Purpose and Scope

The purpose of this report is to describe the analytical procedure for a new direct-injection (DI) LC/MS/MS method for determining PFAS compounds in matrix-modified centrifuge supernatant of water samples and to provide data to validate method performance. The validation studies include bias and variability, MDL, and holding time. All experiments were performed in four water matrices (reagent water [RW], groundwater [GW], surface water [SW], and treated municipal wastewater effluent [WWE]), except for the frozen holding time experiment, which omitted groundwater. Bias and variability were evaluated by analyzing replicate samples (n=7) spiked at seven different concentrations for each compound. If a specific compound could not be detected in low-concentration spikes, bias and variability were reported at fewer than seven concentration levels. MDLs were determined according to National Water Quality Laboratory (NWQL) quality management system requirements. Advancing Standards Transforming Market’s (ASTM’s) DQCALC software was used to calculate Currie’s critical level (Lc) within laboratory values as the MDL. Laboratory reporting levels were assigned as twice the MDL (ASTM International, 2010, 2013; Foreman and others, 2021). These ASTM standards were not revised or reapproved within an 8-year window and were withdrawn from active status. ASTM has confirmed that this has no bearing on the validity of the described processes, and that it remains acceptable to use and reference both of these historical standards (James R. Morgan, ASTM, written commun., June 2025). Stability was studied at 4 degrees Celsius (°C) with compound recoveries evaluated after 7-, 28-, and 90-day storage intervals and at −4 °C with recoveries evaluated after 90 days to determine maximum holding times.

Previous Studies

Sample collection, isolation, and sample cleanup must be considered before any discussion of instrumental analysis, particularly for this class of compounds because of their ubiquitous presence in the environment and potential sorption to sampling containers (Martin and others, 2004; Taniyasu and others, 2005). Fluorinated materials used in the analysis, such as filtration and extraction media (Arp and Goss, 2009) or tubing and seals in the HPLC instrumentation (Yamashita and others, 2004), can also cause analytical artifacts. Procedural losses can occur during sample preparation, but these are more easily compensated for using IDQ, or in ideal circumstances, standard addition quantification (SAQ). Both IDQ and SAQ can involve additional expense compared to internal- or external-standard quantification, because IDQ requires the acquisition of stable-isotope analogs of target compounds and SAQ requires the analysis of multiple spiked replicate samples to yield a single analytical result. Therefore, the interlaboratory comparison of van Leeuwen and others (2009) demonstrated that SAQ may generate more robust results in the absence of exact isotopic analogs, but when cost is considered, IDQ appears to be a reasonable alternative.

Another important consideration in the development of PFAS methods is the stability of PFAS compounds in stored samples and standards. In a study of 29 PFAS compounds in four matrices (bottled water, surface water, two wastewater treatment plant [WWTP] effluents) at three temperatures (20 °C, 4 °C, −20 °C), 10 compounds showed changes in concentration during storage in at least one experimental treatment (Woudneh and others, 2019). Most of these changes were observed in 20 °C treatments, indicating that refrigeration or freezing is at least somewhat effective at mitigating the interconversion of PFAS compounds. In fact, no changes to compound concentration were observed in any matrix for frozen samples (−20 °C). For refrigerated (4 °C) samples, no changes were observed in bottled water samples, but in surface water, 4 compounds had substantial decreases in concentration and 2 compounds increased. Likewise, in refrigerated WWTP effluents, 5 compounds decreased in concentration and 3 compounds increased. Woudneh and others (2019) demonstrate that, although the conventional wisdom that PFAS compounds are particularly stable in laboratory storage conditions is true, there are exceptions that must be considered when determining acceptable holding times.

One of the first extraction techniques for PFCA and PFSA utilized weak anion exchange SPE prior to derivatization gas chromatography/mass spectrometry (GC/MS) analysis (Moody and Field, 1999). The method was refined and applied to an LC/MS/MS quantification method by Taniyasu and others (2005). Other methods compared by van Leeuwen and others (2009) utilized SPE phases of octadecyl alkyl (C₁₈), OASIS-HLB, OASIS-WAX, styrene-divinylbenzene, or liquid-liquid extraction with methyl tert-butyl ether (MTBE) as a solvent. Methods that used a cleanup step generally involved washing the extraction media with a weaker solvent than was used to elute the compounds of interest.

Some methods have eliminated offline preconcentration as more sensitive mass spectrometers have become available and online SPE has become more feasible. Schultz and others (2006) utilized online SPE of a 500 microliter (μL) sample aliquot on a guard column followed by LC/MS/MS analysis. Backe and Field (2012) used large-volume injection of 900 μL samples and analysis of several PFCAs, PFSAs, and steroid estrogens to determine the need for SPE concentration techniques in general. Methods developed at the NWQL for pesticides (Sandstrom and others, 2015) and pharmaceuticals (Furlong and others, 2014) show that, with modern instrumentation, DI can achieve reasonable MDLs without the large-volume injection that required SPE for earlier methods. This has also been shown by Weber and others (2017).

