Risk-Benefit Analysis of Seafood Consumption: A Review


  • Rosalee S. Hellberg,

    1. Author Hellberg is with the U.S. Food and Drug Administration, Pacific Regional Laboratory Southwest, 19701 Fairchild, Irvine, CA 92612, U.S.A. Author DeWitt is with Oregon State Univ. Seafood Research and Education Center, 2001 Marine Blvd., Astoria, OR 97103, U.S.A. Author Morrissey is with the Oregon State Univ. Food Innovation Center, 1207 NW Naito Parkway, Portland, OR 97209, U.S.A. Direct inquiries to author Morrissey (E-mail: michael.morrissey@oregonstate.edu).
    Search for more papers by this author
  • Christina A. Mireles DeWitt,

    1. Author Hellberg is with the U.S. Food and Drug Administration, Pacific Regional Laboratory Southwest, 19701 Fairchild, Irvine, CA 92612, U.S.A. Author DeWitt is with Oregon State Univ. Seafood Research and Education Center, 2001 Marine Blvd., Astoria, OR 97103, U.S.A. Author Morrissey is with the Oregon State Univ. Food Innovation Center, 1207 NW Naito Parkway, Portland, OR 97209, U.S.A. Direct inquiries to author Morrissey (E-mail: michael.morrissey@oregonstate.edu).
    Search for more papers by this author
  • Michael T. Morrissey

    1. Author Hellberg is with the U.S. Food and Drug Administration, Pacific Regional Laboratory Southwest, 19701 Fairchild, Irvine, CA 92612, U.S.A. Author DeWitt is with Oregon State Univ. Seafood Research and Education Center, 2001 Marine Blvd., Astoria, OR 97103, U.S.A. Author Morrissey is with the Oregon State Univ. Food Innovation Center, 1207 NW Naito Parkway, Portland, OR 97209, U.S.A. Direct inquiries to author Morrissey (E-mail: michael.morrissey@oregonstate.edu).
    Search for more papers by this author

  • Disclaimer: The views presented in this article do not necessarily reflect those of the Food and Drug Administration.


Abstract:  Seafood, defined here as marine and freshwater fish and shellfish, is recognized as a healthy food choice because it is a low-fat protein source that provides long-chain omega-3 fatty acids important for early development along with eye and heart health. However, seafood is also known to contain certain contaminants, such as methylmercury and persistent organic pollutants, which can have harmful effects on human health and development. In order to limit exposure to contaminants while maximizing the benefits of seafood consumption, a number of quantitative and qualitative risk-benefit analyses have been conducted for seafood consumption. This review paper provides a brief background on risk-benefit analysis of foods, followed by a discussion of the risks and benefits associated with fish consumption. Next, risk-benefit analyses are reviewed in an historical context. While risk-benefit analysis consists of three main elements (that is, assessment, management, and communication), this review will primarily focus on risk-benefit assessments. Overall, most studies have found that the benefits far outweigh the risks among the general population, especially when a variety of fish is consumed at least twice per week. However, for certain populations (for example, pregnant women and young children) a more targeted approach is warranted in order to ensure that these groups consume fish that are low in contaminants but high in omega-3 fatty acids. The potentially harmful unintended consequences of risk-benefit communication on the general population and certain groups are also discussed.


Probably no food category has lent itself more to a risk-benefit comparison than seafood. While much of this has been played out in the popular press, the science of risk-benefit analysis has steadily improved and has helped decision-makers and government agencies develop consumption guidelines that influence individual consumption patterns as well as food policy. Seafood, defined here as fish and shellfish from marine or fresh water, farmed or wild, has been an important part of the human diet for a long time and now represents 16.7% of the global population's animal protein intake. In 2009, worldwide production of seafood reached 145 million metric tons (MMT), divided between capture fisheries at 90 MMT and aquaculture at 55.1 MMT (FAO 2009). Intake of seafood and long-chain fatty acids of marine origin has been associated with many benefits, such as reduced risk of coronary heart disease (CHD) and improved neurodevelopment; however, there are also several contaminants present in seafood, such as methylmercury (MeHg) and persistent organic pollutants (POPs), that have been associated with adverse health effects. In order to address this dilemma, numerous studies have been published examining various aspects of the risks and benefits of seafood consumption (see Appendix A). Health organizations worldwide have also released guidance values and advisories to help consumers manage risks and maximize benefits. The purpose of this review is to provide a historical overview of risk-benefit analysis work that has been carried out thus far regarding seafood consumption. This will include a discussion of the identification and characterization of the health effects of seafood and risk-benefit assessments related to seafood. Risk-benefit management and communication advisories will also be discussed. In order to familiarize the reader with the common terms and methods that will be discussed, an initial overview of risk-benefit analysis of food is provided.

Methods of Risk-Benefit Analysis of Food

Risk analysis is a well-established field that is comprised of 3 major elements: risk assessment, risk management, and risk communication (FAO/WHO 1997). The risk assessment aspect involves hazard identification, hazard characterization (that is, dose-response assessment), exposure assessment, and risk characterization. Risk management and communication are often performed by public health organizations and include risk evaluation, option assessment, option implementation, monitoring and review, and dissemination of information (FAO/WHO 1997). Traditionally, analysis of risks in foods has been carried out by toxicologists, while nutritionists have evaluated the benefits making a combined risk-benefit analysis difficult (Fransen and others 2010). When a food exerts both positive and negative effects on health, it becomes important for risk managers to be able to assess both the benefits and the risks (Barlow and others 2010). Despite the importance of risk-benefit analysis of foods, there is currently no scientific consensus regarding the underlying methodology and general principles for this type of analysis (EFSA 2006; Barlow and others 2010; Fransen and others 2010). However, there is increased interest and recognition regarding the importance of risk-benefit analysis of foods, and scientists at the European Food Safety Authority (EFSA) have proposed a framework based on risk analysis that incorporates risk-benefit assessment, risk-benefit management, and risk-benefit communication (EFSA 2006).

An EFSA Scientific Committee recently presented a guidance document outlining a methodology for risk-benefit assessments of food (Barlow and others 2010). They recommended that risk-benefit assessments be comprised of 3 elements: risk characterization, benefit characterization, and risk-benefit comparison (Figure 1). The Committee outlined 4 steps on the benefit assessment side that were designed to mirror those involved in risk assessment: identification of positive health effects/reduced adverse effects, characterization of those health effects (dose-response assessment), exposure assessment, and characterization of benefits. The identification of risks or benefits involves the description of adverse or positive health effects of the food or food components based on human observational or occupational studies, animal model studies, or in vitro studies. During health effect characterization, a dose-response curve is developed that describes the relationship between these effects and intake of the food or food component. This step often results in the establishment of a health-based guidance value, such as a tolerable daily intake (TDI) for a hazardous compound or a reference daily intake (RDI) for a beneficial compound. The intake of the compounds from foods is estimated during the exposure assessment step, based on food consumption data obtained from food frequency questionnaires, food diaries, food surveys, and so on. Next, risks or benefits can be characterized by combining the dose-response relationship with exposure assessment to evaluate the probability of a health risk or benefit occurring in response to intake of a specific food or food compound. The results of the separate risk and benefit assessments can then be compared in the final stage of risk-benefit assessment, termed "risk-benefit comparison." This assessment should be carried out for the general population and for any subpopulations that show increased sensitivity to the food or food component(s) being evaluated. All risk-benefit assessments should include recognition of the assumptions, uncertainties, strengths, and weaknesses of the method. Recently, another approach to risk-benefit assessment of foods was presented by Fransen and others (2010). The authors were in general agreement with the approach proposed by EFSA scientists, with the exception that their tiered approach inserted a number of "stop" points, where a decision is made as to whether or not the assessment should be continued, and they recommend conducting the exposure assessment prior to characterization of health effects. The advantage of conducting an exposure assessment early on is to first determine whether or not current levels of exposure (or lack of exposure) are of concern prior to carrying out an extensive literature search and dose-response modeling during the health effects characterization step.

Figure 1.

–Major steps in risk-benefit assessment, as recommended by the EFSA Scientific Committee and based on the discussions of the EFSA scientific colloquium on risk-benefit analysis of foods. Risks and benefits are first assessed separately and then the results of the 2 assessments are compared in the final stage of risk-benefit assessment, termed “risk-benefit comparison.” Figure modified from Barlow and others (2010).

In addition to providing a framework for risk-benefit assessment of foods, EFSA scientists also outlined 3 different levels of risk-benefit assessments that might be considered in a stepwise approach: (1) separate assessments for risks and benefits to determine whether the risks far outweigh the benefits or vice versa, (2) assessments of risks and benefits using common metrics that express risks and benefits in the same unit and can be directly and quantitatively compared, and (3) assessments of risks and benefits using composite metrics (Barlow and others 2010). Some examples of common metrics include estimates of the proportion of the population that is not within the health-based guidance value or estimates of the incidence of disease or mortality occurring at a certain exposure level and the impact of changing that exposure level. On the other hand, composite metrics, such as disability-adjusted life years (DALYs) and quality-adjusted life years (QALYs), are designed to integrate increases or decreases in 2 or more of the following components: morbidity, mortality, disease burden, and quality of life (Barlow and others 2010). QALYs were originally introduced in 1976 and are an important part of health interventions and economic evaluations by regulatory agencies. They allow, as part of a cost-effectiveness analysis, the ability to assess the improvement in quality-adjusted life expectancy that may be obtained through specific health intervention (Ponce and others 2000; Sassi 2006). Similarly, DALYs are used to quantify the burden of disease across a population, essentially measuring the gap between the current level of health and an ideal health situation (WHO 2010). These tools have been applied in risk-benefit analysis of seafood consumption, as discussed in a later section (Ponce and others 2000; Cohen and others 2005a; Guevel and others 2008). Consistent with the stepwise approach proposed by EFSA scientists, the risk-benefit assessment portion of this paper will be organized to first present qualitative research comparing risks and benefits, followed by studies that used common metrics, and then studies using composite metrics to examine the risks and benefits of seafood and its components.

Identification and Characterization of the Health Effects of Seafood

Positive or reduced adverse health effects of seafood

Interestingly, fish was initially recognized as a healthy food choice, because it is a low-fat protein source, and not for its health-promoting lipids (Anderson and Wiener 1995). High-fat diets were associated with an increased risk of CHD and some cancers, and the National Research Council (NRC) recommended substituting fish for fatty meats and whole-milk dairy products as a way of reducing fat and cholesterol intake (NRC 1989). Early studies with the Greenland Inuits (Bang and Dyerberg 1972, 1980) and Danish men (Kromhout and others 1985) suggested an association between fish consumption and reductions in death from CHD. For example, Kromhout and others (1985) reported a 44% reduction in CHD death for men consuming 1 to 2 servings (100 to 200 g) of fish per week compared to men not eating fish. Research over the next few decades supported the association between fish consumption and reductions in CHD incidence and mortality (Whelton and others 2004; König and others 2005), and identified two omega-3 polyunsaturated fatty acids (PUFAs) that have a major role in these beneficial effects: eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) (Simopoulos 1991; Wang and others 2006; Mozaffarian 2008). EPA and DHA cannot be synthesized in substantial amounts by the human body and must be obtained through the diet, with the main source being seafood, especially fatty fish (Williams and Burdge 2006). Another omega-3 fatty acid found in plant-based oils, alpha-linolenic acid (ALA), is a precursor to EPA and DHA but is converted at very low rates in the human body (<10%) (Williams and Burdge 2006). In addition to reduced CHD risk, a number of other health benefits have been strongly associated with EPA/DHA intake and fish consumption and have been characterized in large-scale epidemiological studies, reviews, and meta-analyses (Kris-Etherton and others 2002; IOM 2005; Mozaffarian and Rimm 2006). These include reduced risk of other cardiovascular disease outcomes such as sudden death (Wang and others 2006; Mozaffarian 2008) and stroke (He and others 2004a; Bouzan and others 2005), increased duration of gestation (IOM 2005), and improved visual and cognitive development (Fleith and Clandinin 2005; IOM 2005; Brenna and Lapillonne 2009). Some other health effects that have been associated with fish consumption and EPA/DHA include alleviation of colitis (Hudert and others 2006) and rheumatoid arthritis (Kremer 2000), reduced cognitive decline, dementia, depression, and suicides (Morris and others 2005; Sontrop and Campbell 2006; Calon and Cole 2007; van Gelder and others 2007; Hibbeln 2009), and decreased likelihood of macular degeneration (SanGiovanni and others 2007). Seafood also contains a number of vitamins (for example, A, B-complex, and D) and minerals (such as selenium, iodine, iron, and zinc) that have been linked to various health benefits (Delange and Lecomte 2000; Rayman 2000; WHO 2002). Selenium in seafood has been evaluated for its potential to exert protective effects in terms of reducing accumulation of mercury in fish (Paulsson and Lundbergh 1989) and in humans (Seppänen and others 2000). Dietary intake of selenium has also been reported to be inversely related to MeHg-induced adverse health effects in rats, although a human study into this relationship was inconclusive (Choi and others 2008; Ralston and others 2008).

