Article dans une revue
Health risk assessment with multiple reference indices
Abstract : Background Conducting a risk assessment can be challenging, especially when dealing with several reference indices, which could lead to conflicting conclusions between studies. The common approach is to use a reference index from a single source based on the risk assessor's preference. Objectives To propose an approach for constructing a multi-reference index-based aggregated risk estimate using mathematical objectivity to reflect all of the available information. Methods The aggregated risk estimate based on multiple reference indices (AREMRI) results from the weighted linear combination of risk distributions that were obtained with each reference index available. The weights were calculated using the degree of agreement among the reference index-based risk distributions. The approach is illustrated through three practical cases of benzene inhalation cancer risk assessment using inhalation unit risks (IURs) from six different regulatory agencies. Results The degrees of agreement between the reference index-based risk distribution, obtained with the six IURs, ranged from 0.7 to 92%. The highest weights were attributed to reference index-based risk distributions that had the highest degree of agreement with the maximum number of other reference index-based risk distributions. Regardless of the practical case considered, the AREMRI risk distribution resulted in the third highest risk compared to the six single risk distributions. Conclusion Our approach can be useful in the presence of several reference indices by providing useful insights, consistency and direct comparisons between studies to support better-informed risk assessment and management decisions. This approach can shed some light on some of the uncertainties associated with the toxicological reference values in a risk assessment associated with the toxicological reference values. If the uncertainty is large, more detailed evaluation of the toxicological reference values would be needed.