Review and analysis of personal-ambient ozone measurements

Ambient ozone measurements are often used as surrogates for personal exposures. Due to the limited number of central ozone monitors and varying personal behavioral patterns, some level of variability between ambient and personal exposures is expected. Low-cost sensors and different ways to capture personal activity patterns are being developed as an effort to improve the accuracy of exposure assessment. However, it is still most common to use the traditional approach of using unadjusted ambient concentrations as surrogates for personal exposures. To our knowledge, there has not been a meta-analysis that summarizes the findings from studies that investigated the differences between personal and ambient ozone. We conducted a literature search in PubMed and Science Direct for peer-reviewed studies reporting at least one of the following in a numeric format: 1) personal-ambient measurements, 2) personal-ambient slopes, or 3) personal-ambient correlations to identify and summarize existing studies that investigated personal and ambient ozone concentrations. Twenty-two articles met inclusion criteria and were included in our review. Ambient concentrations almost always overestimated personal exposures. A meta-analysis of slopes showed an overall personal-ambient slope of 0.21 (95% CI: 0.15, 0.27) with high heterogeneity (97%) across studies. The correlations between personal and ambient ozone varied dramatically across subjects from a strong positive (0.77) to a moderate negative correlation (-0.43). Our study found that ambient measurements are not accurate representations of personal exposure, while the magnitude of exposure measurement error varied across studies. Different sources of ozone and how they contribute to true exposure levels for individuals in complementary ways need to be better addressed. The effort to better understand the impact of traditional exposure assessment on the risk estimates must be emphasized along with efforts to improve the current exposure assessment approaches to provide context for interpreting the results from ozone epidemiological studies. Implications: The traditional approach of using ambient ozone measurements as surrogates for personal exposures is likely to result in exposure misclassification, which is a well-recognized source of bias in epidemiological studies. There are efforts to characterize the differences between ambient and personal ozone measurements, though, to our knowledge, there has not been a meta-analysis that summarizes the findings of different studies. Better understanding of the pattern and magnitude of exposure error for ambient and personal ozone can provide directions for future studies and context for interpreting the results from ozone epidemiological studies.