The impact of air pollution and weather on cardiovascular events: The importance of time scale and historical air quality improvement

Biondi-Zoccai et al.¹ describe in this issue the impact of environmental pollution and weather changes on the daily incidence of ST-elevation myocardial infarction (STEMI) based on retrospective analysis of data obtained between 2017 and 2019 in three high-volume centres with primary percutaneous coronary intervention services in Lazio, the second most populated region of Italy. The descriptive analysis of air pollutants in this paper indicates that the air quality parameters in Lazio in this recent period compare unfavourably with many other European countries and with the US, but are drastically better than in countries like India and China. The European Union has developed standards and objectives for a number of air pollutants, which also provides a framework to interpret the degree of recent air pollution in Lazio. The incidence rate of STEMI in this report by Biondi-Zoccai et al.¹ using the time series method was associated with several air pollution parameters and weather parameters in bivariate analysis.

Count-based data contain events that occur at a certain rate. Two approaches for the analysis of daily exposure and count-based data are time series analysis and case-crossover design.^2–4 Recently, a case-crossover analysis demonstrated a higher risk of out-of-hospital cardiac arrest associated with short-term exposure to air particulate matter (PM)₁₀ concentration in the Large Metropolitan Milan Area.⁵ It is critical to understand which conclusions can and cannot be drawn from these types of analyses.^1–5 The time series approach as applied by Biondi-Zoccai et al.¹ specifically investigates very short-term associations between environmental exposures and health outcomes. In time series regression, methods should be applied to control for seasonal patterns and for long-term trends in both the exposure and outcome data.⁶ By definition, a trend is a long-term increase or decrease in the level of the time series or, in other words, a systematic change that does not appear to be periodic. These systematic patterns over time are often present in time series, which may result in correlations between exposure and outcome that, generally, are unlikely to represent causal relationships.⁶ Moreover, many health, pollution, and weather parameters show seasonal variations over the course of the year, which can also occur even in the absence of a causal relationship.⁷

The focus of time series analysis is, therefore, on high-frequency fluctuations in the health outcome and in the exposure. Associations over these very short timescales are more likely to reflect true causal relationships.⁷ The effect of exposure on the event rate may be immediate or may be delayed by one or several days. Exploring too many lag relationships carries the risk of identifying non-causal relationships that have occurred by chance,⁷ as may have occurred in the analyses described in Table 1S in the report of Biondi-Zoccai et al.¹

The Poisson discrete probability distribution can be applied to model the number of times an event occurs in an interval of time (rate) provided that a series of assumptions is being met. The expected value (mean) for a Poisson distribution is λ. When λ is constant, stochastic variability around the expected count λ occurs and the probability that y events occur can be calculated based on the probability mass function corresponding to the specific value of λ. On the other hand, the Poisson regression model is applied for a non-constant λ. In Poisson regression, one assumes that the value of λ is influenced by a vector of explanatory variables. The classical Poisson regression model is a log-linear model whereby the logarithm of λ equals a linear combination of the explanatory variables of the model. The exponential of a particular regression coefficient is the estimated incidence rate ratio associated with the relevant variable. In general, these effects of short-term variability of air pollution on day-to-day variability in cardiovascular outcomes are small over realistic exposure ranges. In a meta-analysis of 34 studies, short-term PM_2.5 exposure increased the incidence rate for acute myocardial infarction by 2.5% per 10 mg/m³ increase (incidence rate ratio 1.025; 95% confidence interval 1.015 to 1.036).⁸

