Causal pathways of toxicity may depend on the duration of exposure
High energy physics experiments attempt to re-create conditions immediately after the big bang, by searching back in time for evidence of elusive short lived particles. The linked study by Bhaskaran and colleagues (doi:10.1136/bmj.d5531) reports findings from a case crossover analysis of very short term effects of air pollution on myocardial infarction in England and Wales. This state of the art analysis tests the epidemiological (as opposed to physical) limits of detection.1
The effects of air pollution on death and morbidity from cardiorespiratory disease on daily and longer time scales are now well established.2 The largest effects are seen after years or decades of chronic exposure, as shown by cohort studies.3 However, such studies are logistically demanding, requiring high quality data on spatial patterns of exposure and long follow-up of large populations or linkage of routinely collected datasets.4 For this reason, most epidemiological studies have analysed the acute effects of air pollution, using daily time series of exposure to air pollution and health outcomes available from routinely collected data. These studies have confirmed that even at the relatively low levels of exposure typically seen in high income countries, urban air pollution has subtle but measurable and consistent effects on mortality and morbidity in the subsequent days.5
Time series approaches to studying the effects of air pollution at daily time scales have important limitations. Until recently, technical and data limitations have prevented the extension of these methods to shorter (such as hourly) or longer (weekly) effects. Because the effects of short term exposure are typically subtle, large populations must be studied to achieve sufficient statistical power, particularly if the focus is on specific health outcomes. The potential confounding effects of factors that vary with time, such as ambient temperature, need to be allowed for. Estimates of exposure that are based on measurements at urban monitoring stations may not be representative of exposures experienced indoors, while travelling, or while exercising.
Urban air pollution consists of a complex (and often dynamically changing) mixture of pollutants; daily average concentrations of individual pollutants tend to be correlated, and it can be difficult to separate out the effects of individual pollutants. Particulate matter is thought to have important effects on health, but the impact of particle size and other characteristics on toxicity is unclear.6 7
Bhaskaran and colleagues tackle many of the limitations described above.1 They assembled data on nearly 80 000 diagnoses of myocardial infarction from 15 cities and assessed the effect of changes in exposure to air pollution from several hours to several days before onset. Exposure to air pollution was higher one to six hours before the event than in matched control periods. There was, however, no overall increase in risk associated with exposure over the three days before admission to hospital. As the authors state, this suggests that ischaemic events that would have occurred soon were merely brought forward by a few hours.
Epidemiological evidence of the effects of hourly exposure to air pollution on cardiovascular events is mixed. In a recent review, four of five studies reported significant associations between hourly exposure and myocardial infarction.8 However, only one of these studies analysed data over sufficiently long periods to exclude short term temporal displacement as the main explanation for the results.9 Several daily time series studies suggest that exposure to air pollution increases the risk of myocardial infarction over longer exposure periods (typically, the day of admission and the previous day).10 11 In those studies and in a multi-city study of mortality,12the effects were not primarily the result of short term displacement of health events.
How should Bhaskaran and colleagues’ results be interpreted? The prevalence or severity of cardiovascular disease might be lower in England and Wales than elsewhere. Alternatively, the estimates of exposure to air pollution might be less representative than in other settings. Either of these factors would reduce the power of the study to detect a true effect, although neither would explain the temporal pattern of the results. Despite the strengths of the study, it is possible that a true effect was missed because of imprecise measurements and inadequate statistical power.
In general, low levels of acute exposure to air pollutants can have minor effects in healthy people and trigger more serious events only in sensitive subpopulations. However, as shown by cohort studies, over several years even low levels of exposure to air pollution can reduce life expectancy substantially. An increase in the prevalence of chronic cardiorespiratory conditions is probably the most important mechanism for the reduction in life expectancy associated with exposure to air pollution over years or decades. Studies of very short term exposure may improve understanding of the early physiological mechanisms of the effects of air pollution, but they are not the best basis for developing policy on the control of air pollution. Given other evidence that exposure to air pollution increases overall mortality and morbidity, the case for stringent controls on pollutant levels remains strong.