Meta-analysis of environmental health data.

Most airborne contaminants have minimal health effects at typical ambient levels, and this is especially true for contaminant levels in the Arctic. Among multiple studies of similar health effects, usually only one or two of the studies show statistically significant effects. Only after an attempt to quantitatively combine the data might some consistency emerge or might the sources of inconsistency be explored. Methods for combining the data from such studies are presented. The methods include combining P-values, combining effect measures using fixed effects models, and combining effect measures using random effects models. Examples of the methods applied to several pollutants, including lead and oxides of nitrogen, are given. Some discussion of the implications of these findings to contaminant levels found in the Arctic is presented.

Duke Scholars

Author Victor Hasselblad Biostatistics & Bioinformatics, Division of Biostatistics