Evaluation of a national data set for insights into sources, composition, and concentrations of per- and polyfluoroalkyl substances (PFASs) in U.S. drinking water

There was a statistically significant difference between the groundwater and surface water sample sets in PFSA/PFAS ratios (Mann-Whitney U test; p<0.01).

“There was a statistically-significant difference between the groundwater and surface water sample sets (Mann-Whitney U test; p<0.01).”

Odds ratios for all PFAS pairs ranged from 46 for PFOS and PFNA to 876 for PFOS and PFHxS.

“Odds ratios ranged from 46 (PFOS and PFNA) to 876 (PFOS and PFHxS).”

Significant positive correlations in concentrations occurred for six PFAS pairs: PFOS/PFOA, PFOS/PFHxS, PFOS/PFHpA, PFOA/PFHxS, PFOA/PFHpA, and PFBS/PFHpA.

“significant (p<0.05) positive correlations in concentrations of six PFAS pairs (PFOS/PFOA, PFOS/PFHxS, PFOS/PFHpA, PFOA/PFHxS, PFOA/PFHpA, and PFBS/PFHpA) did occur.”

There was a high degree of association between most pairs of PFASs.

“there was a high degree of association between most pairs of PFASs.”

Hierarchical cluster analysis of samples with one or more PFAS detections produced clusters that support bivariate concentration correlations.

“Resulting clusters support results of bivariate concentration correlations.”

Rates of occurrence and concentrations of individual UCMR3 PFASs were higher in groundwater than surface water, except for PFBS.

“rates of occurrence and concentrations of the individual UCMR3 PFASs were also higher in groundwater vs. surface water with the exception of PFBS (Figure S.5)”

Groundwater represented 72% of all samples where PFAS detections occurred.

“groundwater was predominant in UCMR3 data, representing 72% of all samples where PFAS detections occurred”

The number of detections per compound ranged from 19 for PFNA and PFBS to 379 for PFOA.

“The number of detections of each compound varied from 19 (PFNA and PFBS) to 379 (PFOA, Figure S.1).”

61% of groundwater samples had total PFSA to total PFASs ratios greater than 0.5.

“For example, 61% of groundwater samples had total PFSA to total PFASs ratios greater than 0.5.”

PFAS detections ranged from 0.01 μg/L for PFHpA to 7 μg/L for PFOS.

“PFAS detections ranged from 0.01 (PFHpA) to 7 μg/L (PFOS).”

From 2013-2015, 64 public water systems exceeded the non-enforceable LHA of 0.07 μg/L for PFOS and PFOA.

“From 2013–2015, 64 PWSs exceeded the non-enforceable LHA of 0.07 μg/L for PFOS and PFOA.”

Relationships based on UCMR3 data exhibit significant variability (R2=0.00-0.89).

“The relationships based on the UCMR3 data exhibit a significant degree of variability (R2 = 0.00-0.89, Figure S.2)”

Odds ratios between PFASs and other UCMR3 contaminants were lower (1.2-14.2) than those among PFAS pairs.

“Odds ratios were lower than those associated with PFAS pairs (1.2-14.2).”

In samples with 3-5 PFASs detected, 75-100% exceeded the LHA.

“In samples with 3-5 PFASs detected, 75-100% exceeded the LHA.”

Mann-Kendall analysis of quarterly total PFAS detection rates indicated an increasing trend over time (p=0.03).

“Mann-Kendall analysis of quarterly total PFAS detection rates indicated an increasing trend over time (p=0.03).”

Multiple PFAS ratios in UCMR3 were consistent with ratios measured in groundwater at AFFF-impacted sites.

“AFFF poses particular relevance to groundwater because it is primarily land applied leading to potential for subsurface migration, and as discussed, multiple PFAS ratios in UCMR3 were consistent with ratios measured in groundwater at AFFF-impacted sites.”

In 598 samples with detectable PFASs, about half contained mixtures of two or more PFASs.

“In 598 samples with detectable levels of PFASs, approximately 50% contained mixtures of 2 or more PFASs (Figure 1).”

