Annex 8: The health effects of particulate air pollution

What is the evidence base for health effects?

In issues of public health, significance assessments are made of the scientific evidence by multidisciplinary groups of independent scientists who are regarded as having the necessary expertise to interpret the individual studies. In assessing the strength of evidence for health effects, account is taken of issues such as publication bias, likelihood that an observed association could be explained by chance, study group selection and measurement bias, confounding102, consistency of the studies’ results, and biological plausibility of the effects. All studies are evaluated according to the same criteria, and findings of similar strength and validity are weighted when drawing conclusions.

The many studies of the health effects of air pollution have been assessed by a number of expert independent groups who have published their reviews (e.g. WHO 2003, 2005, 2006; Brook et al 2004; US EPA 2004; Pope and Dockery 2006; UK Committee on the Medical Effects of Air Pollutants 2009). These reviews include epidemiological, controlled human exposure, animal and in vitro103 mechanistic studies as each of these approaches have strengths and weaknesses.

Epidemiological studies estimate the effects of real life PM exposure and include susceptible groups. However exposure is generally to a complex mixture, it is difficult to accurately estimate exposure and studies are limited by their ability to control for the range of other factors that may also affect health.  In controlled human exposure and animal studies, the exposure is to PM alone and well characterised. Human PM exposure studies provide the most direct evidence of a causal relationship between PM and health effects but are limited because they generally deal with short term mild, reversible effects, and a small number of individuals. Animal studies are limited by possible species differences, applicability of the various animal models of human disease and often having to extrapolate from the higher PM levels used to lower ambient concentrations.

Epidemiological studies are able to show whether an association exists between health effects and PM, but it is more difficult for the studies to establish that the relationship is causal. It is possible that a statistically significant relationship is in fact due to some unidentified factor that is correlated with PM concentrations. A relationship is more likely to be causal when epidemiological results are replicated by similar findings in different studies in many locations under different exposure conditions and the results are supported by laboratory studies. In the case of PM, numerous studies have found a similar magnitude of effect for mortality and have linked it with cardiopulmonary morbidity (illness), and there is increasing controlled human exposure and animal evidence regarding the health effects associated with exposure to PM.

Broadly, studies can be separated into those that have investigated the health effects of short-term (acute) or long-term (chronic) air pollution. Studies of the health effects of short-term exposure are more numerous and are generally investigated by time series analyses of population changes in health outcomes (e.g. mortality, exacerbation of symptoms, hospital admissions, and health-care visits) in relation to daily PM concentration variations. Findings of single-city studies have been strengthened by large multi-city studies e.g. the National Morbidity, Mortality, and Air Pollution Study in the United States (US), and the Air Pollution and Health: a European Approach (APHEA-2) study. Panel studies which follow small similar groups for a short period and record health outcomes, exposures and potential confounders for each individual at different times, and case crossover104 studies have added to the time-series studies.

Cohort studies assess health risk in relation to long-term average PM pollution usually by comparing people living in different locations that differ in the average chronic ambient concentrations and mixtures of air pollutants. An advantage of such studies is the availability of individual information, which allows adjustment for potential confounders such as smoking status and occupation.

The first large prospective cohort study that demonstrated heath effects from long-term air pollution exposure was the Harvard Six Cities study by Dockery et al (1993) which followed up about 8,500 adults for 11 to 14 years. This was complemented by the American Cancer Society (ACS) cohort. Subsequently there have been extensions to the ACS study and publication of further cohort studies in the US and elsewhere e.g. Canada, Netherlands, France.

Historically there were criticisms about the validity of the data from the first main studies, in particular whether the results could be explained by confounders, which were not considered in the original analyses, or the results were sensitive to particular analytical methods. In response, the US Health Effects Institute, an independent research agency, reanalysed both the time-series and cohort studies. The reanalyses confirmed the original results of the time-series studies and established the methodological validity of time-series methods (Samet et al 2000) and the original results of the cohort studies and established that the PM-mortality effect estimates were robust to alternative analytical methods and inclusion of additional potential confounders (Krewski et al 2000). Further analysis of the ACS cohort, which takes account of 18 years of follow up has recently occurred (Krewski et al 2009).

Since publication of the first studies in the early 1990s showing health effects at PM concentrations previously thought to pose no public health risk there has been a significant reduction in uncertainty about the observed findings.

What are the health effects?

The health effects, which occur following inhalation of PM are predominantly respiratory and cardiovascular but more recent evidence includes pregnancy-related outcomes such as low birth weight. Outcomes range from subclinical functional changes (e.g. reduced lung function) to symptoms, impaired activities (e.g. school or work absenteeism), doctors’ or emergency room visits, hospital admissions, and death (Table 1). The proportion of the population affected by less severe outcomes (eg, symptoms, reduced lung function) is much larger than that affected by more severe outcomes such as hospital admissions and death. It is usually the more susceptible groups who experience the more severe outcomes.

