Quantification
After mass spectrometry, we used Sciex Analyst software (Version: 1.6; Framingham, MA) to integrate peaks manually. To determine analyte concentrations, we first calculated peak area ratios between the target compound and the matched IS. We then used the peak area ratios of calibrants to interpolate linear regressions, establishing calibration curves (Supplementary Tables 3 and 4). For the biphenyl data, where there were high analyte concentrations in spiked and pregnant samples, we created separate high and low calibration curves to target relevant concentrations.
We defined each analyte’s observed reporting limit (RLobs) as the lowest calibrant used in the calibration curve. We calculated the limit of detection (LOD) by adding the mean of the three blanks with three times the standard deviation of the blanks. We compared the RLobs and LOD and used the higher of the two as the reporting limit (RL). For calculations using analyte concentrations, we substituted values below RL with RL/2.
We determined accuracy with the percent analyte recovered in the spike-recovery experiment and used relative standard deviation (RSD) to assess precision. We calculated percent recovery by comparing recovered analyte mass in spiked samples with expected analyte mass, using the following equation:
\begin{equation} Percent\ Recovery=\frac{\text{ab}}{cb+de}\ \times 100\%\nonumber \\ \end{equation}
where ‘a’ is the analyte mass fraction (ng/g) measured in a spiked sample, ‘b’ is the mass (g) of that spiked sample, ‘c’ is the mean analyte mass fraction (ng/g) measured in endogenous (i.e., un-spiked) samples, ‘d’ is the analyte mass fraction (ng/g) in the calibrant used to spike the sample, ‘e’ is the mass of the spike (g). We considered percent recovery between 70 % and 120 % and relative standard deviation (RSD) values below 15 % acceptable.