Positive associations occurred between PFAS detections and other UCMR3 contaminants including 1,4-dioxane, HCFC-22, hexavalent chromium, chlorate, and 1,1-dichloroethane.

“Positive associations also occurred between detection of PFASs and other UCMR3 contaminants, including 1,4-dioxane, HCFC-22, hexavalent chromium, chlorate, and 1,1-dichloroethane.”

Among samples with detections, average PFAS concentrations ranged from 0.025 μg/L for PFHpA to 0.18 μg/L for PFBS.

“Within detections (i.e. non-detects eliminated), average PFAS concentrations ranged from 0.025 (PFHpA) – 0.18 (PFBS) μg/L.”

The strongest odds ratios were between PFASs and 1,4-dioxane.

“Strongest odds ratios were between PFASs and 1,4-dioxane.”

When detected, median total PFAS concentrations were higher in small PWSs (0.12 μg/L) than in large PWSs (0.053 μg/L).

“when detected, median total PFAS concentrations were higher in small PWSs (0.12 μg/L) than in large (0.053 μg/L).”

The median fraction of PFCAs comprising total PFASs in UCMR3 surface water systems was 100%.

“The median fraction of PFCAs comprising total PFASs in UCMR3 surface water systems was 100% (i.e. the majority of surface water samples with detections contained PFCAs but not PFSAs).”

The UCMR3 dataset used in this study contained analytical results for 36,139 samples from more than 5,000 public water systems.

“UCMR3 efforts yielded a dataset of 36,139 samples containing analytical results from >5,000 public water systems (PWSs).”

Approximately 14 million people are served by 153 public water systems that exceeded the Vermont PFOA regulation.

“approximately 14 million are served by the 153 PWSs that exceeded the Vermont PFOA regulation (Table 1).”

PWSs sourced from surface water were more likely to be dominated by PFCAs.

“In contrast, PWSs sourced from surface water were more likely to be dominated by PFCAs.”

Among samples with measurable PFAS concentrations, average total PFAS concentration increased log-linearly with the number of compounds detected (R2=0.97).

“Considering only samples with measureable PFAS concentrations, there was a log linear increase (R2=0.97) in average total PFAS concentration with increasing number of compounds detected.”

Large PWSs (>10,000 customers) were 5.6 times more likely than small PWSs (≤10,000 customers) to exhibit PFAS detections.

“Large PWSs (>10,000 customers) were 5.6 times more likely than small PWSs (≤10,000 customers) to exhibit PFAS detections”

MRLs for PFCAs (PFHpA, PFOA, PFNA) were 0.01-0.02 μg/L, lower than MRLs for PFSAs (PFBS, PFHxS, PFOS) at 0.03-0.09 μg/L.

“MRLs of 0.01-0.02 μg/L for PFCAs (PFHpA, PFOA, and PFNA) were lower than MRLs of 0.03 – 0.09 μg/L for PFSAs (PFBS, PFHxS, and PFOS).”

The median percentage of PFSAs in UCMR3 groundwater samples with PFAS detections was 74%.

“This range is higher than the median of 74% PFSAs observed in UCMR3 groundwater samples with PFAS detections, possibly reflecting different AFFF formulations, associated precursor transformation, or contributions of additional PFAS sources such as fluoropolymer manufacturing that may lead to higher fractions of PFCAs.”

Over 8 million people are served by 124 UCMR3 public water systems with PFOA and PFNA measured above New Jersey regulations.

“over 8 million people are served by the 124 UCMR3 PWSs with PFOA and PFNA measured in excess of New Jersey regulations”

The ratio of total PFSAs to total PFASs was higher in groundwater samples than in surface water samples.

“The total PFSA to total PFASs ratio exhibits a clear shift towards higher percentages of PFSAs in samples from groundwater relative to samples from surface water (Figure S.6).”

Average total PFAS concentration was higher in groundwater (0.21 μg/L) than in surface water (0.09 μg/L), p<0.0001.

