Our study demonstrated that air pollution is associated with morning PEF in patients with moderate asthma on well-defined asthma medication regimens. The significant, negative associations were observed in all the three asthma medication regimens. These results support our hypothesis that air pollutants adversely affect the health of asthmatics despite medication use.
To our knowledge, asthma clinical trial data have not been used with respect to air pollution, and this study is the first attempt to explore the interaction between air pollutants and PEF in patients with persistent asthma with well-defined asthma medication regimens from multiple asthma clinical centers. There are several strengths in this study. First, although there are 154 subjects in this study, the sample size provides 14,919 person-days of observations, which makes our study one of the largest pollution/asthma panel studies. Second, it was possible to make full use of (1) more specific and more precise medication information, (2) extensive and high-quality health outcome measures, and (3) precise documentation of asthma status. Lastly, successful identification of more homogeneous asthmatic subpopulations from the SOCS asthma clinical trial may also facilitate addressing issues such as asthma pathogenesis.
Several studies were conducted to evaluate if asthma medications were related to differences in response to air pollution. The findings from these studies are contradictory. For example, significant positive associations were found between air pollution and PEF in asthmatic children on respiratory medication (Roemer et al., 1999
). Hiltermann et al. (1998)
, however, reported that stratification by steroid use did not affect the magnitude of the observed associations between the prevalence of shortness of breath with NO2
, and black smoke. Contrary to this observation, Peters et al. (1997)
reported that medication use was not a confounder but attenuated the assocations between particulate air pollution and PEF in asthmatics. Furthermore, Delfino et al. (2002)
found enhancement of exposure-response relationships in asthmatics not taking anti-inflammatory medications. The disagreements already discussed are probably due to the major difficulties involved in the studies, which stem largely from the participants who are living in uncontrolled environments. A traditional epidemiological study is difficult to collect precise infomation about various asthma medications. Lack of precise asthma medication information may bring serious misclassification, thus biasing health effect estimation. It is also difficult to collect accurate asthma status. The difference in asthma severity status may significantly contribute to the disagreement of estimated air pollution health effects.
Close analyses suggest nonhomogeneity of health effects of different pollutants by different medication regimens. In other words, there is no clear evidence that the pollution effects on PEF are modified by these medications because pollutant-specific effects for each medication group are not consistently greater or lesser than those of the placebo group, across pollutants. This nonhomogeneity may indicate that an individual pollutant interacts with asthma medication regimens through a different major pathway/mechanism. However, Koutrakis et al. (2005)
suggested that ambient gaseous pollutants were not well correlated with personal measures of the same gas but were reasonably correlated with personal PM2.5
measures in Baltimore, MD, and Boston. Thus, the studied ambient gaseous pollutants might just be surrogates for particulates or particulates from different sources in this study. Unfortunately, this study is limited to addressing the study question, and further studies with improved exposure assessment are needed.
There are several limitations involved in this study. First, data are not available on a number of factors that are known or suspected to affect respiratory health, such as indoor air pollution sources. Nevertheless, these factors are not likely to vary with daily pollution exposures, and thus the estimated pollution effects are unlikely to be confounded significantly by these factors (Dockery & Brunekreef, 1996
; Pope, 1998
). Second, exposure misclassification is clearly applicable to this study because stationary air monitoring data were used as a surrogate of personal exposure. While there was only a surrogate of personal exposure, it was as close as possible to participants by using the monitoring station closest to their home ZIP codes. Given the 2-yr duration of the study, it was impossible to equip the whole panel with personal samplers, especially since multiple pollutants were studied. Third, the exposure unit is day, and the PEF was measured upon awakening in the morning, which would lead to exposure classification. Delfino et al. (2002)
showed that hourly pollution data captured adverse effects of exposure to ambient air pollutants better than 24-h averages. Unfortunately, hourly pollution data are not available for this study and further effort is needed in the future. Lastly, aeroallergens and viral respiratory infections are potential confounders. However, previous studies reported no confounding of outdoor fungi and pollen (Delfino et al., 1996
; Anderson et al., 1998
; Lewis et al., 2000
). In addition, it is possible, at least partially, to remove such confounding by temporal filtering of the regression model analyses, since the levels of aeroallergens generally follow seasonal and sub-seasonal weather patterns (Lewis et al., 1991
). As to the viral respiratory infections, published literature reported that controlling for respiratory infections did not substantially alter the associations between pollutants and health outcomes (Neukirch et al., 1998
). Findings from Peters et al. (1997)
further confirmed the “no confounding” findings just described. Even so, supplement analyses were conducted by restricting data analyses to the nonepidemic winter seasons, and no significant change of the estimated effects was found (results not shown).