Kavlock R J, Allen B C, Faustman E M, Kimmel C A
Developmental Toxicology Division, U.S. Environmental Protection Agency, Research Triangle Park, North Carolina 27711, USA.
Fundam Appl Toxicol. 1995 Jul;26(2):211-22. doi: 10.1006/faat.1995.1092.
Recently, most attention on the application of benchmark dose (BMD) techniques to toxicology data has focused on quantal measures of response. Before the advantages of the BMD approach can be exploited in the risk assessment process, it is important that continuous measures of response also be modeled appropriately. In this study, we examined a variety of approaches to estimating BMDs for a change in fetal weight following chemical exposure from a total of 85 developmental toxicity experiments. We modeled the change in the mean fetal weight of a litter in response to treatment using a continuous power model, as well as reductions in the weight of individual fetuses within litters (defined as falling below a preset level) using a log-logistic model which incorporates litter size as a covariable and considers intralitter correlations. For the litter-based approach, several methods of defining a benchmark effect (BME) were considered, including a percentage change in mean litter weight, a change in mean litter weight relative to variability in the control group, and a reduction in the mean litter weight to some point on the control group distribution curve. For the fetus-based approach, we examined several BME options on the cumulative frequency distribution of the control fetuses for defining a low weight fetus and calculated several levels of additional risk. BMDs for four litter-based BMEs (a difference of 5% in mean fetal weight, a decrease to the 25th percentile mean weight of control litters, a decrease in the mean weight by 2 standard errors, and a decrease of 0.5 standard deviation units) and two fetus-based BMEs (a 5% added risk of weighing less than the 5th percentile of control weights and a 10% added risk of weighing less than the 10th percentile) showed strong similarities to each other and to statistically derived NOAELs. In addition to providing comparison with the NOAEL as a reference value, these analyses provided confirmation of the advantages of the BMD approach over the NOAEL in terms of the influence of dose spacing and dose selection. Combined with our previous analyses of quantal endpoints of fetal effects, this information provides a firm basis upon which to implement the benchmark dose concept in developmental toxicity risk assessments.
最近，基准剂量（BMD）技术在毒理学数据应用方面的大部分关注都集中在反应的定量测量上。在风险评估过程中充分利用BMD方法的优势之前，对反应的连续测量进行适当建模也很重要。在本研究中，我们从总共85个发育毒性实验中，研究了多种估计化学物暴露后胎儿体重变化的BMD的方法。我们使用连续幂模型对一窝胎儿平均体重的变化进行建模，以反映治疗效果，同时使用对数逻辑模型对窝内单个胎儿体重的降低情况（定义为低于预设水平）进行建模，该模型将窝大小作为协变量，并考虑窝内相关性。对于基于窝的方法，考虑了几种定义基准效应（BME）的方法，包括平均窝体重的百分比变化、相对于对照组变异性的平均窝体重变化，以及将平均窝体重降低到对照组分布曲线上的某个点。对于基于胎儿的方法，我们在对照胎儿的累积频率分布上研究了几种BME选项，以定义低体重胎儿，并计算了几个额外风险水平。四种基于窝的BME（平均胎儿体重相差5%、降低到对照窝第25百分位数的平均体重、平均体重降低2个标准误、平均体重降低0.5个标准差单位）和两种基于胎儿的BME（体重低于对照体重第5百分位数的额外风险增加5%、体重低于对照体重第10百分位数的额外风险增加10%）的BMD彼此之间以及与统计学推导的无观察到有害作用水平（NOAEL）有很强的相似性。除了与NOAEL作为参考值进行比较外，这些分析还证实了BMD方法在剂量间距和剂量选择影响方面优于NOAEL的优势。结合我们之前对胎儿效应定量终点的分析，这些信息为在发育毒性风险评估中实施基准剂量概念提供了坚实的基础。