Although ambient PM exposure poses smaller relative risks105 for cardiovascular disease than risk factors such as obesity or tobacco smoking, the absolute number of people affected is very high because the entire population is exposed, and represents a large health burden.
Small shifts in the population average of physiological measures such as lung function also mean substantial changes in the number of people with clinical conditions.

Outcomes reported to be associated with PM exposure

  • Physiological changes e.g. ↓ lung function, ↑ heart rate, ↓ heart rate variability, blood coagulation factors, inflammatory mediators, blood vessel reactivity, blood pressure, blood vessel structure
  • Low birth weight
  • Infant (especially post-neonatal) mortality
  • Respiratory symptoms e.g. cough
  • Exacerbations of asthma, chronic bronchitis
  • School/work absenteeism
  • Respiratory mortality
  • Cardiovascular mortality
  • Myocardial infarction (heart attack)
  • Stroke
  • Cardiac arrhythmia (abnormal heart rhythm)
  • Lung cancer
  • Reduced lung growth
  • Reduced life expectancy
  • Respiratory and cardiovascular disease medication use, hospital admissions, emergency department visits, primary care visits

Sources: Kunzli and Tager 2005; WHO 2005, 2006; Pope and Dockery 2006; Heinrich and Slama 2007.
A threshold below which effects do not occur is not apparent in the studies. Hence the exposure-response relationship is regarded as essentially linear (i.e. increasing PM exposure is associated with increased frequency of effects).

Most time-series studies have estimated relative risks of mortality of less than 1% per 10 µg/m3 increase in PM levels over the preceding days. A similar finding was reported in Christchurch (Fisher et al 2007). A recent combined analysis of time-series data from 124 North American and European cities estimated a 0.2 to 0.6% increase in mortality for a 10 µg/m3 increase in 24 hour average PM10 concentration depending on the assumed lag period between exposure and death and the analytical methods used (Samoli et al 2008). Effect estimates appear to be similar in developed and developing countries.

In New Zealand long-term effects were studied using modelling of the spatial distribution of air pollution in Auckland and Christchurch. The annual average exposure in each census area unit within the urban areas was estimated and the association between these exposure estimates and annual mortality adjusted for potential confounders such as age and smoking was analysed. The annual mortality increase per 10 µg/m3 increase in annual PM10 was 6% in Auckland and 8% in Christchurch (Fisher et al 2007).

Cohort studies have shown that long-term exposure has a much greater effect on public health than short-term exposure with relative risks of mortality of around 4-10% for the same change in PM pollution on a long term basis.

The extended ACS cohort study is the largest and most extensively analysed study currently available and considered the best single source of effect estimates (COMEAP 2009).  It found average adjusted relative risks of mortality associated with a 10 µg/m3 increase in PM2.5 of 6% for all cause, 9% for cardiopulmonary and 14% for lung cancer (Pope et al 2002). The effect of PM10 on all cause mortality was weaker. Cohorts with smaller spatial scale analyses report higher relative risks. Reasons for variability in the results among the cohort studies may include variations in exposure assessment, composition and toxicity of the air pollution mixture, population exposure or underlying susceptibility within the population, different time periods of exposure, and different patterns of change in long term concentrations. The COMEAP (2009) concluded that to take account of uncertainty associated with the best estimate of the published relative risks for mortality for a 10 µg/m3 increase in PM2.5, values of 1% and 12% could be used in sensitivity analysis.

Who are most affected?

The frequency of occurrence of a health effect associated with PM exposure depends on the extent of exposure and factors that determine susceptibility. These include individual characteristics such as age, health status, time-activity patterns (e.g. time spent outside) and genetic makeup, and neighbourhood characteristics such as proximity to major roads.

Understanding of the impact of PM on children’s health is incomplete. However, children are particularly susceptible due to immature lungs, incomplete metabolic systems, immature defence mechanisms, high respiratory infection rates, and a higher respiration rate and therefore higher intake per unit of body weight, and activity patterns which can lead to higher exposure and higher doses reaching the lungs.

The WHO Regional Office for Europe Working Group (WHO 2005) concluded that evidence was sufficient to infer a causal relationship between PM and post-neonatal106 respiratory mortality; adverse effects on lung function development (reversible acute lung function deficits and chronically reduced lung growth rates and lower function levels); aggravation of asthma and increased prevalence and incidence of cough and bronchitis. Evidence was suggestive of a causal relationship between PM and low birth weight.

There is increasing evidence that supports the possibility that much of the morbidity and mortality related to air pollution in children occurs via interactions with respiratory infections, which are common among children. Asthma is also common among New Zealand children.

