 1). Several studies have examined the intraspecific variability in both aquatic and terrestrial organisms, and conclude that a safety factor of 10 is appropriate for bounding the variability of within species responses.

Intraspecific variability also includes test-to-test variation or differences between laboratories. This results from variations in the strain of organisms used for testing, deviations from specifications in the physical or chemical properties of the test media (water or soil), and other operator-dependent effects. For standard aquatic test organisms, there is a two- to fivefold difference between laboratories in acute toxicity (LC50) values when the same chemical is tested in the same species. Intralab replication of the same test by the same researchers usually results in only a twofold variation. Generally, this variability is taken into account as part of the intraspecies safety factor, and no additional safety factor is applied.

Interspecies Extrapolation

The most common use of safety factors is to extrapolate data from tested species to other, nontested organisms to provide protection for at least 95% of all species. Unfortunately, while there is a general pattern to relative sensitivity of species (particularly within particular classes of chemicals), the same species will not be the most (or least) sensitive for all chemicals. Therefore, it generally is not possible to a priori pick the most sensitive species for testing. Examination of data from pesticide studies where a large number of chemicals have been tested in the same animal species suggests that the maximum difference in species sensitivity following dietary (chronic) exposure is about sevenfold for birds and fourfold for mammals, and acute lethal doses vary no more than tenfold (birds) or 20-fold (mammals). These data provide support for use of an interspecies wildlife safety factor for pesticide risk assessments of 10-20 when using acute data or 5 when using chronic (dietary) exposures. However, the range in response to nonpesticide chemicals is greater (likely due to the greater range of modes of action of these substances), with 95% of the results being within a factor of 50. Bayesian statistical analyses currently are being investigated as a possible way of using this prior knowledge of relative sensitivities to make predictions about interspecific responses to new chemicals that may result in more reliable safety factors.

For terrestrial plants, a safety factor of 2 will capture 80% of the total potential variability among genera within a single family. However, most extrapolations are done across families or orders which appear to require a safety factor of at least 15 to capture 80% of the variability. Acute toxicity for freshwater aquatic organisms, on the other hand, can vary over 5 orders of magnitude (100 000-fold). This larger variability likely is due to inclusion of multiple classes of organisms in the database, including plants, vertebrate fishes, and invertebrates. Variability within a single order ranges from one- to fourfold differences. As with other taxonomic groups, no single species is always the most sensitive to all chemicals.

For taxonomic groups where the range of relative sensitivities is unknown (e.g., reptiles, amphibians, soil invertebrates, or saltwater species), the lowest LC50 (or LD50) value (or other measured endpoint) would be used and a larger safety factor applied to that value. In these instances, the safety factor generally ranges from 10 to 1000, depending upon how many species have been tested, whether there are at least three different trophic levels represented (a plant, a primary consumer, and a carnivore), and if the studies are acute (i.e., very short) or chronic (i.e., long-term) exposures.

If more than eight species are tested, then a statistical distribution oftheir LC50s (or another benchmark toxicity value) is derived and the point on the curve that represents the LC50 of the most sensitive 5% of the species is determined (Figure 2). This is discussed in more detail below (see the section titled 'Alternative methods for incorporating uncertainty').

Figure 2 Species sensitivity distribution, showing the cumulative density function of the response of all tested species. The dotted lines on either side of the solid curve represent the 95% confidence interval. The toxicity threshold value is taken as either the 5th percentile on the mean (solid line) curve, or the 5th percentile from the upper confidence interval (dotted line) curve.

Figure 2 Species sensitivity distribution, showing the cumulative density function of the response of all tested species. The dotted lines on either side of the solid curve represent the 95% confidence interval. The toxicity threshold value is taken as either the 5th percentile on the mean (solid line) curve, or the 5th percentile from the upper confidence interval (dotted line) curve.