In terms of dose-response relationships, meta-analyses have reported that reduced risks for CHD were found to occur with intakes of at least 250 mg/d of EPA+DHA, and maximum benefits were found with about 500 mg/d of EPA+DHA (Mozaffarian and Rimm 2006; Harris and others 2008; Harris and others 2009). With regard to neurodevelopmental benefits of seafood intake during pregnancy, one observational cohort study reported beneficial effects to children when maternal seafood intake exceeded 340 g (about 3 to 4 servings) per week (Hibbeln and others 2007) and others have reported increased infant cognition scores when maternal weekly seafood intake was 1 serving or more (Daniels and others 2004), 1.5 to 3.5 servings (Oken and others 2008a), or greater than 2 servings (Oken and others 2005; Oken and others 2008b), compared to mothers not consuming seafood.

Based on the identification and characterization of the health benefits of seafood and EPA/DHA intake, a number of health-based guidance values and nutritional recommendations have been developed by various health organizations (Table 1). Scientific committees from organizations such as the Institute of Medicine (IOM), the Food and Agriculture Organization of the United Nations (FAO)/World Health Organization (WHO), and the Dietary Guidelines Advisory Committee (DGAC) of the U.S. Department of Agriculture (USDA) and the U.S. Department of Health and Human Services (HHS) have recommended consumption of 2 servings per week of a variety of seafood (IOM 2005; DGAC 2010) or about 250 mg of EPA + DHA per day (FAO/WHO 2008; DGAC 2010). The American Heart Association (AHA) recommends that all adults consume fish (particularly fatty fish) twice a week and that patients with CHD obtain 1000 mg of EPA+DHA per day (Kris-Etherton and others 2003). Other organizations, such as the International Society for the Study of Fatty Acids and Lipids (ISSFAL 2004) and the French Food Safety Agency (AFFSA 2010) have a recommended intake of 500 mg of EPA + DHA per day for adults. The FAO/WHO expert consultation also recommended specific EPA + DHA intakes for pregnant or nursing women of 300 mg/d, at least 200 mg of which should be DHA (FAO/WHO 2008).

Table 1. Health guidance values for beneficial compounds worldwide.
Compound Guidance term Guidance value Target population Organization/reference
  1. *of which at least 200 mg/d should be DHA.

  2. 1.  (a) Healthy adults; (b) general adult population; (c) adult males and nonpregnant/nonlactating adult females.

  3. 2. Adult men.

  4. 3. Adult women.

  5. 4. (a) Adult pregnant and lactating females; (b) pregnant women; (c) pregnant and lactating women.

  6. 5. Patients with cardiovascular disease.

  7. 6. See Appendix B for abbreviations/acronyms.

DHA + EPARI500 mg/d1a ISSFAL 2004
 RDI250 mg/d1c FAO/WHO 2008
 RDI300 mg/d*4a FAO/WHO 2008
 RI1000 mg/d5AHA (Kris-Etherton and others 2003)
 800 mg EPA + DHA1bNATO (Simopoulos 1989)
DHA + EPA/seafoodRITwo, 113-g servings of seafood per week, equivalent to about 250 mg/d1bDGAC 2010
Omega-3 fatty acids from fishAI450 mg/dAllHCN (Netherlands 2006)
Long-chain omega-3 fatty acidsRDA500 mg/d 250 mg DHA and 500 mg EPA + DHA1b 4bAFFSA (AFFSA 2010)
FishTwo servings of fish per week, especiallyfatty fish1bAHA (Kris-Etherton and others 2003)
PUFA500 mg/d long-chain PUFA1bADA (Kris-Etherton 2007)
ALA: 10% EPA + DHA1.6 g/d approximately 10% EPA + DHA 1.1 g/d approximately 10% EPA + DHA2 3 IOM 2005a
n-3 PUFAs1% to 2% of energy/d1bWHO (WHO/FAO 2002)
DHA200 mg DHA4cWAPM (Koletzko and others 2007)

Hazards and negative health effects of seafood

Despite the demonstrated positive health effects of seafood, there are also several potential hazards (Table 2) that have been found in seafood, including pathogens, marine toxins, environmental pollutants, and heavy metals (Yasumoto and Murata 1993; Plessi and others 2001; Storelli and others 2003; Iwamoto and others 2010). Although the greatest risk to human health is from pathogens in seafood, this can be overcome through proper cooking, handling, and storage; therefore, negative health effects from seafood pathogens are not discussed here. The hazards that are generally considered in risk-benefit assessments are heavy metals, especially mercury, and POPs. Mercury is released into the environment from both natural and anthropogenic sources and is converted into MeHg by aquatic microorganisms (Rasmussen and others 2005). MeHg can bioaccumulate through the aquatic food chain, resulting in higher levels in large, predatory species such as sharks and swordfish, and the most common form of human exposure to MeHg is from fish consumption (Gunderson 1995). A correction factor of 0.85 can be applied to measurements to account for the proportion of MeHg versus total Hg in seafood. MeHg was first recognized as a hazard in seafood as a result of large-scale industrial poisonings in the 1950s in Minamata Bay, Japan, where recorded mercury levels in the local seafood reached 36 ppm (Harada 1995; Eto 2000). A series of 3 additional large-scale poisonings occurred in Iraq between 1955 and 1972 due to the consumption of wheat seeds coated with alkylmercury fungicide, where mercury levels in maternal hair ranged from 18 to 598 ppm (Bakir and others 1980; Marsh and others 1987; WHO 1990; Watanabe and Satoh 1996). The events of Minamata Bay and Iraq demonstrated the detrimental effects that organic mercury can have on the nervous system and the heightened sensitivity of the fetus to high levels of exposure. While the mothers were only slightly affected, they gave birth to infants with severe neurological problems (Watanabe and Satoh 1996). Many of the victims had hair mercury levels above 50 ppm and fish and shellfish showed a range from 5.61 to 35.7 ppm of Hg. Normal exposure levels to MeHg in the United States are at a much lower level; for example, a mean mercury level in hair of 0.38 ppm has been reported for U.S. women of childbearing age consuming 3 or more servings of fish per month (McDowell and others 2004), and mercury levels in most U.S. commercial fish are below 0.5 ppm (FDA 2009c).

Table 2.  –Health guidance values for contaminants worldwide.
Compound Guidance term Guidance value Target population Organization/reference
  1. aOriginally expressed on a weekly or monthly basis; bAdjusted to TDI here based on 30 d/mo; cOnly in regards to Aroclor 1016, 1248, and 1254; dEscolar, orange roughy, marlin, fresh and frozen tuna, shark, and swordfish; eFor predatory fish species; fWithdrawn in 2010, no new PTWI established; gPer gram muscle meat of fish and fishery products, excluding eel; iIn fish.

  2. 1.  (a) All; (b) general population; (c) all (with special groups in mind); (d) all potentially exposed populations.

  3. 2.  Pregnant women and women who might become pregnant within a year.

  4. 3.  Children and women of childbearing age.

  5. 4.  (a) For protection against nondevelopmental adverse effects (that is, increased cancer risk); (b) For protection of developing male reproductive system; (c) For protection of developmental effects.

  6. 5.  For noncancer effects.

  7. 6.  For neoplastic effects.

  8. 7.  See Appendix B for abbreviations/acronyms.

MeHgRfD0.1 μg/kg bw/d1cEPA 2001; Rice and others 2003
 PTDI0.2 μg/kg bw/d; 0.47 μg/kg bw/d3; 1bHC (Dabeka and others 2004)
 PTWI0.23 μg/kg bw/d*1cFAO/WHO (JECFA 2003)
 MRL0.3 μg/kg bw/d1d ASTDR 1999
 GL0.23 μg/kg bw/da; 0.47 μg/kg bw/da2; 4a SACN/COT 2004
Hg in seafoodML0.5 ppm, 1.0 ppm for six fishd1HC (2007)
 EU ML0.5 ppm; 1.0 ppme1 EC (2006)
 AL1.0 ppm1 FDA (2011)
CadmiumPTWI0.83 μg/kg of body wt/da JECFA 2010
LeadPTWI3.6 μg/kg of body wt/daf JECFA 2010
HexachlorobenzeneTDI0.17 μg/kg bw/d; 0.16 μg/kg bw/d5; 6 IPCS 1997
Dioxins and dl PCBsPTMIb2.33 pg TEQ/kg body wt/da JECFA 2001
 PTWI2 pg TEQ/kg body wt/da4b SCF 2001a; SCF 2001b
 GL2 pg TEQ/kg bw/d; 8 pg TEQ/kg bw/d4b; 4a SACN/COT 2004
Dioxins (PCDD/F)iEU ML4 pg TEQg1 EC 2006
 TDI1 pg TEQ/kg bw/d FAO/WHO (JECFA 2003)
Dioxins and dl PCBsiEU ML8 pg TEQg1 EC 2006
PCBsTDI0.13 μg/kg bw/d1HC (Health Canada 2007)
 RFDc0.02 μg/kg bw/d1cEPA 1999
 CO MRLc0.02 μg/kg bw/d1ATSDR 2009
 TDI0.02 μg/kg bw/d (0.01 for iPCBs)4cAFSSA 2007
PCBsiTLg2.0 ppm1FDA/EPA (FDA 2011)

In order to examine the effects of low-level, prenatal exposure to MeHg through dietary consumption of seafood, several large-scale epidemiological studies were conducted in the 1980s and 1990s (Grandjean and others 1997; Crump and others 1998; Davidson and others 1998; Myers and others 2003; Daniels and others 2004; Wieslaw and others 2006); however, the outcomes of these studies have been inconsistent (Spurgeon 2006). The 2 most notable are the Faroe Islands study, n= 917 (Grandjean and others 1997), and the Seychelles Child Development Study, n= 711 (Myers and others 1995; Davidson and others 1998; Myers and others 2003), which monitored families for over 10 y, but reported conflicting results. For the purpose of developing health guidance values for MeHg intake, the maternal hair-mercury concentration equivalent to a no-observed-effect level (NOEL) or benchmark dose lower confidence limit (BMDL) for neurobehavioural effects has been calculated for each study. The calculated values were 12 ppm for the Faroe Islands studies and 15.3 ppm for the Seychelles Islands (JECFA 2003). Interestingly, while both populations are known for their high-fish diets, the results of the Faroe Islands study showed some adverse effects of mercury exposure on neuropsychological dysfunctions (such as with language, attention, and memory); whereas no adverse effects were observed in the Seychelles study, with the exception of one outlier. Inhabitants of the Faroe Islands are also known to consume pilot whale, which has been found to contain high levels of mercury and polychlorinated biphenyls (PCBs) in edible tissues (Juhlshamn and others 1987; Longnecker and others 2003).

