Many consumer products contain lipophilic, poorly soluble ingredients representing large-volume substances whose aquatic toxicity cannot be adequately determined with standard methods for a number of reasons. In such cases, a recently developed approach can be used to define an aquatic exposure threshold of no concern (ETNC_aq; i.e., a concentration below which no adverse affects on the environment are to be expected). A risk assessment can be performed by comparing the ETNC_aq value with the aquatic exposure levels of poorly soluble substances. Accordingly, the aquatic exposure levels of substances with water solubility below the ETNC_aq will not exceed the ecotoxicological no-effect concentration; therefore, their risk can be assessed as being negligible. The ETNC_aq value relevant for substances with a narcotic mode of action is 1.9 μg/L. To apply the above risk assessment strategy, the solubility in water needs to be known. Most frequently, this parameter is estimated by means of quantitative structure/activity relationships based on the log octanol–water partition coefficient (log K_OW). The predictive value of several calculation models for water solubility has been investigated by this method with the use of more recent experimental solubility data for lipophilic compounds. A linear regression model was shown to be the most suitable for providing correct predictions without underestimation of real water solubility. To define a log K_OW threshold suitable for reliably predicting a water solubility of less than 1.9 μg/L, a confidence limit was established by statistical comparison of the experimental solubility data with their log K_OW. It was found that a threshold of log K_OW = 7 generally allows discrimination between substances with solubility greater than and less than 1.9 μg/L. Accordingly, organic substances with a baseline toxicity and log K_OW > 7 do not require further testing to prove that they have low environmental risk. In applying this concept, the uncertainty of the prediction of water solubility can be accounted for. If the predicted solubility in water is to be below ETNC_aq with a probability of 95%, the corresponding log K_OW value is 8.