The EPA has developed multiple methods for PFAS determination that have been approved. The discussion in this report is limited to the two most common methods that are offered by multiple laboratories and in widespread use. Method 537.1 is approved for the determination of 18 PFAS compounds in drinking water only (Shoemaker, and Tettenhorst, 2018), and Method 1633 (EPA, 2021) has been approved for the determination of 40 PFAS in water and other matrices including solids and tissues. All 18 compounds in Method 537.1 are determined by the method detailed in this report; Method 1633 determines 10 compounds not determined by this method (this method determines 4 compounds that are not determined by Method 1633).

The major differences between this method and the EPA methods are: (1) separate determination of branched isomers for PFHxS and PFOS (this method); (2) use of IDQ (this method and Method 1633); (3) applicability to water matrices other than drinking water (this method and Method 1633); (4) use of offline SPE (EPA Methods 537.1 and 1633); and (5) use of direct sample injection (this method). Direct sample injection is considered a major advantage of this method compared to Methods 537.1 and 1633 because it can reduce sample handling and the opportunity to introduce bias, reduce sample volume requirements and associated shipping and storage costs, and eliminate the need for filtration, a potential source of negative bias to results. The tradeoff is that processing a smaller volume can make lower detection limits more difficult to achieve.

Analytical Approach

Although the primary purpose of this report is to document the version of this method as implemented at the NWQL, this section also discusses the development process and thus includes some historical information for context.

Coding and Nomenclature Used by the National Water Quality Laboratory and the National Water Information System

The sections “Method Number, Schedule, and Code,” “Scope and Application,” and “Nomenclature” summarize the naming schemes, scope, and nomenclature applicable to this report’s method. This information is utilized to identify PFAS and to distinguish results generated by this method from those generated by other methods that may also be present in NWIS.

Safety Considerations

All procedures that require the use of solvents, such as the preparation of mobile phases, calibration standards, and samples for centrifugation, are done in a fume hood. Eye protection, nitrile gloves, and protective clothing should be worn when handling reagents, solvents, and standards.

Liquid waste produced during sample preparation and analysis, including solvents used for rinsing glassware, unused mobile phase mixtures, and mobile phase waste eluting from the HPLC instrument, is collected in thick-walled plastic carboys. Solid waste, including analytical vials containing samples, standard mixtures, and methanol, is stored in glass carboys. All liquid and solid waste is stored and disposed of according to NWQL waste-handling procedures.

Interferences

Interferences causing either positive or negative bias to analytical results have been well documented in the analysis of PFAS compounds (Rodowa and others, 2020). It is crucial that glass and polytetrafluoroethylene (PTFE, also known as Teflon) are eliminated from sample containers, HPLC vials, caps, cap liners or septa, pipet tips, sampling pumps, tubing, seals, and any other material that will come into direct contact with samples. The disposable materials used in this method are polypropylene, high-density polyethylene (HDPE), or similar nonfluorinated polymer plastics whenever possible. Anionic functionalities in PFAS compounds may sorb to glass surfaces, resulting in a negative bias, although glass containers may be used for concentrated standards stored in an organic solvent such as methanol. Target analytes, such as PFOA and 4,8-dioxa-3H-perfluorononanoate (DONA), may be used in the synthesis of PTFE and potentially leach from PTFE-based materials.

Commercial HPLC systems typically contain PTFE in tubing, seals, solvent degasser membranes, and other components and must be modified to minimize contamination (Anumol and others, 2017). The LC/MS/MS instruments used for this analysis are modified to remove as much PTFE as possible; PTFE is replaced with stainless steel, polyether ether ketone (PEEK), or another nonfluorinated replacement. The HPLC instrument is fitted with a delay column to separate target compounds derived from instrument components from target compounds present in samples. The delay column is placed in the HPLC flow path immediately before the sample injector. Any target compounds present in elution solvents, pumps, pump seals, solvent degasser, or other components upstream from the delay column must pass through two HPLC columns, but target compounds derived from the sample only pass through the instrument’s analytical column. This results in a delay in the contaminant peak, if such a peak is present. When instrument contamination is present, two peaks for PFOA, DONA, or other compounds may be observed; the earlier eluting peak, which has the same retention time as the exact analog IDS, if present, is the peak of interest, and the later eluting peak should not be quantified.

Higher mass PFAS compounds (primarily PFAS with 10 or more carbon atoms) may also sorb to container surfaces in an aqueous sample matrix, causing a negative bias. This can be mitigated by adding methanol to samples to aid compound desorption. For this method, 45–55 percent methanol is added to each sample prior to centrifugation. The methanol fraction varies depending on the volume of the sample submitted. If a subsample must be taken from an aqueous sample (for example, if sample vials are submitted more than half full), an alternate sampling protocol, described later in the report, is applied.

Other sources of interference that are less well documented may occur, including ion suppression or enhancement from matrix components such as ionic strength or dissolved organic carbon. One major advantage of using IDQ is that isotope recovery is used to correct calculated concentrations for such interferences, which may be unpredictable and sample specific. These interference issues can also affect results during field sampling processes before samples are submitted to the laboratory.

Equipment and Supplies

The equipment and supplies listed below were tested during development of this specific method. Specific brand names and catalog numbers are shown as examples used for method development, but equivalent vendors can be used where appropriate.