A third cohort study also evaluated scholastic and psychological tests administered to 6- and 7-y-old children in New Zealand, although their findings were highly influenced by a single child whose mother's mercury level was more than 4 times that of any other women in the study. When this single outlier was removed, the BMDL for mercury hair concentration was determined to range from 7.4 to 10 ppm, and a mercury effect was found in 6 out of 26 tests administered. However, when the outlier was included, there was no significant association between mercury hair concentration and any of the tests. The authors had no reasons to believe the outlier's test results or mother's hair mercury levels were flawed and could not conclude which set of data was nearer to the truth (Crump and others 1998). An epidemiological study conducted more recently in the United States with 135 mother-infant pairs reported that infant cognition was favorably associated with increased maternal fish intake; however, higher maternal hair mercury levels were associated with lower infant cognition (Oken and others 2005). Overall, the highest infant cognition scores were obtained when women consumed 2 or more servings of fish per week, but they had hair mercury levels ≤1.2 ppm.

In addition to neurodevelopmental effects, mercury has also been examined in terms of its adverse cardiovascular effects (Salonen and others 1995; Ahlqwist and others 1999; Hallgren and others 2001; Guallar and others 2002; Yoshizawa and others 2002; Virtanen and others 2005), but the results of these studies have also been inconsistent. For example, Guallar and others (2002) reported that toenail mercury levels were 15% higher in 684 male patients with myocardial infarction compared to controls, while Yoshizawa and others (2002) reported no significant association between mercury levels in toenails and risk of CHD for 470 male health professionals with documented cases of CHD, compared to controls. According to Stern (2005), the strongest basis for a formal risk assessment of the cardiovascular effects of MeHg is provided by the study of Salonen and others (1995) who reported a 2-fold increased risk of acute myocardial infarction and a 2.9-fold increased risk of cardiovascular death for Finnish men with hair levels of MeHg ≥ 2.0 ppm.

Focusing on management of risks associated with neurodevelopmental effects, health-based guidance values have been developed for MeHg by a number of organizations worldwide. The U.S. Environmental Protection Agency (EPA) developed a reference dose (RfD) for MeHg of 0.1 μg/kg body weight (bw)/d, which includes a 10-fold uncertainty factor (Rice and others 2003). RfD is defined as the daily intake that is likely to be without appreciable risk of deleterious effects during a lifetime. The RfD for MeHg was originally calculated based on the Iraq wheat seed poisonings (Marsh and others 1987) and was later re-assessed taking into consideration epidemiology studies from the Faroe Islands, Seychelles Islands, and New Zealand (Rice and others 2003). Other guidance values put in place by various health organization are listed in Table 2. In terms of risk management of MeHg concentrations in seafood, the U.S. Food and Drug Administration (FDA) set an action level of 0.5 ppm in 1969, which was adjusted to 1.0 ppm in 1979 based on data from the Minamata Bay poisoning disaster. This level corresponds to a daily intake of 0.5 μg/kg bw/d (NRC 2000; Burger and others 2005). Health Canada has set tolerance limits of 0.5 ppm for most fish, with the exception of 1.0 ppm for escolar, orange roughy, marlin, fresh and frozen tuna, shark and swordfish (Health Canada 2007); and the European Commission (EC) has set tolerance limits at 0.5 ppm for most fish and 1.0 ppm for predatory fish species (EC 2006).

In 1994, the U.S. FDA advised pregnant women and women of childbearing age who may become pregnant to limit consumption of shark and swordfish to no more than one meal per month due to the mercury levels in these fish (ASTDR 1999). Other consumers were advised to limit regular consumption of these species to about 200 g per week and to limit consumption of fish with mercury levels below 0.5 ppm to 400 g per week. With regard to the top 10 species consumed in the U.S., consumption advice was considered unnecessary because of the low levels of mercury in these species and because few people eat more than the suggested weekly limit of 2.2 pounds (1 kg) for the general contamination level of these species. In 2001 the U.S. FDA updated a previous advisory and singled out infants, small children, pregnant or nursing mothers, and women who may become pregnant to completely avoid shark, swordfish, king mackerel, and tilefish; while stating that, in general, consumers should limit their consumption of all fish to no more than 340 g per week. This advisory was found to reduce target consumers’ consumption by 15% to 30% or 1.4 servings per month, and it was associated with a 21.8% net drop in canned fish consumption among households with young children (Oken and others 2003; Schwarz and others 2007; Shimshack and others 2007). In 2004, the U.S. FDA released a joint advisory with the U.S. EPA that targeted pregnant women, women who might become pregnant, nursing mothers, and young children (FDA 2004a). These groups were advised to eat up to 340 g (2 average meals) of seafood per week of a variety of fish and shellfish that are lower in mercury, such as shrimp, canned light tuna, salmon, pollock, and catfish. As part of the 340 g per week, they were advised to eat only up to 170 g (1 average meal) of albacore (“white”) tuna, due to slightly higher levels of mercury in this species compared to canned light tuna. They were also advised not to eat 4 types of fish due to higher levels of mercury: shark, swordfish, king mackerel, and tilefish. Metals other than mercury such as lead, manganese, chromium, cadmium, and arsenic may also be found in seafood, although studies have shown that seafood does not seem to be a main route of exposure to these metals. The geographic source of the fish is always a factor and consuming a wide variety of seafood will reduce the risks involved with the trace amount of metals found in different species and areas (IOM 2005).

POPs, such as PCBs and dioxins, have also been identified as hazards that are present in seafood. PCBs are a group of 209 congeners that were widely used in a variety of commercial and industrial applications (Ross 2004). In the late 1960s and 1970s, concerns over the health effects of PCBs were raised due to poisoning events resulting from the consumption of PCB-contaminated food in Japan (Masuda and others 1996) and Taiwan (Ross 2004). Due to these events and the persistence of PCBs in the environment, these chemicals were banned in the United States in 1979. Dioxin is a collective term for a group of toxic chemicals with a similar mechanism of toxicity; it includes 7 congeners of polychlorinated dibenzodioxins (PCDDs) and 10 congeners of polychlorinated dibenzofurans (PCDFs). There are also 12 congeners of dioxin-like polychlorinated biphenyls (dl PCBs) that the U.S. EPA includes within the term dioxin (EPA 2010), but they can be listed separately (SCF 2001a). In Europe, indicator PCBs (iPCBs) are a set of 7 PCB congeners used to estimate total PCBs and have shown to be a good predictor of dl PCBs (Babut and others 2010). For the purposes of this paper, the term dioxin will refer to the 17 congeners of PCDDs and PCDFs, and the groups of dl PCBs and iPCBs will be referred to separately. Although the environmental levels of PCBs and dioxins have been declining, these compounds are widely distributed and current exposure levels remain a concern. PCBs and dioxins can bioaccumulate and have been characterized by the EPA as likely human carcinogens. In addition to the potential carcinogenic effects of PCBs and dioxins, noncancer effects, including changes in hormone levels and fetal development, have been observed at levels of about 10 times above the normal background exposure (EPA 2010). The most toxic dioxin congener, 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), was classified as a known human carcinogen in 1997 and accounts for about 10% of the total background dioxin risk (EPA 2010).

The toxicity of the various PCB and dioxin congeners is calculated relative to the toxicity of TCDD and is expressed as a toxic equivalent factor (TEF), as established by WHO (Schecter and others 2001). Most human exposure to dioxins and PCBs is through the diet, specifically from animal fats found in meats, seafood, and dairy products. To calculate the toxicity of a food due to PCBs and dioxins, the level of each congener in the food is multiplied by its TEF and the sum is reported as the total dioxin-like toxic equivalency, or TEQ. As shown in Table 2, limits for dioxin exposures range between 1 and 4 pg TEQ/kg bw/d for protection against developmental effects (JECFA 2001; SCF 2001b; SACN/COT 2004) and for protection against increased cancer risk, a limit has been set at 8 pg TEQ/kg bw/d (SACN/COT 2004). In general, adult men and women have daily TEQ intakes of 2.4 and 2.2 pg/kg bw, respectively. About 9% of dietary exposure is from fish and shellfish (Schecter and others 2001), primarily from fish caught in fresh waters, estuaries, and near-shore coastal waters rather than the open ocean. The EC has established a maximum level for dl PCBs and dioxins in seafood of 8 pg TEQ per g seafood (ppt TEQ) (EC 2006) and the U.S. FDA has set a tolerance level for PCBs in seafood of 2.0 ppm (FDA 2011). A review of numerous studies reporting levels of PCBs and dioxins in a variety of fish and shellfish reported ranges of 0.5 to 100 ppb for PCBs and 0.2 to 17 ppt TEQ for dioxins (Mozaffarian and Rimm 2006), while international averages of dl PCBs and dioxins in fish, cephalopods, and crustaceans have been reported to range from 0.15 to 3.0 ppt TEQ (JECFA 2001; Storelli 2008).

The greatest risk to human health from POPs occurs when fish is harvested recreationally from contaminated waters and consumed in large amounts by subsistence anglers, pregnant women, and young children. While some studies have reported an association between prenatal exposure to PCBs and dioxins and childhood neurodevelopmental problems (Jacobson and others 1990; Jacobson and Jacobson 1996; Patandin and others 1999; Grandjean and others 2001; Ribas-Fito and others 2001; Schantz and others 2003; Stewart and others 2003; Nakajima and others 2006; Stewart and others 2008), others have not found an effect (Daniels and others 2003; Gray and others 2005). For example, Grandjean and others (2001) reported an association between umbilical cord PCB concentration and deficits in 2 to 3 out of 17 neuropsychological tests among a cohort of 435 Faroese children. Interestingly, PCB-associated deficits only occurred in children that also had higher levels of mercury exposure, indicating a possible confounding effect. A cohort from the Great Lakes area in the United States was examined for the relationship between prenatal PCBs and Neonatal Behavioral Assessment Scale (NBAS) performance in babies born to women consuming Lake Ontario fish (n= 156). The authors concluded consumption of these contaminated fish is associated with lower IQ in children (Lonky and others 1996; Stewart and others 2000; Stewart and others 2003; Stewart and others 2008). On the other hand, in a series of studies published by Daniels and others (2003) and Gray and others (2005) involving over 1200 mother-child pairs, there was no relationship between prenatal PCB exposure and mental or psychomotor scores in 8-mo-old infants or IQ scores at 7 y of age.

EPA has developed a set of guidelines to assess the cancer risks of certain compounds, such as POPs, in the population (EPA 2000, 2005). When making risk assessments, the EPA assumes there is no “safe” lower threshold for carcinogenic compounds, signifying any exposure may pose some cancer risk. This cancer risk is expressed as a cancer slope factor (CSF) or cancer potency factor (CPF) with units of risk expressed in mg carcinogenic agent/kg bw/d exposure. The initial data are dose-response data acquired during epidemiological studies or chronic animal bioassays. These data are then extrapolated to represent the lower doses encountered by the general public. The EPA has only identified CSFs for compounds with enough data to justify development of a value, such as arsenic, polycyclic aromatic hydrocarbons (PAHs), PCBs, and dioxins/furans. CSFs can be calculated for both oral and inhalation exposure; as well as unit risks, for example, data specific for contaminants in a specific medium (air or water, for example) are expressed as risk per one unit of concentration of the contaminant. Cancer potency is determined by fitting available dose-response data to standard cancer risk extrapolation models. The CSF is then estimated as the slope of the linear extrapolation to the origin drawn from the 95% lower confidence limit in the low-dose region. This provides a higher estimate of risk and the actual risk can be much lower than or as low as zero. The CSFs can then be useful in calculations of daily consumption limits. Most CSFs for different compounds can be found on EPA's Integrated Risk Information System (http://www.epa.gov/IRIS/).

