Consolidated octanol/water partition coefficients: combining multiple estimates from different methods to reduce uncertainties in log KOW

The octanol/water partition coefficient (KOW) is a key parameter for assessing the fate and effects of chemicals. It is a metric of their hydrophobicity, related to uptake and accumulation in organisms and specific tissues, and distribution in water, soil and sediments. The log KOW can be determined experimentally, more often it is calculated. Variability may be due to properties of the substances, different experimental methods, or different computational approaches with different domains of applicability. The objective of the present study is to derive coherent log KOW estimates with known variability by (1) estimating multiple log KOW values by different methods for diverse chemicals to exemplify their variabilities, (2) analysing the variabilities of log KOW estimates by underlying methods and for different chemical classes, and (3) recommending approaches to obtain reliable and robust log KOW estimates for hazard and risk assessment. Comparative analyses were based on 231 case study chemicals representing diverse chemical classes, such as POPs, PCB, PAH, siloxanes, flame retardants, PFAS, pesticides, pharmaceuticals, surfactants, etc. The variability of up to 36 log KOW values per substance, determined experimentally or estimated by different computational approaches, is 1 log unit and more across the entire log KOW range from <0 to >8. No systematic pattern is evident. Different methods for deriving log KOW perform sometimes better and sometimes worse for different chemicals. None of the methods (experimental or computational) is consistently superior and any method can be the worst. Iterative consensus modelling combines multiple estimates by a weight-of-evidence (WoE) or averaging approach for scientifically valid and reproducible log KOW estimates with known variability. Consolidated log KOW, being the mean of at least 5 valid data obtained by different independent methods (experimental and computational), are a pragmatic way to deal with the variability and uncertainty of individual results. While this approach does not solve any of the problems about “correctly” determining log KOW, it does limit the bias due to individual erroneous estimates. Consolidated log KOW are robust and reliable measures of hydrophobicity, with variability mostly within 0.2 log units.