Ultraviolet light inactivation of protozoa in drinking water: a Bayesian meta-analysis

To assess the dose of UV light needed to achieve specified levels of Giardia spp. cysts and Cryptosporidium spp. oocysts inactivation in drinking water, a Bayesian meta-analysis is used to analyze experimental data from several studies. Of the 20 studies identified by an extensive data collection effort, 14 (five reported experiments on Giardia and nine on Cryptosporidium) were selected for analysis based on a set of criteria. A substantial amount of the log inactivation data are reported as greater than a given inactivation level (i.e., censored data). The Bayesian hierarchical modeling approach used in this study not only properly addresses the common concerns in a meta-analysis but also provides a robust method for incorporating censored data. Different statistical models will result in different estimates of the UV doses needed to achieve a specific inactivation level. The Bayesian approach allows us to present the uncertainty in terms of risk, which is better suited for supporting US EPA in developing regulations.