Probabilistic estimation of the dietary exposure to Cadmium in Japan in 2019-2021 using two-dimensional Monte Carlo simulation after Bayesian handling of left-censored data

Abstract

Due to significant environmental cadmium (Cd) pollution incidents in Japan, probabilistic estimation of dietary Cd exposure is necessary for more accurate risk assessment. Although the total-diet (TD) method is effective for evaluating dietary exposure, "non-detected" values increase uncertainty. To better estimate dietary Cd exposure distribution in the average Japanese population, we applied two-dimensional Monte Carlo simulation (2D-MCS) with Bayesian estimation (BE) to TD samples. A total of 140 representative TD samples from 14 food groups across 10 regions were collected over three years, and Cd concentrations were measured using ICP-MS. BE was used to estimate Cd concentration distributions considering left-censoring and food consumption distributions were modeled using zero-inflated models. We conducted 2D-MCS by generating 50 random numbers from 2000 Monte Carlo samples obtained via BE. Estimation uncertainty increased with higher non-detection rates. Mean dietary Cd exposure was estimated at 0.351 ± 0.028 μg/kg/day. The highest exposure contribution (30.2 ± 18.3 %) came from FG1 (rice and rice products). The hazard quotient, compared to the 7.0 μg/kg/week guideline derived by the Food Safety Commission of Japan, was 0.351, with a 0.6 % probability of exceeding 1. This method effectively handles left-censored data and could serve as a standard for exposure assessment.