The Epidemiological Pitfalls of Attribution in Environmental Health: Illustrations from Air Pollution

Environmental health per se is a misnomer, literally, it may mean health of the environment but as a subcategory of public health, it is used more to denote the effect of the environment on health. Most noises are made about the adverse impact of poor environment on human health with greater emphasis on air pollution. Through the process of rapid industrialization, urbanization, and extensive use of biomass fuel, we found a rapid decline in ambient environment in our living spaces, outdoors as well as indoors. Concerns were raised initially about the effect of poor environment on the occurrence of chronic respiratory illnesses being lead by chronic obstructive pulmonary diseases (COPDs) and acute exacerbation of bronchial asthma. Researchers started looking beyond cigarette smoking as a risk factor for COPD.[1] Research publications associating air pollution with chronic respiratory illnesses started crowding the journal spaces. As understanding of air pollution advanced, the effect was explored beyond the respiratory system and gradually it was suggested that more systems are as harmfully affected by the polluted air as the respiratory system. As of now, this association has covered illnesses such as cancer, metabolic diseases, adverse maternal and birth outcomes, developmental defects, cognitive impairment, and central nervous system diseases.[2] The scope is widening with ever-expanding list of diseases that find air pollution as a risk factor. This raises an epidemiological concern of attribution. The research paradigm has its limitations as most of the associations are reported based on cross-sectional studies that at best show an empirical association. Even if risk attribution is quantified using odds ratio as an effect measure, it predominantly remains confined to bivariate analysis and that is a major epidemiological hazard of attribution research. For diseases such as acute cardiovascular events, pregnancy outcome-related adverse effects, diabetes mellitus, and anemia, a list of large number of putative risk factors already exists, and adding air pollution to that list without adjusting for the known risk factors and confounders does not serve the purpose of understanding the etiopathogenesis of these illnesses. It further clouds the prevention, control, and treatment paradigm as eventually, we become clueless as to how much improvement will reduction in air pollution bring to reducing the burden of the diseases in question in given population sets without addressing other known risk factors. Cohort studies studying the effect of pollution are very few and difficult to conduct because of the slow and insidious pathogenesis of most of the listed diseases; and real-time personal exposure assessment is equally challenging because of the limited availability of methods of accurate personal exposure assessment.[3] Further, air pollution is a continuous exposure risk factor with immense variability as the population moves through different exposure zones in a 24-h cycle and seasonal variations in exposure are equally diverse. The inaccurate quantification of exposure to air pollution leads to uncertainty in health risk estimates.[4] Most of the environmental health researchers such as other specialist researchers also carry the tubular vision of establishing air pollution as the single biggest culprit to push the agenda of air pollution control. There is nothing wrong in it but such a linear approach clouds the scientific clarity of cause–effect relationship that is way more complex due to exposure to other risk factors, the presence of confounders, and practically nonexistent knowledge about the effect modification resulting from interaction of several known risk factors,[5] thus weakening the risk attribution. Another important dimension that should be looked into is the scarcity of data. We are aware that access to good quality data about the health status of the population in India in particular and developing countries in general is a huge challenge. In addition, even the air pollution data are available only at a very macrolevel as exposure monitoring is not yet granular enough to help us build an exposure matrix in diverse land use settings.[6] Our understanding of air pollution attribution to disease risk is limited by data analysis carried out in developed countries where the quality and quantity of data is far superior and accurate but exposure levels are much lower. Because the relationship between air pollution and mortality is steeper at lower levels of exposure, such as in the USA, linear extrapolations from these low PM2.5 concentrations to the higher concentrations that exist in India would overestimate its impact.[6] For example, linear extrapolation of risk estimation calculated with exposure to PM2.5 in the exposure range of 5–30 μg/m3 to the population exposed to 180 μg/m3–400 μg/m3 will unrealistically inflate the burden of air pollution-related diseases. The prediction models developed based on these data sets need to be subjected to sensitivity analysis on populations living in highly polluted regions of the world. As the major focus of air pollution research is focused on the concentration of PM2.5 in ambient air and most of the models have tried to calculate the increase in risk per 10 μg/m3 increase in PM2.5 concentration, it misses out on an equally important, if not more, dimension that is chemical speciation of the particulate matter. Whether it is a heavy metal with known toxicity or it is a corrosive salt or a polycyclic aromatic hydrocarbon, and their fractional composition, to name a few, will determine the role of the pollutant in the pathogenesis of the underlying disease. The etiological hypothesis is not only confounded by other risk factors but also by socioeconomic status. It is evidenced that populations with lower socioeconomic status are more likely to be exposed to higher levels of pollution than those with higher SES, some researchers have found it to be a confounder while others have proven it to be an effect modifier.[7] Thus, there are several loose ends and unanswered aspects of environmental factors that contribute to disease causation and risk attribution. This can be done only after mounting cohort studies specific to a given population, that measure the effect of chronic exposure to high concentrations of air pollutants.[8] Instead of hurrying into drawing conclusions about DALY losses and excess mortality attributed to air pollution, we need to tighten the loose ends and develop more robust epidemiological evidence-based attribution of environmental risk factors to the causation of myriad disorders and not fall prey to the mass media-created shock effect of air pollution on human health.