Instrumentation

For the LC/MS/MS system, samples were analyzed using an Agilent model 6495A or 6495CA triple quadrupole mass spectrometer equipped with electrospray ionization with Agilent jet stream and ion funnel technology. The tandem mass spectrometer (MS/MS) was coupled to an Agilent 1200 series LC system consisting of a binary pump (part G4220A), secondary pump (part G7112B), 1290 FlexCube (part G4227A), thermostatted column compartment (part G1316C), and 1290 multisampler (part G7167B). The secondary pump and FlexCube are configured to allow online solid-phase extraction and not strictly required for this method. Qualitative and quantitative analyses were performed using the Agilent MassHunter data system (version B10.00).

Reagents

Reagents used specifically for this laboratory method include neat (undiluted) reagents, solutions that are obtained from external vendors, and solutions that are prepared in the laboratory as specified.

Standards

Analytical reference standards and isotopically labeled standards were obtained from Wellington Laboratories (Ontario, Canada). Standards were purchased as either multiple compound or single-compound solutions rather than as neat material.

Sample Collection, Shipment, and Holding Times

Samples must be collected using techniques designed to collect a representative, unbiased sample. The National Field Manual for the Collection of Water Quality Data (https://www.usgs.gov/mission-areas/water-resources/science/national-field-manual-collection-water-quality-data-nfm) describes protocols and provides guidelines for USGS personnel who collect water-quality data. Specific guidance for the collection of PFAS samples is in development.

There are precautions that must be taken for this method that may differ from those for other organic chemical methods. PFAS compounds, such as PFOA, other PFCAs, and DONA, may be used in the synthesis of PTFE. Therefore, it is critical that PTFE-containing materials are not used in the collection of PFAS samples. This extends to sample containers, tubing, and anything else that may come in direct contact with water samples. Although it is prudent to avoid the use of sample collection items that may have previously contained PFAS compounds, contamination from such items has not been documented when appropriate protocols are used during sampling (Rodowa and others, 2020).

In addition, PFAS compounds are prone to adsorption to surfaces and interfaces, which can result in analyte losses and negative bias on results. For that reason, sample collection protocols in this method have been streamlined to eliminate filtration steps and minimize steps that involve sample handling and transfer between containers.

Sample Preparation

Upon arrival at the laboratory, samples are frozen at −4 °C if they will not be analyzed within one week. Otherwise, samples are stored at 4 °C in a refrigerator until they are analyzed. Before analysis, samples, IDS, IIS, and standard solutions are brought to room temperature. All samples, calibration standards, and QC standards that are to be analyzed as part of the same instrument run must be processed at the same time. Each instrument run can include as many as three prep batches containing as many as 22 environmental and QC samples. Each prep batch consists of one laboratory reagent water blank, one laboratory reagent water spike (RWS), as many as 18 environmental samples, one LMS (denoted “MSPK” in the worklist shown in table 7), and one replicate (denoted “DUP” [duplicate]). The LMS and replicate samples are selected randomly from the 18 environmental samples in the batch using one of the replicate microcentrifuge tubes submitted with each sample (table 7).

The reagent water blank is prepared by adding 1,000 μL of reagent water to a new microcentrifuge tube. The RWS is prepared by adding 1,000 μL of reagent water and 25 μL of 10,000 ng/L prep spiking standard to a new microcentrifuge tube. For each sample, customers provide three environmental sample replicates. The LMS is prepared by weighing and recording the weight of one of these environmental sample replicates and adding 25 μL of 10,000 ng/L PFAS spiking standard. After preparation is complete, these QC samples are processed identically to the environmental samples in the batch. Preparation of blanks and spikes can be completed the day before the batch is processed.

Each centrifuge tube containing an environmental or QC sample is weighed, and the weight is recorded in the batch spreadsheet. Then, 63 μL of IDS in methanol is added to each tube with additional methanol as needed to bring the final solution to approximately 50 percent methanol. Each tube is capped and mixed for 30 seconds on a Vortex mixer and placed on the centrifuge. The samples are centrifuged for 15 minutes at 13,300 revolutions per minute. After samples are removed from the centrifuge, 10 μL of IIS solution is added to an HPLC vial for each sample followed by 940 μL of the centrifuge supernatant. When this is complete, the remaining sample in each centrifuge tube is discarded, the tube is allowed to dry, and the weight of the dry tube is recorded. This allows the exact volume of water in the original sample to be calculated. As many as three prep batches, processed on the same day, can be included in an analytical run with one set of calibration, LOQ, and TPC standards, and with CCV and CCB samples bracketing each prep batch (table 7).

Overview

Samples for analysis by DI-LC/MS/MS in 1-mL polypropylene analytical vials contain a mixture of approximately 50 percent water and 50 percent methanol, IDS compounds, and IIS. An aliquot of 20 μL is injected into the LC/MS/MS system. Analytes are separated on a reversed-phase, porous-shell analytical column with a water-methanol-ammonium acetate gradient elution. A C₁₈ reversed-phase column (delay column) is placed immediately upstream of the injector to separate any PFAS contamination internal to the instrument from the PFAS derived from the sample. Such internal PFAS contamination elutes after the peak of interest. After chromatographic separation, compounds undergo negative electrospray ionization followed by multiple-reaction monitoring of two precursor-to-product ion transitions. There is only one suitable transition available for PFBA and PFPeA, so these compounds are analyzed without a confirmatory ion. Data analysis is performed using Agilent MassHunter software (https://www.agilent.com/en/promotions/masshunter-mass-spec), and results are processed using custom Microsoft Excel reports developed by the NWQL to summarize results and quality control information and to ensure reported data are qualitatively and quantitatively accurate.

