Abstract
The health effects of air pollution are difficult to characterize because of the large number of air pollutants present in the atmosphere and the relatively small contribution of their health effects compared to all other causes. In addition, air pollution does not affect all people in the same way. Some persons are more sensitive than others: for example, those suffering from asthma, chronic obstructive pulmonary disease (COPD) or cardiovascular problems, the elderly, and children. Also, some individuals are more vulnerable than others: those include, for example, workers and residents who tend to be in locations where air pollution exposure is greater than average. This chapter describes first how adverse health effects of air pollution can be identified and quantified using toxicological and epidemiological studies. Next, methods commonly used to conduct health risk assessments related to air pollution are presented. The use of such information to set up air quality regulations is presented in Chapter 15.
The health effects of air pollution are difficult to characterize because of the large number of air pollutants present in the atmosphere and the relatively small contribution of their health effects compared to all other causes. In addition, air pollution does not affect all people in the same way. Some persons are more sensitive than others: for example, those suffering from asthma, chronic obstructive pulmonary disease (COPD) or cardiovascular problems, the elderly, and children. Also, some individuals are more vulnerable than others: those include, for example, workers and residents who tend to be in locations where air pollution exposure is greater than average. This chapter describes first how adverse health effects of air pollution can be identified and quantified using toxicological and epidemiological studies. Next, methods commonly used to conduct health risk assessments related to air pollution are presented. The use of such information to set up air quality regulations is presented in Chapter 15.
12.1 Identification and Characterization of Health Effects
12.1.1 Toxicology and Epidemiology
Two large categories of health studies are used to understand and quantify the health effects of air pollution. Those are toxicological studies and epidemiological studies. The former are conducted in laboratories under controlled conditions and are useful to study specific pollutants. The latter are field studies that estimate the adverse health effects of air pollution in the ambient environment and aim to isolate the components of the individual air pollutants using statistical methods. These two types of studies present advantages and shortcomings and it is preferable to obtain results from both types of studies in order to properly characterize the health effects of an air pollutant. Brief descriptions of toxicological and epidemiological studies are presented in Sections 12.1.2 and 12.1.3, respectively. Next, the approaches used to combine the results obtained from such distinct types of health effect studies are summarized in Section 12.1.4.
12.1.2 Toxicological Studies
The term “toxicological studies” is used here to represent an ensemble of studies characterized by controlled experimental conditions. However, these studies may be extremely different: for example, in vitro versus in vivo studies. Furthermore, in vivo studies may be conducted with human volunteers or animals, such as rats, mice, monkeys or dogs (e.g., Wichers Stanek et al., 2011).
In vitro studies allow one to formulate hypotheses concerning the toxicity of a chemical substance and also to investigate the mechanisms leading to adverse health effects at the cellular or even molecular level (e.g., Devlin et al., 2005). For example, the Ames test allows one to characterize the mutagenicity of chemical substances and, therefore, to identify the substances that could be carcinogenic, since most carcinogenic substances are mutagenic (Ames et al., 1975). These in vitro studies are conducted with cell cultures in the laboratory. In the case of atmospheric particles, for example, cell cultures of the epithelium of the lung are used to understand how particle size may affect the transfer of those particles in such tissues. The oxidative stress of cells and tissues, which could lead, for example, to the production of inflammatory mediators, has also been studied in vitro with cell cultures.
In vivo studies conducted with human volunteers typically involve exposing several individuals to various concentration levels of an air pollutant in order to determine at which concentration level some adverse health effect is observed. These studies are performed in a clinical laboratory and the exposure conditions are, therefore, well controlled. However, such studies are limited to short exposure durations (referred to as acute exposure). In addition, sensitive individuals are generally not exposed, because of the significant health risks that could be associated with their exposure to moderate and high air pollutant concentrations.
Toxicological studies conducted on animals offer several advantages: more hypotheses may be tested, the reproducibility of the results can be tested, longer exposure durations can be used (referred to as chronic exposure), and sensitive animals can be exposed. Furthermore, exposure can be continued until death of the animal so that autopsies may then be performed to better understand the biological phenomena involved. Therefore, more information may be obtained with animal studies than with human studies, but the study protocols must be defined so that animal suffering is minimized. However, the results obtained from animal studies must be extrapolated to humans. Models are used to perform these extrapolations, but there are many associated uncertainties, particularly when the physiological functions of the animal differ significantly from those of humans. For example, the dosimetry of fine particles in the respiratory system may be different in some animals (e.g., rodents, monkeys, dogs) and humans. Therefore, the extrapolation of such animal studies to humans would be adversely affected by this different dosimetry if it were not explicitly taken into account. Indeed, particles deposit within the respiratory system and only a fraction penetrates deeply to reach the lungs and deposit there. These particle deposition processes depend on particle size, on the characteristics of the airflow within the respiratory system, and on the surfaces available for deposition (see Chapter 11). Figure 12.1 illustrates the efficiency of particle deposition in the human respiratory system as a function of particle size. Ultrafine particles (diameter <0.1 μm) are those that deposit the most in the lungs. Coarse particles (diameter >2.5 μm) deposit in the upper part of the respiratory system (nose, mouth) and, therefore, do not penetrate deeply into the respiratory system. Fine particles (diameter between 0.1 and 2.5 μm) deposit little in the upper part of the respiratory system and, therefore, may reach the lungs; however, they do not deposit significantly within the lungs compared to ultrafine particles.
