I want to preface this blog post by saying I am not an infectious disease epidemiologist, and although I have four years of experience in epidemiology research, most of that has been focused on chronic disease epidemiology.
This has been an overwhelming and saddening time for all of us. As an epidemiology student, I’ve had some friends and family members ask me about my personal and professional thoughts on COVID-19, and I had to be completely honest – everything I know about infectious disease modelling, calculating the mortality rate for a novel virus, and when we might start flattening the curve, I’ve learned from online sources in the past month. I’ve refrained from writing about the pandemic because the world didn’t need another “expert” adding to the noise online (see some real expert sources I’ve been following to stay up-to-date below). Truth be told, I am in the same information boat as many of us. But, this week I came across some interesting new science presenting the relationship between air pollution and COVID-19, and as someone studying the health effects of air pollution for my PhD thesis, that is a topic I can speak about:
A research group at Harvard University recently published a pre-print (meaning the work has not yet been verified by the peer-review process) on the association between air pollution and COVID-19 (1). In a perfect world, this kind of work would have taken months, if not years to put together, but obviously, this was not possible during the current pandemic. However, that doesn’t mean that I can’t provide a fair critique of the work based on what we do know:
What Were They Studying?
The research group was studying whether long-term (2000-2016) average exposure to a specific air pollutant, PM2.5 (one of the most commonly studied air pollutants) was associated with COVID-19 deaths across 1,783 counties in the US. The researchers hypothesized that higher air pollution exposure leads to increased COVID-19 death rates in counties. We know that constant exposure to air pollution can damage lungs overtime and lead to many respiratory diseases such as asthma, COPD, hypertension, diabetes, and other chronic diseases that have been closely linked to increased risk of death from COVID-19 infection.
How Did They Study It?
Researchers collected satellite data on PM2.5 levels and COVID-19 mortality data across 3,000 counties in the US. After excluding counties for which sufficient data was not available, the analysis was done on 1,783 counties. It is important to note that the study units were counties and not people (see my previous post to learn more about ecologic studies and what they can and can’t teach us).
What Did They Find?
Researchers found that for every 1-unit increase in PM2.5 levels, the COVID-19 mortality rate increased 15% (95% confidence interval: 5%-25%). If true, this is a huge effect size! For comparison, a similar analysis found that a per-unit increase in PM2.5 leads to a 0.73% increase in all-cause mortality in the US (1).
Were There Any Issues with Their Analysis and Interpretation?
Yes. Foremost, I think we need to be cautious in our interpretation of any scientific work that has not been peer-reviewed. However, these are extraordinary circumstances and it is clear why the authors may have thought the work needed to be publicly available before formal peer review.
The authors stressed that air pollution is known to lead to the same pre-existing conditions that can increase risk of COVID-19 mortality (heart and lung disease, heart attacks, asthma, lung inflammation, coughing, difficulty breathing). So, it is likely that the relationship between air pollution and COVID-19 mortality in counties could be explained by the fact that air pollution leads to these same pre-existing conditions. But, a more important scientific question would have been to study the effect of air pollution exposure on COVID-19 mortality that isn’t due to air pollution’s relationship with pre-existing conditions. This question is a lot tougher to answer, and is unlikely to be something that could be assessed before the pandemic is over as it would require individual - not county - level data, and a more complex study design.
The authors also did not control for social distancing measures in specific counties (and how well they might be working), as well as the size of the outbreak in different counties (due to unreliable testing), both of which could have influenced their results.
There are also some general issues with interpreting data from ecologic studies that I’ve discussed in a previous post. We might be in danger of committing the ecologic fallacy, which occurs when we use ecologic data (such as from counties) to make inferences about exposure-outcome relationships in people.
Another point to keep in mind has to do with the type of air pollutant that was studied. PM2.5 is a catch-all term for any particles in air that are smaller than 2.5 micrometers. This includes dust, organics, carbon, industry sources, and some emissions from vehicles. Importantly, this means that the composition of PM2.5 can be very different depending on where you live (2,3). If you live in a region with many diesel cars and where everyone uses wood stoves for heating, the composition of the PM2.5 you are exposed to is going to look different than from a place where everyone drives electric cars and uses electric heating sources. Therefore, even if there is a strong relationship between air pollution and COVID-19 mortality in the US, this does not mean that this relationship will be the same where you live. There are examples in the air pollution literature of different health effects of PM2.5 based on region due to the composition of PM2.5 (4,5).
