PAH contamination in shellfish: modelling to estimate exposure

Abstract

Polycyclic aromatic hydrocarbons (PAH) are known carcinogens and are abundant in the environment and foodstuffs. Currently the majority of PAH research focuses on benzo[a]pyrene (BaP), although a much greater range of PAH are known to have detrimental effects to human health. Monitoring a large number of PAH is expensive, time consuming and analytically demanding, yet there is currently no clear basis for determining which PAH should be monitored to give an indication of overall exposure. A thorough statistical examination of the relationships between different PAH in different foodstuffs has not previously been carried out. Using a test dataset of homogenised edible flesh from shellfish samples as a case study a modelling process using principal components analysis regression is proposed to determine which PAH subset (from a total of 27 monitored PAH) should be assessed as indicators for general PAH exposure. Multivariate ordination and clustering show that PAH concentrations of compounds of similar chemical structure can be highly correlated in the samples, e.g. the five ringed isomers PAHs benzo[b]fluoranthene, benzo[j]fluoranthene and benzo[k]fluoranthene. The model selection process determined which subsets of PAH can be used to predict the presence and abundance of other PAHs in shellfish samples. Models were more accurate in predicating PAHs concentrations of PAH where concentrations were measured above the limit of detection (LoD). PAH with values below the LoD were harder to predict accurately. The current analysis highlights that laboratories should focus on the following PAHs BaP, benzo[a]anthracene, benzo[g,h,i]perylene, phenanthrene, benzo[g,h,i]fluoranthene, chrysene, benzo[k]fluoranthene, benzo[b]fluoranthene and fluoranthene when analysing shellfish samples. Focussing monitoring on this group of PAH may give a better indication of overall PAH content of samples that the summed PAH indicator methods currently adopted.