Figure 12.1. Fraction of particles deposited in different regions of the human respiratory system as a function of particle size for different breathing intensity levels (light, normal, and heavy).
The advantage of toxicological studies is that a specific pollutant may be studied under controlled and reproducible conditions. In the case of human studies, the concentration levels above which an adverse health effect is observed can be quantified for short-term exposure of healthy individuals. In the case of animal studies, chronic exposure studies may be conducted and higher concentration levels may be used. After autopsy, detailed information on the causes of death may be obtained, which could not be obtained otherwise.
Shortcomings of toxicological studies include the fact that the concentrations used are generally greater than those observed in the ambient atmosphere and that the effect of an individual air pollutant may differ from its effect in the presence of other air pollutants. In addition, although some studies may be conducted with sensitive animals, that is typically not the case with sensitive human individuals.
12.1.3 Epidemiological Studies
In an epidemiological study, the objective is to quantify the statistical relationship between the concentration of a pollutant and an adverse health effect (Rothman, 2012). The large number of pollutants present in the ambient air makes the characterization of the effects of a single pollutant difficult because of interference by other pollutants that may have similar effects. These are called confounding factors. Other confounding factors are present due to personal behavior (e.g., smoking) or environmental conditions (e.g., heat wave, cold weather). In addition, the adverse health effects due to air pollution are generally small in terms of excess relative risk (a few %), which makes their quantification in the presence of confounding factors even more difficult. Therefore, the statistical analysis plays a major role in epidemiological studies, because it must be sufficiently powerful and robust to isolate and quantify the relationship between the pollutant concentration and the corresponding adverse health effect.
Epidemiological studies provide quantitative results that are statistical associations between a pollutant concentration and an adverse health effect. However, they do not provide any information on a cause-effect relationship. Therefore, cause-effect relationships must be obtained from toxicological studies. In addition, there is a large number of uncertainties associated with the pollutant concentration measurements, the exposure of the population or individuals, confounding factors, etc., which must be taken into account when presenting the results of an epidemiological study. For example, individuals may be exposed to pollutants outdoor and indoor (at home, at work, etc.). Pollutant concentrations may vary considerably among those various microenvironments, for example, because of indoor pollution sources, various indoor penetration rates for outdoor pollutants, loss processes for indoor pollutants by deposition on walls, furniture, etc. (e.g., Abt et al., 2000). Outdoor ambient pollutant concentrations are generally used in epidemiological studies, because they are the most readily available via air quality monitoring networks. Therefore, the characterization of the actual exposure of the subjects is approximate if exposure to indoor air is significantly different from exposure to outdoor ambient air. All these uncertainties must be quantified and the results of an epidemiological study are presented with error bounds (for example, the 95 % confidence interval; i.e., the interval that has a 95 % probability of including the true value).
There are several types of epidemiological studies. Two categories of epidemiological studies that are the most widely used in air pollution are briefly described next. They are (1) longitudinal studies, where the adverse health effects are analyzed as a function of air pollution levels that vary with time and (2) cross-sectional studies, where health risks are quantified in terms of their spatial variation.
Longitudinal Studies
In a longitudinal study, the occurrence of adverse health effects is observed as a function of time and correlation with exposure to air pollution is analyzed. Generally, ambient air pollutant concentrations are used as a surrogate for exposure.
Longitudinal studies may be ecological studies. Then, the adverse health effects are studied for a whole population (hospital admissions, deaths …) as a function of air pollutant concentrations. A latency period for the health effect (one day or a few days) may be used to account for the fact that the health effect may not be maximum shortly after exposure to the air pollution, but instead after some time. In an ecological study, there is no information on the specific exposure of individuals to the air pollution (some may have been exposed to greater air pollutant concentrations than others, depending on their activity, location of residence, etc.). Therefore, the statistical power of the ecological study must be sufficiently large to smooth out the exposure variability.
Longitudinal studies may follow a cohort of individuals during a specific period. In such cases, the exposure of the individuals can be characterized with some level of accuracy. A cohort study is more expensive than an ecological study since it requires collecting additional information on the cohort members, but it provides more accurate data in terms of the individual exposure and specific health effects.