Qualitative Risk-Benefit Assessments

Incorporating the many risks and benefits that have been associated with seafood consumption into a risk-benefit assessment is a challenging task. All studies involving characterization of health effects have a certain degree of uncertainty and variation that must be taken into account. Also, inconsistencies among the findings of several studies that have investigated the same health endpoint further complicate the issue. In order to provide a comprehensive assessment that includes consideration of the numerous health endpoints associated with seafood, several large-scale, qualitative risk-benefit assessments have been conducted. Many of these have resulted in specific seafood consumption recommendations for the general population and for target populations. For example, at the request of the National Oceanic and Atmospheric Administration (NOAA), the Institute of Medicine (IOM) of the National Academies put together an expert committee to examine relationships between the risks and benefits of seafood in order to help consumers make informed choices (IOM 2005). The committee carried out a 4-part qualitative protocol to assess and balance the risks and benefits of seafood, involving (1) identification of the magnitude of the risks and benefits, (2) identification of risks and benefits that are important enough to be included in the balancing process, (3) evaluation of the changes in risks and benefits associated with different consumption patterns, and (4) balancing of the risks and benefits to arrive at specific consumption guidelines. In order to identify the health benefits of seafood and EPA + DHA, the committee examined numerous studies and other systematic reviews by the Agency for Health Research and Quality (AHRQ) (Balk and others 2004; Schachter and others 2004; Wang and others 2004; Schachter and others 2005). Positive health effects were examined for a number of areas, including benefits to women during and after pregnancy; duration of gestation and birth weight; infant and child development; and cardiovascular disease, cardiovascular mortality, and all-cause morbidity and mortality. Overall, the greatest evidence for health benefits was related to reduced risk of cardiovascular disease, greater duration of gestation, and improved visual and cognitive development associated with maternal seafood or EPA/DHA intake. However, the committee also noted that the average amounts of seafood consumed by the U.S. population are below levels suggested by many authoritative groups to achieve positive health effects. In terms of hazards associated with seafood consumption, the committee conducted an extensive evaluation of the health effects of MeHg and other metals, such as cadmium; POPs; microbiological agents; seafood allergens; and naturally occurring toxins. Based on their review of the evidence, the committee concluded that among the potential chemical contaminants in seafood, MeHg poses the greatest concern for adverse health effects, whereas the risk associated with POPs remains uncertain, and risks associated with microbial hazards, allergens, and toxins are persistent, yet more controllable. Due to uncertainties in terms of the ability of the committee to quantify risk-benefit tradeoffs based on the available evidence, a qualitative risk-benefit assessment was conducted. After balancing the evidence with regard to cardiovascular benefits and developmental risks and benefits of seafood, 4 target populations were identified and specific consumption guidelines were developed for these populations. Briefly, the committee stated that healthy adolescent and adult males and females (who will not become pregnant) may reduce their risk for cardiovascular disease by consuming seafood regularly, making sure they chose from a variety of seafood if consuming more than 2 servings per week. Similar guidelines were given for adult males and females who are at risk of cardiovascular disease, with the addition that this group may benefit from including seafood high in EPA and DHA. The committee stated that the other 2 target populations, females who are or may become pregnant or who are breast-feeding and children up to 12 y of age, may benefit from consuming seafood, especially selections high in EPA and DHA, and they can safely consume up to 340 g of seafood per week including up to 170 g of albacore tuna, but that they should avoid large predatory species such as shark, swordfish, tilefish, or king mackerel.

Another in-depth qualitative risk-benefit assessment of seafood consumption was conducted by Mozaffarian and Rimm (2006) from the Harvard School of Public Health and Harvard Medical School. The assessment involved examination of scientific publications, government reports, systematic reviews and meta-analyses related to 4 categories: (1) association between intake of fish or fish oils and reduced risk of cardiovascular events and mortality, (2) effects of MeHg and fish oil on early neurodevelopment, (3) association between MeHg exposure and negative cardiovascular or neurologic effects in adults, and (4) health risks of PCBs and dioxins in fish. The authors focused on human health studies from randomized trials and large prospective studies and, when possible, conducted meta-analyses to better characterize risks and benefits. In terms of cardiovascular outcomes, the assessment showed that intake of about 250 mg/d EPA + DHA, or 1 to 2 servings/wk of fish high in these fatty acids, was associated with risk reductions of 36% for CHD death and 17% for total mortality. However, at higher intakes, there was little additional risk reduction, suggesting a threshold for maximum cardiovascular benefits of 250 mg/d EPA + DHA. For developmental outcomes, the assessment revealed the importance of DHA for cognitive and visual development, but also showed the potential negative effects of low-level MeHg exposure on cognitive development. Based on these findings and dose-response relationships, the authors recommended that women of childbearing age and nursing mothers should consume 2 servings of seafood per week, with limited intake of selected species that show elevated mercury levels. The assessment of MeHg and cardiovascular outcomes in adults revealed uncertainty in terms of the health effects of low-level exposure among adults. Therefore, Mozaffarian and Rimm (2006) suggested choosing from a variety of seafood and noted that adults consuming ≥ 5 servings/wk should limit intake of high-mercury fish species. Levels of PCBs and dioxins were reported to be relatively low in fish and the health benefits of seafood intake were found to outweigh any potential adverse effects from these contaminants. However, women of childbearing age were advised to check regional risk advisories prior to consuming locally caught freshwater fish. Overall, the authors concluded that the benefits of fish intake exceed the potential risks for major health outcomes among adults, and for women of childbearing age, the benefits of modest fish intake, excepting a few selected species, also outweigh the potential risks. They also cautioned that avoidance of fish due to confusion surrounding the risks and benefits could lead to suboptimal neurodevelopment in children and thousands of additional deaths annually from CHD. Although this risk-benefit assessment has been criticized for discounting the negative health effects of MeHg on cardiovascular disease in adults (Stern 2007), studies examining this topic have reported conflicting results, as discussed in an earlier section, and according to Mozaffarian and Rimm (2006); even in studies reporting negative effects of MeHg, the net effect of seafood consumption was still positive.

In Europe, FAO and WHO held an expert consultation in January 2010 to compare the risks and benefits of seafood consumption, based on a request for scientific advice from the Codex Alimentarius Commission (JECFA 2010). The consultation examined the levels of specific nutrients and contaminants (that is, MeHg, dioxins, dl PCBs, and furans) in seafood, as well as the scientific literature regarding the risks and benefits of seafood consumption. This information was used to consider risk-benefit assessments for certain health endpoints, including sensitive populations. The consultation compared the benefits of fish consumption related to neurodevelopment and prevention of cardiovascular disease with the risks of fish consumption related to neurodevelopment, cardiovascular disease, and cancer. Overall, the consultation recognized that fish was a source of energy, protein, and other important nutrients, such as DHA and EPA, and that consumption of fish, especially oily fish, lowers the risk of CHD mortality among the general adult population. They did not find convincing evidence of CHD risks from MeHg and determined that the potential cancer risks of dioxins and dl PCBs were well below established CHD benefits. With regard to neurodevelopmental effects of omega-3 intake and MeHg exposure, the consultation concluded that in most cases maternal fish consumption lowers the risk of suboptimal neurodevelopment compared to no maternal fish consumption. Further, when maternal dioxin and dl PCB intake is below the provisional tolerable monthly intake (PTMI) of 70 pg/kg bw/mo established by the JECFA, neurodevelopmental risk was determined to be negligible; however, when intake exceeds the PTMI, the risk may no longer be negligible. Finally, the consultation outlined a series of recommendations, including an emphasis on the benefits of eating seafood in terms of reducing CHD mortality and improving neurodevelopment, as well as the development and evaluation of risk management and communication strategies related to seafood consumption that both minimize risks and maximize benefits.

In the United States, the 2010 DGAC, which was established jointly by the Secretaries of the USDA and the U.S. Department of Health and Human Services, compared the risks and benefits of seafood and recommended specific guidelines for both seafood and EPA + DHA intake. Several health endpoints were considered, including the association between EPA + DHA and risk of CHD, the association between maternal intake of EPA + DHA and health outcomes in infants, as well as an overall comparison of the risks and benefits of seafood consumption. Twenty-five studies published since 2004 were examined in terms of CHD benefits associated with seafood and EPA + DHA, including 6 systematic reviews/meta-analyses (He and others 2004b; Whelton and others 2004; König and others 2005; Mozaffarian and Rimm 2006; Wang and others 2006; Mozaffarian 2008). DGAC found consistency among these studies and made the conclusion that moderate evidence shows that about two 113-g servings per week of seafood (about 250 mg EPA + DHA per day) is associated with a reduced risk of cardiac mortality from CHD or sudden death in individuals with and without cardiovascular disease. In their evaluation of developmental benefits, the DGAC considered several expert opinions, including the Cochrane Database Systematic Review (Makrides and others 2006), American Dietetic Association (ADA) Evidence Analysis (Kaiser and Allen 2008), and the European Union Perinatal Lipid Intake Working Group assessment (Koletzko and others 2007), as well as a background paper that reviewed 23 trials (Brenna and Lapillonne 2009), and 9 additional papers discussing the effects of omega-3 fatty acids on breast milk composition and infant health outcomes. Overall, the DGAC concluded that moderate evidence indicated an association between increased maternal intake (during pregnancy and lactation) of long-chain omega-3 fatty acids, especially DHA from at least 2 servings of seafood per week, and increased levels of DHA in breast milk as well as improved infant health outcomes, such as visual acuity and cognitive development. In their overall assessment of the risks and benefits of seafood, the DGAC examined 9 studies that have explored a wide range of potential health effects along numerous health endpoints: 3 quantitative risk-benefit assessment studies (Guevel and others 2008; Sioen and others 2008; Ginsberg and Toal 2009), 4 cross-sectional studies (Rawn and others 2002; Huang and others 2006; Dewailly and others 2007; Verger and others 2008), 1 meta-analysis (Gochfeld and Burger 2005), the systematic review from Mozaffarian and Rimm (2006), and the report from IOM (2005). Based on their review of these studies, the DGAC concluded that moderate, consistent evidence suggests that the health benefits associated with 2 servings per week of cooked seafood outweigh the potential risks from MeHg and POPs exposure, even among women of childbearing age, pregnant and nursing women, and children ages 12 and under. In agreement with the IOM report, the DGAC stated that consumers can safely eat at least 340 g of a variety of seafood every week, provided that they also follow federal and local advisories and limit consumption of large predatory fish. Overall, the DGAC felt that encouragement of seafood consumption in the United States is justified, as intake levels continue to be below those recommended for health by the Committee and by the IOM.

Quantitative Risk-Benefit Assessments

Assessments based on common metrics

Risk-benefit assessments with common metrics include the use of single outcome measures, like incidences of mortality, morbidity, or exceeding/not meeting health-based guidance values. These types of assessments are easy to comprehend and have limited data needs, but they must be interpreted with caution because they usually only provide some of the information of interest (Fransen and others 2010). Further, many of these studies do not combine the risks and benefits into a net health outcome, as do composite metrics, but can be a useful step in a full risk-benefit analysis by presenting the information in the same units for comparison. Risk-benefit assessments that have been carried out for seafood using common metrics are reviewed in this section.