“the average total PFAS concentration was higher (p<0.0001, Mann-Whitney U test) in groundwater (0.21 μg/L) than in surface water (0.09 μg/L, Figure 3).”

The frequency of detection for all UCMR3 PFASs was higher in large PWSs than in small PWSs.

“The frequency of detection for all UCMR3 PFASs was higher in large PWSs than in small (Figure S.5).”

Mann-Kendall analysis indicated an increasing trend in overall PFAS detection rates over time (p=0.03).

“For overall PFAS detections, the slope of the best-fit regression line was increasing but not statistically different from zero (p=0.26), while Mann-Kendall analysis indicated an increasing trend in detection rates over time (p=0.03).”

Small, groundwater PWSs had higher total PFAS concentrations (median = 0.18 μg/L) relative to all other detections (median = 0.053 μg/L) (p=0.01).

“However, these small, groundwater PWSs were also characterized by higher total PFAS concentrations (median = 0.18 μg/L) relative to all other detections (median = 0.053 μg/L) (p=0.01).”

Water samples supplied from large PWSs were 5.6 times more likely than small PWSs to have a detection of one or more PFASs (p<0.0001; 95% CI 2.9-10.8).

“Based on a categorical analysis of sample data for total PFASs (sum of the 6 UCMR3 PFASs), water samples supplied from large PWSs were 5.6 times more likely than small PFASs to have a detection of one or more PFASs (p<0.0001; 95% confidence limits of 2.9 - 10.8 on the odds ratio).”

The detection frequency for at least one PFAS in samples from small systems was 0.3% (9 of 2830 samples).

“The detection frequency for ≥1 PFASs in samples from small systems was low (9 of 2830 samples, or 0.3%).”

PFOA detection rates increased over time (p=0.01).

“Similar results were obtained for analysis of individual detection rates for PFOA (increasing, p=0.01) and PFOS (probably increasing, p=0.1).”

Only 8 samples with PFAS detections were collected at PWSs that were both small and sourced from groundwater.

“Only a small set of 8 samples with PFAS detections were collected at PWSs that were both small and sourced from groundwater (i.e. the surrogate for private wells).”

When detected, total PFAS concentrations were higher in small PWSs (average = 0.30 μg/L) than in large PWSs (average = 0.17 μg/L) (p=0.026).

“However, when detected, total PFAS concentrations were higher (p=0.026, Mann-Whitney U test) in small PWSs (average = 0.30 μg/L) than in large (average = 0.17 μg/L) (Figure 3).”

Transient detections of PFASs at concentrations near the MRL were observed in some PWSs.

“Additionally, transient detections of PFASs at concentrations near the MRL were observed in some PWSs (Supplementary data, Section S.2, Figure S.8), raising questions regarding the influence of compound MRLs on the representativeness of UCMR3 data.”

PFOS detection rates probably increased over time (p=0.1).

“Similar results were obtained for analysis of individual detection rates for PFOA (increasing, p=0.01) and PFOS (probably increasing, p=0.1).”

PFNA's low detection frequency cannot be attributed to its MRL because its 0.02 μg/L limit equals PFOA's, the most frequently detected compound.

“The low frequency of detection for PFNA cannot be attributed to the MRL as the limit of 0.02 μg/L was the same as that of PFOA, the most frequently detected compound.”

The composition of total PFASs in surface water versus groundwater is largely consistent with the types of releases more prevalent in each type of source water.

“The composition of total PFASs in surface water vs. groundwater is largely consistent with the types of releases more prevalent in each type of source water.”

The higher median PFSAs at AFFF-impacted sites (82-96%) compared to UCMR3 groundwater (74%) may reflect different AFFF formulations, precursor transformation, or additional PFAS sources such as fluoropolymer manufacturing.

“This range is higher than the median of 74% PFSAs observed in UCMR3 groundwater samples with PFAS detections, possibly reflecting different AFFF formulations, associated precursor transformation, or contributions of additional PFAS sources such as fluoropolymer manufacturing that may lead to higher fractions of PFCAs.”

