Statistical Derivation of a Fold-Change Cut-Off for Non-Targeted Differential Screening Studies Based on Analytical Method Variability

Chemical characterization of smoke-free products relies on both validated quantitative methods for specific analytes and screening methods such as non-targeted differential screening (NTDS). The latter approach identifies compositional differences between complex samples without prior knowledge of the differential compounds. However, for a fraction of compounds with variable concentrations between two samples in an NTDS study, the observed differences are due in part to intrinsic variability in the aerosol collection and analytical methods. As a result, a subset of the reported compounds—particularly those with low fold changes—are specious and not reproducible between studies. The goal of this work was to derive an appropriate fold-change cut-off value to limit false-positive discoveries to a reasonable number.

A dataset was acquired by applying seven liquid chromatography-mass spectrometry and gas chromatography-mass spectrometry methods to aerosol samples produced by the Tobacco Heating System (THS) 3.0., and this dataset was augmented with simulated data points. A fold-change cut-off value of 1.25 was calculated, such that if a THS product was subjected to a comparison to itself, only 2.5 % of all analytical features would be above the cut-off value.

Considering only compounds increased by > 25 % in the sample vs. the comparator improves the statistical and scientific relevance of NTDS results and shifts the focus to compounds with differences that have greater toxicological relevance.