摘要目的 为提高评估精度并与国际食品安全风险评估技术接轨,构建中国膳食暴露评估非参数概率模型.方法 利用我国膳食调查、污染物监测数据及相应的人口学资料建立膳食消费量和化学污染物浓度经验分布.通过蒙特卡洛(Monte Carlo)模拟和自助法(Bootstrap)抽样获得人群膳食暴露变异度和不确定度.其中,膳食量数据和人口学数据来源于2002年中国居民营养与健康状况调查24 h膳食回顾法收集的22 567个家庭66 172人连续3 d调查共计193 814个人日、1 810 703条数据.污染物监测数据为2000-2006年全国14个省或地区食品污染物监测网以及2005-2006年海关出口农产品监测数据,包括重金属、农药,以及霉菌毒素(如黄曲霉毒素)等135种污染物,涉及499种食物,共计487 819条数据.结果 构建了包括重金属、农药及部分毒素的我国人群膳食暴露非参数概率评估模型,得到不同污染物膳食暴露量分布的指标统计量和95%可信区间.对7～10岁儿童乙酰甲胺磷膳食暴露评估显示,城乡儿童膳食暴露量的中位数分别为1.77μg·kg~(-1)·d~(-1)和2.48μg·kg~(-10·d~(-1),其95%可信区间分别为(1.59～2.06)μg·kg~(-1)·d~(-1)和(2.33～2.80)μg·kg~(-1)·d~(-1).结论 构建的非参数概率模型可量化暴露评估中的变异度和不确定度,提高了膳食暴露评估精度.

abstractsObjective To establish a non-parametric probabilistic model for evaluation of Chinese dietary exposure and to improve the assessment accuracy while integrating into the global risk assessment on food safety.Methods Contamination data was from the national food contamination monitoring program during 2000-2006 ,including heavy metals,pesticides and mycotoxins,amounting to 135 contaminants with 499 commodities and 487 819 samples.Food consumption data was obtained from the national diet and nutrition survey conducted in 2002 with three consecutive days by 24-hour recall method,and 66 172 consumers were included.Monte Carlo simulation was applied to derive the intake distribution,and the uncertainty of each percentile was estimated using the Bootstrap sampling,Results Different nonparametric probabilistic models for dietary exposure evaluation on heavy metals,pesticides and some of the toxins were established for Chinese people,and intake distributions with 95% confidence intervals of these contaminants were estimated.Taking acephate as an example,the results of its model shows that,for the 7-10 year-old children,the median dietary exposure in urban and rural areas were 1.77μg·kg~(-1)·d~(-1) and 2.48μg·kg~(-1)·d~(-1) respectively,with a 95% confidence interval of (1.59-2.06)μg·kg~(-1)·d~(-1) and (2.33-2.80)μg·kg~(-1)·d~(-1) respectively.Conclusion The non-parametric probabilistic model can quantify the variability and uncertainty of exposure assessment and improve the assessment accuracy.