National Ambient Air Quality Objectives For Ground-Level Ozone- Summary - Science Assessment Document

10. Risk Characterization

The purpose of the Risk Characterization is to evaluate the weight of evidence presented in the Science Assessment Document to determine whether or not the findings support a causal association between ground-level ozone and the noted effects. Uncertainties in the data are discussed, and the research needed to fill the gaps in scientific understanding is described. Populations that are most at risk from exposure to ozone are identified. Once a causal association is established, the next step in the Risk Characterization is to provide an estimation of population risks at current (and future) ambient ozone concentrations. This analysis will be developed in an Addendum to the Science Assessment Document.

10.1 Materials

All materials for which literature results were available experienced deleterious impacts due to ozone. Cracking, fading or erosion occurs in elastomers, textiles, textile dyes, artists' dyes, paints, metals and stone. Although the information presented is consistent in identifying the qualitative nature of the impacts, quantification of the impacts is problematic for a number of reasons, most notably the difficulty in separating the effects of ozone from those of other pollutants and/or environmental factors. Also, different methodologies employed by different investigators have complicated the task of trying to derive appropriate quantitative relationships between ozone and different materials There is no dispute, however, that ozone does impact materials in an adverse manner. There are clear causal mechanisms to account for the interaction of ozone at the molecular level with organic materials, and for the synergistic impacts of ozone and other pollutants on inorganic materials.

For regulatory purposes, what is needed is a better characterization of concentration-response relationships. To date, there have only been a few studies that have provided this type of information. Clearly this task is complicated by the synergistic nature of the reactions of ozone both with other pollutants and other environmental variables. Given the lack of relevant information for defining effect levels or concentration-response relationships, no further quantitative assessment of the impacts on materials in the Canadian environment is possible at this time.

10.2 Vegetation

No other phytotoxic air pollutant has been studied as intensively as ozone. From the date of its first causal linkage with crop damage in the U.S. in 1958 to the present time, there have been literally thousands of concentration-response studies conducted throughout North America, Europe, Australia and numerous other countries. Based on a plethora of scientific evidence, it is now universally accepted that ozone is the most damaging of all air pollutants affecting vegetation, with many regions worldwide experiencing sufficiently high ozone levels to impair the growth and yield of sensitive plants. Certain agricultural crop species are continually observed to be more susceptible to ozone, regardless of where they are grown. Visible foliar injury has been observed in crops grown in British Columbia, Ontario, Québec and New Brunswick, providing empirical evidence that crops are being affected at ozone concentrations currently experienced in Canada. Studies of crop yield impacts in Canada are, however, extremely limited. Potential impacts are suggested by comparing observed ambient ozone levels to the LOAELs identified for crops and trees. Data from 1980-93, from non-urban sites across Canada and in bordering U.S. states, are used here to show the distribution of maximum 3 month SUM60s. The LOAEL range for crops is presented in the Figure as a reference point; the LOAEL range for trees is slightly less than that for crops (4,400 - 6,600 vs. 5,900 - 7,400 ppb-h respectively) (Figure 10 (SAD Figure 14.1). It is apparent from this graph that:

• Ontario and, to a lesser extent, Québec and New Brunswick are the provinces most severely impacted by levels of ozone in a range where significant crop and/or tree-growth reductions are possible.
  
• Potentially damaging levels of ozone have been experienced periodically in several other areas of Canada: Nova Scotia, Manitoba, Saskatchewan, Alberta and British Columbia.

Recently, significant attention has been directed to the study and assessment of ozone impacts on vegetation with the intent of developing air quality criteria (e.g. in the U.S., Europe and Canada). Based on these activities, the following conclusions can be made concerning the weight of evidence behind the current state of knowledge.

• The toxicological evidence for ozone impacts provides a plausible mechanism for effect, though there remain areas that are poorly understood (for example, the mechanism of response of stomatal conductance to the presence of ozone, and the physiological nature of resistance to ozone). However, there is no significant debate over the general toxicological mechanisms of ozone impact.
• The research community is comfortable with the use of a cumulative exposure index to relate ambient ozone air quality to yield or biomass losses in both agricultural crops and forest species, and to acute foliar injury in crops.
• The selection of the most appropriate cumulative exposure index (e.g., SUM60, AOT40 etc.) appears to be a decision, which requires a subjective interpretation of the available science, accommodation of existing uncertainties in the experimental databases, and awareness of the forum in which such indices will be applied. In this regard, there appears to be significant comfort among expert panels in North America with selection of the SUM60 index.