Final Thoughts
The authors of this work did a good job of providing timely research on the impact that air pollution exposures could be having on COVID-19 mortality in the US. They’ve also made their data and code available for the public to use (rare in epidemiological research)!
On March 26th, 2020, while COVID-19 headlines dominated the news, the Environmental Protection Agency quietly announced a major relaxation of environmental pollution rules due to the COVID-19 pandemic. This news was both worrying and disappointing. I will not pretend to know a lot about American politics and policy, but it is possible that the industry will continue polluting based on the relaxed guidelines even once the pandemic is over. This work on air pollution and COVID-19 highlights why it is so important to consider the health effects of environmental pollution, even during a global pandemic.
Overall, I am impressed by the work produced by the researchers under such a time constraint. However, I think we still need to be cautious about interpreting scientific work before it has been reviewed. Thorough science takes time.
There are some gaps in the research produced, but, the authors did make the effort of providing the data and code for others to reproduce, and hopefully improve on their work. I also worry about the quick uptake of this work by major news outlets such as The New York Times and National Geographic before the peer-review could be completed.
I look forward to seeing if any edits are made on the paper after the peer-review process, and how this work will be used by other epidemiologists to build on our knowledge of the impact air pollution can have on our COVID-19 burden.
References:
1. Wu, X., Nethery, R. C., Sabath B. M., Braun. D., Dominici F. Exposure to air pollution and COVID-19 mortality in the United States. medRxiv 2020.04.05.20054502; doi: https://doi.org/10.1101/2020.04.05.20054502
2. Bell, M. L., Dominici, F., Ebisu, K., Zeger, S. L., & Samet, J. M. (2007). Spatial and temporal variation in PM2. 5 chemical composition in the United States for health effects studies. Environmental health perspectives, 115(7), 989-995.
3. Franklin, M., Koutrakis, P., & Schwartz, J. (2008). The role of particle composition on the association between PM2. 5 and mortality. Epidemiology (Cambridge, Mass.), 19(5), 680.
4. Atkinson, R. W., Kang, S., Anderson, H. R., Mills, I. C., & Walton, H. A. (2014). Epidemiological time series studies of PM2. 5 and daily mortality and hospital admissions: a systematic review and meta-analysis. Thorax, 69(7), 660-665.
5. Hamra, G. B., Guha, N., Cohen, A., Laden, F., Raaschou-Nielsen, O., Samet, J. M., ... & Loomis, D. (2014). Outdoor particulate matter exposure and lung cancer: a systematic review and meta-analysis. Environmental health perspectives.
My top COVID-19 News Sources*:
The New York Times Coronavirus Deaths Chart (presented on a log scale): https://www.nytimes.com/interactive/2020/03/21/upshot/coronavirus-deaths-by-country.html
Plotting the number of deaths viruses the number of new cases is a less biased indicator of the burden of COVID-19 because testing rates have been very different around the world. Also, plotting the deaths on a log-scale allows me to easily see the regions of the world where the death rate is slowing down.
Epidemiological Summary of COVID-19 Cases in Canada:
Figure 1 is useful for visualizing the epidemic curve in Canada. The X axis shows the date of illness onset instead of the date reported positive, which is useful to see due to COVID-19’s long incubation period.
Canada's COVID-19 Info App:
Coronavirus: Why you must act now (blog post by Tomas Pueyo):
This is not a source that is constantly updated, and is maybe a bit out-of-date as many governments have taken promising actions to fight COVID-19. But, this blog post was shared and endorsed by many epidemiologists, and I found the presentation of information to be extremely helpful in understanding why we need to flatten the curve, and the implications of not doing so.
Chief Public Health Officer of Canada’s (Dr. Teresa Tam) Twitter Account: https://twitter.com/CPHO_Canada
The Globe and Mail’s Coronavirus Coverage:
*These are my personal preferences. I use these sources because they are regularly updated, and I feel like I get the right amount of information daily without overwhelming myself.