Separation by Liquid Chromatography

Environmental samples, QC samples, and calibration standards in 2-mL HPLC vials are placed in the sample tray of the autosampler. The sample tray is maintained at room temperature and not chilled because lower temperatures increase the tendency of PFAS to sorb to vial surfaces. A 20 μL aliquot is injected into the HPLC to begin the separation.

The HPLC parameters are shown in table 8, and a representative chromatogram is shown in figure 1 with separation of branched and linear PFHxS (fig. 2A) and PFOS (fig. 2B). An Agilent Poroshell 120 PFP HPLC column (3.1 mm×100 mm×2.7 micrometers [μm]) is used for separation. An Agilent Zorbax Eclipse (3.1 mm×50 mm×5.0 μm or similar C₁₈ column) is used as the delay column and is placed immediately upstream of the injector and downstream of the JetWeaver mixer. A guard column is not used in this method. The mobile phase consists of 10 millimoles per liter (mmol/L) ammonium acetate in water (A) and methanol (B). The column temperature is 40 °C, the flow rate is 0.4 mL/minute, and the HPLC pump gradient (table 9) increases from 10 percent B to 95 percent B during the 23-minute run time. The initial composition of the gradient has less methanol than the sample injection solution does. This composition can result in the broadening of early eluting peaks, but the high methanol content in the vials is necessary to ensure that the less-polar, late-eluting peaks do not sorb to vial walls. The gradient program was optimized to ensure the separation of the branched and linear isomers of PFHxS and PFOS. The isomers of each compound share common fragmentation patterns, so chromatographic separation is the best way to distinguish them for separate quantification.

Analysis with Dynamic Multiple-Reaction Monitoring

Dynamic MRM conditions were developed and optimized using procedures similar to those used for DI-LC/MS/MS analysis of pesticides (Sandstrom and others, 2015). Except for PFBA and PFPeA, two unique precursor-to-product ion transitions are monitored for each compound. In negative electrospray conditions, these low molecular weight compounds (PFBA and PFPeA) only produce one fragment ion in sufficient quantities for use in MRM analysis. Ion transitions were chosen on the basis of relative signal strength and comparison of selected ions to published PFAS methods. The ion source parameters are shown in table 10; MRM conditions for the 34 target compounds, 20 isotopically labeled IDS compounds and surrogate standards, and the isotopically labeled IIS compound are shown in table 11. The MRM transitions and HPLC gradient were optimized before the ionization parameters so that ionization conditions could be developed according to solvent conditions present at the time of each compound’s elution. Several changes to the solvent gradient were made during method development, and after each change, ionization parameters were reoptimized.

Mass Calibration

The mass spectrometer is tuned and calibrated using a calibrant mixture and procedures specified by the instrument manufacturer. Before every analytical run, a Checktune routine is initiated to verify that the instrument calibration is acceptable. Checktune is a procedure run in the Agilent MassHunter software that verifies the instrument tuning is still valid and retuning of the instrument is not necessary. If the Checktune report is not satisfactory, the Autotune procedure, also built into the Agilent MassHunter software, is used to optimize ion transmission and electron multiplier voltage. Even if the calibration remains stable and the Checktune passes, Autotune is run at least every 3 months. Typically, these procedures are run only in the negative electrospray mode used by this method, but the Autotune procedure should be run in negative and positive modes when the vendor provides preventative maintenance on the instrument up to three times annually.

Sample Analysis Sequence

Analytical batches are composed of a sequence of calibration, QC, and environmental samples from one or more prep batches and are constructed as shown in table 7. All the QC sample types discussed in this section are prepared by the NWQL analyst as described in table 7. When the same type of QC standard is injected multiple times while preparing a batch, a separate vial is prepared for each injection; multiple injections from the same vial are not advised. Before instrument calibration, wash blanks are used to verify that the instrument is not contaminated from previous analysis. A mid-level (100 μg/L) standard (INSTCHECK-100-1) is injected to verify the presence and adequate response of peaks for each analyte. The INSTCHECK-100-1 is evaluated immediately after injection so the run can be paused before any more injections if additional instrument cleaning or maintenance is required. A CCB is also analyzed before the calibration curve and denoted as CCBPRE. Making these injections before calibration ensures that instrument conditions have stabilized before any samples that will be used in data analysis are analyzed.

Next, a series of 10 standards ranging in concentration from 1 to 1,000 ng/L is analyzed in ascending order of concentration. After the last calibration standard, another CCB standard (CCBPOST) is analyzed to verify that there is no measurable carryover from the high concentration standard, and a TPC standard, made with solutions sourced separately from the calibration standards, is analyzed. Carryover is rarely, if ever, observed after the 1,000 ng/L standard, and in-house data processing scripts flag any sample results greater than 1,000 ng/L so the analyst can investigate any signs of carryover. As many as three prep batches, each bracketed by a CCB and CCV standard, may be analyzed. At the end of the analytical batch, a series of low-concentration LOQ standards is analyzed to verify that instrument sensitivity has been maintained throughout the course of analysis. The calculated concentration must be between 70 and 130 percent of nominal for CCVs and 50 and 150 percent for LOQs. Each block of environmental samples must be bracketed by satisfactory CCVs; in the case of CCV failure, samples must be reanalyzed or data for affected analytes appropriately qualified. For LOQ failures, reporting levels are adjusted to match the concentration of the lowest-passing LOQ.