An early risk-benefit assessment of seafood by Anderson and Wiener (1995) used a risk tradeoff analysis method to compare the potential cancer risk of eating fish to the potential CHD risk of not eating fish. The cardiovascular risk of not eating fish was derived from the results of Kromhout and others (1985), described earlier, where consumption of 1 to 14 g fish/person/d was estimated to reduce the risk of CHD by 36% and consumption of 15 to 29 g fish/person/d was estimated to reduce the risk of CHD by 44%. The cancer-related risk of eating fish was estimated by calculating potential exposure to 6 carcinogenic compounds with an allowable concentration limit set by the FDA and a CSF determined by the EPA: PCBs, dioxins, dichlorodiphenyltrichloroethane (DDT), chlordane, dieldrin, and heptachlor. The analysis assumed that all fish consumed were contaminated with all 6 compounds at levels matching the FDA limits of 2.0 ppm (PCBs), 50 ppt (dioxins), 5.0 ppm (DDT), and 0.3 ppm (chlordane, dieldrin, and heptachlor). The CSFs associated with exposure to compounds at these levels were calculated, assuming an average body weight of 70 kg and chronic exposure over a 70-y lifespan. The results of this analysis revealed that the lifetime risk of getting cancer for someone eating 1 g of contaminated fish per day over the course of 70 y was estimated at 5.0 × 10−4 (5 in 10000) and consumption of 20 g of contaminated fish per day (1.4, 100-g servings/wk) for 70 y was estimated to increase cancer risk to 1 in 100. The overall risk for cancer for the average American is 25%, so based on these predictions consumption of 20 g/d of contaminated fish would increase the overall cancer risk to 26%. However, the cancer risk of eating 20-g fish/d in the United States based on the actual levels of contaminants in fish has been estimated to be much less (0.75 in 10000) than the risk of eating contaminated fish calculated using the previous assumptions (IOM 1991). On the other hand, at the time of this study the average total risk of CHD mortality was 35% and the average fish consumption was 15 g/d. The results of Kromhout and others indicated that reducing fish consumption from 15 to 29 g/d to 0 g/d would increase the risk of dying from CHD by about 66%, meaning that the total risk of dying from CHD would be predicted to increase from 35% to 58%. Based on the results presented here, the cardiovascular benefits of fish consumption were determined to far outweigh the carcinogenic risks, even when fish contain all 6 carcinogenic contaminants at the FDA limits and all cancers are assumed to be fatal.

Several studies have been carried out assessing the levels of organic contaminants in farmed and wild salmon and potential risks to human health (Hites and others 2004a; Hites and others 2004b; Foran and others 2005a; Huang and others 2006), with resulting recommendations to limit consumption of salmon to only 0.2 to 8 meals per month, depending on the harvest location. However, salmon is a rich source of omega-3 fatty acids and reducing consumption to these levels would be expected to concomitantly increase the risk of cardiovascular disease and mortality, as discussed above. To better understand the overall health effects of altering consumption levels of farmed and wild salmon, risk-benefit assessments have been conducted by Foran and others (2005b) and Dewailly and others (2007).

Foran and others (2005b) considered the cancer and noncancer risks associated with exposure to organic contaminants when enough salmon is consumed to provide 1 g of EPA + DHA per day. Levels of organic contaminants and omega-3 fatty acids in farmed Atlantic salmon and wild Pacific salmon were obtained from previous work (Hites and others 2004a). Risks were calculated based on an average body weight of 70 kg and continuous exposure to contaminants over a 70-y lifespan. Noncancer risk was calculated for 14 contaminants (including PCBs, hexachlorobenzene (HCB), chlordane, dieldrin, heptachlor, MeHg, and others) and was deemed acceptable when the ratio between the cumulative exposure level for all contaminants and the cumulative RfD for all contaminants was <1. The acceptable cancer risk level resulting from the cumulative analysis of 11 contaminants with CSFs (including PCBs, 2,3,7,8-TCDD, HCB, chlordane, dieldrin, heptachlor, and others) was set at 1 × 10−5, which is the midpoint of the acceptable range established by the EPA (1 × 10−4 to 1 × 10−6). At consumption levels of two 100-g servings per week, the cumulative noncancer risk from organic contaminants was predicted to remain within acceptable levels, with the greatest benefit coming from wild salmon. On the other hand, salmon consumption at these levels was associated with an increased cancer risk, with a cumulative risk of 8 × 10−5 (that is, 8 in 100000) for wild salmon and 2.4 × 10−4 for farmed salmon. Based on these results, Foran and others (2005b) made recommendations to limit consumption of farmed and wild salmon to 4 or fewer meals per month, depending on the source of the fish. However, when the health outcomes of consuming 1 g EPA + DHA per day were calculated, the number of lives that would be saved from CHD mortality was reported to be about 7100 per 100000 people, which outweighs the potential risk of cancer by a factor of 300 (farmed salmon) to 900 (wild salmon). Additional age-adjusted analysis by Mozaffarian and Rimm (2006) indicated that the CHD benefits outweigh the cancer risks by 100- to 370-fold for farmed salmon and 300- to more than 1000-fold for wild salmon. These factors may be further increased if salmon is consumed at levels that provide about 250 mg EPA + DHA/d (that is, about 150 g of wild salmon per week or 100 g of farmed salmon every 2 wk), which would be predicted to provide similar protection against CHD mortality as 1 g EPA + DHA/d while reducing lifetime cancer risk from salmon consumption by about 75%. Further, the levels of organic contaminants that form the basis of the risk calculations included skin and disregarded the losses of PCBs that can occur during cooking, so the actual risks of eating cooked salmon are predicted to be less than that calculated by Foran (Santerre 2010).

The CSF method used by Foran and others (2005b) to calculate cumulative cancer risk has been criticized as not being appropriate for the type of carcinogens found in salmon (that is, epigenetic carcinogens) because it assumes a linear, nonthreshold relationship between risk and low-dose exposure (Dewailly and others 2007). Instead, Dewailly and others used health-based guidance values for MeHg, PCBs, and dioxins to examine the reproductive and developmental risks associated with consumption of enough salmon to provide 500 mg EPA + DHA per day. Atlantic salmon and rainbow trout samples were collected in Quebec, Canada, and levels of contaminants and fatty acids were analyzed in skinless, raw fillets with all subcutaneous or mesenteric fat removed. The levels of EPA + DHA determined for farmed salmon in this study were about 4-fold less than those reported by Foran and others (2005b), and the results indicated that daily intake of 58 to 69 g (or four to five, 100-g servings per week) of farmed salmon or trout would provide 500 mg EPA + DHA. For risk-benefit assessment, the authors considered the contaminant exposure associated with consumption of two 180-g meals per week of farmed Atlantic salmon, which provides about 440 mg EPA + DHA/d. Calculations were carried out based on 20- to 39-y-old women with an average body weight of 60 kg. The predicted exposure to contaminants at this consumption rate was below the tolerable intakes established in Canada and by the FAO/WHO in all cases, with levels of 0.015 μg/kg bw/d for MeHg, 0.012 μg/kg bw/d for PCBs, and 0.070 pg TEQ/kg bw/d. Although not all dl PCBs were measured in this study, total TEQ intake was estimated to be 0.28 pg TEQ/kg bw/d, which was based on the assumption that dioxins make up about 25% of the total TEQ exposure. This predicted exposure level is well below the FAO/WHO tolerable intake of 2.33 pg TEQ/kg body wt/d. Farmed trout showed similar (for MeHg) or lower (for PCBs or dioxins) levels of contaminants as compared to farmed salmon and consumption of 360 g/wk would not be expected to result in excessive exposure. Overall, the authors concluded that two 180-g meals per week of farmed salmon or trout available in North American markets would be expected to provide sufficient levels of EPA + DHA without concern over the putative health risks.

Combining exposure assessments with health-based guidance values. A number of studies have assessed the risks and benefits of seafood by comparing exposure assessments with health-based guidance values for seafood (for example, TDI, RfD, or recommended intake). Although this type of study does not allow for a full quantitative risk-benefit assessment, it can serve as a valuable first step in identifying whether a complete assessment is warranted (Fransen and others 2010). These studies are generally carried out using either a deterministic (“worst case” scenario) or probabilistic approach (Cardoso and others 2010). The latter approach allows for an estimate of the probability of the target population that is at risk of either exceeding a maximum safe limit or not reaching a recommended intake level. The modeling of probability distributions takes into account the variability of the data and this approach has been increasingly used to assess contaminants in foods. It is important to note that probability estimates are highly dependent on the tail behavior of the distributions, since health-based guidance values tend to be at the upper or lower range of most individual intakes, and the statistical tools used for these assessments must be highly rigorous and reliable (Cardoso and others 2010). Some important variables to take into account in both deterministic and probabilistic studies are the food intake data and the levels of contaminants and nutrients in these foods. Food intake data are generally obtained through food consumption surveys such as food frequency questionnaires, 24-h recall dietary surveys, and one- to seven-day food diaries (Barlow and others 2010), while contaminants and nutrients in foods are generally obtained through databases, such as the one described by Sioen and others (2007a) which pools data from different publications. However, seafood is a complex food group with many different categories and species, and food intake data are generally not specific with regard to the seafood species consumed. Because various seafood species are known to contain different levels of nutrients and contaminants, exposure levels estimated for seafood in these types of assessments may be associated with a high degree of uncertainty. Studies that have compared exposure assessments with health-based guidance values to assess risks and benefits have been carried out specifically for consumers in countries such as France (Crépet and others 2005; Leblanc and others 2006; Verger and others 2008; Pouzaud and others 2009), Spain (Domingo and others 2007a), The Netherlands (van der Voet and others 2007), and Belgium (Sioen and others 2008), as well as on a broader scale for consumers across Europe (Cardoso and others 2010) and worldwide (Sioen and others 2009).

Several studies have been devoted to risk-benefit assessments for French consumers living in coastal regions with high fish consumption. One such assessment, called the CALIPSO study, was delegated by the General Food Directorate to the French Institute for Agronomy Research (INRA) (Leblanc and others 2006). The CALIPSO study examined dietary patterns among some 1000 French consumers living in 4 coastal regions with high seafood consumption, evaluated blood and urinary biomarkers associated with nutrient and contaminant intake for about half of the study participants, and determined levels of nutrients (omega-3 fatty acids) and contaminants (6 trace elements and 3 categories of POPs) in a variety of seafood sampled in the study regions. The study participants were limited to adults that consumed seafood at least twice per week and consumption data were obtained with a food frequency questionnaire that listed 82 fishes, mollusks, crustaceans, and seafood-based dishes. Omega-3 fatty acid intakes were compared to the French Recommended Daily Allowance (RDA) of 400 to 500 mg/d for adults and exposure levels to trace elements and POPs were compared to provisional tolerable intakes established by the JECFA, when available. The results indicated that consumption of at least 2 servings of fish, including some oily fish, per week would allow consumers to obtain the RDA of omega-3 fatty acids. Male and female study participants in the age range of 18 to 64 y consumed a weekly average of about 630 to 640 g of fresh and frozen fish, 260 to 270 g of mollusks and crustaceans, and about 270 to 310 g of other types of seafood, making total seafood consumption equivalent to about twelve, 100-g servings per week. EPA + DHA intake exceeded 500 mg/d for 84% of the study participants, with an average level of 1240 ± 960 mg/d (Bemrah and others 2008). However, most of the types of seafood that contributed strongly to omega-3 fatty acid intake also accounted for greatest exposure to POPs, particularly salmon, mackerel, and sardine. In terms of risks, the study found that only individuals in the highest consumption group had a nonnegligible risk of exceeding the maximum limits for MeHg, Cd, dioxins, and PCBs. The average exposure to POPs among all study participants was 18.7 ± 19.6 pg TEQ/kg bw/wk for dioxins and dl PCBs, 0.04 ± 0.06 μg/kg bw/wk for iPCBs, and 2.2 ± 1.8 ng/kg bw/d for PBDEs. Study participants had a 39% probability of exceeding the provisional tolerable weekly intake (PTWI) for dioxins and dl PCBs and 72% probability of exceeding the PTWI for iPCBs. The average exposure level to MeHg from seafood was 1.5 ± 1.2 MeHg/kg bw/wk, with some 34% of the study participants exceeding the PTWI established by JECFA (1.6 μg MeHg/kg bw/wk). None of the study participants exceeded the PTWIs established for organic tin, arsenic, and lead, and only 8.5% exceeded that established for Cd. In terms of blood levels of these trace elements, most subjects (94% to 97%) had levels at or below the standard level, and women of childbearing age had average MeHg levels of 2.3 to 3.4 μg/L, which is well below the highest exposure level at which adverse effects do not occur to the fetus (56 μg/L). Even individuals consuming up to 4.5 kg seafood per week with predicted exposure levels of 9.6 μg/kg bw/wk had maximum blood MeHg levels of 18 μg/L. When blood levels were converted to weekly exposure for women of childbearing age, the data indicated an average exposure of 0.4 ± 0.3 μg/kg bw/wk, as compared to a predicted average exposure of 1.3 ± 0.9 μg/kg bw/wk based on consumption and contamination data for the same group of women. Overall, based on the study results, the authors concluded that the general population should consume at least 2 servings of fish, especially oily fish, per week and pregnant and nursing women should limit consumption of predatory fish to once per week.

