Does the Effect of PM(10) on Mortality Depend on PM Nickel and Vanadium Content? A Reanalysis of the NMMAPS Data

BACKGROUND: Lack of knowledge regarding particulate matter (PM) characteristics associated with toxicity is a crucial research gap. Short-term effects of PM can vary by location, possibly reflecting regional differences in mixtures. A report by Lippmann et al. [Lippmann et al., Environ Health Perspect 114:1662–1669 (2006)] analyzed mortality effect estimates from the National Morbidity, Mortality, and Air Pollution Study (NMMAPS) for 1987–1994. They found that average concentrations of nickel or vanadium in PM(2.5) (PM with aerodynamic diameter < 2.5 μm) positively modified the lag-1 day association between PM(10) and all-cause mortality. OBJECTIVE: We reestimated the relationship between county-specific lag-1 PM(10) (PM with aerodynamic diameter < 10 μm) effects on mortality and county-specific nickel or vanadium PM(2.5) average concentrations using 1987–2000 effect estimates. We explored whether such modification is sensitive to outliers. METHODS: We estimated long-term average county-level nickel and vanadium PM(2.5) concentrations for 2000–2005 for 72 U.S. counties representing 69 communities. We fitted Bayesian hierarchical regression models to investigate whether county-specific short-term effects of PM(10) on mortality are modified by long-term county-specific nickel or vanadium PM(2.5) concentrations. We conducted sensitivity analyses by excluding individual communities and considering log-transformed data. RESULTS: Our results were consistent with those of Lippmann et al. However, we found that when counties included in the NMMAPS New York community were excluded from the sensitivity analysis, the evidence of effect modification of nickel or vanadium on the short-term effects of PM(10) mortality was much weaker and no longer statistically significant. CONCLUSIONS: Our analysis does not contradict the hypothesis that nickel or vanadium may increase the risk of PM to human health, but it highlights the sensitivity of findings to particularly influential observations.