Study subjects
Cohort 1 – EuroMix human blood samples. A total of 215 samples were analysed; 143 subjects were studied from T1. Of these, 72 subjects were also studied at T2, forming a paired sample which were collected 2 to 3 weeks apart. These samples were provided by the National Institute of Public Health (NIPH), Norway, via the EU project 'European Test and Risk Assessment Strategies for Mixtures' (EuroMix, 633172-2), funded by the H2020 programme. The participants had completed a diary of their weighed food record and a cosmetic use diary on both study days. A food frequency questionnaire (FFQ) (only on the first day) and a questionnaire for socio-demographic and lifestyle characteristics (only first day). Details of the sample collection, data collection, registration and processing are published elsewhere. The study was approved by the Regional Committees for Medical and Health Research Ethics (REK ID no. 2015/1868), and all participants provided written informed consent.
Cohort 2 – A total of 20 commercially available sera samples from the US were purchased from BioIVT. Ten serum samples were from individuals with records of taking fluoxetine at different dosages, while the other 10 were from individuals with no known use of fluorinated pharmaceuticals. These samples were analyzed for EOF and 25 selected PFAS and fluoxetine and its active metabolite (xeproxetine).
Target compounds
The 64 PFAS analyzed in Cohort 1 included perfluoroalkane sulfonic acids (PFSAs; C2-10, C12), perfluoroalkyl carboxylic acids (PFCAs, C2-C14, 16, 18), fluorotelomer carboxylic acids (FTCAs, 3:3, 5:3, 7:3), fluorotelomer unsaturated carboxylic acids (FTUCAs, 6:2, 8:2, 10:2), fluorotelomer sulfonic acids (FTSAs, 4:2, 6:2, 8:2), fluorotelomer phosphoric acid monoester (monoPAPs, 6:2, 8:2, 10:2), fluorotelomer phosphoric acid diester(diPAPs, 6:2, 8:2, 10:2), perfluorinated phosphoric acids (PFPAs, C6, C8, C10) and perfluorinated phosphinic acids (C6/C6, C6/C8, C8/C8), perfluoroalkane sulfonamides (C4, C6, C8, Methyl-C4, Methyl-C6, Methyl-C8, Ethyl-C8), fluoroalkane sulfonamidoacetic acid (C8, methyl-C8, ethyl-C8), perfluoroethylcyclohexane sulfonic acid (PFECHs), 3H-perfluoro-3-[(3-methoxy-propoxy)propanoic acid (ADONA), hexafluoropropylene oxide dimer acid (GenX), chlorinated polyfluorinated ether sulfonic acids (Cl-PFESAs, 6:2, 8:2), perfluoro-3,6-dioxaheptanoic acid (3,6-O-PFHpA), perfluoro-4-oxapentanoic acid (PF4OPeA). The 25 PFAS studied in Cohort 2 included PFSAs (C2-10, C12), PFCAs (C2-C14), FOSA, 6:2 FTSA, and GenX.
Target fluorinated pharmaceuticals. Fluoxetine and seproxetine (one of the active metabolites of fluoxetine) were measured in sera samples of Cohort 2.
Sample pretreatment
Serum samples (0.5 mL for Cohort 2 and 2 mL for Cohort 1) were extracted at pH 4 using a modified ion-pair extraction method. The sample extracts were split for EOF and targeted PFAS analyses. Additionally, the samples from Cohort 2 were underwent another round of extraction following the same procedure described above to capture fluoxetine and seproxetine, except that the pH was adjusted to 11 with ammonia solution before extraction. The sample extracts were also split for EOF, PFAS, fluoxetine and seproxetine analyses. The reason for extracting samples at pH 4 first was for comparison with other EOF studies in human blood samples with the same extraction method; the reason for extracting samples from Cohort 2 at pH 11 was that fluoxetine and seproxetine showed better recoveries at this pH; data are shown in the QA/QC section.
Instrumental analysis
The ultrashort-chain compounds (C2-C3 PFCAs and C1-C3 PFSAs) were measured using an supercritical fluid chromatography (SFC) system (Waters Ultra Performance Convergence Chromatograph, UPCC; Waters Corporation, Milford) coupled to a XEVO TQ-S (Waters Corporation) tandem mass spectrometer (MS/MS) using CO2 and MeOH with 0.1% ammonia as mobile phases with a Torus DIOL analytical column (3.0 × 100 mm, 1.7 μm). The majority of target PFAS (≥C4) were quantified using an Acquity UPLC coupled with a Xevo TQ-S MS/MS (both from Waters Corporation). Separation was achieved with a C18 BEH column (2.1 × 100 mm, 1.7 μm); mobile phases were MeOH and a 30/70 (v/v) mixture of MeOH and water with 2 mmol/L ammonium acetate and 5 mmol/L 1-methylpiperidine as additives. Two novel PFAS (ADONA and hexafluoropropylene oxide-dimer acid (HFPO-DA)) were measured using a Waters Acquity UPLC system coupled with a XEVO TQ-S micro MS/MS; the column and mobile phase were the same as shown before.
The EOF analysis was carried out on the combustion ion chromatography (CIC) system. Details of the method is available elsewhere. Briefly, the CIC consisted of a Methrom IC (Herisau, Switzerland) connected to an autosampler and combustion module from Analytic Jena (Konrad-Zuse-Straße, Germany). An ion exchange column (Metrosep A Supp5-150/4) was used for separation of fluoride using an isocratic elution method with carbonate buffer with 64 mmol/L sodium carbonate and 20 mmol/L sodium bicarbonate, and the fluoride was measured with a conductivity detector.
Quality Assurance and Quality Control measures
The sera samples in Cohorts 1 and 2 were stored at -80 oC at NIPH and BioIVT. Sera samples were packed with dry ice during the transportation to ORU, and they were stored in -20 oC in ORU before analysis.