An approach used to minimize the cost associated with a longitudinal study, while taking into account the exposure and health effects of specific individuals, is to conduct a case-control study. A case-control study is particularly useful in cases where the health risks related to air pollution are low compared to those of other causes. Identifying and quantifying comparatively low health risks in a cohort requires following a large number of individuals in order to have sufficient statistical power; therefore, large resources are needed and the associated costs may be significant. In a case-control study, the individuals that show adverse health effects are identified first: they are the “case” group. Next, individuals with similar behavior, but who do not show those adverse health effects are identified. These individuals must be at least as many as those of the case group; they are generally selected to be in greater number to obtain a larger statistical sample. These individuals are the “control” group. The exposures of the case and control groups to air pollution are then estimated retrospectively.
There are two major differences between a cohort study and a case-control study. In the cohort study, the exposure of the individuals constituting the cohort are estimated first and the adverse health effects are identified next and correlated with the estimated exposure. In the case-control study, the individuals with adverse health effects are identified first and the exposures of individuals with and without the health effects are estimated next. The other difference is that a case-control study uses a smaller number of individuals than a cohort study. Thus, a case-control study is less costly than a cohort study. However, its statistical power is less and the results are less robust. In addition, exposure is estimated, but is generally not documented with precision. Therefore, the incidence rates (see definition in the section on Definitions of Various Terms) of the health effect cannot be determined, because there is no information on the population corresponding to the selected individuals with the adverse health effects. Therefore, absolute risks cannot be calculated and only relative risks are estimated. The result of such a calculation in a case-control study is called an odds ratio (OR): the ratio of the odds of an adverse health effect for exposed individuals and the odds of the same adverse health effect for non-exposed individuals (see the calculation of the odds ratio in the section on Definitions of Various Terms).
Cross-sectional Studies
In a cross-sectional study, the difference in adverse health effects is analyzed between two populations that have different exposures to air pollution. A cross-sectional study may be conducted, for example, as an ecological study at the level of urban areas. Then, the exposure of the population in each urban area may be estimated from air pollutant concentrations measured by the air quality monitoring network and the adverse health effects may be estimated from hospital admissions (morbidity studies) and death certificates (mortality studies). Cross-sectional studies may also be conducted with smaller areas and populations, such as populations living at different distances of a major roadway or industrial site. Cross-sectional studies pertain to chronic exposure to air pollution.
Definitions of Various Terms Used in Epidemiology: Incidence, Risk, and Odds-ratio
The main terms used in epidemiology to quantify adverse health effects are defined further in this section. In the examples provided, one refers to exposed and non-exposed groups for the sake of clarity. Actually, air pollutants are always present in some amount in the atmosphere, but at different concentration levels. Therefore, epidemiological studies that pertain to air pollution compare groups exposed to air pollution levels that are significantly different, rather than groups exposed to zero and non-zero air pollutant levels. The results of air pollution epidemiological studies are presented either in terms of relative risk or odds-ratio per increment of air pollutant concentration; for example, for an increment of 10 μg m−3 or 10 ppb. An example of results from epidemiological studies conducted to characterize the adverse health effects due to fine particles, PM2.5, in terms of concentration differences of 10 μg m−3 is shown in Figure 12.2.
Figure 12.2. Relative risk for the statistical association between premature death (all causes) and an increase in PM2.5 concentration of 10 μg m−3 from the meta-analysis of Hoek et al. (2013). The relative risk is noted ES for effect estimate. The 95 % confidence intervals are indicated in parentheses and by the solid black lines; the weight given to each study is listed in % in the far-right column and illustrated by the gray squares. The result of the meta-analysis is indicated by the diamond. See Hoek et al. (2013) for details on the different studies and the method used for the meta-analysis.
Incidence: The incidence rate, IH (or simply incidence), is defined as the number of individuals who develop an adverse health effect (NH) divided by the duration of the exposure to the pollutant for all individuals in the cohort (TN):
(12.1) If all individuals (total number: NT) have been exposed over the same duration DE:
(12.2) Risk: The health risk, RH, is defined as the incidence rate multiplied by the exposure duration:
The health risk is therefore a unitless fraction. If the exposure duration is identical for all exposed individuals:
(12.4) The excess relative risk, ER, of an air pollutant is defined as the excess risk due to the pollutant (i.e., the difference in the risk for two distinct levels of exposure to the pollutant, RH1 and RH0) divided by the risk for the lowest level of exposure to the pollutant (RH0):
(12.5) The relative risk, RR, may be written in terms of a risk ratio or an incidence ratio (assuming identical exposure durations):
(12.6) Therefore:
As mentioned previously, the excess relative risk is typically calculated for an increment in pollutant concentration, rather than with respect to a zero concentration of the pollutant.