Other studies conducted among French consumers have reported that women of childbearing age had a 3% to 5% probability of exceeding the PTWI for MeHg, based on data from exposure assessments (Crépet and others 2005; Verger and others 2007). In order to examine methods for reducing MeHg exposure, Crépet and others (2005) assessed the probabilistic effects of 5 different risk management scenarios (assuming 100% compliance): (1) no change in consumption patterns, (2) remove all predatory fish above 1.0 ppm MeHg and all other fish above 0.5 ppm MeHg from the market, (3) remove all fish exceeding 0.5 ppm MeHg from the market, (4) remove 12 species of predatory fish from the market, or (5) restrict consumption of predatory fish to an exact number of portions per week. Current fish consumption and exposure levels were obtained from a previous survey (French INCA survey) that used a 7-d food log obtained from a nationally representative sample of the French population. Crépet and others (2005) utilized data from 1945 male and female adults and 848 children within this data set, combined with mercury levels reported in previous studies for 89 individual seafood items. Average consumption of seafood varied for each age group, with children consuming 174 g/wk and adults consuming 285 g/wk. Mean body weights were 19 kg for children ages 3 to 6 y, 29 kg for children ages 7 to 10 y, and 58 kg for women of childbearing age. Based on these data, under scenario 1 the probability of exceeding the PTWI for MeHg was 4.4% for women of childbearing age and 6.7% for children (12.6% for ages 3 to 6 y; 5.0% for ages 7 to 10 y). Scenarios 2 and 3 did not significantly reduce exposure levels among children, but scenario 3 (removing all fish above 0.5 ppm) did significantly reduce probability of MeHg exposure among women of childbearing age to 0.6%. In scenario 4, where 12 predatory species are removed from the market, the authors reported significant reductions in MeHg exposure for women of childbearing age (0% probability of exceeding the PTWI for MeHg) and children ages 3 to 6 y (2.8% probability), but not among children aged 7 to 10 y (1.5% probability). Under scenario 5, the authors suggested an exact number of portions of predatory fish per week that would allow target groups to remain under the PTWI for MeHg. For example, a recommendation that women of childbearing age limit consumption to two, 170-g portions (or 340 g) of predatory fish per week or to 255 g per week if also consuming nonpredatory fish. Overall, the authors suggested that risk management options that provide advice on food consumption, such as scenario 5, are more efficient compared to additional restrictions for MeHg levels in fish. However, a later study examining the effects of a risk-benefit advisory among French consumers reported that the advisory did not lead to a decrease in predatory fish consumption, but it did result in a significant decrease in overall fish consumption (Verger and others 2007).

A subsequent study compared the exposure levels of PCBs and dioxins to intake levels of omega-3 fatty acids among 401 French fish consumers in western coastal areas (Verger and others 2008). The consumption data used in this study were obtained previously in the fish advisory study, which utilized a 1-mo food diary detailing types and frequency of seafood consumed for 195 men and 206 women (Verger and others 2007). The men had an average body weight of 75 kg and ate seafood 3.3 times per week, while the women had an average body weight of 62 kg and ate seafood 2.9 times per week. Exposure levels were calculated using data from the French Ministry of Agriculture and Fisheries. Overall, 20% to 30% of the target population was estimated to exceed the PTWI established by the Scientific Committee on Food (SCF) for dioxins and dl PCBs of 14 pg TEQ/kg bw/wk. About 25% of the total exposure was from PCDDs and PCDFs and the remaining 75% was from dl PCBs. On the benefit side, about 60% of the target population obtained the recommended intake of 500 mg/d long-chain omega-3 fatty acids. When risks and benefits were compared, only 41% of the study participants had an optimal balance of meeting the RDA of omega-3 fatty acids while remaining below the PTWI for dioxins and dl PCBs: 19% of individuals met the RDA but exceeded the PTWI, while another 38% of individuals were below the PTWI but were also below the RDA. The authors concluded that consuming the RDA for omega-3s of 500 mg/d through seafood consumption was compatible with the threshold for dioxins and dl PCBs, but that consumers with omega-3 intakes above 1500 mg/d from seafood consumption were likely to also be exceeding the PTWI for dioxins and dl PCBs.

Pouzaud and others (2009) assessed seafood consumption patterns and intake of MeHg and omega-3 fatty acids among 161 pregnant French women living in a coastal region with high-fish consumption. The authors used the food frequency questionnaire from the CALIPSO study to obtain seafood consumption data for women at both 12 and 32 wk of pregnancy. Portion sizes were estimated based on a catalog of photos presented to the study participant. At both time points, hair samples were obtained for MeHg testing and body weights were recorded. At week 12, participants had a mean seafood consumption of 322 g/wk and an average body weight of 60 kg, compared to a mean seafood consumption of 309 g/wk and average body weight of 73 kg at week 32. The mean dietary exposure to MeHg from seafood was not significantly different at the 2 time points, with an overall range of 0.6 to 0.7 μg/kg bw/wk. There was also no significant difference in the hair MeHg concentrations across the 2 time points, which ranged from 0.1 to 3.7 ppm, with a mean of 0.8 ppm. Overall, about 5% of the women were exceeding the PTWI for MeHg, similar to results of previous studies on women of childbearing age (Crépet and others 2005; Verger and others 2007), and about 50% of the women were not obtaining the RDA of 500 mg/d for long-chain omega-3 PUFAs. The authors used a cluster analysis tool to group the study participants into 5 different categories related to fish consumption and exposure levels, and found that only women consuming a high proportion of fatty fish meet the RDA for omega-3 fatty acids without exceeding the PTWI for MeHg.

Domingo and others (2007a) estimated dietary exposure to DHA + EPA and chemical contaminants among Spanish consumers. The authors measured fatty acids, metals (Hg, Cd, Pb), and organic pollutants (dioxins, dl PCBs, polybrominated diphenyl ethers (PBDEs), polychlorinated diphenyl ethers (PCDEs), HCB, polychlorinated naphthalene (PCNs), and (PAHs) in the edible portions of the top 14 species consumed in Spain. Daily consumption rates of these 14 species were calculated for a 70-kg male based on consumption data obtained previously and a standard meal size of 227 g. The average EPA + DHA intake was determined to be 244 mg/d, which is very close to the level recommended by FAO/WHO (250 mg/d). The estimated intakes of total Hg (0.14 μg/kg bw/d), Cd (0.02 μg/kg bw/d), and Pb (0.03 μg/kg bw/d) were all below the provisional tolerable intakes established by the FAO/WHO for these compounds (Table 2). When the correction factor of 0.85 is applied to the total Hg intake, the MeHg intake can be estimated at 0.12 μg/kg bw/d, which is below the PTWI but slightly above the RfD established by the U.S. EPA. The estimated intakes of dioxins and dl PCBs (0.54 pg TEQ/kg bw/d) as well as HCB (0.16 ng/kg bw/d) were all below the PTWIs established by FAO/WHO for noncarcinogenic effects. The total intake of 7 carcinogenic PAHs was associated with an increased cancer risk of 0.27 × 10−6 (that is, 2.7 incidences of cancer per 10000000 people) resulting from chronic exposure over a 70-y life span, based on EPA CSFs. Tolerable intake limits have not been established for PBDEs, PCDEs, and PCNs, which had estimated exposure levels of 0.30, 0.56, and 0.02 ng/kg bw/d, respectively. The authors did not report the percentage of the target population that may be at risk from excess exposure to contaminants or deficient intake of EPA + DHA. In order to help consumers remain below exposure limits for carcinogenic and noncarcinogenic effects, the authors developed recommended monthly fish consumption guidelines for the top 14 fish in Spain. They determined the greatest noncarcinogenic risk to be from MeHg exposure, and recommended limiting consumption of tuna to 2 meals per month and swordfish to 0.5 meals per month. The greatest carcinogenic risk was determined to be from PAHs and dioxins. To remain below PAH exposure limits, recommended consumption levels were calculated to be between 0.5 (clam/mussel/shrimp) and 4 (hake/red mullet/sole/cuttlefish/squid) meals per month, depending on the species. To remain below dioxin exposure limits recommended consumption levels were calculated to be between 1 (red mullet) and 16 (hake or cuttlefish) meals per month, depending on fish species. However, the authors did not compare the risks of reducing fish consumption to these levels in terms of the concomitant decreased intake of EPA + DHA that would occur and the subsequent increases in risk of cardiovascular disease and mortality. In a companion paper, Domingo and others (2007b) presented an interactive risk-benefit online tool that allows the consumer to input their weight, meal size, and consumption frequency in order to calculate their intake of EPA + DHA and exposure to metals and organic pollutants from the 14 seafood types examined above.

A study from The Netherlands reported the development of a probabilistic model to calculate simultaneous exposure to multiple compounds from food and to predict different dietary scenarios (van der Voet and others 2007). The model was used to assess long-term intake of EPA + DHA, dioxins, and dl PCBs from a total diet perspective, as well as predict the effects of replacing other types of food in the diet with seafood. Dietary patterns were derived from the Dutch National Food Consumption Survey of 1997/1998, in which body weight and food intake were recorded for 6250 Dutch individuals using a 2-d food diary that included amount and frequency of consumption. The authors considered 18 food types in the model, 11 of which were fish/shellfish, and combined levels of dioxins, dl PCBs, and EPA + DHA in these foods with the dietary information to estimate total intake of these compounds. In order to compare the FAO/WHO TDI for dioxins and dl PCBs with the Health Council of The Netherlands adequate intake (AI) for EPA + DHA (450 mg/d), the AI was expressed in terms of body weight for a 65-kg individual (7 mg/kg bw/d). Based on 500 random samples from 10000 Monte Carlo simulations, the results of the dietary analysis showed that in most cases (98% to 99%) the EPA + DHA intake was below the body weight-adjusted AI and the dioxin and dl PCB exposure was below the TDI established by FAO/WHO. Only about 2% of the population was above the body weight-adjusted AI and below the TDI for dioxins and dl PCBs. In addition to calculating the percentage of the sample population that is meeting health-based guidelines, the authors also examined the probable effects of replacing beef and pork consumption with salmon, eel, or a mixture of fatty fish (salmon, eel, herring, and mackerel) at levels of 10%, 25%, 50%, and 100%. The base rate for frequency of fatty fish consumption reported in the food diaries was 5.3%, compared to 68% for beef, 74% for pork, and 27% for chicken. Overall, the best scenario in terms of meeting health-based guidance values was found to be substitution with 10% to 25% salmon or a mix of fatty fish. At higher percentages (50% to 100%), there was a lower frequency of maintaining dioxin and dl PCB levels below the TDI. At 10% replacement with these fish categories, about 50% of the population reached the AI for EPA +DHA and only 1.3% of the population exceeded the dioxin and dl PCB limit, whereas at 25% replacement, more than 90% of the sample population was able to reach the AI for EPA + DHA, while less than 5% were predicted to exceed the limits for dioxins and dl PCBs. On the other hand, substitution of beef and pork with 25% eel was predicted to result in about 99% of the population reaching the AI, but also about 11% would be exceeding the TDI for dioxins and dl PCBs. The authors pointed out that this analysis was meant to illustrate the use of the statistical model, and that a more complete analysis should be carried out that considers uncertainties, alternative data sets, and additional dietary scenarios. The use of a total diet model, as presented here, allows for a better overall picture of the dietary intake of certain compounds and food replacement scenarios may be useful in developing risk management and communication strategies.