Use of Isotope-Dilution Standards

The concentrations of PFAS in this method are determined by IDQ, which is described in Foreman and others (2012) in relation to NWQL’s steroid hormone method. As shown in Foreman and others (2012), IDQ performs better than traditional internal standard quantitation because it uses isotopic analogs of target compounds that more effectively compensate for matrix suppression, incomplete recovery during sample preparation, and other potential issues.

As of the time of this report, this method uses 20 labeled IDS compounds to provide an exact isotopic analog for 20 of the 34 PFAS measured (PFHxS-¹³C₃ and PFOS-¹³C₈, carbon 13-labeled analogs of PFHxS and PFOS, are not considered exact analogs for branched isomers of PFHxS and PFOS). The remaining 14 PFAS are quantified relative to a structurally similar and closely eluting IDS (table 2). Exact analogs of other target compounds may be added to the method as they become commercially available. For example, during the method validation experiments presented in this report, GenX (HFPO-DA) was quantified relative to PFOS-¹³C₈ but since December 2020, GenX has been quantified relative to the newly available HFPO-DA-¹³C₃ (perfluoro-2-propoxy-¹³C₃-propanoate) standard.

In addition to their function as internal standards that improve quantitative accuracy, IDS compounds are used as surrogates to provide an estimate of absolute recovery for each compound through preparation and analysis. To achieve this, the IDS compounds are themselves quantified relative to an IIS; the IIS is not itself quantified, and no IIS data is ever reported. The IIS is another isotopically labeled PFAS (PFPeA-¹³C₃), which is added to samples at the time of vialing and thus is not subject to procedural losses during sample preparation. During method validation, PFPeA-¹³C₃ was not available and PFOS-¹³C₈ was used as the IIS. This was effective but not ideal due to PFOS-¹³C₈’s structural differences from method analytes. The IIS is used only in the calculation of IDS recoveries and as a possible indicator of a failed injection or other major analysis failure.

Qualitative Determination

PFAS in unknown samples can be positively identified by evaluating mass-to-charge ratio, compound retention times (RTs), qualifier ion ratios (QIRs), SNR, peak-shape metrics, and other qualitative parameters reported by the MassHunter software. The QIR is the ratio of the integrated peak area of Q ion and the qualifying ion (q) for each compound. Expected values of retention times and ion ratios are determined for each analytical batch from values in that batch’s calibration standards. Although MassHunter software and in-house Microsoft Excel scripts may automate and assist in integrating peaks and determining if qualitative identification criteria have been met, the final decision of whether a compound has been identified remains with the analyst.

Retention Times

The acceptance criteria for retention times are set for each batch of samples on the basis of relative retention times (RRTs) of the calibration standards. The RRT is defined as the ratio of the RT of an analyte to the RT of its associated IDS. The default tolerance for RRT variability is set at ±1 percent; if any of the calibration standards deviate by more than the tolerance, the RRT tolerance is increased for that batch only to the lowest integer value (percent) that results in all the calibration standards meeting the criteria.

Qualifier Ion Ratios

The acceptance ranges for QIRs are based on general guidelines published by the European Union (EU) for the identification of veterinary drug residues in food (European Commission, 2002). The EU requirements allow for wider tolerances for higher values of QIR (for example, when the difference in response between Q and q is large), but for this method, all tolerances have been set as ±30 percent of the average QIR for calibration standards within each batch. The 2002 European Commission decision was amended in 2021, and the new requirements (±40 percent) are less restrictive than the requirements of this method (European Commission, 2021). Although these acceptance criteria do not perfectly agree with EU requirements, this method is in line with the EU published literature on PFAS analysis and uses additional qualitative criteria not used in the EU guidelines.

Sometimes the integrations of Q or q peaks by the MassHunter software are performed incorrectly because of noise peaks, dips, or other artifacts, leading to QIR failure, and manual adjustments to the peak integrations might be required. Manual integrations must follow best practices to accurately represent the true peak areas and not be influenced by meeting QC criteria at the expense of correctly representing the peak areas. A useful approach for avoiding unconscious bias is to perform any adjustments to peak integrations before considering if qualitative identification criteria are acceptable, avoiding unwarranted adjustments that bring these criteria into compliance.

Signal-to-Noise Ratios

For each constituent peak, Q and q ions are required to have an SNR greater than 3 for detection in an environmental sample to be reported. In calibration standards, results may be used to calculate calibration curves without meeting the SNR requirement for the q ion if the calculated result for any nonqualified standard meets residual criteria in the resulting curve (±30 percent for concentrations above the MRL and ±50 percent for concentrations below the MRL). When Q is more responsive than q, these nonqualified detections in the calculation can improve the accuracy of the calibration at the low end of the standard curve. This exception applies only to calibration standards with a known concentration and to the inclusion of those responses in the calculation of the calibration curve.