The etiologic fraction of risk (EFR) is defined as the difference between the risks for two distinct levels of exposure to the pollutant (the current level of exposure of the population to the pollutant and a lower level of exposure, typically considered safe) divided by the risk of the population exposed to the current level of the pollutant (i.e., it is the health risk fraction that is due to the pollutant):
(12.8) If RR ≈ 1, then: EFR ≈ ER.
Odds ratio: In the case of case-control studies, the case group (individuals showing an adverse health effect) includes a subgroup exposed to the pollutant, which is represented here by a number of individuals aN, and a subgroup not exposed to the pollutant (or exposed to a significantly lower concentration level of that pollutant), represented by a number of individuals bN. The incidence rates are respectively:
(12.9) There is no information available to define TN1 and TN0. Therefore, they must be estimated. If one selects in the control group an exposed subgroup with a number of individuals cN and a non-exposed subgroup (or a subgroup exposed to a lower concentration level) with a number of individuals dN, such that:
(12.10) Then:
(12.11) The odds ratio is the ratio of the odds that the exposed individuals show an adverse health effect (i.e., the ratio of the probability to show an adverse health effect and the probability not to show any adverse health effect), (aN / cN), and the odds that the non-exposed individuals show this same adverse health effect, (bN / dN). Therefore:
However, this relationship is verified only if the control group represents correctly the statistics of the whole population to which the control group belongs. If the control group represents only approximately the population exposure, then:
As an example, Table 12.1 displays the results of an epidemiological case-control study where benzene is the pollutant of interest and the adverse health effect is child acute myeloblastic leukemia.
Table 12.1. Contingence table showing some aspect of a case-control study for exposure to benzene in close proximity to on-road traffic and incidence of child acute myeloblastic leukemia. Source: Houot et al. (2015).
| Exposure to benzene (concentration level in parentheses) | Acute myeloblastic leukemia | |
|---|---|---|
| Yes (cases) | No (controls) | |
| Yes (>1.3 μg m−3) | aN = 59 | cN = 3 238 |
| No (<1.3 μg m−3) | bN = 33 | dN = 2 909 |
The odds ratio is: OR=aN dNbN cN=59×290933×3238= 1.6
The 95 % confidence interval given by Houot et al. for this odds ratio ranges from 1.0 to 2.4. The study by Houot et al. (2015) is actually more complete and includes an analysis of the distances between the exposed individuals and the roadway and of the road segment lengths. The whole set of results suggest some association between benzene exposure and the incidence of child acute myeloblastic leukemia.
12.1.4 Analysis of Adverse Health Effects
Determination of the Cause-Effect Relationship
A list of the various aspects used to determine a cause-effect relationship was initially proposed by Hill (1965) for epidemiological studies. The U.S. Environmental Protection Agency (EPA) has extended this list to include toxicological studies (i.e., clinical human exposure studies, toxicological animal studies, and in vitro studies). It is as follows:
Consistency: An inference of causality is strengthened when a pattern of elevated risk is observed across several independent epidemiological studies. However, statistical significance is not the only factor to be considered, as discrepancies among epidemiological studies may result from different exposures, confounding factors, and statistical power of the studies.
Strength of the observed association between exposure and health effect: Large and precise adverse health effects increase confidence that the observed statistical association is not random or due to experimental biases.
Specificity of the statistical association: A given pollutant leads to a specific adverse health effect (it is, however, seldom the case).
Temporal relationship between exposure and the health effect: The effect occurs after the exposure.
Biological gradient: The effect increases with the exposure level.
Biological plausibility: The demonstration of a biological process linking the exposure to the effect reinforces the possibility of a relationship between exposure and effect.
Coherence: Qualitatively similar results between clinical or laboratory (toxicological) studies and epidemiological studies reinforce the interpretation of a causal relationship.
Experimental evidence: It is provided by the epidemiological studies.
Analogy: Analogies with other chemical compounds having similar chemical structures may suggest some type of health effect.
It is generally accepted that the determination of an adverse health effect due to a pollutant requires that several of these factors be verified (but not necessarily all of them). In the case of air pollutants, this goal is generally attained by combining toxicological and epidemiological studies. The following aspects predominate:
Plausibility of the health effect due to the air pollutant: this plausibility requires that a biological mechanism be identified that can relate the inhalation or ingestion of the pollutant to an observed adverse health effect. This plausibility is typically demonstrated using toxicological studies.
