Cumulative Risk Meets Inter-Individual Variability: Probabilistic Concentration Addition of Complex Mixture Exposures in a Population-Based Human In Vitro Model

Abstract

Although humans are continuously exposed to complex chemical mixtures in the environment, it has been extremely challenging to investigate the resulting cumulative risks and impacts. Recent studies proposed the use of “new approach methods,” in particular in vitro assays, for hazard and dose–response evaluation of mixtures. We previously found, using five human cell-based assays, that concentration addition (CA), the usual default approach to calculate cumulative risk, is mostly accurate to within an order of magnitude. Here, we extend these findings to further investigate how cell-based data can be used to quantify inter-individual variability in CA. Utilizing data from testing 42 Superfund priority chemicals separately and in 8 defined mixtures in a human cell-based population-wide in vitro model, we applied CA to predict effective concentrations for cytotoxicity for each individual, for “typical” (median) and “sensitive” (first percentile) members of the population, and for the median-to-sensitive individual ratio (defined as the toxicodynamic variability factor, TDVF). We quantified the accuracy of CA with the Loewe Additivity Index (LAI). We found that LAI varies more between different mixtures than between different individuals, and that predictions of the population median are generally more accurate than predictions for the “sensitive” individual or the TDVF. Moreover, LAI values were generally <1, indicating that the mixtures were more potent than predicted by CA. Together with our previous studies, we posit that new approach methods data from human cell-based in vitro assays, including multiple phenotypes in diverse cell types and studies in a population-wide model, can fill critical data gaps in cumulative risk assessment, but more sophisticated models of in vitro mixture additivity and bioavailability may be needed. In the meantime, because simple CA models may underestimate potency by an order of magnitude or more, either whole-mixture testing in vitro or, alternatively, more stringent benchmarks of cumulative risk indices (e.g., lower hazard index) may be needed to ensure public health protection.

New approach methodologies to address population variability and susceptibility

“…NAMs have been applied to assess additional factors that contribute to variability and susceptibility such as sex [16,17], life stage [16,18,19], and comorbidities [20], including rare diseases [21]. Moreover, NAMs have the potential to incorporate complex mixtures and cumulative exposures [22][23][24]. Probabilistic methods can incorporate variability into predictions and have been used to derive reference dose estimates [25] and points of departure [23].…”

Dosing Methods to Enable Cell-Based In Vitro Testing of Complex Substances: A Case Study with a PAH Mixture

“…Cell-based studies allow for high-throughput testing of both defined [ 10 ] and environmental [ 11 ] mixtures. Analyses of dose reconstruction of mixture effects based on in vitro data have concluded that model-based predictions of mixture effects from the individual components are often under-predictive [ 10 , 12 ].…”

Invited Perspective: Uneven Progress Addressing Population Variability in Human Health Risk Assessment

“…Recent attempts have been made at investigating the impact of chemical mixtures, the results of which suggest that, although TD variability in response to mixtures appears no greater than that of its most variable component, component-based additivity models tend to underestimate mixture variability. 25 , 26 However, understanding the contributions of nongenetic host factors to variability—which can range widely across epigenetics, life stage, background disease, and nonchemical stressors 27 —remains a largely unmet challenge.…”