Ratios of PFOS to PFOA and PFHxS to PFOA in UCMR3 appear more consistent with AFFF-impact scenarios than with other release types for which data are available.

“Ratios of PFOS and PFHxS to PFOA in UCMR3 appear more consistent with AFFF-impact scenarios as opposed to other release types for which data are available (Figure 2b and Figure S.3).”

• AssumptionRests on an assumption: Concentration ratios of PFAS pairs in UCMR3 can be compared with ratios from AFFF-impacted groundwater, WWTP effluent, and fluoropolymer manufacturing to infer source types.

Bivariate and multivariate analyses of PFAS composition suggested PWSs reflect impacts due to firefighting foam use and WWTP effluent as compared to other source types for which data were available.

“Bivariate and multivariate analyses of PFAS composition suggested PWSs reflect impacts due to firefighting foam use and WWTP effluent as compared to other source types for which data were available.”

• AssumptionRests on an assumption: Concentration ratios of PFAS pairs in UCMR3 can be compared with ratios from AFFF-impacted groundwater, WWTP effluent, and fluoropolymer manufacturing to infer source types.

Environmental samples with higher percentages of PFCAs have been attributed to sources including fluoropolymer manufacturing, landfills, and WWTP effluent.

“As discussed, environmental samples with higher percentages of PFCAs have been attributed to sources including fluoropolymer manufacturing, landfills, and WWTP effluent (Ahrens et al. 2009; Arvaniti and Stasinakis 2015; Bach et al. 2017; Dauchy et al. 2012; Houtz et al. 2016; Oliaei et al. 2013; Weber et al. 2017), which all have relevance to surface water scenarios.”

• cites Ahrens et al. 2009 sourceNot checked yet

  Not checked in the free run.

• cites Arvaniti and Stasinakis 2015 sourceNot checked yet

  Not checked in the free run.

• cites Bach et al. 2017 sourceNot checked yet

  Not checked in the free run.

Using data from two AFFF-impacted sites for the six UCMR3 PFASs, the median percentage of PFSAs in the sum of 6 UCMR3 PFASs was 82-96% in groundwater.

“Using data from two of these sites for the six UCMR3 PFASs (i.e. eliminating other PFASs measured at these sites for which data are not available in UCMR3), the median percentage of PFSAs in the sum of 6 UCMR3 PFASs was 82-96% in groundwater (Houtz et al. 2013; Hull et al. 2017).”

• cites Houtz et al. 2013 sourceNot checked yet

  Not checked in the free run.

• cites Hull et al. 2017Not checked yet

  Not checked in the free run.

Total PFAS concentrations in groundwater-based drinking water supplies in UCMR3 were more likely to contain mixtures dominated by PFSAs.

“These data confirm that total PFAS concentrations in groundwater-based drinking water supplies in UCMR3 were more likely to contain mixtures dominated by PFSAs.”

PFBS's high average concentration and low detection frequency are partly explained by its 0.09 μg/L MRL, the highest among the six PFASs.

“Elevated average concentration and potentially the low frequency of detection of PFBS are partially a function of the 0.09 μg/L MRL, which was the highest of the 6 PFASs studied (Table 1).”

The separate clustering of PFBS and PFNA suggests influence of both separate source types and environmental distribution.

“Collectively, the separate clustering (i.e. weaker relationships) of PFBS and PFNA suggests the influence of both separate source types and environmental distribution.”

USEPA's UCMR3 program screened six PFASs in U.S. drinking water between 2013 and 2015.

“The United States Environmental Protection Agency (USEPA) completed nationwide screening of six perfluoroalkyl substances in U.S. drinking water from 2013–2015 under the Third Unregulated Contaminant Monitoring Rule (UCMR3).”

PFSAs are primarily associated with AFFF, mist suppressants, and surface protection products.

“PFSAs are primarily associated with AFFF, mist suppressants, and surface protection products (Favreau et al., 2017).”

• cites Favreau et al., 2017 sourceNot checked yet

  Not checked in the free run.