Overall, the best experimental databases currently, for developing statistical relationships between agricultural crop yield losses and ambient ozone concentrations, are the open top chamber (OTC) studies from the United States and Europe. The statistical relationships observed have often been weak as a result of the limited sample sizes (number of crops and cultivars tested) and opportunities for replication. However, the yield responses have been consistent with those expected from the toxicological understanding of ozone impacts. Given that these OTC experiments restrict the composition of the input air, it is possible to screen out other potentially confounding air pollutants. Therefore, it can be stated with reasonable confidence that the observed yield losses are due to ozone impacts.

Based upon the weight of evidence discussion, it is concluded that ozone is a probable and likely cause of agricultural crop and forest impacts ranging from acute foliar injury symptoms to chronic exposures resulting in yield and biomass losses. Further, there is sufficient information available to identify LOAELs, albeit LOAEL ranges rather than unique numbers are identified for both crops and trees. That the information base is sufficient for such a conclusion is the collective opinion of vegetation effects experts. Therefore, there is a risk to vegetation (including agricultural crops, forest species and horticultural species) at ambient ozone concentrations currently experienced in Canadian urban and rural areas. Quantification of that risk through characterization of exposure - response relationships for Canadian species is currently constrained by the lack of information on exposures in non-urban areas (i.e. rural ambient ozone data).

Recommendations for improving the scientific understanding of ozone impacts on vegetation have been summarized in sections 8.4 and 14.2.7 of the Science Assessment Document. Most limiting is the dearth of information on experimental exposures of Canadian agricultural and forest species, grown under Canadian field conditions, at realistic (i.e. near ambient) ozone levels. The analyses in this assessment are to a large extent based upon experimental data collected in the United States and Europe. While the theoretical understanding of ozone vegetation impacts and ambient characteristics of ozone in Canada support the use of this information, it is important that experimental work be carried out to assess exposure - response relationships for Canadian crops and forest species under Canadian climatic and pollutant conditions. In the meantime, it is recommended that when species specific response information is required, the individual LOAELs, presented in Tables 1 and 2 (SAD Tables 8.9 and 8.11) for agricultural crops and forest species respectively, be used. For a conservative estimate of the concentration of ozone above which effects on vegetation can be expected, the LOAEL ranges of 5900-7400 ppb-h (crops) and 4,400-6,600 ppb-h (trees) (3 month SUM60 values) identified in Section 6.2 above should be used.

10.3 Birds and Mammals

The effects of concern for both birds and mammals concern impacts on the respiratory system. Though the avian lung-air sac respiratory system may predispose birds to greater sensitivity there is insufficient information at present to make any predictions concerning relative sensitivities. Also, there was insufficient information in the literature to develop concentration-response relationships and very limited information on which to base effect levels. This precludes any quantitative analysis of the risks to birds and mammals across Canada from exposure to ozone at current ambient concentrations.

10.4 Human Health

Weight of evidence for adverse health effects.7 Weight of Evidence for Ozone as a cause of adverse respiratory health effects

There are several reasons for weighting the epidemiological studies more than the controlled human exposure studies or animal toxicological studies when evaluating levels of exposure that result in adverse health effects.

1. Epidemiological studies addressing the acute and chronic health effects in human populations involve those combinations of environmental conditions and activity levels present under real-world conditions of ozone exposure. This real-world relevance is an advantage over animal and controlled human exposure studies.
2. Urban populations are highly heterogeneous, including individuals who encompass a large range of susceptibilities, disease status, and exposures. Their responses cannot be predicted from classical animal toxicology or even controlled human exposure (clinical) studies. Population (ecological) studies using very large administrative databases are likely to capture a greater range of responsiveness, including those responses from the tail end of the distribution curve.
3. Population-based epidemiological studies are predominantly time-series studies, which are longitudinal and use a single population as its own control, and thus are less vulnerable to inter-population bias.
4. Since the purpose of such time-series epidemiological analyses is usually to help set ambient standards that will ultimately be monitored at central stations (fixed ambient monitors, FAM), the use of FAM data in the original epidemiological studies simplifies the standard-setting process (i.e., thereby avoiding any extrapolation between individual exposures measured as part of research activities and FAM concentrations employed in standards-compliance monitoring).

On the other hand, because of using FAM data, a significant drawback associated with ecological studies is the lack of knowledge about which individuals in the population are responding to a given ozone concentration, i.e., what were the ozone exposures of the individuals who were hospitalized, visited the emergency department, or died. Nevertheless, personal monitoring studies conducted in Canada and in the US have demonstrated that mean personal exposures to ozone have the same temporal trend as the FAM concentrations, suggesting that ozone data from FAM can adequately represent population exposure.

