Particulate matter and lung function growth in children: a 3-yr follow-up study in Austrian schoolchildren

To the Editor:

In the May 2002 issue of the European Respiratory Journal, Horak et al. 1 reported a study on lung function growth in children and air pollution. Their findings were interpreted as showing a significant adverse effect of particulate matter exposure on lung function growth. I was struck by the reported magnitude of the effect. For each 10 µg·m⁻³ of particles with a 50% cut-off aerodynamic diameter of 10 µm (PM₁₀) (i.e. about the difference between the most and least exposed community in the study), the effect on forced expiratory volume in one second (FEV₁) growth was estimated to be −84 mL·yr⁻¹ and for midexpiratory flow between 25 and 75% of forced vital capacity (MEF_25–75) no less than −329 mL·s⁻¹·yr⁻¹. This is approximately equivalent to 50% of the normal annual increase in FEV₁ in children in this age range, and it is equivalent to almost twice the normal annual increase in MEF_25–75 2, 3. In other words, the results, if true, suggest that the most highly exposed children are actually losing their MEF_25–75. Inspection of an earlier paper from the same study 4 shows even larger estimated effects of ozone on lung function growth. The effect on FEV₁ was given as −0.29 mL·day⁻¹·parts per billion (ppb)⁻¹, which translates, over the studied range in ozone concentrations of ∼20 ppb, to an effect of 20×365×−0.029=−212 mL·yr⁻¹. For mean maximal expiratory flow the effect was given as −0.076 mL·day⁻¹·ppb⁻¹, equivalent to 20×365×−0.076=−555 mL·s⁻¹·yr⁻¹. Such values are again approximately equivalent to, or even much larger than, the total expected yearly increase in these indices in growing children.

What could be the explanation for these counterintuitive statistical results? One possible explanation is that in their analysis of change in lung function, the authors adjusted for initial level of lung function which made them run into the eternal problem of regression to the mean 5. Since Oldham 5 (and others before him) raised this issue there has been no shortage of papers and admonitions in the literature warning of this problem 6–12. The problem is introduced in observational studies when a risk factor (in this case, air pollution) is associated with the initial level of the outcome variable of interest (in this case, lung function). Adjustment for initial level then leads to spurious associations between air pollution and lung function change. The presented material does not allow judgement as to whether such associations between initial lung function level and air pollution were present in the data. It is of interest that the authors note that there was no difference in lung function growth between places of residence, despite the clear differences in PM₁₀ (this paper) and ozone 4 that existed between the communities. Other work based largely on the same population has suggested graphically that lung function growth is also essentially the same between high, medium and low ozone communities 13.

PM₁₀ levels were clearly higher in all communities in winter than in summer, and one explanation for the findings could also be confounding by respiratory infections and pollen, which may have depressed lung function results in the spring measurements (which were related to the previous winter PM₁₀ levels) relative to the autumn lung function measurements. In the end, it is clearly important to know whether residence in communities with PM₁₀ levels that, even in the most “polluted” site were, by all existing standards, actually quite low, adversely affects lung function growth. I believe that a more straightforward analysis of the data would need to focus on the maximum period that was observed as the interest is much more in long-term impact than in variations by season. At the end of the discussion, the authors briefly present such an analysis, again adjusting for initial level of lung function, and suggesting similarly strong effects of PM₁₀, which seems contradictory to their statement that there were no differences in lung function growth between communities.

Simple graphical presentations, such as employed in the Children's Health Study from southern California 14, would probably provide a clearer insight. The California study, spanning a range in particles with a 50% cut-off aerodynamic diameter of 10 µm five times the range in the Austrian study, estimated that after 4 yrs of living in the most polluted community, forced expiratory volume in one second would be 3.4% lower than after 4 yrs in the least polluted community. It is my expectation that a more pertinent analysis of the Austrian data would remove most, if not all, of the suggested very large effects of air pollution on lung function growth.