Method Validation and Quality Control Criteria

All the analytical methods, as well as the sample treatments were validated before being applied to the real samples. This means the reported methods had to satisfy certain pre-established criteria in terms of precision, accuracy, sensitivity and relative response factors. All the methods were validated at, at least, two levels of the calibration standards. Absolute and relative recoveries, precision and accuracy were evaluated. Linear intervals and limits of detection and quantification were established according to these results. Table 20 to Table 23 summarise the main validated parameters.

Relative Response Factors

Six calibration points were used with concentrations ranging from 10 to 1,000 ng/mL (concentration in samples can be found in Table 20 to Table 23). The response of each calibration was checked for all the native standards by the calculation of the relative response factor (RRF), defined as the instrument response for a unit amount of target pollutant relative to the instrument response obtained for the same amount of the mass labelled standard. It can be calculated as shown in Equation 02, where ANAT is the area of the native standard, AIS is the area of the mass labelled standard, C is the concentration of the native standard and C is the

concentration of the mass labelled standard. Calculation of RRFs for each of the standards comprising the multi-point calibration should reveal them to be essentially identical for each concentration level. The relative standard deviation (% RSD) of RRFs for a given target compound should not exceed 10 %.

However, the use of a single internal standard for the quantification of more than one native standard (MeFOSA-IS for MeFOSA and EtFOSA, MeFOSE-IS for MeFOSE and EtFOSE, and MeFOSAA-IS for FOSAA, MeFOSAA and EtFOSAA) resulted in RRF values exceeding this 10 % RSD. The difference in the masses is responsible for the difference in the analytical response. Despite this variation between the native and the mass labelled standard, the results from the calibrations were consistent during the study and remained within 25 % of the original calibration RRFs, so these values were still accepted.

Equation 02 𝑅𝑅𝐹 =^{𝐴𝑁𝐴𝑇}

𝐴𝐼𝑆 𝑥 ^𝐶𝐼𝑆

𝐶𝑁𝐴𝑇

Precision and Accuracy

Blank and spiked samples were used in all validation exercises, together with previously reported samples or standard reference materials (SRM) when available.

For dust samples, cleaned sodium sulphate (for blanks), SRM 2585 developed by the National Institute of Standards and Technology (NIST) and dust samples from the UK and Australia were used. For liquid food samples, methanol (for blanks) and an in-house spiked mixture of water:milk:juice (80:10:10) were used. For solid food samples, diatomaceous earth (for blanks), different brands of spiked dry cat food

and inter laboratory fish samples were used. For blood samples, methanol (for blanks) and spiked calf serum were used.

Each sample was prepared in triplicate (as a minimum) and on two different days.

Percent recoveries (% Rec) (Equation 03 for solid samples, and Equation 04 for liquid samples) and relative standard deviation (% RSD) (Equation 05) both intra and inter day were calculated in order to evaluate accuracy – defined as closeness to the true value – and precision – defined as statistical variability related to reproducibility and repeatability- of the validated methods. In the equations below, AIS is the area of mass labelled standard in every sample; ANAT is the peak area of target pollutant in every sample, RRF is the relative response factor for the target pollutant (see Equation 02), MIS is mass of mass labelled standard added to sample (pg), SS is the sample size (g), VV is the final sample extract volume in the vial (mL), VLS is the initial volume of liquid sample (mL) and σn-1 is the standard deviation of the analysed batch of samples.

Equation 03 𝐶𝑜𝑛𝑐 (𝑛𝑔/𝑔) =^{𝐴𝑁𝐴𝑇}

𝐴𝐼𝑆 𝑥 ¹

𝑅𝑅𝐹 𝑥^𝑀𝐼𝑆

𝑆𝑆

Equation 04 𝐶𝑜𝑛𝑐 (𝑛𝑔/𝑚𝐿) =^{𝐴𝑁𝐴𝑇}

𝐴𝐼𝑆 𝑥 ¹

𝑅𝑅𝐹 𝑥 𝐶𝐼𝑆 𝑥 ^𝑉𝑉

𝑉𝐿𝑆

Equation 05 % 𝑅𝑆𝐷 = ^{𝜎 𝑛−1}

𝐴𝑣𝑒𝑟𝑎𝑔𝑒𝑥 100

Limits of Detection and Quantification

The instrumental detection limit (LOD) was defined as the quantity of the analyte providing a signal to noise ratio of 3:1 and it was calculated by extrapolation of the lowest concentration standards (10 mg/mL) in the calibration standards injected during the validation process for each analyte. In the majority of cases, the blank concentrations were not expected to exceed 30 % above the LOD, but in cases where the blanks contained concentrations above this level, the blank concentration would be used as the LOD. The sample detection limit or limit of quantification (LOQ) was determined as the lowest measurable concentration in the extracted sample, with respect to the LOD, final extract volume (VFE), volume of final extract injected (VFEI), sample size (SS) and percentage of mass labelled standard recovery (% IS Rec), and is calculated as shown in Equation 06.

Equation 06 𝐿𝑂𝑄 =^{𝐿𝑜𝐷 ×𝑉𝐹𝐸}

𝑉𝐹𝐸𝐼 ×𝑆𝑆𝑥 ¹⁰⁰

𝐼𝑆 𝑅𝑒𝑐𝑜𝑣𝑒𝑟𝑦 (%)

Mass Labelled Standards

The mass labelled standards (IS) were added to every sample in a specified amount of 200 ng/mL in vial to all the native standards, blanks and samples. They are commonly used in analytical chemistry to avoid recovery corrections when calculating concentrations in samples. Recovery values between 30 and 150 % were accepted. The acceptance of this wide interval is mainly due to matrix effect reasons, which could enhance or suppress heavily the intensity of the target analytes. Their signal to noise ratio was also measured and a minimum value of 20:1 was required.

Blanks

Two types of blank samples were defined: field blanks and matrix blanks. Field blanks were used to evaluate the lack of cross contamination during the sampling, storage and sample manipulation steps and they were usually solvents or inorganic reagents. Meanwhile, matrix blanks were used (when possible) to better mimic the real sample composition, as well as to evaluate the lack of cross contamination during the sample treatment steps.

Field blank and matrix blanks were prepared and run five consecutive times each at the beginning of every validation sequence. Internal standards were also added to them at the beginning of the sample treatment. Values below 30 % of the LOD were accepted for field blanks. For matrix blanks, concentrations below 30 % of the LOQ were accepted if the field blanks prepared together with them met their previously mentioned criteria and the RSD was below 30 %. Matrix blanks were later used as spiked samples during the analysis of real samples when a suitable SRM was not available.