Statistical models that allow for estimation of health effects of air pollution are well developed when only one pollutant is considered. Similarly, co-pollutant models have often been examined, which typically consider two pollutant variables from ambient sources. In this proposed work, exposures to two air pollutants from differences sources (one ambient, one mainly non-ambient) will be modeled. Specifically, these pollutants are ambient particulate matter less than 2.5 microns in diameter (PM2.5), and environmental tobacco smoke (ETS). Individually, exposures to both pollutants have been shown to have negative impacts on health. However, interactive effects can only by examined by including these variables in the same model. Preliminary studies have suggested that for albuterol medication use, there is a significant negative interaction between the two pollutants, meaning that effect of two pollutants together is not as strong as the sum of the individual effects (i.e., the effect of exposure to only ambient PM2.5 plus the effect of exposure to ETS only). Before certain plausible explanations about why this occurs are developed, the model needs to be refined and results verified. In particular, the models need to more properly account for the longitudinal (repeated measures) component of the data, with verification by repeating analysis for each of three years of collected data. Since direct measures of pollutant exposures are difficult or impossible to obtain, other surrogate measures such as cotinine or fixed outdoor monitor concentrations need to be utilized. Using these, plus more unbiased measures of exposures from selected samples, regression calibration can be applied to yield estimates that are in similar and meaningful units of measurement. Regression calibration with two variables, including interaction, has only been recently proposed, and thus carrying out the actual analysis will require development of relevant computer programs to carry out the analysis. This program construction and data analysis will be the scope of the first aim. The remaining aims involve nonparametric regression with two variables for data that involve repeated measures, and monotone smoothing of the resulting mean function. These will involve some methods development (for the monotone estimation), and development of computer programs to carry out both. Collectively, the models will allow for more precise estimation of health effects of ETS and ambient PM2.5 exposure. Although certain limitations exist, such as establishing definitive causal effects, these analyses should be a first step in understanding how outdoor air pollution affects health of children, when ETS is examined as an effect modifier (both short term and long term).
The interactive acute health effects of environmental tobacco smoke (ETS) and ambient fine particulate matter (PM2.5) exposure on health of children with asthma will be analyzed. The results will allow for a better understanding about how children with lower or higher ETS exposure react to PM2.5 from outdoor sources.
| Strand, Matthew; Sillau, Stefan; Grunwald, Gary K et al. (2014) Regression calibration for models with two predictor variables measured with error and their interaction, using instrumental variables and longitudinal data. Stat Med 33:470-87 |
| Rabinovitch, Nathan; Silveira, Lori; Gelfand, Erwin W et al. (2011) The response of children with asthma to ambient particulate is modified by tobacco smoke exposure. Am J Respir Crit Care Med 184:1350-7 |