Quality-Assurance and Quality-Control Criteria

General guidelines for acceptance criteria for QC samples, calibration standards, and other QC procedures relevant to the NWQL (as provided in Maloney [2005]) were used for this method. Acceptance criteria (that is, batch QC criteria) for most QC types are based on IDS-adjusted calculated concentrations of target PFAS and set as 100±30 percent standard deviation (SD) of expected concentration (100±50 percent for LOQ standards). These QC types include RWS, CCV, LOQ, and TPC standards. The use of IDQ for the calculation of concentrations generally improves the accuracy and precision of results; however, there can be instances when procedural losses during sample preparation and analysis are substantial and can impede accurate data reporting. Therefore, each IDS is used as both an internal standard for the calculation of concentrations of target PFAS and a surrogate, calibrated relative to the IIS. Calculation of surrogate recovery provides a sample-specific estimate of procedural losses, matrix effects, and so on, for each of the 20 IDS compounds.

There are two related effects of low IDS recovery that must be accounted for: (1) when compounds are detected, the correction factor applied (implicitly) due to low IDS recovery increases, resulting in increased variability in calculated results for target PFAS; and (2) when compounds are not detected, MRL values calculated from fortified samples with higher IDS recovery may overestimate method sensitivity. For that reason, when there is low IDS recovery, results may be censored or qualified, or MRLs may be raised (table 12).

Results from this method are sometimes qualified with remark codes, value-qualifier codes, or comment fields using NWIS water-quality database conventions when QC acceptance criteria are not met or if problems or deviations in workflows could affect the accuracy of the result. A list of value-qualifier codes that may be used in the reporting of results from this method can be found in table 11 of Appendix A of the NWIS user’s manual (Dupré and others, 2013).

Wherever possible, the application of the codes is automated by custom Microsoft Excel scripts that base their actions on analytical results obtained from MassHunter data processing software. Although scripts may be used to aid the analyst to make qualification decisions, the final decision regarding data qualification is made by a trained analyst.

Concentrations of PFAS in each sample are determined using the calibration curves only after an analyst has reviewed the MRM chromatograms and confirmed that the qualitative identification of each analyte is valid. No quantitative results are reported without this qualitative confirmation of a detection. If the qualitative identification criteria (described in the sections of this report titled “Retention Times,” “Qualifier Ion Ratios,” and “Signal-to-Noise Ratios”) and batch QC criteria (described in the section of this report titled “Quality-Assurance and Quality-Control Criteria”) are met, concentration data are reported to the NWQL’s laboratory information management system. Concentrations below the MDL are reported with a result-level remark code of “E” (estimated or having a higher degree of uncertainty) and an appropriate NWIS value-qualifier code to reflect increased uncertainty in the magnitude of the quantitative result (for example, the “b” value was extrapolated at the low end, and the “t” value was extrapolated below the MDL). The “E” remark should not be interpreted to reflect uncertainty in analyte identification. Qualitative identification and quantitation are distinct and separate processes, and all confirmed detections must satisfy the same qualitative identification criteria.

Evaluation Criteria for Validation Experiments

The evaluation criteria for the validation experiments were that bias must be within 30 percent of the expected value (IDS-adjusted recovery between 70 and 130 percent) and variability less than or equal to 25-percent SD in spikes at a concentration equivalent to a mid-range calibration standard (145 or 516 ng/L). No PFAS were removed from the method for failing to meet these bias and variability requirements. All summary data in this report are tabulated as mean and absolute SD values. Note that absolute SD values of recoveries have units of percent; this does not imply that those values are relative SDs.

For MeFOSA-M and EtFOSA-M, recoveries in reagent water were low and variable (table 13; figs. 1.31 and 1.32 of app. 1), but performance was acceptable in the environmental matrices tested (tables 14–16; figs. 1.31 and 1.32 of app. 1), so those compounds were retained in the method and will likely be reported to customers and NWIS with an estimated remark code (“E”).

Assessment of Suitability for Direct Injection Liquid Chromatography/Tandem Mass Spectrometry Method

Two compounds underwent initial testing and were not included in the final version of the method. Both N-MeFOSE and N-EtFOSE had only one available MRM transition, and in both compounds the transition represented a very common fragmentation pattern that was not considered sufficiently diagnostic for inclusion in the method. These compounds were only included in the initial development of LC/MS/MS tests and were not tested in validation or other matrix testing. These compounds are reported by other published methods using the same ion transitions deemed unsuitable here. Data users should be aware of the potential risk associated with relying on N-MeFOSE and N-EtFOSE results produced by unit-resolution LC/MS/MS methods; inclusion of such compounds could lead to results affected by misidentification with other compounds that have similar retention times or to bias from an overlapping signal.

An additional five compounds (perfluorotridecanoate [PFTrDA], perfluorotetradecanoate [PFTeDA], N-MeFOSA-M, N-EtFOSA-M, and 11-chloroeicosafluoro-3-oxaundecane-1-sulfonate [11Cl-PF3OUdS]) had poor recovery or variability statistics during early phases of developing this method. Preliminary experiments indicate that this is likely because of a high affinity for container walls and other surfaces or interfaces in a 10 percent methanol, 90 percent water solution. Before the method validation experiments, the solvent composition during centrifugation and sample transfer steps was changed to a higher organic content (50 percent methanol, 50 percent water). Performance was improved significantly, and all five PFAS are included in the final version of the method.

