中国膳食暴露评估模型软件开发及验证  

Development and verification of Chinese dietary exposure evaluation model software

作　　者：刘沛[1] 李靖欣[1] 孙金芳[1] Jianping Xue 陈炳为[1] 张宏[1] 余小金[1] 王灿楠[1] 袁宝君[3] 马永建[3] 田子华 

机构地区：[1]东南大学公共卫生学院流行病与卫生统计学系,南京210009 [2]:National Exposure Research Laboratory (NEAL) US Environmental Protection Agency (EPA) [3]江苏省疾病预防控制中心 [4]江苏省农产品检测中心

出　　处：《中华预防医学杂志》2010年第3期204-208,共5页Chinese Journal of Preventive Medicine

基　　金：国家自然科学基金（30671763）;国家“十一五”科技支撑计划（2006BAK02A01）

摘　　要：目的开发具有我国自主知识产权的膳食暴露评估模型软件，并按国际通用标准对其概率评估模型进行合理性和计算正确性验证。方法基于我国膳食调查及食品污染物监测数据，采用统计分析系统（SAS）编程技术建立膳食暴露评估模型计算模块和人机对话界面。使用全国2～7岁儿童膳食暴露概率评估模型的结果与点估引吸江苏省金湖地区2～7岁儿童双份饭研究结果相比较的方法验证模型的合理性。通过随机抽取10000人天的膳食暴露数据与@Risk软件计算结果相比较的方法验证软件外部计算的正确性；以模型计算的2～3岁儿童食物消费量及铅污染残留浓度数据的均数漂移为指标验证本软件内部计算的正确性。结果成功开发了包括多种膳食暴露评估模型并具有易于操作用户界面的中国膳食暴露评估模型软件。在合理性验证方面，概率评估模型结果低于点估计，其中黄瓜的乙酰甲胺磷膳食暴露点估计为4．78μg·kg^-1·d^-1，概率评估的P99．9为0．39μg·kg^-1·d^-1；同时概率评估高于双份饭结果，儿童铅膳食暴露概率评估P95为11．08μg·kg^-1·d^-1，双份饭P95为5．75μg·k^1·d^-1。在计算正确性验证方面，概率评估模型计算结果与@Risk结果接近，乙酰甲胺磷膳食暴露概率评估计算的P95为4．27μg·kg^-1·d^-1，@Risk计算的P95为4．24μg·b^-1·d^-1；消费量及污染物的均数漂移表现为以零为中心的随机误差分布，漂移范围为0．05％～11．9％。结论开发的中国膳食暴露评估模型软件计算正确，结果合理，对提升我国膳食暴露评估技术水平具有现实意义。Objective To develop the dietary exposure evaluation model software accredited of Chinese intellectual property rights and to verify the rationality and accuracy of the results from the probabilistie model in Chinese dietary exposure evaluation model software according to international standards. Methods The software of SAS was used to build various evaluation model based on the data from Chinese dietary survey and the chemical compound in food surveillance and to design an operation interface. The results from probabilistic dietary exposure model for children 2 - 7 years old were compared with that from duplicate portion study of 2 - 7 years children dietary exposure in Jinhu, Jiangsu province in order to analyze the rationality of model. The results from probabilistic model of dietary exposure were compared with the results from @Risk software to verify the correction of the probabilistie model by using the same data of randomly selected 10 000 study subjects from national dietary survey. While,the mean drift was used as an internal index to illustrate the accuracy of the computation. Results Chinese dietary exposure evaluation software was developed successfully. On the rationality, the results from probabilistic model were lower than that from the point estimation （e. g. , cucumber： the result of point estimation of acephate was 4.78 μg·kg^-1·d^-1 ,while the results of probabilistic model which was 0. 39 μg·kg^-1·d^-1）. Meanwhile the results from probabilistic model were higher than the results of duplicate portion study （ on the P95 ,the result of prohabilistic model of Pb exposure in children was 11.08 μg·kg^-1·d^-1 ,while the results of duplicate portion study was 5.75 μg·kg^-1·d^-1）. On accuracy, the results from @Risk and the probabilistic model were highly consistent （on the P95 ,the result of probabilistic assessment of acephate diet exposure was 4. 27 μg·kg^-1·d^-1,while the results of duplicate portion study was 4. 24 μg·kg^-1·d^-1） , and the mean drift was of ran

关 键 词：模型 统计学 膳食调查 软件设计 软件确认 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论] R155.5[自动化与计算机技术—计算机科学与技术]