Purpose Product developers using life cycle toxicity impact assessment models to understand potential impacts of material substitutions face serious challenges related to large data demands and high uncertainty. This motivates greater focus on model sensitivity toward input parameter variability, particularly in the context of emerging contaminants like engineered nanomaterials (ENMs), to guide future efforts in data refinement and design of experiments. This study presents a Monte Carlo tool designed for use with USEtox 1.0 that allows researchers to rapidly prioritize data needs according to influence on characterization factors (CFs).Methods Using Monte Carlo analysis we demonstrate a sensitivity-based approach to prioritize research through a case study comparing aquatic ecotoxicity CFs for the ENM C60 and the vitamin B derivative niacinamide, two antioxidants used in personal care products. We calculate CFs via 10,000 iterations assuming plus-or-minus one order of magnitude variance for fate and exposure-relevant inputs. Spearman Rank Correlation Indices are used for all variable inputs to identify parameters with the largest influence on CFs, which we prioritize for data refinement and future experimental investigation. Based on the importance of aggregate multi-species toxicity (average log EC50) and studies suggesting solvent residues may yield erroneous toxicity estimates, we recalculate C60 CFs omitting all studies using solvents in sample preparation.Results and discussion For emissions to freshwater, the C60 CF is log-normally distributed with a geometric mean of 280 and geometric standard deviation (GSD) of 2.1 PAF m 3 day/kg compared to 2.6 with a GSD=1.8 PAF m 3 day/kg for niacinamide. C60 CFs are most sensitive to varied suspended solids partitioning coefficients (Kpss) and average log EC50, whereas variation of other substance parameters has comparatively little effect on model results. Insufficient experimental evidence hampers to revise assumptions for Kpss, and we suggest prioritizing future experiments that elucidate C60 interactions with suspended solids. Recalculating C60 CFs without toxicity studies that use solvents reduces the geometric mean by more than a factor of ten. This reinforces the importance of thorough characterization of released ENMs, in this case regarding the presence of solvent residues.
ConclusionCalculating stochastic CFs allows sensitivity-based prioritization of data needs and future experiments, which is particularly helpful in the context of emerging contaminants like C60. Researchers can conserve resources and address parameter uncertainty by applying our approach when developing new or refining existing CFs for the inventory items that contribute most to toxicity impacts. The Monte Carlo tool can be applied to current toxicity characterization models like USEtox and is freely available at