Method Detection Level Determination

The method described in this report uses MDLs to define the sensitivity of analysis of each compound and RLs to provide batch- and sample-specific estimates of minimum reportable concentrations. The procedures for determining these parameters are described below.

Performance of Isotope-Dilution Standards

The 20 IDS compounds that were used as internal standards for target PFAS were also treated as surrogates, and their recoveries are reported. From time to time, new, labeled PFAS compounds may become commercially available. Because using exact-analog IDS compounds is preferable to using closely related PFAS for IDS, when new, isotopically labeled PFAS become available, they might be verified and implemented in this method without a need for formal revalidation. When a new, exact-analog IDS is added to the method, the results should be calculated using both the old, nonexact analog IDS and the new IDS to verify that recovery and variability are comparable or improved in batch QC that does (for example, RWS and laboratory matrix spikes) or does not (for example, CCVs and LOQs) undergo the sample preparation procedure.

During the Method Validation Study

During method validation, 19 IDS compounds were used, one of which (perfluorooctane sulfonate-¹³C₈ [PFOS-¹³C₈]) was also used as the IIS to estimate surrogate recovery for the other 18 IDS compounds. For that reason, surrogate recovery information for PFOS-¹³C₈, corrected to itself, is calculated as 100 percent for all samples. Effectively, surrogate recovery information is provided only for the other 18 IDS. Table 17 shows a comparison of IDS surrogate recovery in 275 batch QC samples that did not undergo centrifugation and vial transfer steps with 397 prep QC and environmental samples that did undergo those two steps.

Average IDS recovery was acceptable in all cases, ranging from 83.2 to 105.6 percent in prep QC and environmental samples and from 94.8 to 104.7 percent in batch QC samples. For every compound, the error bars for the two data treatments showed substantial overlap. In both treatments, the SD did not exceed 25 percent of recovery for 17 of 19 compounds, with 4:2 FTS-¹³C₂ (1H,1H,2H,2H-perfluoro-1-[1,2-¹³C₂]hexanesulfonate) and PFTeDA-¹³C₂ (perfluoro-n-[1,2-¹³C₂]tetradecanoate) showing slightly more variability (maximum SD of 33.4 percent of recovery). Variability limits are enforced on analyte results and not on surrogate recoveries.

Current Method Operation

Two new isotopically labeled compounds have been added to the method since the validation study. The new IIS, PFPeA-¹³C₃ (perfluoro-n-[¹³C₃]pentanoate), can be distinguished by the mass spectrometer from the IDS, PFPeA-¹³C₅ (perfluoro-n-[¹³C₅]pentanoate). With this change, PFOS-¹³C₈ is no longer used as an IIS, and its surrogate recovery can be reported. In addition, HFPO-DA-¹³C₃ has been procured, providing one additional compound with an exact analog IDS.

Table 18 shows summaries for samples analyzed after the implementation of the new IIS and IDS compounds (results for 1,010 batch QC samples and 1,536 prep QC and environmental samples are summarized). Mean surrogate recoveries are not demonstrably different from those observed during method validation, ranging from 87.1 percent to 108.7 percent in prep QC and environmental samples and from 92.4 percent to 102.2 percent in batch QC samples. In fact, recovery and variability of some IDS compounds are slightly improved (all SDs less than 20 percent), which is noteworthy because these samples comprise a much wider range of matrix types than the four used during validation. This comparison shows the utility of allowing the addition of new IDS compounds as they become available.

Reagent Water

The assessment of bias and variability in RW used the same data as the MDL determination. Data for all eight concentration levels are summarized in table 13 (also refer to figs. 1.1–1.68 of app. 1).

At 103 ng/L (spike level L5 in table 13), 31 of 34 compounds demonstrated acceptable bias (88.5 to 105.9 percent recovery) and variability (maximum SD of 20.8 percent recovery). Spike level L5 is at least twice the MDL for 32 of 34 PFAS (MDLs for N-MeFOSA-M and N-EtFOSA-M are 61.5 and 81.0 ng/L, respectively) and avoids the increased variability found closer to the high end of the calibration curve. For most PFAS, variability increases markedly below each compound’s MDL (figs. 1.35A–1.68A of app. 1). For the remaining three compounds, N-MeFOSA-M (figs. 1.31A and 1.65A of app. 1) and N-EtFOSA-M (figs. 1.32A and 1.66A of app. 1) are to always be reported with the “E” remark code to indicate that the results are estimated or have a higher uncertainty. For the third compound, the data from level L5 may not be representative of the overall performance because linear PFOS (PFOS-L) bias and variability statistics appear more in line with the other 31 compounds at other concentration levels (table 13; app. 1, figs. 1.18A and 1.52A). All three of these compounds (N-MeFOSA-M, N-EtFOSA-M, and PFOS-L) have markedly improved performance in the SW, GW, and WWE matrices, which contributes to the decision to retain N-MeFOSA-M and N-EtFOSA-M in the method.