During every batch of sample extraction, two procedural blank consisting of ultrapure water and two sera blank consisting of fetal bovine serum (FBS) were included to keep track of possible contamination. If no detectable level of PFAS was found in the procedure blank, the method detection limit (MDL) was defined as the lowest point of calibration curve after taken into account of sample concentration factor. Otherwise, the MDL was defined as the average of the detectable concentrations plus three times of the standard deviation (SD). No detectable concentrations of PFAS were found in the procedure blanks (n=24). The MDLs for PFAS ranged from 0.01 to 3.5 ng/mL.
Quantification of the 64 targeted PFAS and fluoxetine and seproxetine were performed based on internal calibration using corresponding mass-labelled standards; for those target analytes that did not have corresponding mass-labelled standards, the homologue closest in retention time was used for quantification.
Detectable concentrations of EOF were observed in the procedure blanks of 1 batch of extraction which resulted in a higher MDL for that batch (8.9 ng F/mL). The MDL for EOF in the remaining batches was 7.5 ng F/mL. Detectable concentrations of EOF were found in the serum blank samples (n= 22) with an average of 19.2 ng F/mL with a SD of 1.9; two serum blank samples were lost during sample analysis because of instrumental error. No detectable concentrations of PFAS were observed in the serum blank and these detectable EOF results in the serum blank were used to evaluate repeatability.
Before analyzing samples from Cohorts 1 and 2, a spike recovery test using FBS was performed to evaluate the recoveries of different PFAS. The recoveries ranged from 26% to 88%, with a maximum standard deviation (SD) of 21%. The spike recovery results for fluoxetine and seproxetine were 3% (SD: 1%) and 1% (SD: 1%), respectively, at pH 4, and 83% (SD: 5%) and 74% (SD: 5%), respectively, at pH 11.
During each batch of extraction, two quality control (QC) samples (SRM1957 – non-fortified human serum) were also extracted to evaluate both the recovery and repeatability of the analysis. The recoveries of the certified PFAS ranged from 63% to 79%, with a maximum relative standard deviation (RSD) of 18%. The measured EOF in the SRM samples was 12.7 ng/mL F with an RSD of 5.3%.
Data treatment
Since the measured concentrations in this investigation were not recovery-corrected for both PFAS and EOF, the measured PFOS and PFOA concentrations were compared with previously published PFAS results from the EuroMix study from T1, where the data are presumed to be 100% as the samples had been recovery-corrected. The comparison showed that the average recoveries of PFOS and PFOA in the current investigation were 48% (SD 17) and 47% (SD 16), respectively, when compared to the EuroMix study. Irrespectively of the recovery, all data were used for the mass balance analysis to estimate the amount of unrecognized EOF. However, samples with recoveries lower than 30% were excluded from the temporal analysis of EOF between T1 and T2. Further, mean estimated halving time for PFOS and PFOA have been found to be 5.3 years and 2.7 years, respectively. Based on this, paired samples that showed changes in concentrations of PFOS or PFOA greater than 25% between T1 and T2 (13 pairs) were removed from the temporal analysis.
For fluorine mass balance analysis, the measured PFAS concentrations of all targeted analytes (ng/mL PFAS) were converted to respective fluoride-equivalent concentrations (ng F/mL) using the following formula:
C_F=n_F\times\frac{MW_F}{MW_{PFAS}}\times C_{PFAS}
where CF is the concentration of fluoride (ng F/mL) coming from the compound, nF is the number of fluorine atoms in an analyte molecule, MWF is the molecular weight of fluorine, MWPFASi is the molecular weight of the analyte i, and CPFASi is the concentration of analyte i (ng/mL PFASi). Similarly, the measured concentrations of fluoxetine and seproxetine were also converted into fluoride-equivalent concentration (ng F/mL) for comparison. When concentrations of analytes in the targeted analysis were below MDL, zero was assigned for them for mass balance analysis.
Investigation on factors that may affect the proportion of unrecognized PFAS in blood
Blood samples from the Cohort 1 were accompanied by information on sex, age, height, weight, dietary habits, and frequency of personal care product use. Information on dietary habits include information on amounts of different food types consumed such as bread, grain, cakes, potato, vegetables, fruit/berries, meat, fish, egg, dairy, cheese, butter/oil, sweets, beverages and other foods from previous publication. Information on the use of personal care products (PCP) included data on how many occasions the following items had been used during the 24h study periods for study T1 and T2: shower gel, shampoo, conditioner, deodorant, facial cleanser, facial moisturiser, body lotion, mouthwash, toothpaste, perfume, lip gloss/balm, foundation, hand cream, hair styling, eye make-up, rouge powder and hand soap. To explore if the proportion of unrecognized EOF in the samples was related to consumption of specific foods or use of certain PCPs, multivariable linear regressions (MLRs) were performed in the same manner as in the previous publication. A three-step approach was used: Step 1) A univariate linear regression was performed between the proportion of unrecognized fluorinated chemicals as the dependent variable and each food or PCP group, age, and gender of the participants as independent variables to determine if there was a significant difference (p < 0.05) in consumption or use between males and females. The outcome decided if the MLR should be performed separately for males and females. Step 2) A second univariate linear regression between the proportion of unknowns as a dependent variable and each food or PCP group as the independent variable was performed to establish if the targeted food or PCP group was contributing to the proportion of unknowns. Age and gender were included as covariates. The food and PCP groups from these linear regressions with a p-value < 0.2 were included in the final MLR models for each chemical. Step 3) The MLR was performed with the proportion of unknowns as the dependent variable and by including the independent categorical variables with a p-value below 0.2 from step 2 as the independent variables.