Getting in Shape for the World’s Leading Environmental Risk Factor

Air pollution is the single most important environmental risk factor for global mortality, with recent estimates of 8.9 million deaths annually attributable to ambient particulate matter air pollution (1). The fine particulate matter fraction (<2.5 μm in diameter, PM_2.5) from anthropogenic sources such as fossil fuel and power generation is by far the most important contributor, and the component with the most extensive links (2). Evidence from animal models, human panel studies, time-series analysis, and longitudinal cohort studies show a clear relationship between PM_2.5 exposure and cardiovascular (CV) events, including mortality. Multiple pathophysiologic pathways mediate the effects of PM_2.5 on the CV system, including inflammation, thrombosis, vasoconstriction, and insulin resistance (2). PM_2.5 levels have come down substantially over the last 2 decades in the Americas, with the majority of the United States at levels below legislated U.S. National Ambient Air Quality Standards (NAAQS), set at PM_2.5 <12 μg/m³. At these levels, the air quality in most parts of the United States and Canada is downright pristine, compared with annual averages of >50 μg/m³ experienced in parts of the Middle East, India, and China.

The majority of the studies on ambient PM_2.5 and health outcomes have been in environments where PM_2.5 levels have been well below the annual average of <35 μg/m³ (this value corresponds to Interim Target 1, the highest level of 4 sequentially lower targets, defined by the World Health Organization to goal-set and gauge progress). These studies form the basis of the integrated exposure response (IER), used in the Global Burden of Disease to evaluate the health burden of PM_2.5 (3). The IER interpolated PM_2.5 concentrations not only from ambient air pollution studies, but also from studies involving other sources of PM_2.5 (household air pollution, secondhand smoke, and tobacco smoking) into a single curve, from which risk estimates for PM_2.5 air pollution were obtained, across a range of PM_2.5 concentrations (Figure 1A). To do this, the IER assigned concentrations of PM_2.5 to each type of exposure on an equivalent μg/m³ basis, assuming that risk is determined by the 24-h PM_2.5 inhaled dose, regardless of the exposure source. An updated model (known as the global estimates of mortality model) (1), based on which recent estimates of 8.9 million deaths annually were derived, includes data exclusively from ambient air pollution cohorts (Figure 1B). However, this model is limited by the available exposure data up to 84 μg/m³, with an extrapolation for levels above this upper limit, effectively “shaping” the curve at these high levels (Figure 1B).

Figure 1. Exposure-Response Curve for PM_2.5.

Exposure-response curve for fine particulate matter fraction (PM_2.5) Exposure-response curve for fine particulate matter fraction (PM_2.5) for (A) ischemic heart disease mortality using integrated exposure-response (IER) model (adapted with permission from Burnett et al. [3]) and (B) all-cause mortality using the global estimates of mortality model (GEMM) (adapted with permission from Burnett et al.[1]). Black line shows estimates based on data, purple dashed line represents data extrapolation, red dotted line refers to the PM_2.5 levels studies by Liang et al., and the blue dotted lines represent 95% confidence interval of the relative risk estimates.

In this issue of the Journal, Liang et al. (4) fills in the gaps in available models at higher levels of exposure (>100 μg/m³). This study is an important addition to the growing amount of published data on CV health risk associated with air pollution. The investigators should be commended on utilizing a robust air pollution assessment method, a large prospective and geographically dispersed cohort, and a long duration of follow-up. This study is indeed among the few studies that have investigated long-term outcomes associated with air pollution and is the largest study to date from China. The study in particular expands our understanding of the shape of the concentration-response curve of PM_2.5 and CV outcomes. In their study, Liang et al. (4) linked 116,972 adults without CV disease with a modeled PM_2.5 exposure using residential address. The PM_2.5 concentrations were estimated in 1-km grids, utilizing satellite-derived data and machine learning models to estimate monthly PM_2.5. The investigators employed PM_2.5 as a time-varying covariate and Cox proportional hazard models to estimate the hazards of association between PM_2.5 and CV events. Mean exposure levels of PM_2.5 during the study period was 59.3 μg/m³ (interquartile range: 42.9 to 75.5 μg/m³), which is nearly 5× the NAAQS average annual limit of <12 μg/m³. The investigators report an association between ambient PM_2.5 and incident CV events (acute coronary syndrome and stroke) and CV mortality, after adjustment for traditional CV risk factors. The shape of this association was curvilinear with a steep increase in hazard for CV events and mortality at higher PM_2.5 levels. In contrast to the IER and global estimates of mortality model models, which suggest that there is flattening of the exposure-response curve at higher levels, Liang et al. (4) show that the relation between exposure and CV disease and mortality at higher levels may not flatten, but continue with a steeper relationship at PM_2.5 levels >60 μg/m³. The investigators estimate that every 10-μg/m³ decrement in PM_2.5 would result in >1.2 million CV events and >300,000 CV deaths prevented in mainland China alone.

Why is this important? These epidemiologic data are essential to estimate the true health impact of air pollution and corresponding measures of limiting emissions. It has been previously suggested that in countries with very high exposures (e.g., India and China), significant lowering of PM_2.5 exposures may be needed to see a measurable impact on health outcomes and mortality, given the hypothesized plateau effect (5). The current study emphasizes that policies and efforts to reduce PM_2.5 exposure in highly polluted densely populated areas may in fact be very effective in improving population health, in that even small decreases of 10 μg/m³ in annual exposure, could save as much as 300,000 lives annually. Comparatively speaking, the benefits of a mere 10-μg/m³ reduction in PM_2.5 would be identical to the benefit of 23 mg/dl (1 mmol/l) of low-density lipoprotein cholesterol reduction.

What does this study mean for clinicians and patients? We already know that PM_2.5 air pollution is a major preventable risk factor for CV disease and mortality, particularly in susceptible populations. This study highlights the importance of higher risk at high levels of exposure, such as those in India and China, where daily PM_2.5 concentrations frequently reach 500 to 1,000 μg/m³. Clinicians should take air pollution into consideration when evaluating risk for CV diseases with additional measures to incorporating air pollution metrics into CV risk assessment. Additionally, consideration should be given to personal interventions to mitigate harmful effects of air pollution (6). For example, high-efficiency home air filters and personal air cleaners are effective in lowering indoor exposures (6,7). Personal N-95 respirators, effective in reducing exposure to PM_2.5 may be recommended during periods of high pollution (6). However, the efficacy of such measures to reduce risk is currently limited to surrogate measures of CV risk with no evidence that these measures reduce CV events (2). We and others (8) have argued for large-scale clinical trials to test the hypothesis that vulnerable populations at risk for acute coronary syndrome may benefit from measures to mitigate personal exposure. Until these data become available, we may be reliant on the implicit assumption that lowering air pollution with personal intervention is likely of health benefit across the exposure spectrum. Given the continued robust association of PM_2.5 with overall mortality even at levels below the NAAQS, as suggested by recent studies and the fact that areas in the United States are starting to experience increases in air pollution due the policies of the current Trump administration, these findings may provide additional reasons to motivate studies of personal mitigation in vulnerable subsets and in “reshaping” policy (9–11).