Re: An ecological analysis of long-term exposure to PM_(2.5) and incidence of COVID-19 in Canadian health regions

摘要

We read with interest this paper on air pollution in Canada (Stieb et al., 2020) and its possible contribution to an increased incidence of COVID-19. In this ecological study, the authors used as the unit of observation very broad geographical regions defined administratively (referred to as "health regions"), which vary dramatically in size and constitution within and between Canadian provinces. They then made use of data on incident cases of COVID-19 in these 111 regions (73,390 cases until the end of May 2020). To these grouped counts, they juxtaposed, in a negative binomial statistical model, past satellite observations of fine particulate matter (PM_(2.5)) over a 17-year period (2000-2016) that had a spatial resolution of 1 × 1 km. These were then averaged to produce a summary exposure measure for each region. A number of ecological covariates were included in the model. They found that a 1 μg/m~3 increase in PM_(2.5) increased the incidence of COVID-19 by 7% that they indicated "was several fold larger per unit PM_(2.5) than hazard ratios typically observed in cohort studies of mortality". The authors indicated that due to the study design "the findings should be interpreted with caution", and they also discussed some limitations due to the "coarseness" of these regions as well as cross-level bias ("ecological fallacy"). Nevertheless, it is our view that the limitations of their data are so severe that they do not advance public health as their risk estimates are neither credible nor interpretable. The basis for this conclusion is our recent detailed methodological review of mortality studies of both SARS and COVID-19 that we published in October 2020 in Environmental Health Perspectives, where we concluded that all studies of associations between these infectious diseases and environmental factors are seriously biased (Villeneuve and Goldberg, 2020).