A case for “natural experiments” in urban health

Posted on August 25, 2013


My previous post dealt with a study I wasn’t overly impressed with, so to speak. Studies on “dirty electricity” to date are just really not very good. Actually they are pretty much useless. Have a look at my discussion of a recent paper here.
So it is time to put a good study in the lime lights.
I really like a recent publication by Boogaard et al. published in the September 2013 issue of Epidemiology, entitled “Respiratory Effects of a Reduction in Outdoor Air Pollution Concentrations” (link to abstract and full text pdf). Yes, I know, it is a Dutch study…but this is just a coincidence. It could have been conducted anywhere…in fact similar studies have been conducted elsewhere.
What I like about this study is that this study demonstrates the impact of a reduction in urban air pollution in the same community, rather than using existing concentration-3055359796_2ffc0eeb53response functions from observational epidemiology. Essentially this can be compared to the well-established clinical trials in which some individuals are given a new drug and others are given a placebo (or better, the “normal”/standard drug) and biological markers are measured in both groups before and after the intervention. This comparison is not 100% perfect, since for example randomization of subjects over the interventions is not random and the “trial” is not well controlled, but nonetheless is provides an actual estimate of the impact of an intervention; in this case policy implementations that prohibited old heavy duty vehicles to enter low emission zones, while in one zone a traffic recirculation plan as also implemented.
The important difference with other epidemiological study designs is that this can provide estimates of the actual population health benefit of the policy and one can thus say: “if urban pollution concentrations are reduced by this much, this will be the population health benefit”.
The alternative approach is that the number of people with a certain disease (or certain markers of health (or disease)) are compared between groups of people with different exposure levels, and that one assumes that if the concentrations for a certain group of people with high exposure can be reduced to the lower concentrations measured in another group, then the pattern of disease (or markers of disease/health) will change and will mimic that of the lower exposed population. Or, as the authors state “…making implicit assumptions that causal effects will apply “in reverse” when exposure is reduced.” The problem with the latter is that this is not necessarily true, but in any way it cannot be proven.  This of course, can cause problems when trying to convince policy makers to spend a lot of money on policy interventions to reduce air pollution levels.
So anyway, good idea this study, but what was actually done and what did it find?
All residents of 12 locations in the Netherlands were invited to participate in a study in which air pollution and respiratory health were measured in 2008 (before) and again after new policies were implemented (2010). These 12 locations where air pollution (PM10, PM2.5, soot, NO2 and NOx) were measured included 8 busy urban streets in five cities that would be affected by the new policies and 4 suburban locations where the policies were very unlikely to have any impact (the control areas). Participants did not live more than 500 meters away from the measurement locations and during two periods following the measurement periods (2009 and 2011, respectively) their respiratory health (more specifically FVC, FEV1, MMEF and PEF) as well as exhaled NO as a marker of airway inflammation were measured. Ah yes, and of course they got a questionnaire on both occasions (its epidemiology, after all…).
3054524009_7ea7157f67Unfortunately, the response rate was low (10%) and of these 77% completed both examinations. However, this still resulted in a total of 661 participants of 4 years and older.
The measurements showed that implementation of the policies did have some effect on average concentrations of the pollutions, but that overall the effect was small (~10-25%); for example PM10 levels reduced from 26.6 to 23.0 ug/m3, PM2.5 from 16.0 to 11.7 ug/m3 and NOx from 69.6 to 63.2 ug/m3. Correspondingly, the effects on measures of respiratory health (FVC and airway resistance) as well as exhaled NO were also relatively small (~1%).
Interestingly, there was one location where a more drastic policy was implemented (traffic recirculation aimed at reducing concentrations at hotspots) and average reductions were higher for all pollutants but PM10 (again for PM10, PM2.5 and NOx, these were -4.3ug/m3, -7.6 ug/m3 and -39.5 ug/m3) than in the other seven locations (-3.1, -5.1 and -6.4 ug/m3, respectively) or the control locations (-4.6, -1.7 and 5.3 ug/m3). Correspondingly, FVC had improved by as much as 6% on average and FEV1 by 3.5%.
Based on their data, the authors concluded that small improvements in respiratory function were observed, but that these are likely to be clinically insignificant for healthy subjects. However, they further argued that small shifts in the population mean may results in a more substantial shift in the fraction of subjects with low lung function. In other words, the observed shift in average response will likely have an effect on susceptible individuals such as the elderly, asthmatics and children. Indeed, they refer to another study (link) that showed that a 3.5% decrement in mean FVC in children was accompanied by an odds ratio of 2.5 for FVC below 85% predicted.
As I said, I like this study. There are of course a couple of scientific “hums”, “buts” and “hmmms” (so have a closer look at the actual paper), but as a study design I think it is great. It is great to hear that policies to reduce urban pollution do have actual health benefits, rather than just theoretical, and that some of these effects are reversible. Importantly though, better interventions lead to more pronounced health benefits, so hopefully that will convince policy makers to spend a bit more time designing good and effective interventions.
Here in Manchester there are plans to do something to reduce air pollution concentrations, especially in the “Oxford road corridor” where the University is located and which is “cleverly” combined with the busiest bus route in Europe. With a bit of luck, people from the local city council involved in this read this blog….otherwise, I can always accidentally forget a couple of copies of the paper next time I visit city hall…