Indiana University, The School of Public & Environmental Affairs, 1315 E. Tenth St, Bloomington, IN 47405, USA.
Sci Total Environ. 2012 Aug 15;432:57-64. doi: 10.1016/j.scitotenv.2012.05.074. Epub 2012 Jun 18.
The study of stressor interactions by eco-toxicologists using nonlinear response variables is limited by required amounts of a priori knowledge, complexity of experimental designs, the use of linear models, and the lack of use of optimal designs of nonlinear models to characterize complex interactions. Therefore, we developed AID, an adaptive-iterative design for eco-toxicologist to more accurately and efficiently examine complex multiple stressor interactions. AID incorporates the power of the general linear model and A-optimal criteria with an iterative process that: 1) minimizes the required amount of a priori knowledge, 2) simplifies the experimental design, and 3) quantifies both individual and interactive effects. Once a stable model is determined, the best fit model is identified and the direction and magnitude of stressors, individually and all combinations (including complex interactions) are quantified. To validate AID, we selected five commonly co-occurring components of polluted aquatic systems, three metal stressors (Cd, Zn, As) and two water chemistry parameters (pH, hardness) to be tested using standard acute toxicity tests in which Daphnia mortality is the (nonlinear) response variable. We found after the initial data input of experimental data, although literature values (e.g. EC-values) may also be used, and after only two iterations of AID, our dose response model was stable. The model ln(Cd)ln(Zn) was determined the best predictor of Daphnia mortality response to the combined effects of Cd, Zn, As, pH, and hardness. This model was then used to accurately identify and quantify the strength of both greater- (e.g. AsCd) and less-than additive interactions (e.g. Cd*Zn). Interestingly, our study found only binary interactions significant, not higher order interactions. We conclude that AID is more efficient and effective at assessing multiple stressor interactions than current methods. Other applications, including life-history endpoints commonly used by regulators, could benefit from AID's efficiency in assessing water quality criteria.
生态毒理学家使用非线性响应变量研究应激物相互作用受到事先所需知识量、实验设计的复杂性、线性模型的使用以及缺乏最佳非线性模型设计来描述复杂相互作用的限制。因此，我们开发了 AID，这是一种供生态毒理学家使用的自适应迭代设计方法，可更准确、高效地检查复杂的多重胁迫相互作用。AID 结合了一般线性模型的威力和 A-最优标准，以及一个迭代过程：1）最小化事先所需知识量，2）简化实验设计，3）量化个体和相互作用效应。一旦确定了稳定的模型，就会确定最佳拟合模型，并量化单个和所有组合（包括复杂相互作用）的胁迫方向和大小。为了验证 AID，我们选择了污染水系统中常见的五种共同存在的成分、三种金属胁迫物（Cd、Zn、As）和两种水化学参数（pH 值、硬度），使用标准急性毒性试验进行测试，其中水蚤死亡率是（非线性）响应变量。我们发现，在初始输入实验数据后，尽管也可以使用文献值（例如 EC 值），并且仅经过两次 AID 迭代后，我们的剂量反应模型就稳定了。ln(Cd)ln(Zn) 模型被确定为预测 Cd、Zn、As、pH 值和硬度联合作用对水蚤死亡率的最佳预测因子。然后，我们使用该模型准确地识别和量化了加性（例如 AsCd）和弱于加性相互作用（例如 Cd*Zn）的强度。有趣的是，我们的研究发现只有二元相互作用是显著的，而不是更高阶的相互作用。我们得出结论，AID 在评估多重胁迫相互作用方面比当前方法更高效、更有效。其他应用，包括监管机构常用的生命史终点，都可以从 AID 评估水质标准的效率中受益。