Risk-benefit assessment in Belgium was carried out by Sioen and others (2008) using a probabilistic model to assess simultaneous exposure to PBDEs and omega-3 fatty acids exclusively due to fish consumption. Consumption data for some 800 Belgian fish consumers representative of the Belgian adult population with respect to age and region was obtained from a SEAFOODplus food frequency questionnaire in 2004 and body weights were incorporated based on previously determined age and sex distributions for the Belgian population. These individuals consumed an average of 216 ± 204 g seafood per week, as compared to the general Belgian population, which consumes an average of 168 g/wk. Based on a 100 to 150 g serving size, these levels are similar to the recommendations of the Belgian Health Council to consume fish 1 to 2 times per week. Levels of EPA + DHA and PBDEs were obtained for 10 fish commonly available on the Belgian market (for example, cod, salmon, tuna, saithe, and sole) using previously published data and exposure levels for 4 different dietary scenarios were calculated: (1) base consumption (216 g/wk of a variety of 10 fish), (2) consumption of 150 g/wk of cod (lean fish) and 150 g/wk of salmon (fatty fish), (3) consumption of 300 g/wk salmon, and (4) consumption of 150 g/wk salmon and 150 g/wk herring (also a fatty fish). Monte Carlo simulations were used to estimate the variability of the intakes in terms of consumption, body weight, and concentration of the contaminants and nutrients in fish. EPA + DHA intake was adjusted for body weight and was compared to a health-based guidance value of 9.7 mg/kg bw/d (derived from a dietary reference intake of 681 mg/d for a 70-kg individual consuming 2046 kcal/d). The results of the analysis revealed that scenario 1 was associated with the lowest mean intakes of both EPA + DHA (3.54 mg/kg bw/d) and PBDEs (0.85 ng/kg bw/d), while consumption of 2 servings per week of salmon in scenario 3 allowed for the highest ratio of EPA + DHA (11.9 mg/kg bw/d) to PBDEs (1.28 ng/kg bw/d). The replacement of 1 serving of salmon with herring in scenario 4 also provided elevated levels of EPA + DHA (9.6 mg/kg bw/d) compared to scenarios 1 and 2, but led to slightly higher intake of PBDE (2.4 ng/kg bw/d). While there is no established tolerable intake for PBDEs, the lowest observed adverse effect level associated with this group of compounds has been reported to be 0.6 mg/kg bw for the penta-BDEs (Darnerud 2003; Siddiqi and others 2003). Overall, increased fish consumption was associated with increased intake of EPA + DHA and PBDEs, with the greatest benefit: risk ratio being from consumption of 2 servings of salmon per week.

In a subsequent study, Sioen and others (2009) used the probabilistic approach described above to assess exposure to several nutrients (EPA + DHA, vitamin D, iodine) and contaminants (MeHg, iPCBs, dioxins + dl PCBs) on a global scale. The authors used consumption data for 7 different seafood categories gathered by the Global Environment Monitoring System (WHO 2007), which reports average food consumption for 13 regional "cluster" diets representing 183 countries worldwide. General levels of nutrients and contaminants were obtained from databases developed in previous studies (Sioen and others 2007a, b) and an intake assessment was performed for all 13 cluster diets with probability distributions fitted for each seafood category, nutrient, and contaminant. Exposure levels were based on the general adult population, with a mean body weight of 60 kg for most regions and 55 kg for individuals from Asian countries. The highest seafood intake was reported for Cluster L, which included Japan, Korea, Philippines, Madagascar, and others, with 69.0 g/person/d, followed by Cluster F (the Nordic-Baltic countries; 49.2 g/person/d), and Cluster G (Afghanistan, China, India, Thailand, and others; 45.0 g/person/d). The 2 clusters with the highest seafood intake had the highest intakes of EPA + DHA (400 to 600 mg/person/d), iodine (30 to 40 μg/person/d), and vitamin D (3 to 4 μg/person/d). Differences in the dietary patterns between regions were reflected in the types of contaminants present: Clusters L and F consume relatively high levels of pelagic fish and had the highest exposure to MeHg (about 200 to 300 ng/kg bw/d), whereas Cluster G, which consumed a greater proportion of freshwater fish, cephalopods, crustaceans, and mollusks, had lower exposure to MeHg (about 100 ng/kg bw/d) but the highest exposure to iPCBs (about 60 ng/kg bw/d) and dioxins + dl PCBs (about 3.3 pg TEQ/kg bw/d). In order to combine data for nutrients and contaminants, the intake of EPA + DHA was divided by the dietary reference intake (DRI) of 500 mg/d and graphed against the exposure to either MeHg or dioxins + dl PCBs divided by the TDIs established by the JECFA. These plots revealed that most clusters were not exposed to contaminants above the tolerable exposure levels; however, they also were not obtaining sufficient EPA + DHA to meet the DRI, with intake levels of about 100 to 300 mg/d. The only cluster diet (Cluster L) that obtained 500 mg/d of EPA + DHA also exceeded the TDIs for MeHg and dioxins + dl PCBs. Cluster G exceeded the TDI for dioxins + dl PCBs and had a daily EPA + DHA intake of about 200 to 300 mg, while cluster F was just below the DRI for EPA + DHA and had exposure at levels around the TDIs for both MeHg and dioxins + dl PCBs. When the authors compared the mean exposure levels for each cluster to the tolerable intakes for MeHg and dioxins + dl PCBs established by the Scientific Advisory Committee on Nutrition/Committee on Toxicity out of the United Kingdom (SACN/COT) to protect against nondevelopmental health problems (Table 2), none of the clusters exceeded the guidance values. There are several uncertainties of this study that could influence the estimated nutrient and contaminant exposure levels. For example, food consumption data were gathered by dividing food availability for a given country by the total population and it tends to overestimate consumption by about 15%. Also, the nutrient and contaminant data were based on seafood tested in Europe or North America and do not represent regional seafood consumed by some cluster diets. In conclusion, the authors noted that the benefits outweigh the risks of seafood consumption when the focus is on nondevelopmental effects and they called for a more in-depth international study that would include local nutrient and contaminant concentration data.

Cardoso and others (2010) examined seafood consumption patterns across 8 European countries (Germany, France, United Kingdom, Italy, Spain, The Netherlands, Portugal, and Iceland) and calculated the probability of exceeding the tolerable intake for MeHg or being deficient in EPA + DHA for consumers in each country. The authors took into account the 5 most-consumed types of seafood for each country and calculated per capita weekly consumption, assuming that two-thirds of the seafood weight was edible and average body weight was 60 kg. Because detailed consumption surveys were not used, log-normal distributions were constructed to reflect seafood consumption patterns among different consumers, including individuals that do not regularly eat seafood, and the values of each distribution curve were randomly sampled using a sample size of 10000 with the Monte Carlo method. As was the case with Sioen and others (2009), the nutrient and contaminant data were not representative of the entire study population, but rather were based on seafood collected in Portugal. The probabilities of exceeding the weekly reference intakes for MeHg and EPA + DHA from consumption of each seafood species were calculated using a tail-estimation estimator for most cases and a plug-in estimator for the few cases with high probabilities. Total per capita seafood consumption for the 8 countries ranged from 140 g/wk in the United Kingdom to 630 g/wk in Iceland, while consumption of the top 5 species examined in this study for each country ranged from 80 g/wk in the United Kingdom to 390 g/wk in Iceland. Based on total seafood consumption, the probability of exceeding 500 mg/d of EPA + DHA was estimated at 0.3% for the United Kingdom, 2.0% for Italy, 12.4% for Germany and The Netherlands, 20.3% to 24.2% for France and Iceland, 61.1% for Spain, and 66.0% for Portugal. Although Spain and Portugal consume less per capita seafood (210 and 290 g/wk, respectively) than Iceland, these 2 countries include sardines among the top 5 species, which are a rich source of EPA + DHA. The probability of exceeding the PTWI established by JECFA for MeHg based on total seafood consumption was below 5% for most countries and reached 6.7% for Portugal and 9.6% for Iceland. Among the top 5 fish consumed in Iceland, the highest probabilities of exceeding the PTWI for MeHg based on a single fish species were with tuna (0.50%) and haddock (0.34%), which had consumption levels of 69 and 158 g/person/wk, respectively. However, exceeding the PTWI based on exclusive consumption of either of these fish would require about five 100-g servings of tuna per week (7 times the current consumption levels) or fourteen 100-g servings of haddock (9.2 times the current consumption levels). The results of this study highlight the fact that selecting fish high in EPA + DHA and low in MeHg will improve the benefit:risk ratio related to seafood consumption.

Assessments using combined risk-benefit dose-response models. In 2009, the U.S. FDA issued 2 draft reports examining the health outcomes of seafood consumption with the purpose of providing additional scientific information to help address concerns over risks and benefits of commercial seafood in the United States (FDA 2009a, b). In one report, the beneficial effects of seafood consumption and omega-3 fatty acids for certain neurodevelopmental and cardiovascular endpoints were summarized (FDA 2009b) and in the other report, a quantitative risk-benefit assessment was conducted for seafood consumption (FDA 2009a). The risk-benefit assessment was focused on 3 health endpoints: (1) fetal neurodevelopment, (2) risk of fatal CHD, and (3) risk of fatal stroke. The consideration of both the benefits and the risks of seafood in the same quantitative analysis was a novel approach for the FDA, which has historically focused on quantifying the risk but not the countervailing benefits of a particular food. Current levels of U.S. fish consumption (that is, amounts and species) were estimated based on 3 sources of data: a 3-d food survey conducted by the U.S. Department of Agriculture between 1989 and 1991 (USDA 1993), the 30-d National Health and Nutrition Examination Survey (NHANES) conducted in 2001 to 2002 (CDC 2004), and market share data on consumable commercial fish in 2005 from the National Marine Fisheries Service (NMFS). This information was combined with MeHg concentrations in different fish species, as reported by FDA, EPA, and NMFS, to calculate current levels of MeHg exposure from fish consumption. The negative effects of MeHg on fetal neurodevelopment were modeled based on verbal development measurements primarily in children from the Iraq MeHg wheat poisoning event (Marsh and others 1987), with some data from the Seychelles Islands study (Myers and others 1995), while dose-response relationships regarding the positive effects of fish consumption were modeled using data from the UK cognitive development study (Daniels and others 2004). The effects of prenatal MeHg exposure and fish consumption were combined to estimate a net effect of IQ size equivalents in offspring using several hypothetical dietary scenarios involving women of childbearing age (15 to 45 y). IQ size equivalents are IQ points based on Z-Score conversions, which are statistical tools that measure the size of an effect and facilitate the comparison of results from different models. At current consumption levels (about 5% of women eating ≥340 g of fish/week, 95% of women eating <340 g/wk), there is an estimated net neurodevelopmental benefit equivalent to 0.225 IQ point per child, with 99% of the population likely to have a net benefit on IQ size equivalents (in excess of 1 IQ point per child for about 5% of population), 0.9% of the population likely to have no net effect, and 0.1% of the population likely to experience a net negative effect, equivalent to about 0.04 IQ point. In a scenario where women do not change their current consumption amounts, but instead eat only fish low in MeHg (≤0.12 ppm), an increase equivalent to 0.02 IQ point was predicted. Another scenario in which 100% of women were eating exactly 340 g of fish per week resulted in a net benefit equivalent to 0.57 IQ point. On the other hand, if women that are currently eating more than 340 g/wk were to decrease their consumption to this level, a net decrease equivalent to IQ point of 0.01 was predicted. Overall, these results indicate the greatest net neurodevelopmental benefit for pregnant women with increased fish consumption, especially fish that are low in MeHg.