North Carolina recently adopted a drinking water health goal of 0.14 μg/L for the PFOA replacement product perfluoro-2-propoxypropanoic acid (PFPrOPrA, trade name GenX).

“The state of North Carolina recently adopted a drinking water health goal of 0.14 μg/L for the PFOA replacement product perfluoro-2-propoxypropanoic acid (PFPrOPrA, trade name GenX) (NCDHHS 2017), so it is evident that concerns regarding PFASs in the environment extend beyond the UCMR3 PFASs.”

• cites NCDHHS 2017Not checked yet

  Not checked in the free run.

PFCAs in prior studies of WWTP comprised a median value of 75% of total PFASs.

“PFCAs in prior studies of WWTP comprised a median value of 75% of total PFASs (Ahrens et al. 2009).”

• cites Ahrens et al. 2009 sourceNot checked yet

  Not checked in the free run.

There is moderate evidence of increasingly frequent detection rates over the course of UCMR3 monitoring for PFAS overall and for PFOA and PFOS individually.

“This provides moderate evidence of increasingly-frequent detection rates over the course of UCMR3 monitoring for PFAS overall and for PFOA and PFOS individually.”

Aqueous PFOS measurements were 0.02-0.06 μg/L in a sample with known concentration 0.023 μg/L, with RSD 29%.

“Similarly, aqueous PFOS measurements were 0.02-0.06 μg/L in sample with a known concentration 0.023 μg/L, resulting in a RSD of 29% (van Leeuwen et al. 2009).”

• cites van Leeuwen et al. 2009 sourceNot checked yet

  Not checked in the free run.

Private, state, and local monitoring efforts have identified PFAS impacts in small PWSs and private wells not involved in UCMR3.

“Private, state, and local monitoring efforts have identified PFAS impacts in small PWSs and private wells not involved in UCMR3 (e.g., Hoosick Falls, NY and Bennington, VT).”

Use of replacement PFASs that may serve as precursors after release may contribute to consistent or increased detections of PFASs.

“Additionally, use of replacement PFASs that may serve as precursors after release into the environment may contribute to consistent or increased detections of PFASs.”

Temporal trends in UCMR3 data for PFOA and PFOS do not provide evidence that the 2002 phase-out of PFOS and gradual phase-out of PFOA beginning in 2006 had an impact on detection rates during UCMR3 monitoring.

“Temporal trends in UCMR3 data for PFOA and PFOS do not provide evidence that the 2002 phase-out of PFOS and gradual phase-out of PFOA beginning in 2006 had an impact on detection rates during the period of UCMR3 monitoring.”

In prior studies of WWTP effluent, 88% of PFAS detections were below the MRLs for the respective UCMR3 compounds.

“In prior studies of WWTP effluent, 88% of PFAS detections were below the MRLs for the respective UCMR3 compounds (Ahrens et al. 2009).”

• cites Ahrens et al. 2009 sourceNot checked yet

  Not checked in the free run.

The UCMR3 dataset does not fully represent the composition of total PFASs present in U.S. drinking water.

“Collectively this suggests that the UCMR3 dataset does not fully represent the composition of total PFASs present in U.S. drinking water, and that future monitoring of additional PFASs may be warranted.”

If small groundwater systems are considered a proxy for private wells, then private wells are 4.5 times less likely to detect one or more PFASs relative to other systems.

“If small groundwater systems are considered a proxy for private wells, then private wells are 4.5 times less likely to detect one or more PFASs relative to other systems.”

• AssumptionThe authors state this as an assumption: Data from small public water systems using groundwater are representative of private wells.

Concerns regarding PFASs in the environment extend beyond the UCMR3 PFASs.

“The state of North Carolina recently adopted a drinking water health goal of 0.14 μg/L for the PFOA replacement product perfluoro-2-propoxypropanoic acid (PFPrOPrA, trade name GenX) (NCDHHS 2017), so it is evident that concerns regarding PFASs in the environment extend beyond the UCMR3 PFASs.”