Clinical (controlled human exposure) studies provide valuable information about the threshold of a specific effect at ambient exposure concentrations of a single pollutant or of a mixture of pollutants, but not of the complex mixtures experienced in most locations. Clinical studies are valuable in providing quantitative information on the response to ozone in healthy individuals and in individuals with pre-existing respiratory disease, such as asthma and COPD. Data from controlled human exposure studies show that respiratory patients may form a prime target group for the adverse effects of ozone. Direct and conclusive information on susceptible groups cannot be obtained from the epidemiological studies based on population responses as captured in large administrative databases. Results on increased airway responsiveness, lung function changes, symptomology and airway inflammation after known doses of ozone have all been obtained from controlled human exposure studies, and provide direct evidence for the links between ozone exposure and health effects observed in the epidemiological studies. The disadvantage of the clinical studies reviewed in this document is that these studies used small sample sizes and short exposure duration, and when examining pulmonary compromised individuals, evaluated only those persons with mild conditions. The results, therefore, are not necessarily representative of the general population. No tissue injury has been tested for concentrations less than 80 ppb ozone for healthy subjects, or less than 120 ppb ozone for asthmatics. These drawbacks limit the use of clinical data for predicting the health effects in a general population and for obtaining a LOAEL or NOAEL.

Animal studies are valuable in elucidating cellular changes and mechanisms of action of ozone. They are particularly useful in studying possible effects of long-term exposures, because controlled human studies over a long time frame are impractical, and epidemiological studies on chronic endpoints are as yet too few to draw conclusions regarding chronic effects. The animal studies were the first to demonstrate the link between ozone exposure and immunotoxicity mediated through the detrimental effect on alveolar macrophages and lymphocytes, thus impairing defense mechanisms in the lung. Acute death was seen when animals were treated with ozone (400 ppb for 3 hours) and Streptococcus zooepidemicus bacteria. The findings that infection-related illnesses are associated with peak ozone exposure in the epidemiological studies are thus supported as biologically plausible.

There have been difficulties in extrapolating data from animal studies to humans. Comparative dosimetric studies have provided evidence that humans receive four to five times more ozone in their lower airways than rats do when given the same dose. Following deposition in the deep airways, animals and humans have been shown to have a similar tissue dose-effect relationship (lung injury), on a per unit body weight basis. On this basis, the doses used in rats are clearly relevant to the concentrations encountered by human populations at current ambient levels of ozone. So far there has only been a limited number of comparative dosimetric studies carried out, which precludes establishing a human LOAEL or NOAEL for ozone using animal toxicological data.

In summary, the controlled exposure/clinical studies and the animal toxicity studies provide a coherent picture of ozone-induced inflammation of the respiratory tract, triggering of hyper-responsive bronchi in asthmatics and others, and destruction of cells involved in the immune defence system of the lung. It is plausible that these responses initiate a cascade of effects progressing from reduced activity, absences from work/school and physician visits, to Emergency Department visits, hospitalizations, and even death as detected in the epidemiological studies.

On causality

Epidemiological studies do not themselves provide data to elucidate biological mechanisms that would explain the observed associations. Associations found in epidemiological studies between ozone and health effects may reflect chance, bias or cause. The criteria first described by Hill (1965) and modified by succeeding epidemiologists are used in this document to assist building a case for causality.

Strength of Association

The magnitudes of associations seen in all the epidemiological studies, although seemingly small, are statistically significant in many cases, and represent large numbers of people and important impacts on public health, since most of the population is exposed.

Consistency

An association of ozone pollution with population health effects was found by many investigators in cities across North America, in Central and South America, and in Europe, with these locations including a variety of pollutant mixes, ozone levels, weather, and socio-economic status of the populations.

Specificity

Many of the recent publications reviewed in the epidemiological sections have used some form of statistical technique to correct the cyclic impact of seasonal and weather factors on mortality, morbidity and ozone concentrations. By using regression analyses, researchers were able to differentiate the impacts of co-occurring pollutants on health endpoints. Data from the available studies demonstrate that the effects on mortality and morbidity ascribed to ozone are independent of the effects of the ambient co-pollutants.

It should be noted that some of the studies did not report the potential confounding effects of other pollutants on ozone, as these studies focused on the health risk of particulate matter. Several studies did not consider other pollutants, or considered only a limited number of co-pollutants.

Temporality

A logical temporal relationship exists, with ozone exposure followed by increased health effects. Negative lag times (health endpoints occurring before ozone changes) were investigated, and were found not to be associated with the respiratory conditions.

Concentration-response relationships

A concentration-response relationship of mortality and respiratory hospitalization was observed from very low ambient levels up to much higher concentrations in many of the studies. Dose-response relationships have been demonstrated in controlled human exposure studies for a number of spirometric variables as well as some symptoms. There have been no models established for airway inflammation.