Coherence between toxicological studies and epidemiological studies: this coherence is established, for example, via similar effects being obtained for concentrations of the same order of magnitude. The effects obtained in an epidemiological study will generally be obtained for lower exposure levels than those of toxicological studies, because in the former sensitive individuals will be exposed, whereas such sensitive individuals would have been eliminated from a clinical study. In addition, an epidemiological study may show an effect for chronic exposure, whereas only acute exposure can be studied in clinical studies. Coherence is, therefore, qualitative or in some cases semi-quantitative.
Consistency among epidemiological studies: Consistency is obtained when several epidemiological studies lead to similar results. These studies may give results that are quantitatively different, but it is inappropriate if some studies show a significant effect while several others show no effect.
The results of an analysis of adverse health effects (combining toxicology and epidemiology) may be grouped as follows in terms of causation between the health effect and the exposure to the pollutant (EPA, 2008b):
Sufficient to conclude that there is a causal relationship
Sufficient to conclude that there is a likely causal relationship
Suggestive but not sufficient to conclude that there is a likely causal relationship
Inadequate to conclude that there is or not a causal relationship
Suggestive of the lack of a causal relationship
A regulation is generally proposed in the first two cases (the precaution principle is implicitly applied in the second case).
Table 12.2 summarizes the analyses of relationships between pollutant exposure and health effects conducted by the U.S. EPA for air pollutants regulated in the United States and Europe. It appears that there is sufficient evidence for causal relationships concerning adverse health effects (morbidity) for chronic exposure to lead (Pb) and benzene, and acute exposure to sulfur dioxide (SO2), nitrogen dioxide (NO2), and ozone (O3), and both acute and chronic exposures to fine particles (PM2.5). On the other hand, evidence is less certain (likely or only suggestive of a likely causal relationship) for acute and chronic exposure to carbon monoxide (CO), and chronic exposure to NO2 and O3. The evidence is inadequate for chronic exposure to SO2. For inhalable coarse particles (i.e., those particles of diameter between 2.5 and 10 μm) and ultrafine particles (those particles with a diameter less than 100 nm), the evidence is only suggestive of an adverse health effect for acute exposure and is inadequate for chronic exposure.
Table 12.2. Adverse health effects of air pollutants regulated in the United States and in Europe. Sources: U.S. Environmental Protection Agency; EPA, 2008a, 2008b, 2009, 2010, 2013a, 2013b, 2016.
| Cause-effect relationshipa | |||
|---|---|---|---|
| Pollutant | Duration | Morbidityb | Mortalityb |
| Lead (Pb) | Long term | Causal (children: cognitive deficit; adults: hypertension, cardiovascular diseases, reproduction …) | Causal (cardiovascular diseases) |
| Carbon monoxide (CO) | Short term | Likely (cardiovascular diseases) | Suggestive |
| Long term | Suggestive (child birth and development) | None | |
| Sulfur dioxide (SO2) | Short term | Causal (respiratory system) | Suggestive |
| Long term | Inadequate | Inadequate | |
| Nitrogen dioxide (NO2) | Short term | Causal (respiratory system) | Suggestive |
| Long term | Suggestive | Inadequate | |
| Ozone (O3) | Short term | Causal (respiratory system), likely (cardiovascular diseases) | Likely |
| Long term | Likely (respiratory system) | Suggestive | |
| Fine particles (PM2.5) | Short term | Causal (cardiovascular diseases), likely (respiratory system) | Causal |
| Long term | Causal (cardiovascular diseases), likely (respiratory system), suggestive (reproduction, cancers) | Causal | |
| Coarse particles (PM10-2.5) | Short term | Suggestive (cardiovascular diseases, respiratory system) | Suggestive |
| Long term | Inadequate | Inadequate | |
| Ultrafine particles (PM0.1) | Short term | Suggestive | Inadequate |
| Long term | Inadequate | Inadequate | |
| Benzene | Long term | Causal (lower number of lymphocytes) | Causal (leukemia) |
(a) Definitions of the pollutant exposure-health effect relationships:
Causal: Sufficient to conclude that there is a causal relationship with relevant pollutant exposures
Likely: Sufficient to conclude that a causal relationship is likely to exist with relevant pollutant exposures
Suggestive: Suggestive but not sufficient to conclude that there is a likely causal relationship
Inadequate: Inadequate to conclude that there is or not a causal relationship
None: Suggestive of the lack of a causal relationship
(b) Effects are mentioned in parentheses if the information is available.
Causal relationships are conclusive for deaths (mortality) due to chronic exposure to lead (cardio-vascular diseases) and benzene (leukemia) and due to both acute and chronic exposures to fine particles. There is sufficient evidence of a likely causal relationship (mortality) due to acute exposure to ozone. However, the experimental data are only suggestive of a causal relationship (mortality) for acute exposure to CO, SO2, NO2, and inhalable coarse particles, and for chronic exposure to ozone. For other exposure durations and other pollutants, the data are inadequate to conclude to a relationship between exposure and mortality.