Acute-to-Chronic Comparisons

Many toxicity studies are conducted over a very short time period (generally a few hours); these are known as acute toxicity studies. However, in nature, organisms frequently are chronically (i.e., continuously) exposed to pollutants. For aquatic organisms, it is frequently possible to establish the relationship between the toxicity thresholds following acute exposure and those following long-term (days to months), chronic exposures. An acute-to-chronic ratio (ACR) can be derived by dividing the acute LC50 by the chronic NOAEC. Because many more species are subjected to acute toxicity studies than to chronic ones, the ACR derived from a few tested species may be used to estimate the chronic no effect level for many other (nontested) species. It is important to note that this extrapolation is done almost exclusively for aquatic organisms and that ACRs have not been generated for plants or wildlife. This ACR often is similar across species for the same chemical, although in some cases it may vary significantly (up to 20 000-fold differences), particularly when using data from multiple chemicals. Thus, the preferred approach is to use an experimentally derived ACR for a particular chemical, taking the average of the values from all the various species that have been tested. If the number of species for which an ACR has been measured is less than 10, then a safety factor of 20-40 may be applied to the acute toxicity threshold to estimate a chronic effects level; if it is greater than 10, then no safety factor is applied.

LOAEC-to-NOAEC Extrapolations

To ensure adequate protection of species of concern without expending more money, time, or resources than is necessary, risk managers would like to know the maximum amount of a chemical to which organisms can be exposed without showing any significant adverse effects. If this threshold is incorrectly estimated to be higher than what really occurs, then organisms may continue to be affected even after corrective or preventive actions are put in place whereas if the estimate is too low then unnecessary mitigation expenses might be incurred. Therefore, it is important to estimate the toxicity threshold as accurately as possible and to apply a safety factor that will err on the side of ecosystem protection. It is now commonly accepted that nearly all contaminants have a threshold of effect, although thresholds for cancer initiations may be so low as to be considered nonexistent. For the purposes of ecological risk assessments, however, cancer endpoints generally are not considered and all chemicals are assumed to have measurable thresholds.

Most risk assessments are based on the chronic NOAEC. This is the highest tested concentration of a chemical that causes no statistically significant response of the test organisms under a particular study design. However, this piece of information frequently is not available, and only the lowest concentration that did cause a statistically significant effect (the LOAEC) is reported. It is difficult to extrapolate NOAECs from the LOAECs for most chemicals, as often there is an insufficient amount of information to reach any definitive conclusion. Many studies report unbounded values, that is, either all test organisms showed some response (an unbounded LOAEC) or no test organisms responded (an unbounded NOAEC). Furthermore, even studies that result in calculation of both an NOAEC and an LOAEC cannot definitively define the true threshold value. The NOAEC and LOAEC values are generated using an analysis of variance statistic that is dependent upon study design factors such as number of organisms in each test concentration, exposure levels tested, and variability of the organism responses. Given these design dependencies, different NOAEC or LOAEC values may be generated for the same organism-chemical combinations simply by redesigning the study. Even assuming that the resulting values are true representations of no or low effects, they frequently are at least an order of magnitude apart and the actual concentration where at least some organisms begin to show measurable responses to chemical exposure lies at some unknown distance between the two. Therefore, the convention has evolved to estimate the no effect level (NEL) as the geometric mean of the NOAEC and LOAEC values. Again, this is most frequently applied in aquatic risk assessments, but there have been some similar applications to soil organism tests (e.g., soil invertebrate or plant toxicity testing).

Because of the inherent biases in derivation of the NOAEC and LOAEC data, there are no reliable statistics to estimate what the usual ratio of these two values might be for a variety of species across chemical classes. Therefore, an alternative approach has been suggested to use dose-response functions (Figure 1) rather than the analysis of variance approach, and develop benchmark doses rather than NOAEC-LOAEC values. The NEL threshold is then estimated by applying a 100-fold safety factor to the acute LC50 (or LD50) value (i.e., dividing it by 100) or a tenfold safety factor if the benchmark dose is based on a chronic endpoint. While this has begun to be accepted in some risk assessment applications, especially for terrestrial wildlife, the NOAEC-LOAEC approach remains standard practice in many others. Safety factors may be applied to the NEL and often are based on professional judgment. Depending upon the number of studies of a particular chemical and the range of species tested, safety factors may range from 10 to 1000.

