Statistical methods for airway and alveolar nitric oxide in environmental epidemiology

Abstract

Background: The fractional concentration of exhaled nitric oxide (FeNO) has been used as a non-invasive biomarker of airway response to inhaled environmental exposures. FeNO measured at multiple flow rates can now be partitioned into proximal (airway) and distal (alveolar) sources using statistical models based on mathematical models of lower respiratory tract physiology. There is strong potential to improve our understanding of inflammatory mechanisms induced by inhaled exposures by studying effects on airway and alveolar NO, but current statistical methods may not be well-suited to epidemiologic studies. Aims: Evaluate the statistical properties of two methods to quantify exposure effects on airway and alveolar NO: (1) standard two-step approaches (estimate airway and alveolar NO separately for each person, and then relate airway and alveolar NO estimates to the exposure) and (2) unified non-linear mixed effects model approaches novel to this application. Methods: We simulated 500 sets of 500 multiple flow FeNO datasets similar to those collected in 1640 children participating in the southern California Children’s Health Study (2 measurements of FeNO each at flow rates 30, 50, 100, and 300 ml/s). FeNO was generated by adding random error to theoretical FeNO from a two-compartment model in which alveolar or airway NO was a function of a continuous exposure. Results: Two-step approaches were underpowered compared to mixed-effects approaches. In one simulation, the power to detect the exposure effect was: (a) 0.56 and 0.80 for alveolar NO, and (b) 0.72 and 0.81 for airway NO, respectively for the 2 methods. Both approaches had negligible bias. Conclusions: When studying airway and alveolar NO, a mixed-effects approach can considerably improve power to detect determinants, particularly for alveolar NO. Improved methods are highly relevant for studying compartmental airway response to ambient air pollution exposure, and will be implemented in the Children’s Health Study.