Nevertheless, even when the nature of the associations between air pollution/weather parameters and STEMI in the study by Biondi-Zoccai et al.¹ is causal, fluctuations in the daily incidence rate do not provide information on whether STEMI was advanced by days, weeks, months, or years. Mechanisms that may underlie the association between air pollution and ischemic cardiovascular diseases include endothelial barrier dysfunction and disruption, pro-inflammatory effects, prothrombotic pathways, oxidative stress, and autonomic balance favouring sympathetic tone.⁹ In relation to STEMI, air pollution could promote the formation of a pathological substrate for the event or could be the final trigger for the event or both. Therefore, a harvesting effect – in other words, a displacement by only a few days – could theoretically explain variations in the daily rate of STEMI events. The extent of mortality displacement by air pollution has been analysed in time series studies by Schwartz 20 years ago, by using time scales of increasing lengths.¹⁰ If deaths are displaced by a few days only, little association between weekly or monthly averages of air pollution and a multi-day average of daily deaths would be expected. If, however, death is advanced by substantially more days, the relation would remain, or even become stronger if additional damage occurred as a result of cumulative exposures beyond that caused by the daily fluctuations. In the study by Schwartz,¹⁰ the percentage increase in all deaths associated with a 10 µg/m³ increase in PM_2.5 rose significantly from 2.1% to 3.75% as the time window size increased from 0 to 60 days. A similar increase was observed for the percentage increase in ischemic heart disease deaths.¹⁰ Seasonality acts as an impediment to further increase the time-window size. Therefore, the impact of cumulative exposure of air pollution on ischemic heart disease and STEMI in particular, or on life expectancy, can only be estimated in large prospective cohort studies in which data of confounders are controlled for. A particular problem with long-term studies is exposure assessment. A meta-analysis of data obtained in 11 European cohorts concluded that a 5 μg/m³ increase in estimated annual mean PM_2.5 was associated with a 13% increased risk of coronary events (hazard ratio 1.13; 95% confidence interval 0.98 to 1.30), and a 10 μg/m³ increase in estimated annual mean PM₁₀ was associated with a 12% increased risk of coronary events (hazard ratio 1.12; 95% confidence interval 1.01 to 1.25), with no evidence of heterogeneity between cohorts. A study by the American Cancer Society in 2007 showed that each increase of 10 μg/m³ in the mean PM_2.5 concentration was associated with an increased relative risk ratio of 1.18 (95% confidence interval 1.14 to 1.23) for death from ischemic heart disease.¹¹ In a prospective cohort study of 65,893 postmenopausal women without previous cardiovascular disease in 36 US metropolitan areas from 1994 to 1998, with a median follow-up of six years, each increase of 10 μg/m3 of PM_2.5 was associated with an increase in cardiovascular events (hazard ratio 1.24; 95% confidence interval 1.09 to 1.41) and an increase in cardiovascular deaths (hazard ratio 1.76; 95% confidence interval 1.25 to 2.47).¹² The broad confidence intervals highlight the lack of precision of these estimates.

Taken together, the long-term impact of air pollution in prospective cohort studies appears to be greater than the short-term impact in time series studies. Nevertheless, the true impact at population level is uncertain. Estimations of the impact of air pollution in prospective cohort studies are highly dependent on the assumptions made in the modelling process.^13–15 The higher slope in prospective cohort studies may also result from uncontrolled confounding. In addition, air quality in both Europe and the US has improved considerably since these prospective studies were initiated and conducted, which implies that the contemporary exposure range is different compared to the range observed in these historical studies. Annual emissions of PM_2.5 in the UK have fallen by 78% between 1970 and 2018. PM_2.5 emissions in the European Union decreased by 28.5% between 2000 and 2017. The US Environmental Protection Agency reported that annual PM2.5 concentrations decreased by 39% between 2000 and 2018. Long ago, the Great Smog of London in 1952 was likely the worst and most infamous air-pollution event in the history of the UK. Since the mid-1960s, the age-adjusted mortality rates attributed to coronary heart disease levelled off at first, and have since markedly decreased. Although this decline is, in part, explained by significant advances in the management of traditional cardiovascular risk factors and in clinical cardiology, improvement in air quality since 1970 may also have contributed to this tremendous progress. The margin for further progress should be seen in light of these historical changes in both outcome and exposure.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References