Stahl et al. 2023 — Mercury in U.S. river fish fillets, national probability-based assessment
Stahl et al. report on EPA’s National Rivers and Streams Assessment (NRSA) characterization of total mercury, PCBs, and PFAS contamination in fillet tissue of fish from U.S. rivers, using a statistically representative probability-based survey design covering 48,826–79,448 river kilometers in 2013–14 and 66,142 km in 2018–19. All 353 (2013–14) and 290 (2018–19) fillet samples contained detectable mercury and PCBs. Applying EPA’s fish tissue screening levels to national mercury distributions, 23.5–26.0% of the sampled river-kilometer population contained fish with mercury concentrations above levels protective of human health for average fish consumers; this fraction is nearly unchanged between the two study periods, indicating no material national improvement over five years. The study also reports a 10-year temporal comparison using 2008–09 NRSA mercury data (n=541 fillet samples).
Key numbers
- tHg detection rate: 100% (all samples, both study periods)
- Fish tissue screening level exceedances for tHg (for average fish consumption): 23.5% of sampled river length in 2013–14; 26.0% in 2018–19 (no statistically significant change)
- PCB exceedances: 17.3% to 51.6% of sampled river length across screening levels
- PFOS (one PFAS) exceedances: 0.7% to 9.1% of sampled river length
- Target population: rivers ≥5th order (approximately ≥5th Strahler stream order); 353 sites (2013–14), 290 sites (2018–19)
- Sample_n above is the sum of both periods (353 + 290 = 643 composite samples)
- 2008–09 NRSA tHg data (historical comparison): n = 541 fillet samples across national river network
- EPA fish tissue screening levels applied: 300 ng/g for mercury (water quality criterion, protective for average consumers), and stricter levels for subsistence-level fish consumption
Note: The metals scope of this paper is limited to tHg for the HMT&C-relevant analytes. PCBs and PFAS (also measured) are out of scope for the wiki’s metals focus but are relevant to the broader fish-safety literature. This source does not report iAs, Pb, Cd, Ni, Al, Cr, Sn, or U in fish.
Methods (brief)
Probability-based survey design (unequal probability, stratified by state and ecoregion) consistent with NRSA. Fish fillet composite samples. Mercury analyzed by established EPA methods for fish tissue; speciation not performed — values are total mercury (tHg), not methylmercury (MeHg). The study does not report MeHg separately, though the introduction notes that essentially all mercury in fish fillets is in the methylmercury form; users should treat this tHg data as a proxy for MeHg in fish fillet when no speciation data exist, with the understanding that the equivalence is near-complete for muscle tissue but not for whole-body or organ-level samples.
Limitation: tHg only, no MeHg speciation confirmation (though in fillet tissue, tHg ≈ MeHg in practice). Coverage is U.S. rivers ≥5th order — does not include smaller streams, lakes, reservoirs, estuaries, or ocean fish. Composite samples (multiple fish pooled) dampen within-site variability. National distribution estimates are for river kilometers, not for specific fish species or retail categories.
Implications
Certification: U.S. river fish mercury exceedance rates (23–26% of river km for tHg above human-health screening level) establish that a large fraction of U.S. freshwater fish is contaminated at levels warranting advisory. Products using freshwater fish from U.S. rivers carry documented mercury exposure risk.
Courses: The temporal comparison (2008–09 vs. 2018–19) showing no significant reduction in mercury exceedance rates illustrates the persistence challenge; regulatory limits alone have not driven cleanup of the national fish mercury burden.
App: tHg occurrence data for freshwater-fish ingredient: U.S. river fish. The exceedance fractions are useful for population-level risk framing but not directly usable as ppb values; for ppb data, the actual screening levels (~300 ng/g = 300 ppb tHg as the reference point) and the distribution tables in the source should be consulted.