July 31, 2003

PCBs in Farmed Salmon: Methodology


  • Laboratory PCB analysis of farmed salmon fillets
  • Farmed salmon consumption estimates
  • Cancer risk assessment for PCBs in farmed salmon

Laboratory PCB analysis of farmed salmon fillets. Prepared farmed salmon fillets, with skin on and belly flap removed, were purchased from supermarkets in three U.S. cities in May 2003, with samples spanning five countries of origin and ten separate farming companies:

NOTE: Retailers are listed for descriptive purposes only. A store's presence on this list DOES NOT MEAN OR IN ANY WAY IMPLY that consumers should shop elsewhere, or that consumers can avoid PCBs in farmed salmon by shopping elsewhere.

  • Washington, DC: Super Fresh (Camanchaca, Chile)
  • Washington, DC: Safeway (Aqua Chile, Chile)
  • Washington, DC: Wagshal’s (Fjord Seafood, State of Maine)
  • Washington, DC: Whole Foods Market (Iceland Fish Co. Inc., Iceland)
  • Portland, OR: Fred Meyer (Pacific Sea Food Co. Inc., Canada)
  • Portland, OR: Safeway (PanFish Sales, Canada)
  • San Francisco: Safeway (Lus America Food Inc., Canada)
  • San Francisco: Albertson's (supplier unable to ascertain the country of origin)
  • San Francisco: Berkeley Bowl (SalmoCo, Scotland)
  • San Francisco: Ver Burgge (Sterling Company, Canada)

Samples were shipped overnight in coolers on ice to Axys Analytical Services in Sidney, British Columbia for analysis. Laboratory procedures and results are described in detail in Axys (2003), and are presented briefly below.

Samples were stored at -20° C immediately after they were received at the lab and until the samples were processed for analysis. Sample preparation involved cutting each sample into smaller pieces, homogenizing it three times using a medium grinder, and mixing between each time. Samples were analyzed using high resolution gas chromatography and high resolution mass spectrometry techniques in general accordance with 'USEPA Method 1668, Revision A: Chlorinated Biphenyl Congeners in Water, Soil, Sediment, and Tissue by HRGC/HRMS.' Analysis was conducted individually for all 209 PCB congeners. Standard quality control procedures were followed, including analysis of a lab blank and a spiked matrix.

Farmed salmon consumption estimates. EWG has estimated the distribution of salmon consumption across the population using a state-of-the-art seafood consumption record called the National Eating Trends Database, or NETD, compiled from detailed food diaries of 20,000 households across the country, in a survey designed and compiled by the New York-based international marketing data firm called NPD Group. These data are routinely used by the seafood industry in its own market research.

In NETD, men, women, and children maintain detailed food diaries over two-week survey periods, called “waves.” Some people are selected for multiple years to complete longitudinal studies on consumption, with the survey designed such that no geographical, ethnic, or other obvious biases exist in the data.

From these data, we characterized and quantified the variability of salmon consumption across the adult population of the U.S. First, we determined how many adults eat salmon by calculating how many of the 20,300 adults in the database reported eating salmon in any of the waves in which they participated. We then calculated the probability that an adult who did not report eating salmon during any of their sampling waves would have reported eating salmon had they been surveyed more often, using data from individuals with more reporting waves.

Probability-based factors were applied across all adults who did not report eating salmon, to correct for the likelihood that some people were misclassified as “non-salmon eaters” because of insufficient survey waves. People reclassified by this method were also assigned a consumption frequency based on the consumption statistics of people who met two criteria: 1.) had more survey waves than the reclassified individual; and 2.) reported no salmon consumption for at least the number of waves associated with the reclassified person. With these corrections, we were able to estimate the salmon consumption distribution across the adult population.

Cancer risk assessment for PCBs in farmed salmon. EWG estimated cancer risk posed by the consumption of farmed salmon in accordance with standard risk assessment procedures recommended by the Environmental Protection Agency.

Cancer risk for an individual was calculated as the product of cancer potency and chemical dose, normalized by body weight. We used Monte Carlo techniques to simulate measured variability in body weight, salmon serving size, and salmon consumption frequency, essentially creating millions of “modeled people” that allowed us to then compute population-based risk estimates. Model parameters are described below:

  • Body weight, age, and gender: Randomly selected from the more than 9000 participants in CDC’s National Health and Nutrition Examination Survey (NHANES) (1999-2000 data release). Data for adults ages 18 to 70 were used.
  • Portion sizes for salmon fillets: Randomly chosen from data collected between 1994 and 1996 in USDA’s Total Dietary Intake (TDI) survey. Since there were only 124 incidences of salmon fillet consumption in this database, we included another 610 times that baked or broiled fish was consumed by the survey participants. Using the age and gender assigned from the NHANES data, the model selected a portion size from the appropriate subgroup in the TDI data, and maintained this portion size as constant throughout the modeled period for any given modeled individual.
  • Consumption frequency: Systematic assignment of a salmon consumption frequency for an individual surveyed in NPD Group’s “sNational Eating Trends,” with corrections applied to correct for the tendency for the over-prediction of non-salmon eaters inherent in the data, particularly for survey respondents with short total periods of record. As with portion sizes, data selection was limited to the applicable gender. For any given modeled individual, salmon consumption frequency was maintained constant through the exposure duration.
  • Exposure duration: Adulthood, ages 18 to 70.
  • Averaging time to diffuse computed salmon risk across a lifetime: 70 years.
  • Cancer potency: 2 per mg/kg-day (EPA 1999).
  • Exposure concentration: PCB levels in farmed salmon taken from Axys (2003), with probability of selecting a particular farmed salmon in the data set proportional to the distribution of salmon in the U.S. food supply, by the salmon’s country of origin. Throughout a modeled lifetime, PCB concentrations were freshly selected for each exposure event. The model was restricted to computing cancer risk from farmed salmon, and did not account for incremental risks from the low levels of PCBs in wild and canned salmon.
  • Assumed PCB reduction from fat loss during cooking: 40 percent reduction (EPA 2000).


  • Axys Analytical Services Ltd. 2003. Environmental Working Group. PCB congeners analysis; Tissue samples; Axys method: MLA-010 Rev04. Data Package: 4226 L5841-1 to -10. Report from Axys Analytical Services Ltd. to Environmental Working Group. July 2003.
  • Environmental Protection Agency (EPA). 1999. Integrated Risk Information System. Polychlorinated biphenyls (PCBs) (CASRN 1336-36-3). Accessed July 21 2003 at http://www.epa.gov/iris/subst/0294.htm.
  • Environmental Protection Agency (EPA). 2000b. Guidance for assessing chemical contaminant data for use in fish advisories. Volume 2. Risk assessment and fish consumption limits. Third edition. EPA 823-B-00-008. November 2000.