A regulation is generally proposed when the causal relationship is conclusive, whether it is for morbidity or mortality. A regulation may also be proposed when the causal relationship is likely or if the information is suggestive of a causal relationship; however, it will not always be the case and the regulatory context then plays a role (history of regulations, existing similar regulations …). The setup of air pollutant regulations is described in Chapter 15.
Classification of Carcinogenic Chemical Substances
Health agencies such as the International Agency for Research on Cancer (IARC, which is part of the World Health Organization, WHO) and the U.S. Environmental Protection Agency develop lists of carcinogenic and non-carcinogenic substances. Chemical substances may be classified as follows:
Category 1: carcinogenic to humans
Category 2A: probably carcinogenic to humans
Category 2B: possibly carcinogenic to humans
Category 3: not classifiable as to carcinogenicity to humans
Category 4: probably not carcinogenic to humans
For example, benzene, which is regulated as an air pollutant in Europe, is listed in Category 1 (cause of leukemia for chronic exposure).
The list of chemical substances evaluated by the IARC is available at: http://monographs.iarc.fr
The list of chemical substances evaluated by the U.S. EPA is available at: http://www.epa.gov/iris
12.2 Health Risk Assessment
Health risk assessment (HRA) provides quantitative estimates of the health risks associated with exposure to pollutants of an individual or a population. These pollutants may be present in the air or may have been transferred from the atmosphere to other environmental media (water, soil, vegetation, animals …) following atmospheric deposition to an ecosystem. Multi-media modeling is required in the latter case (see Chapters 11 and 13). Therefore, several exposure pathways must be taken into account. Exposure to pollutants present in the atmosphere occurs via inhalation. Exposure to pollutants present in other media may occur via ingestion and/or dermal absorption.
An HRA may be seen as a four-step process as described by the U.S. National Research Council (NRC, 1994) and recommended by the U.S. Environmental Protection Agency (EPA). These four steps are the following:
– Hazard identification: It pertains to the identification of the chemical substances that are likely to lead to adverse health effects either through inhalation or through other exposure pathways after their transfer to environmental media other than the air. Typically, it consists of identifying the chemical substances emitted from a source or present in the environment and assessing whether they are toxic.
– Dose-response assessment: It involves the selection of quantitative relationships between the dose of a chemical substance absorbed by an individual and the corresponding adverse health effects (i.e., the response). They must be defined for each chemical substance identified in the previous step and may be defined for short- and long-term exposure as well as for carcinogenic and non-carcinogenic effects.
– Exposure assessment: It pertains to the quantitative estimate (calculation or measurement) of the exposure of the individuals or population studied to the chemical substances identified in the first step and the calculation of the corresponding doses absorbed by these individuals or the population.
– Risk characterization: It combines the results of the exposure and dose-response assessments for the chemical substances selected in the hazard identification to provide quantitative estimates of the adverse health effects associated with exposure to those chemical substances. This step should also include some discussion or quantitative assessment of the associated uncertainties.
Identifying whether a chemical substance is toxic involves evaluating whether it may lead to adverse health effects for the exposure periods considered. Both carcinogenic and non-carcinogenic risks must be considered. In addition, risks are typically evaluated for acute exposure (i.e., short-term exposure, generally ranging from one hour to one day) and chronic exposure (i.e., long-term exposure, generally ranging from a few months to a lifetime). There are also subacute and subchronic exposures, which correspond to intermediate duration periods; however, those types of exposures are not considered in standard air pollution studies. Several databases exist that provide detailed information that may be used for this hazard identification step. One may mention, for example, the Integrated Risk Information System (IRIS) of the U.S. EPA (www.epa.gov/iris) and the chemical database of the California Office of Environmental Health Hazard Assessment (OEHHA; https://oehha.ca.gov/chemicals). These databases also provide quantitative information useful for the dose-response assessment.
In the case of non-carcinogenic health effects, one typically assumes that a threshold level exists below which there is no adverse health effect. This threshold level is expressed as a reference concentration, RfC (μg m−3), for air pollutants present in the atmosphere (i.e., being inhaled) and as a reference dose, RfD (mg kg−1 day−1), for pollutants that have been transferred to other media (i.e., those being ingested or absorbed through the skin). The use of a threshold may in some cases be an approximation, because such a threshold may not exist or may not have been identified for some air pollutants. It is the case, for example, for fine particles: some sensitive individuals may develop some adverse health effects at very low concentrations.