Laboratory-to-Field Extrapolations

Organisms tested under laboratory conditions may not respond the same as their wild counterparts. Laboratory tests frequently are conducted using highly controlled conditions that are known to be nonstressful to the organism (e.g., most appropriate temperatures, constant and sufficient feeding of nutritionally adequate diets, absence of predators, etc.), although it is possible to purposely simulate environmental stress (e.g., temperature regime changes) or unintentionally introduce novel stresses (e.g., isolation housing of gregarious species or fluorescent lighting wavelengths and flicker rates). In nature, organisms are exposed to multiple stressors, simultaneously and often continuously. Field studies generally are less controlled than laboratory studies, although use of caged animals, potted plants, or careful application of chemicals can standardize many of the study variables. Such experiments generally benefit from less variable and more defined exposure but do not reliably control other environmental stressors. Stress of any kind, whether in laboratory of field, may alter organisms' physiology and therefore change their response(s) to a particular stressor, such as a chemical pollutant. Many studies have been conducted to examine the differences between responses under these two sets of conditions (laboratory and field) with mixed results. About half the time, laboratory studies yielded lower toxicity thresholds and the other 50% of the time organisms in the field were more sensitive. Obviously, this remains an area of high uncertainty, and safety factors of 10 to 100 may be applied to laboratory data to compensate.

Other Safety Factors

Other safety factors are used occasionally by some jurisdictions, but not others. For example, the United States Environmental Protection Agency (US EPA) will sometimes estimate effects to terrestrial organisms by dividing an aquatic toxicity threshold by 10 (based on professional judgment only, with no empirical basis for this particular value). In Canada, toxicity thresholds for sediment-dwelling (i.e., benthic) organisms frequently include a tenfold safety factor, even if information from toxicity tests conducted with sediment organisms is available. This is to account for the potential of ingested chemicals to add to toxicity resulting from gill uptake from water. As mentioned previously, effects thresholds for saltwater organisms may be estimated using data from freshwater organisms and the application of a very large (up to 10 000) safety factor. Britain and other European countries also apply safety factors when estimating changes in population growth rates as a consequence of measured effects to individual organisms. This seems counterintuitive as populations have many compensatory factors that allow some effects to occur to individuals before changes in population growth rates are manifest (as exemplified by hunting or harvest mortality of ubiquitous animals such as deer).

It is common practice to use multiple safety factors when developing a regulatory threshold that is protective of organisms of concern. For example, if toxicity data are available from only two aquatic species tested in short-term studies (e.g., an LC50), then a tenfold safety factor is applied to account for interspecific difference, another tenfold factor (or, in some cases, a 100-fold factor) is used for acute-to-chronic conversion, and yet another tenfold factor may be applied for laboratory-to-field extrapolations. The end result is a total safety factor of 1000 to 10 000 being applied to the measured toxicity data. This drives the acceptable risk threshold very low. While this is appropriate for humans, we need to ask if it is equally appropriate for valued ecological resources (i.e., plants, fish, and wildlife). Toxicity data are based on studies of the effects of chemicals on particular attributes of single organisms, and risk assessments generally provide estimates of potential effects on organism health (including longevity and reproductive ability). However, naturally occurring species depend upon relative fitness of adults and their potential to contribute to the gene pool and successfully raise offspring. Populations can sustain significant mortality or reduced reproductive rates before entering into an inevitable decline. Therefore, it is likely that use of multiple safety factors results in overly protective threshold values. However, how chemical exposures affect species interactions or whether adverse affects to populations of one species (e.g., predators) benefit another (e.g., prey species) remain largely unknown. Competing environmental management goals also add significant debate to required level of protection. For all these reasons, generation of additional data to more accurately estimate the variability in organism responses is highly encouraged as the use of large, and often arbitrary, safety factors can then be avoided.

Was this article helpful?

0 0
10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Responses

  • Anke
    Does a safety factor decrease LC50?
    7 years ago

Post a comment