Dose-response modeling : evaluation, application, and development of procedures for benchmark dose analysis in health risk assessment of chemical substances

Abstract

In this thesis, dose-response modeling and procedures for benchmark dose (BMD) analysis in health risk assessment of chemical substances have been investigated.

The BMD method has been proposed as an alternative to the NOAEL (no-observedadverse- effect-level) approach in health risk assessment of non-genotoxic agents. According to the BMD concept, a dose-response model is fitted to data and the BMD is defined as the dose causing a predetermined change in response. A lower statistical confidence limit on the BMD (the BMDL) has been suggested as the point of departure in the determination of guidance values such as acceptable daily intakes (ADIs). The fact that the BMD corresponds to an explicit response level has been argued as a major advantage since this introduces consistency and suggests that the point of departure for risk assessment will be based on more information. However, this feature also represents one of the challenges associated with the BMD approach, i.e. which level of response should the BMD correspond to? This and related questions have been addressed in the present thesis.

Given the original definition of the BMD as the dose causing a 1 10% increase in the risk for adverse heath effects compared to background, this study has analyzed the impact of model choice in BMD calculations. In the case of quantal data, it is suggested that the BMD is defined as corresponding to risk levels in the range of 5 - 10%.

The introduction of the BMD method for continuous endpoints has merited more discussion, and the present thesis has mainly focused on these issues. A probability based procedure suggested for continuous data has been analyzed in detail. By the definition of a cutoff value denoting adverse response, this approach allows the BMD to be interpreted as corresponding to some risk level, similar to the case for quantal data. While this may be of interest, it was shown that the choice of determination of the cut-off point is of high importance and dictates how the BMD depends on the variance.

In the present thesis, the BMD approach was introduced for neurobehavioral endpoints using 2,2 ,4,4 ,5-pentabromodiphenyl ether (PBDE99) as a model substance. According to proposed methods of BMD analysis for continuous data, a BMDL of about 0.5 mg/kg bw was obtained for PBDE99. In another application, statistical differences in sensitivity between dioxin sensitive Long-Evans (L-E) and dioxin resistant Han/Wistar (H/W) rats following longterm exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin were demonstrated for the first time. Differences between L-E and H/W rats were most pronounced for volume fraction of hepatic foci; L-E rats were approximately 80 times more sensitive than H/W rats. Considering data on body and organ weights, L-E rats were 10-20 times more sensitive than H/W rats. For retinoid parameters, and hepatic CYP1A1 induction, differences between the strains were generally about 5-fold, and associated with a low uncertainty.

Besides analysis and application of suggested procedures for BMD analysis of continuous endpoints, developments have also been proposed. For dose-response relationships that are Sshaped, the BMD may be defined as the dose where the slope of the curve changes the most in the low dose region. It is discussed whether this definition may provide a biological basis for selection of the response associated with the BMD. Considering a conservative scenario, it was shown that the dose where the slope changes the most corresponds to a response in the range of 5 - 10%, if defined as a percentage change in response relative to the magnitude of response. This definition of the BMD may be considered for continuous data in future applications.