In short-term studies, the elderly and people with pre-existing respiratory and cardiovascular disease and Type 2 diabetes were found to be more susceptible to effects of PM on mortality and morbidity. Increases in mortality are greatest among the elderly. Whether increased age per se or the high prevalence of underlying cardiovascular disease and other risk factors explains the increased risk among the elderly is unclear. Asthmatics show increased symptoms, larger lung function changes and increased medication use.
Airway deposition models suggest those with pre-existing respiratory and cardiovascular disease receive higher doses of PM in their airways and lungs compared to healthy people.

Long term studies suggest groups with low socioeconomic or educational status experience increased mortality and morbidity due to air pollution. This may relate to limited access to health care, higher exposures, nutritional status and more risk factors for the health effects of PM. There is no convincing evidence that gender, ethnicity, and other pre-existing cardiovascular risk factors (e.g. smoking, high blood pressure) increase the risk.

How does particulate matter cause health effects?

Generally, larger (coarse) particulate matter (between 2.5 and 10 µm) deposits in the upper airways whereas smaller (fine) particulate matter (< 2.5 µm, PM2.5) deposits in the very small airways deep in the lung. Inhaled very small (ultrafine, < 0.1 µm) particulate matter may enter the blood circulation. Fine particulate matter is more hazardous than coarse particulate matter in terms of mortality and cardiovascular and respiratory outcomes.

Understanding of the biological mechanisms by which PM leads to health effects is incomplete. Several plausible mechanisms including inflammation, oxidative processes, and dysfunction of the autonomic nervous system have been described.

Experimental studies in animals and humans suggest PM may have adverse effects on blood vessels, the heart and respiratory system. Exacerbations of symptoms have been seen in those with pre-existing respiratory and cardiovascular disease following controlled inhalation of PM at concentrations above ambient levels. Studies suggest these effects are due to induction of lung inflammation, alterations in blood viscosity (thickness), oxygen deprivation, and disturbances in heart rhythm.

Is the increase in mortality only among frail people?

One of the questions directed at the PM-mortality findings of the initial daily time-series studies was that increases in PM may only be hastening the deaths of frail people by a few days and hence of limited public health significance. This short-term forward temporal shift in mortality rate is known as mortality displacement (or harvesting).

Under this hypothesis, the increase in mortality during the higher PM pollution days would be immediately followed by lower than expected mortality, which persists until the mortality level comes back to the expected level. If there is short-term displacement, then an association between PM and mortality would only be detected at shorter but not at longer timescales. Different statistical approaches suggest more than short-term displacement of mortality occurs and that longer lag periods up to about 40 days are associated with higher relative risks of cardiopulmonary mortality.

Higher relative risks of mortality are seen in the cohort studies than time-series studies. At current ambient PM levels in Europe, the effect of PM on life expectancy is estimated to be a reduction of one to two years.

What are the public health benefits of reduced particulate pollution?

Benefits of reduced air pollution have been demonstrated with findings of reduced mortality such as after the 1990 coal ban in Dublin, Ireland (Clancy et al 2002) and increased average life expectancy in US cities between 1980 and 2000 (Pope et al 2009).

Clancy et al (2002) reported a reduction in adjusted mortality rates of 5.7%, 10% and 15.5% of total, cardiovascular and respiratory deaths respectively in the three years subsequent to the coal ban.

Pope et al (2009) found increased average life expectancy of 0.61 (+/-0.2) year with a 10 µg/m3 decrease in fine PM concentration at county level in 50 US cities after adjusting for changes in socioeconomic, demographic and proxy smoking indicator patterns. Although the ability to control for potential confounders was limited as the study was population based, previous prospective cohort studies using measures of ambient PM and controlling for individual risk factors have indirectly estimated similar improvements in life expectancy (Krewski et al 2000).

Reduced age-related decline in lung function (Downs et al 2007) and lower rates of reporting respiratory symptoms (Schindler et al 2009) were found on follow up of the Swiss Study on Air Pollution and Lung Disease in Adults (SAPALDIA) cohort in 2002 after subtle decreases in PM10 since 1991.
During the California Children’s Health Study follow up, children who had moved to areas of lower PM10 showed increased growth in lung function whereas children who moved to more polluted communities showed decreased growth in lung function (Avol et al 2001).

102 Confounders are factors that are correlated both with exposure (PM pollution) and with the outcome (mortality). Confounders can increase or decrease the size of the effect estimate between PM pollution and mortality. Studies adjust for them but some residual confounding may remain as it is not possible to adjust for unidentified confounders, and adjustment may be incomplete for identified confounders.

103 In vitro studies are carried out on tissue removed from the body.

104 Case crossover studies – for each case, the distribution of exposures in the pre-event period are compared with the distribution of exposures estimated from a control time period.

105 Relative risk refers to the percentage change in health outcome per unit change in PM concentration.

106 Post-neonatal is the period of infancy from one to 12 months of age.