Validity of using ambient concentrations as surrogate exposures at the individual level for fine particle and black carbon: A systematic review and meta-analysis.

Exposure measurement error is an important source of bias in epidemiological studies. We assessed the validity of employing ambient (outdoor) measurements as proxies of personal exposures at individual levels focusing on fine particles (PM) and black carbon (BC)/elemental carbon (EC) on a global scale. We conducted a systematic review and meta-analysis and searched databases (ISI Web of Science, Scopus, PubMed, Ovid MEDLINE®, Ovid Embase, and Ovid BIOSIS) to retrieve observational studies in English language published from 1 January 2006 until 5 May 2021. Correlation coefficients (r) between paired ambient (outdoor) concentration and personal exposure for PM or BC/EC were standardized as effect size. We used random-effects meta-analyses to pool the correlation coefficients and investigated the causes of heterogeneity and publication bias. Furthermore, we employed subgroup and meta-regression analyses to evaluate the modification of pooled estimates by potential mediators. This systematic review identified thirty-two observational studies involving 1744 subjects from ten countries, with 28 studies for PM and 11 studies for BC/EC. Personal PM exposure is more strongly correlated with ambient (outdoor) concentrations (0.63, 95% confidence interval [CI]: 0.57-0.68) than personal BC/EC exposure (0.49, 95% CI: 0.38-0.59), with significant differences in ṝ (0.14, 95% CI: 0.03-0.25; p < 0.05). The results demonstrated that the health status of participants was a significant modifier of pooled correlations. In addition, the personal to ambient (P/A) ratio for PM and average ambient BC/EC levels were potential effect moderators of the pooled ṝ. The funnel plots and Egger's regression test indicated inevident publication bias. The pooled estimates were robust through sensitivity analyses. The results support the growing consensus that the validity coefficient of proxy measures should be addressed when interpreting results from epidemiological studies to better understand how strong health outcomes are affected by different levels of PM and their components.