Real world-like simulations show efficient predictive power of in vitro skin corrosion tests used as stand-alone and in combination and how can toxicologists take advantage of them

Abstract

Historically, performance of in vitro toxicology test methods has been evaluated and considered for regulatory purposes based on sensitivity & specificity values derived from validation studies. Other indicators are however useful to evaluate in vitro tests such as positive & negative predictive values (PPV & NPV), likelihood ratios (LRs) or odds ratio (OR). These indicators, which are routinely used in diagnostic tests, if adapted and adequately applied to in vitro tests would help determine their realistic predictive power in real world-like situation and help risk assessors know how they can rely on in vitro tests used for their safety assessments. This paper performs a series of simulations considering the actual distribution in ECHA C&L inventory of skin corrosive chemicals to calculate several of these indicators of performance (PPV, NPV, LRs, OR). It shows applied examples of predictive power on EpiSkin™ and SkinEthic™ RHE two validated in vitro skin corrosive tests, explains how to build testing strategies based on these examples, compares so called ‘bottom-up’ and ‘top-down’ approaches, and demonstrates the number of tests required, and how risk assessors can practically take advantage of this.