Biological Plausibility

Ozone has been shown in animal experiments and in controlled human exposure studies to result in inflammation, epithelial cell necrosis, lowered lung function, increased airway reactivity, and increased animal mortality when the animals were subsequently challenged with a bacterial aerosol. Because of its highly reactive nature, ozone has been found to generate reactive oxygen intermediates which may cause cell membrane and macromolecule damage. In addition, asthmatics and individuals with 'hyperreactive' bronchi have been shown to be sensitive to the effects of ozone, with pain on deep inspiration, lowered lung function, increased need for medication, and asthma attacks. The experimental evidence supports the findings of the associations between ozone pollution and increased mortality and respiratory morbidity in epidemiological studies. The progression from impaired respiratory function and tissue injury to the point where medical attention is sought is quite plausible in light of the above.

Although the doses used for animals which produce pathological changes are higher than those seen in ambient air, comparative dosimetric studies have provided evidence that humans receive four to five times more ozone in their lower airways than rats do when given the same dose. Thus, it is reasonable that the doses used in rats are relevant to concentrations currently encountered by people. Furthermore, once ozone is delivered into lower airways, the relationship between the pulmonary tissue dose (normalized to body weight) and the pulmonary injury has been predicted to be in the same linear pattern among rats, guinea pigs, rabbits and humans.

Coherence

For a given increase in ozone concentration, the percentage increase in ED visits was larger than the percentage increase in hospitalizations, and the latter larger than that of total non-accidental mortality. This difference is expected on the basis that an element of choice is involved in the decision to seek medical attention at an ED or doctor's office, while hospitalizations represent only the most serious cases as determined by a physician. In addition, compared to ED visits, the percentage increases in doctors' visits and in days with reduced activity (absences from work or school) were also greater. On balance, data from epidemiological studies provide a coherent picture of an ozone-associated progression of health effects from high numbers of incidents recorded as ED visits, compared to those who were hospitalized and died.

Susceptible Populations

Clinical studies have identified patients with asthma, COPD and allergic rhinitis to be more sensitive to ozone-induced pulmonary function decrements and airway inflammation. These patients may constitute a sub-population susceptible to ozone pollution. These observations are consistent with findings in epidemiological studies indicating that hospitalizations and ED visits due to respiratory illness, especially asthma, increased significantly following elevated ozone pollution. However, the time series studies have not provided enough data to determine if the increased respiratory illness outcomes are due to an exacerbation of existing diseases or new incidences. A few studies on chronic effects have demonstrated ozone-related increases both in cumulative asthma incidence (new cases) and in asthma severity.

Another group at high risk compared to the general population is comprised of individuals whose activities lead to elevated ventilation rates. Cyclists, joggers, walkers, outdoor workers, and children would be included in this category. For example, the lunchtime joggers would be likely to have a high ventilation rate, comparable to the heavily exercising individuals in some of the chamber studies, and be exposed at a time of day when ozone levels are at or near their daily peak in most parts of the country. Consequently, these people are more at risk than sedentary individuals, since increased ventilation rates result in them receiving a larger ozone dose at the target tissue in the lung per unit of time, and a greater possibility of decreased FEV₁, more reporting of symptoms, and a tissue inflammatory injury.

Uncertainties

While time series studies have the advantage of being less biased by differences in indoor-outdoor concentrations within and between microenvironments, in life style, and in variability in daily time-activity patterns compared with cross-sectional studies due to the use of a single population as its own control, we do recognize that bias from exposure misclassification is a concern. This is because time series studies often rely on a single fixed ambient monitor (FAM) to characterize the pollution levels in a given community. The concentration of ozone measured at that FAM is used as a surrogate for personal/population exposure. The impact of this uncertainty is addressed in some studies by averaging the data from several FAM's on an hourly basis to better represent regional population exposure. In the province of Ontario, for example, much of the ozone is the result of broad regional air transport, and high correlations are observed between sites up to a hundred kilometres apart. Thus, studies using data from even a single FAM can provide a quantitatively sound assessment of the population impacts of ozone. Moreover, personal monitoring studies conducted in Canada and in the US have demonstrated that mean personal exposures to ozone have the same temporal trend as the FAM concentrations, which supports the use of FAM ozone data as a good indicator for population exposure.

One of the most difficult issues continues to be the role played by other pollutants (PM, SO₂, NO₂, and CO) in the health effects ascribed to ozone. Some of the available studies did not consider these co-pollutants. The fact that they may be highly correlated with ozone makes the separation of effects difficult in some instances. Nonetheless, the body of evidence amassed to date does justify the conclusion that the observed relationships can be attributed to ozone per se.

In some locations temperature was highly correlated with ozone (R > 0.5-0.6), making it a potential confounder since temperature is itself associated with increased respiratory distress. The method of handling temperature in the statistical analysis is therefore important, with inclusion in the regression appearing to provide the most reliable results. Removal of the temperature effect prior to running the regression with ozone carries with it the risk of removing part of the ozone effect in cases where correlation between these two factors is moderate or high.