A wide variety of micropollutants (MP) of diverse origins is present in waste and surface waters without knowing the effect of their combination on ecosystems and human. The impact of chemical mixtures is poorly documented and often limited to binary mixtures using MP of the same category. Knowing that it is not realistic to test every possible combination found in mixtures, we aimed to develop a new method helping to predict cocktail effects. Six chemicals of agriculture, industry or pharmaceutical origin were selected: cyproconazole, diuron, terbutryn, bisphenol A, diclofenac and tramadol. Individual MP were first used in vitro to determine the concentration at which 10% (Effective Concentration EC10) or 25% (EC25) of their maximal effect on human cytotoxicity was observed. Using an Orthogonal Array Composite Design (OACD), relevant complex mixtures were then tested. Multiple linear regression was applied for response surface modeling in order to evaluate and visualize the influence of the different MP in mixtures and their potential interactions. The comparison of the predicted values obtained using the response surface model with those obtained with the model of independent effects, evidenced that the hypothesis of independence was unjustified. The cocktail effect was further investigated by considering micropollutant response surfaces pairwise. It was deduced that there was a neutralizing effect between bisphenol A and tramadol. In conclusion, we propose a new method to predict within a complex mixture of MP the combinations likely involved in cocktail effects. The proposed methodology coupling experimental data acquisition and mathematical modeling can be applied to all kind of relevant bioassays using lower concentrations of MP. Situations at high ecological risk and potentially hazardous for humans will then be identified, which will allow to improve legislation and policies.

废水和地表水中存在着多种多样的不同来源的微污染物（MP），却不知道它们的组合对生态系统和人类的影响。化学混合物的影响文献很少，通常仅限于使用相同类别MP的二元混合物。知道测试混合物中存在的每种可能组合都是不现实的，因此我们旨在开发一种有助于预测鸡尾酒效应的新方法。选择了六种农业，工业或制药来源的化学品：环丙康唑，地隆，特丁龙，双酚A，双氯芬酸和曲马多。首先在体外使用单个MP来确定观察到的最大MPMP对人细胞毒性作用的10％（有效浓度EC10）或25％（EC25）的浓度。使用正交阵列复合设计（OACD），然后测试相关的复杂混合物。为了评估和可视化混合物中不同MP的影响及其潜在相互作用，将多元线性回归应用于响应面建模。使用响应面模型获得的预测值与通过独立效应模型获得的预测值的比较表明，独立性的假设是不合理的。通过成对考虑微污染物响应面，进一步研究了鸡尾酒效应。推测双酚A和曲马多之间具有中和作用。总之，我们提出了一种新的方法来预测MP的复杂混合物中可能涉及鸡尾酒效应的组合。所提出的结合实验数据采集和数学建模的方法可用于使用较低浓度MP的所有相关生物测定。然后将查明生态风险高，对人类潜在危险的情况，这将有助于改进立法和政策。