Analytical uncertainties in a longitudinal study - A case study assessing serum levels of per- and poly-fluoroalkyl substances (PFAS).

Per- and poly-fluoroalkyl substances (PFAS) are a range of persistent organofluorine contaminants, some of which have been found to accumulate in humans and have long half-lives. In longitudinal studies, when relying on measurements obtained at different points in time, it is critical to understand the associated analytical uncertainties when interpreting the data. In this manuscript we assess precision measurements of serum PFAS analysis in a follow-up study undertaken approximately 5 years after the initial study. These measurements included intra-(n = 58) and inter-batch duplicates (n = 57), inter-batch replicates (n = 58), inter-laboratory replicates (n = 10) and a re-analysis of 120 archived serum samples from the initial study. Average coefficients of variation (CV) for perfluorooctanoic acid (PFOA), perfluorohexane sulfonate (PFHxS) and perfluorooctane sulfonate (PFOS) associated with the reanalysis of archived samples ranged from 4 to 8%, which was greater than the inter- and intra -batch duplicates (<3%), but lower than the inter-laboratory comparison (CV ≥ 10%). Multi-centre analytical capacity in studies increases the variance within the dataset and implementation of variability-measures are useful to refine and maintain comparability. Due to long PFAS half-lives, this variance is an important consideration when deciding appropriate time intervals for sample collections in longitudinal studies, to ensure the difference is greater than the analytical uncertainty.