A statistical procedure has been developed to derive a screening value from an observational study related to the developmental toxicity observed in loggerhead turtle (Caretta caretta) eggs exposed to long chain per- and polyfluoroalkyl substances (PFAS). A dataset of 41 nests in which the hatching rate was inversely correlated with the increase in the PFAS concentration in unhatched eggs was processed via a categorical regression approach. After outliers identification and removal, categorical regression analysis tested the relationships of the outcomes with the following parameters: perfluoro-nonanoic (PFNA), decanoic (PFDA), undecanoic (PFUdA), and dodecanoic (PFDoA) acids; perfluoroctansulfonate (PFOS); polychlorobiphenyls (PCBs) 28, 52, 101, 138, 153, 180; lead (Pb), total mercury (Hgtot), and cadmium (Cd); and other factors, such as "nest site," "clutch size," "incubation duration," and "nest minimum depth," as confounders/modifiers of the hatching rate. Among considered contaminants, PFOS, PFDA, and PFNA only were significant (p ˂ 0.05), as were "nest site," "clutch size," and "incubation duration," confirming their possible role in decreasing the hatching rate of sea turtle eggs. According to a chemical-specific visual strategy, PFOS only showed a typical monotonic dose/response curve, which allowed the identification of provisional hypothetical thresholds of PFOS, 1,386 (CI95 = 1,080-1,692) ng/kg, corresponding to average hatching rates of 93.3% (CI95 = 91.4%-95.2%). Our preliminary results indicate the feasibility of the extrapolation of a screening value from observational studies under the following requirements: (a) individuation of most influencing factors for the outcome; (b) datasets referred to baseline contamination; (c) selection of undisturbed nests; (d) appropriate statistical multivariate methodology. This work aims to contribute to the New Approach Methods (NAMs) to assess PFAS early-stage embriotoxicity in marine biota.
Read full abstract