應用基準劑量推估31種農藥每日可接受攝食量做為健康風險評估之可行性探討

Using the Benchmark Dose Approach to Perform a Risk Assessment of 31 Pesticides

呂水淵(Shui-Yuan Lu) ； 牟為謙(Wei-Chien Mou) ； 陳敏貞(Min-Chen Chen) ； 蔡韙任(Wei-Ren Tsai)

《臺灣農藥科學》 1期 (2016/12) Pp. 24-49

https://doi.org/10.6671/TPS/2016.001.02

摘要

本研究目的為以基準劑量方法研析農藥毒性資料，主要蒐集109 個世界糧農組織、世衛組織農藥殘留專家聯合會議 (Joint FAO/WHO Meeting on Pesticide Residues, JMPR)、歐洲食品安全局 (European Food Safety Authority, EFSA)、以及少數其他相關組織報告或文獻之農藥毒理資料，採用美國環保署開發的基準劑量軟體 (Benchmark Dose Software, BMDS) 分析基準劑量值(benchmark dose, BMD) 與其95% 信賴區間下限值 (95% lower confidence limit of benchmark dose, BMDL)。本研究研析的109 個農藥毒理資料中，只有31 個符合BMDS 分析的條件，然而有2/3 以上的試驗數據無法符合，其主要原因可能為試驗設計不足與動物試驗數據的問題。本研究判定基準劑量相關分析數值的品質則依據BMD/BMDL 比值、模式適合度及基準劑量下限值是否落於處理劑量範圍內，本研究結果顯示符合與不符合BMD/BMDL 比值小於2 分別為65%與35%；模式適合度則依據美國環保署標準，以非連續型數據，模式表現，Akaike 信息準則(Akaike information criteria, AIC) 等評估，數值越小較佳，本研究31 個農藥多數以Log-Logistic模式最佳，AIC 值大約在10 左右；模式適切度，P 值 ( 顯著性) 大約0.90 左右，數值越大較佳；卡方殘差 (scaled residual)，小於絕對值2；此外，符合與不符合基準劑量是否落於處理劑量範圍內分別為94% 與6%。此外，本研究發現大部分農藥推估的BMDL 值大於未觀察到不良效應的最高劑量 (no observed adverse effect level, NOAEL)。本研究結論為BMDS 可做為推估農藥每日可接受攝食量的健康風險評估重要分析工具，因此建議未來新興農藥的動物毒性試驗設計應符合以BMD 方法學分析的精神，以利於未來數據也能以此方法進行風險評估。

關鍵字

基準劑量 ； 農藥 ； 每日可接受攝食量 ； 風險評估

並列摘要

This study analyzed the toxicity of pesticides using the benchmark dose (BMD) approach. For this, we collected toxicological records for 109 pesticides from the Joint FAO/ WHO Meeting on Pesticide Residues (JMPR), the European Food Safety Authority (EFSA), and other related reports or references. We then used software developed by the USEPA to analyze BMD and benchmark dose lower bound (BMDL) values. Among 109 pesticides, only 31 met the requirements for BMD anlaysis; and we therefore could not fit models for approximately two thirds of pesticides. This primarily occurred due to experimental design and effect level. For the 31 pesticides that underwent BMD analysis, criteria for quality determination were based on the BMD/BMDL ratio, goodness of fit (GOF), and the BMD range. The BMD/BMDL ratio should be less than two, and our data was fulfilled and not fulfilled with 65% and 35%, respectively. For the GOF test developed by the USEPA, smaller Akaike Information Criteria (AIC) scores, and larger P values indicate better model performance. Out of the Log-Logistic models that we developed for the 31 pesticides, the best one had an AIC score of around 10 and a P value of around 0.90. In Log-Logistic models, the absolute value of scaled residuals should be less than two. Furthermore, criteria of BMD in the range was fulfilled and not fulfilled for 94% and 6% of pesticides, respectively. We also found that, for most of the 31 pesticides, the BMDL value was greater than the noobserved- adverse-effect-level (NOAEL). Our results show that the BMD approach can serve as a robust tool for pesticide risk assessment. We strongly suggest that the animal toxicity studies which investigate new pesticides should meet the requirements of BMD analysis in order to fully elucidate associated risks.

並列關鍵字

benchmark dose ； pesticides ； acceptable daily intake ； risk assessment