In the case of carcinogenic effects, it is assumed that there is no threshold below which there would be no risk of cancer. Furthermore, it is assumed that the relationship between the effect (i.e., the probability of developing cancer) and the dose (or the exposure concentration in the case of inhalation) is linear. The slope of this relationship between the dose (or exposure concentration) and the cancer risk is called the unit risk factor, UR (μg−1 m3), in the case of inhalation (i.e., for a concentration-response relationship) and a cancer slope factor, CSF (mg−1 kg day), in the case of ingestion and dermal absorption (i.e., for a dose-response relationship). UR corresponds to the additional probability of getting cancer per additional unit concentration of the carcinogenic compound. Similarly, CSF corresponds to the additional probability of getting cancer per additional unit dose of the carcinogenic compound. Units of UR and CSF are the inverse of the units of concentration and dose, respectively.
The reason for using a concentration rather than a dose for inhalation is that there is a single medium to be considered for exposure (namely, the air). Therefore, a default dose/concentration relationship is assumed (typically, an inhalation rate of 20 m3 of air per day) to get the reference concentration and the calculation of the inhalation dose can then be skipped during the HRA. This simplifying approach cannot be used for ingestion because there are several exposure pathways to be considered (drinking water and milk, eating meat and vegetables, etc.), which are associated with different chemical concentrations and ingestion rates. Therefore, the complete dose calculation must be performed for ingestion. Dermal absorption is not a significant exposure pathway in the case of air pollutants.
The exposure assessment involves calculating (or measuring) the concentrations of the chemical substances in the air and, in the case of ingestion and dermal absorption, the dose via multimedia exposure. This step is conducted using the methods described in Chapter 6 to calculate the transport and dispersion of pollutants in the atmosphere. For multimedia exposure, the transfer of air pollutants to other media via dry and wet deposition must be calculated, as described in Chapter 11.
Next, the results of the dose-response assessment and exposure assessment are combined to perform the risk characterization.
The non-carcinogenic health risk is quantified by the hazard quotient, HQ, which is the ratio of the exposure concentration, C (or dose, DH), and the reference concentration, RfC (or reference dose, RfD):
(12.14) The carcinogenic health risk is calculated as an excess individual risk, which is the product of the exposure concentration (or dose) and unit risk factor (or cancer slope factor):
(12.15) The sum of the health risks due to exposure via different exposure pathways may be calculated. For non-carcinogenic health risks, the hazard index, HI, is calculated as the sum of the hazard quotients for inhalation, ingestion, and dermal absorption. Similarly, a cancer risk is calculated for all exposure pathways:
(12.16) where HQi refers to the hazard quotient for exposure pathway i and RH,i refers to the excess cancer risk for exposure pathway i.
Next, the total risk for exposure to multiple pollutants may be calculated as the sum of the risks of the individual pollutants: HIt for non-carcinogenic health risks and RH,t for carcinogenic risks:
(12.17) where the subscript j refers to individual pollutants and Nc is the total number of pollutants (chemicals) included in the HRA. If HIt < 1, there is no non-carcinogenic risk. If HIt > 1, the calculation may be performed again by separating the adverse health effects according to different target organs and values of HIt that are organ-specific are then obtained. If all those organ-specific HIt values are <1, then, one may consider that there is no non-carcinogenic risk.
The acceptable carcinogenic risk varies depending on the application. For cancer risk related to radioactivity, a value up to 10−4 is generally considered acceptable (i.e., a probability of 1 in 10,000 to get cancer during lifetime). For cancer risk due to chemical exposure, an upper value ranging from 10−6 to 10−5 is generally considered acceptable. For example, the U.S. EPA targets an objective of 10−6 for the excess cancer risk due to air pollution; i.e., air pollution should not increase the probability of getting cancer by more than 10−6 over lifetime (i.e., a probability of one in one million). For comparison, the risk for the whole U.S. population to develop cancer is on average 0.4 (American Cancer Society, 2019). However, the fraction of cancer-related deaths is smaller: 0.22 (Xu et al., 2018), because many cancer cases are cured.
The sum of the individual excess cancer risks over the whole population is called the population burden. It is generally supposed to be less than 1 for air pollution: i.e., less than one person in the population should develop cancer due to exposure to air pollution.