The median value of 100% PFCAs in UCMR3 surface water samples is elevated relative to these sources.

“The median value of 100% in UCMR3 surface water samples is elevated relative to these sources.”

Recent efforts at PFAS source control of new or legacy PFASs are not clearly reflected in occurrence rates in drinking water data.

“Based on these results, recent efforts at PFAS source control of new or legacy PFASs are not clearly reflected in occurrence rates in drinking water data.”

A round robin study measured aqueous PFOA concentrations of 0.021-0.083 μg/L in a sample with known concentration 0.025 μg/L, with RSD 32%.

“A round robin study of PFAS analysis measured aqueous PFOA concentrations of 0.021 – 0.083 μg/L in a sample with a known concentration of 0.025 μg/L, resulting in a relative standard deviation (RSD) of 32% (van Leeuwen et al. 2009).”

• cites van Leeuwen et al. 2009 sourceNot checked yet

  Not checked in the free run.

PFASs in surface water discharges are likely subject to dilution in receiving waters before uptake by downstream PWSs.

“PFASs in surface water discharges are likely subject to dilution in receiving waters before uptake by downstream PWSs.”

For states adopting regulatory limits for PFASs at levels less than the LHA, UCMR3 data may not provide adequate information regarding compliance of PWSs with local standards.

“Additionally, for those states (e.g. New Jersey, Vermont) adopting regulatory limits for PFASs at levels less than the LHA, UCMR3 data may not provide adequate information regarding the compliance of PWSs with local standards.”

Low rates of detection in surface water PWSs may be related to PFAS dilution to concentrations below compound MRLs.

“Low rates of detection in surface water PWSs may be related to PFAS dilution to concentrations below compound MRLs (Table 1).”

Per capita PFAS exposure risks are potentially greater at impacted small, groundwater PWSs and private wells compared to other impacted systems.

“These results suggest per capita PFAS exposure risks are potentially greater at impacted small, groundwater PWSs and private wells as compared to other impacted systems.”

• AssumptionThe authors state this as an assumption: Data from small public water systems using groundwater are representative of private wells.

Results in North Carolina and UCMR3 may partially reflect a lack of destructive attenuation mechanisms that would accelerate decreasing PFAS concentrations and detections over time.

“Each PFAS included in UCMR3 is a perfluoroalkyl substance, so results in North Carolina and UCMR3 may partially reflect a lack of destructive attenuation mechanisms that would accelerate decreasing PFAS concentrations and detections over time.”

When selected in 2010, the PWSs sampled in UCMR3 served approximately 79% (248 million) of the US population.

“When selected in 2010, the PWSs sampled in UCMR3 served ~79% (248 million) of the US population (Adamson et al., 2017).”

• cites Adamson et al., 2017 sourceNot checked yet

  Not checked in the free run.

The UCMR3 monitoring program included only a subset of environmentally relevant PFASs and drinking water sources in the U.S.

“The UCMR3 monitoring program included only a subset of environmentally-relevant PFASs and drinking water sources in the U.S.”

Precursor transformation is a potential explanation for the relatively high rate of PFHpA occurrence and increasing trends in quarterly PFAS detection rates in UCMR3 data.

“Additionally, precursor transformation is a potential explanation for relatively high rate of PFHpA occurrence and increasing trends in quarterly PFAS detection rates in UCMR3 data.”

The combination of elevated MRLs and dilution is likely an underlying cause of low surface water detection frequency in UCMR3 relative to groundwater.

“In general, the combination of elevated MRLs and dilution is likely an underlying cause of low surface water detection frequency in UCMR3 relative to groundwater.”

Considering an average RSD of 31% for PFOS and PFOA, drinking water with a true concentration of 0.07 μg/L for the sum of PFOA and PFOS may yield laboratory measured concentrations of 0.049-0.091 μg/L.

“Considering an average RSD of 31% for PFOS and PFOA, these results suggest drinking water with a true concentration of 0.07 μg/L for the sum of PFOA and PFOS may yield laboratory measured concentrations of 0.049-0.091 μg/L.”