Development of a model for robust and exploratory analysis of the rodent brief-access taste aversion data.

The rodent brief-access taste aversion (BATA) model is an efficient in vivo screening tool for taste assessment. A new E(max) (maximum effect attributable to the drug) model was developed and further investigated in comparison with three previously published models for analysing the rodent BATA data; the robustness of all the models was discussed. The rodent BATA data were obtained from a series of experiments conducted with a bitter reference compound, quinine hydrochloride dihydrate (QHD). A new E(max) model that could be applied to both "lick numbers" and "lick ratios" was built and three published models that used lick ratios were employed for analysing the BATA data. IC50, the concentration that inhibits 50% of the maximum lick numbers, quantified the oral aversiveness of QHD. One thousand bootstrap datasets were generated from the original data. All models were applied to estimate the confidence intervals of the IC50s without symmetric assumption. The IC50 value obtained from the new E(max) model was 0.0496 mM (95% CI 0.0297-0.0857) using the lick numbers for analysis, while an IC50 of 0.0502 mM (95% CI 0.0267-0.0859) was acquired with the lick ratios. Except one published model, the IC50 values have a similar range for the 95% CI. The new E(max) model enabled the analysis of both "lick numbers" and "lick ratios" whereas other models could only handle data presented as "lick ratios". IC50s obtained with these two types of datasets showed similarity among all models thereby justified the robustness of the new E(max) model.