A health risk assessment for non-carcinogenic air pollutants may also be conducted by assuming that there is a continuous relationship between the cause and the effect instead of using a threshold reference concentration. For example, given some concentration levels for ozone, fine particles or nitrogen dioxide in an urban area, one may estimate the health risk attributable to that pollution (the etiologic fraction, see Section 12.1.3) in terms of morbidity or mortality. Quantitative dose-response relationships obtained from epidemiological studies (relative risk or odds ratio) are used. Then, one may include cases where no threshold has been observed. For example, in the case where the relative risk, RR, is expressed according to a Poisson regression (commonly used in environmental epidemiological studies, also called log-linear relationship), the incidence rate, IH1, for a given concentration, C1, is written as follows:
(12.18) where IH0 is the incidence rate for a zero concentration and βS is the slope of the concentration-risk relationship. For an epidemiological study where two incidence rates are, IH1 and IH2, which correspond to two different concentrations of the air pollutant, C1 and C2, the relative risk may be written as the ratio of the two incidence rates (RRΔCRRΔC = IH1 / IH2):
where ΔC is the pollutant concentration difference, (C1 – C2), corresponding to the relative risk of the two exposed populations. For example, a concentration difference of 10 μg m−3 may be used to express the values of RR (and OR) obtained from different epidemiological studies on a same scale. The etiologic fraction may be calculated to estimate the health risk due to a given air pollution level. For example, the number of deaths that would be avoided if the annual PM2.5 concentration, C, were reduced to the value recommended by the World Health Organization (WHO), i.e., CWHO = 10 μg m−3, is calculated as follows:
(12.20) Thus, it may be written as a function of the relative risk expressed for a concentration difference ΔC (Equation 12.19):
(12.21) As an example, a relative risk of 1.06 is used here for ΔC = 10 μg m−3 (see Figure 12.2; Hoek et al., 2013), which was used in a HRA performed by the French national public health agency, “Santé publique France” (Pascal et al., 2016). The annual PM2.5 concentration for the urban background is assumed to be 15 μg m−3, which is typical of an urban area such as Paris (Airparif, 2015). The fraction of deaths attributable to PM2.5 concentrations above the value recommended by WHO is calculated as follows:
(12.22) Given an annual death rate in the Paris region of 70,500, fine particle air pollution would correspond to about 2,000 annual premature deaths in the Paris region (a premature death is defined as a death occurring prior to the average life expectancy). Actually, PM2.5 concentrations vary spatially and a more detailed analysis would be needed. The study conducted by “Santé publique France” (Pascal et al., 2016) for the whole metropolitan France provides more detailed information; it estimates that 17,000 deaths would be avoided annually in France if PM2.5 concentrations were reduced to the concentration recommended by WHO. Note that the relative risk used in this study (RR = 1.06) has an associated uncertainty and that the 95 % confidence interval leads to a relative risk ranging between 1.04 and 1.08 (Hoek et al., 2013).
The calculation of a health risk involves a myriad of uncertainties resulting from the dose-response assessment and exposure assessment. Therefore, it is recommended, at the minimum, to discuss those uncertainties or, to the extent possible, to quantify those uncertainties. The U.S. National Research Council provided some recommendations to address uncertainties in an HRA (NRC, 1994, 2009). The most comprehensive approach consists of assigning probability distributions to the various sources of uncertainty and calculating the propagation of those uncertainties through the HRA to obtain a probability distribution of the resulting health risk. In addition, one may want to distinguish between epistemic uncertainties and aleatory uncertainties. Epistemic uncertainties are those that could be reduced with additional information. Aleatory uncertainties are those that cannot be reduced. One example of aleatory uncertainty is the meteorological variability, which affects the pollutant concentrations and, therefore, the exposure assessment. Generally, epistemic uncertainties are simply referred to as uncertainty and aleatory uncertainties are called variability. A comprehensive analysis of the uncertainties in an HRA may treat both uncertainty and variability jointly providing a single probability distribution of the resulting health risk. Alternatively, uncertainty and variability may be treated separately in a two-dimensional analysis, thereby providing more detailed information on the main sources of the overall uncertainty affecting the calculated health risk (e.g., Lohman et al., 2000).
Problems
Problem 12.1 Health effect regulations
An epidemiological study shows a correlation between high concentrations of a chemical compound and death rates of older people during a heat wave. On the other hand, toxicological studies do not show any health effect even at concentrations greater than those observed in the atmosphere. Should this chemical compound be regulated as an air pollutant?
Problem 12.2 Carcinogenic health risk assessment
Calculate the cancer risk due to a benzene concentration of 5 μg m−3, which is the European air quality standard. The benzene unit risk factor, UR, is 2.2 × 10−6 (μg m−3)−1.
Problem 12.3 Carcinogenic and non-carcinogenic health risk assessment
Calculate the non-carcinogenic and carcinogenic health risks due to an annual concentration of hexavalent chromium (Cr(VI)) of 0.003 μg m−3 in the air and a chronic dose by ingestion of Cr(VI) of 2 × 10−6 mg kg−1 day−1. For non-carcinogenic health risk, the chronic reference concentration for inhalation, RfC, is 8 × 10−6 mg m−3 and the chronic reference dose for ingestion, RfD, is 3 × 10−3 mg kg−1 day−1. The unit risk factor, UR, for carcinogenic risk via inhalation is 1.2 × 10−2 (μg m−3)−1.