Benchmark Dose (BMD) approach is the way to calculate the risk at the lower dosage of chemical exposure, applying a mathematical model to the dose-response relationship. Although a number of the guidelines and software have been developed worldwide, the harmonization of BMD application is still undergoing. Before applying in actual risk assessment, it is essential to evaluate whether the BMD and NOAEL approaches give the same range of POD values and how different BMD values the major BMD software would give.
Here, we calculated the lower limit of BMD confidence interval (BMDL) from 201 tumorigenicity data publicized in the pesticide risk assessment reports by the Food Safety Commission of Japan (FSCJ). We applied three well-known BMD software, PROAST, BMDS, and BBMD, to compare their BMDLs to NOAELs and LOAELs and between the recently implemented methodologies such as model averaging (MA) or Bayesian inference.
Our result indicates that the BMD approach gives Point of Departure (POD) similar to the NOAEL approach if the data applied show a clear dose-response relationship. However, most of the datasets that resulted in failed calculation or extremely low BMDLs showed unclear dose-response relationships, such as non-monotonous and sporadic responses. We also noted that the Bayesian inference software gave failed calculation or extreme BMDLs less than the frequentist approaches.