Groundwater

Data for GW are summarized in table 14 and figures 1.1B–1.68B of appendix 1. Three replicates of unfortified GW were analyzed (table 19) with six detections between 2.1 and 14.7 ng/L. Calculated recoveries were adjusted where necessary for PFAS present in the ambient (unspiked) sample, but no ambient PFAS concentrations were high enough to substantially affect the bias and precision calculations at or above 145 ng/L. At a concentration of 145 ng/L (spike level 4, indicated by L4 in table 14), all 34 compounds had acceptable recoveries, which ranged from 83.6 to 101.1 ng/L (median 94.4) percent. Absolute SD ranged from 4.7 to 17.3 ng/L (median 9.8) percent recovery for L4. Performance in GW was markedly better than that observed in RW.

Surface Water

Data for SW are summarized in table 15 and figures 1.1C–1.68C of appendix 1. Three replicates of unspiked SW were analyzed, and because none had detections of PFAS, no ambient correction was necessary to determine recoveries. In spike level four (145 ng/L), indicated by L4 in table 15, all 34 compounds showed acceptable recoveries, ranging from 87.2 to 101.2 (median 94.3) percent. Standard deviation ranged from 4.6 to 20.2 (median 8.2) percent for all PFAS except N-EtFOSA-M, which has a nonexact IDS and an SD of 31.8 percent. It is notable that overall performance in this matrix is better than in RW.

Wastewater Effluent

Data for WWE are summarized in table 16 and figures 1.1D–1.68D of appendix 1. Four replicates of unfortified WWE were analyzed, and 11 compounds were detected at concentrations ranging from 1.8 to 93.3 ng/L. In particular, the amount of compound found in the ambient sample exceeded 30 percent of the fortified concentration for three compounds: PFPeA (93.3 ng/L), PFHxA (perfluorohexanoate, 42.3 ng/L), and PFOA (56.6 ng/L), which could conceivably affect the precision of recovery calculations for the 145 ng/L spike. This did not appear to be the case, as their respective recoveries ±SD were 97.1±15.7, 87.4±10.7, and 114.0±12.1 percent. For spike concentrations less than three times the concentration in the unspiked samples, bias and recovery calculations become less reliable; therefore, method performance in WWE is best characterized at the 145 ng/L level. Overall performance at the 145 ng/L spike concentration was once again better than in RW. Recoveries were in the range of 82.7 to 112.1 percent (median 93.5) and had an SD for 33 of 34 compounds ranging from 6.4 to 23.0 percent (median 11.8), with N-EtFOSA-M (30.3 percent), which has a nonexact IDS, exceeding 25 percent.

Stability Study Design

The stability of PFAS spiked into all four matrices (RW, GW, SW, and secondary WWE) used for the bias and variability studies was investigated over a 90-day period. Including environmental matrices in addition to RW is important because RW represents a simple chemical matrix with minimal opportunity for microbial degradation of compounds. Conversely, matrix components such as dissolved salts and organic matter may enhance physical losses such as sorption to container walls. In short, to understand PFAS stability in environmental samples, it is necessary to test in environmental matrices.

For each matrix, 28 replicate spikes were prepared in 2-mL microcentrifuge tubes and separated into four groups of seven replicates to be analyzed after different holding times. Samples were spiked at a medium-high concentration (413 ng/L nominal) and recovery was assessed at four time intervals. The first set of seven replicates per matrix was processed immediately (t=0), and the remaining 21 replicates were refrigerated and stored at 4 °C until analysis. No centrifugation, addition of IDS solution, or other processing took place until the day each set of replicates was to be analyzed. After the initial set of seven replicates per matrix at t=0 was analyzed, an additional seven replicates were analyzed after 7, 28, and 90 days (t=7, t=28, t=90).

Stability in frozen samples was subsequently investigated over a 90-day period for three of the test matrices; a sufficient volume of the groundwater matrix was not available to complete these tests. Seven replicates each of RW, SW, and WWE were spiked (250 ng/L nominal) and immediately transferred to a freezer, where they were stored for 90 days at −4 °C before analysis. In addition, one unspiked sample of each matrix was analyzed at the beginning (0 days) and end of the 90-day storage period.

Maximum Holding Times

Sandstrom and others (2015) describe two approaches for the calculation of maximum holding times—a statistical calculation specified by the ASTM (ASTM International, 1993), and a practical approach based on the data-quality objectives (DQOs) of the validation experiments. For this report, the goal was to develop a simple rule to specify how long samples can be stored before analysis. Therefore, only the practical approach was applied. A criterion that the lower limit of stability be 70 percent recovery (alternatively, that maximum loss is 30 percent) for all compounds at the maximum holding time was chosen for consistency with the control limits of other method QC parameters (for example, RWS recovery and CCV recovery). For all stability tests discussed in this report, raw (uncentrifuged) samples were stored for a specified amount of time, and centrifugation was done at the time of sample preparation.

Stability Study Results

The results of the refrigerated stability study in RW (table 20), GW (table 21), SW (table 22), and WWE (table 23) are shown in terms of PFAS recovery ±SD. Data from t=7, t=28, and t=90 are not normalized to t=0; rather, the recovery is calculated independently for each sample, and any compound losses are compared to the nominal concentration spiked rather than the measured concentration at t=0. Although there was some variation in PFAS between the four matrices, all compounds in all four matrices had recoveries greater than or equal to 70 percent after 28 days. Likewise, all four matrices had recoveries of individual PFAS compounds less than 70 percent after 90 days. Therefore, there is no need to consider setting different holding times for each matrix or making the holding time for certain matrices shorter than necessary to maintain consistency between samples.