Quantification bias in non-targeted extractables and leachables screening: A Monte Carlo comparison of uncertainty-factor and response-factor corrections

Semi-quantitative non-targeted screening of extractables and leachables (E&L) relies on surrogate standards and is therefore subject to quantification bias driven by the high variability of relative response factors (RRFs). Thus, an uncertainty factor (UF) is used to account for RRF variability, which consequently lowers the analytical evaluation threshold (AET). However, the statistical basis of the ISO 10993-18 formula, UF_ISO= 1/(1 - RSD), with RSD the relative standard deviation of the RRFs, has not been established. The ISO formula omits a 1/μ factor (μ, mean RRF) and is defined only for RSD < 1. A Monte Carlo simulation was developed using the US FDA Chemicals List for Analytical Performance (CLAP) RRF distributions for gas chromatography-mass spectrometry (GC/MS) and for liquid chromatography-mass spectrometry (LC/MS) in positive and negative ionization modes, combined with empirical extractable concentration distributions. Multiple UF- and RRF-based correction strategies were compared across four AET levels and three surrogate concentrations. Without correction, false-negative (FN) rates ranged from 41% to 95% across analytical methods. All three CLAP datasets had RSD > 1, outside the ISO domain. Conditional correction produced simulated FN rates below 3%. These rates should however be considered optimistic, as they assume complete identification and a reliable compound-specific RRF, conditions rarely met in routine screening. Percentile-based UFs bound FNs at the selected threshold without these assumptions, although the most protective levels lowered the AET to values that are not achievable in E&L screening. A method-specific percentile UF, selected in line with the analytical instrumentation limits, is therefore proposed.