Cardiovascular effects in the FDA risk-benefit assessment (FDA 2009a) were assessed using 2 types of models (meta-analysis and pooled analysis) that were developed based on studies that reported the effects of fish consumption, but not omega-3 fatty acids or MeHg, on CHD or stroke fatalities. The CHD meta-analysis model was based on the meta-analysis conducted by He and others (2004b), while the CHD-pooled analysis model included the studies in this meta-analysis as well as 3 additional studies published later (Folsom and Demissie 2004; Nakamura and others 2005; Iso and others 2006). The stroke meta-analysis model was based on another meta-analysis conducted by Bouzan and others (2005). The stroke-pooled analysis model utilized this meta-analysis as well as studies by Mozaffarian and others (2005), Nakamura and others (2005), and 3 additional studies that were analyzed by He and others (2004a). Based on the central estimates of these models, current levels of fish consumption were estimated to be averting approximately 31000 (meta-analysis model) to 40000 (pooled analysis model) deaths per year from CHD and approximately 22000 (meta-analysis) to 25000 (pooled analysis) deaths per year from stroke. Results of dietary scenarios suggested that if all women of childbearing age ate 340 g of fish per week, there would be a predicted decrease of approximately 250 (meta-analyses) to 340 (pooled analyses) deaths per year from CHD and stroke. If there were a 10% decrease in the amount of fish consumed by adult men and older women (> 46 y), there would be a predicted increase of approximately 3500 (pooled analyses) to 4000 (meta-analyses) deaths per year from CHD and stroke; and if all adult men and women were to increase their fish consumption by 50%, there would be a predicted decrease of approximately 11000 (pooled analyses) to 18000 (meta-analyses) deaths per year from CHD and stroke. However, it should be noted that the confidence intervals for the pooled analyses models are, by design, wider than those of the meta-analyses models and therefore they did include a small possibility that current fish consumption is associated with deaths in each age/gender category. Nevertheless, the bulk of the probability distribution did indicate a beneficial effect of fish consumption, making it more likely than not that increased fish consumption leads to a decrease in cardiovascular mortality.

The approach of combining the risks and benefits into a net health effect for a given endpoint was also used in a study examining potential development of species-specific fish consumption advice (Ginsberg and Toal 2009). The concentrations of MeHg and omega-3 fatty acids in 16 species of fish commonly available in the state of Connecticut, U.S.A, were used to predict the developmental and cardiovascular health effects associated with each fish. The cardiovascular dose-response relationship was developed based on the results of Guallar and others (2002), who found that the relative risk for a 1st myocardial infarction increased by 23% per 1 ppm hair mercury and on the findings of Mozaffarian and Rimm (2006), which showed that increasing EPA + DHA intake from 100 to 250 mg/d was associated with a 14.6% decrease in the risk of CHD death. Although these endpoints are a measure of cardiovascular health and they were compared equally in this risk-benefit assessment, CHD mortality includes sudden death and death from myocardial infarction, whereas a first myocardial infarction is not necessarily fatal. The dose-response relationship for neurodevelopmental health was based on the study of Oken and others (2005), who reported a 2.0-point increase in infant visual recognition memory (VRM) score per 100 mg fish oil/d and a 7.5-point decrease in VRM score per 1 ppm hair mercury. However, it should be noted that Oken and others (2005) only observed adverse effects among participants that had hair mercury levels above 1.2 ppm, and the maximum level of hair mercury reported was 2.4 ppm. Interestingly, Ginsberg and Toal (2009) incorporated a threshold of 0.51 ppm hair mercury into the cardiovascular dose-response model as the level below which no adverse effects were evident, but the hair mercury threshold for adverse cognitive effects observed by Oken and others (2005) was not incorporated into the neurodevelopmental dose-response model. The level of mercury exposure resulting from consumption of one 170-g meal was calculated for each fish species and then converted to hair mercury concentration using a one-compartment model (Rice and others 2003). The results showed that when one to two, 170-g meals per week are consumed, the estimated cardiovascular benefits from omega-3 fatty acids outweigh the risks from MeHg for all species except shark, swordfish, and yellowfin tuna. On the other hand, one 170-g meal per week from 9 out of the 16 fish examined was predicted to result in adverse neurodevelopmental effects, even for some low-mercury fish like cod (0.11 ppm) and canned light tuna (0.12 ppm). This is likely due to the relatively low levels of omega-3 fatty acids in these fish and the fact that the hair mercury threshold level for adverse cognitive effects was not incorporated into the model. Consumption of these 9 fish at this level would not be expected to exceed the RfD established by EPA. As expected, low-mercury fish with high levels of omega-3 fatty acids, such as salmon, trout, and herring, exhibited the greatest net benefits to neurodevelopment. The authors used the results of the risk-benefit assessments and acceptable body burdens based on the RfD to suggest species-specific fish consumption advisories. For example, for individuals concerned with neurodevelopmental effects, the authors recommended unlimited consumption (up to one, 170-g meal per day) for 7 of the fish species; limited consumption of canned light tuna and cod (two, 170-g meals per week), limited consumption of 4 other species (that is, canned white tuna, tuna steak, halibut, sea bass, and lobster) to just one, 170-g meal per week, and complete avoidance of shark and swordfish. However, the authors note that the focus of this study was to present a framework for risk-benefit assessment and the uncertainties in the dose-response relationships presented here make the conclusions tentative.

Assessments based on composite metrics

Composite metrics allow for the integration of risks and benefits for multiple health endpoints into a single net health impact. These assessments can combine 2 or more types of common metrics, such as mortality, morbidity, or disease incidence, to quantify the cumulative effect on health. Most studies assessing the risks and benefits of seafood with composite metrics have utilized the QALYs to express the net health outcome (Ponce and others 2000; Cohen and others 2005a; Guevel and others 2008), and 1 study used a monetary value based on standard EPA health benefit transfer figures (Shimshack and Ward 2010). These studies are reviewed here.

Studies using QALYs. The earliest publication that applied QALYs to risk-benefit assessment of seafood was presented by Ponce and others (2000). Benefits of fish consumption were defined as a decrease in myocardial infarction mortality and the risks were defined as neurodevelopmental delays associated with prenatal MeHg exposure. The benefit to dose-response relationship for fish consumption was modeled by logistic regression from summary epidemiological data collected for an adult male population in Chicago, IL (U.S.A.), over a 30-y period, which found an inverse relationship between fish consumption and the risk of death from CHD, especially nonsudden death from myocardial infarction (MI) (Daviglus and others 1997). A Weibul excess risk model was used to develop a dose-response relationship for risks using data on delayed talking incidence among children in Iraq with gestational exposure to contaminated grain (Marsh and others 1987). Fish consumption was examined over a range of intakes (0 to 300 g/d) and at a range of mercury concentrations (0.5 to 2.0 ppm). The net health impact of fish consumption was then calculated for 2 different population groups (n= 100000 individuals/population): (1) all members of a population and (2) women of childbearing age (defined by Ponce and others (2000) as 15 to 44 y) and their offspring. When myocardial infarction mortality was assumed to be equal in severity to delayed talking (that is, starting to talk at 24 mo of age), the net effect of consumption of two, 100-g servings of fish per week with mercury levels of 0.5 to 2.0 ppm was predicted to be positive for the total population (2000 to 5000 QALYs), but negative for the subpopulation of women of childbearing age and their offspring (−250 to 2000 QALYs). The maximum benefits occurred among the total population when more than ten, 100-g servings of fish with 0.5 ppm mercury were consumed per week, with a net gain of about 15000 QALYs. On the other hand, the subpopulation of women and children exhibited a net loss of 250 to 2000 QALYs for 2 servings of fish per week. When myocardial infarction mortality was weighted as more severe than delayed talking, two, 100-g servings of fish per week were linked to a net benefit of about 5000 QALYs for the total population, regardless of the mercury concentration in the fish (0.5 to 2.0 ppm). The results of unequal weighting were reported to continue to result in a net negative health impact for the subpopulation of women and children (QALYs not reported). However, the benefits of fish consumption to neurodevelopment, which would be expected to greatly improve the net health impact, were not considered in this model. Further, the mercury levels for fish used in this model were higher than those in most commonly consumed fish, which generally have concentrations of <0.05 to 0.35 ppm, with the exception of a few large predatory fish, such as swordfish, shark, and king mackerel (FDA 2009c). To improve future analyses, the authors suggested the inclusion of a greater number of health endpoints and scenarios comparing fish-based diets with other types of diets, as well as the substitution of fish with low MeHg levels.

To this regard, a comprehensive risk-benefit assessment was carried out for a range of fish consumption scenarios using composite metrics that incorporated dose-response relationships from 4 different studies (Cohen and others 2005a). An expert panel convened by the Harvard Center for Risk Analysis published a series of 4 papers developing dose-response relationships for fish consumption and CHD mortality (König and others 2005); fish consumption and stroke (Bouzan and others 2005); prenatal DHA intake and cognitive development (Cohen and others 2005b); and prenatal MeHg exposure and cognitive development (Cohen and others 2005c). The dose-response relationships were only developed for health endpoints that were expected to be substantially affected by changes in fish consumption and for which there were sufficient data for a quantitative analysis. In the study determining a relationship between fish consumption and heart disease mortality, including CHD and nonfatal MI, the authors identified 7 observational studies (Kromhout and others 1985; Ascherio and others 1995; Daviglus and others 1997; Albert and others 1998; Oomen and others 2000; Hu and others 2002; Mozaffarian and others 2002) of individuals with no pre-existing CHD for use in the dose-response analysis (König and others 2005). To develop a dose-response relationship between fish consumption and stroke risk, the authors combined relative risk results from 6 studies (5 prospective cohort studies and 1 case-controlled study) (Gillum and others 1996; Orencia and others 1996; Iso and others 2001; Caicoya 2002; He and others 2002; Bouzan and others 2005). The dose-response relationship between prenatal intake of n-3 PUFAs and cognitive development utilized 8 randomized control trials (RCTs) (Agostoni and others 1997; Willatts and others 1998; Lucas and others 1999; Birch and others 2000; Makrides and others 2000; Auestad and others 2001, 2003; Helland and others 2003) comparing cognitive development for children or mothers receiving n-3 PUFA supplementation (Cohen and others 2005b). The dose-response relationship between prenatal MeHg and cognitive effects was determined by aggregating results from 3 major epidemiology studies conducted in the Faroe Islands, Seychelles Islands, and New Zealand (Cohen and others 2005c). The dose-response relationships developed in these studies were then used to calculate the net public health impact of fish consumption patterns related to risk-benefit advisories. The impacts of changes in fish consumption on MeHg exposure, omega-3 fatty acid intake, and servings of fish per week were estimated using a modified version of a previously developed exposure assessment model (Carrington and Bolger 2002; Carrington and others 2004). This model assumes that 10% to 20% of the population does not eat fish, and therefore health impacts due to changes in fish consumption are assumed to affect 85% of the population. Consumption rates for 42 types of fish were estimated using data from the USDA Continuing Survey of Food Intake by Individuals (CSFII) (USDA 1998) and NHANES data from 1999 to 2000 (CDC 2003). Levels of MeHg in fish were obtained from the FDA and NMFS and levels of omega-3 fatty acids were derived from the USDA Agricultural Research Service Nutrient Data Laboratory (http://www.nal.usda.gov/fnic/foodcomp/search/). The average maternal body weight was estimated at 60 kg and the baseline fish consumption was 18.7 g/d (130 g/wk) for women of childbearing age (15 to 44 y) and 23.1 g/d (160 g/wk) for other population members ≥15 y of age. The public health impacts of 5 dietary scenarios were compared to the baseline fish consumption for the U.S. population: (1) women of childbearing age maintain current amounts of fish consumption, but only eat fish with mercury levels ≤ 0.13 ppm, (2) women of childbearing age decrease total fish consumption by 17%, regardless of mercury content, (3) in addition to women of childbearing age, other members of the population also reduce fish consumption by 17%, (4) all females not of childbearing age and all males increase fish consumption by 50%, and (5) all adult females (including those of childbearing age) and males increase fish consumption by 50%. When women of childbearing age maintained current consumption levels but only ate fish lower in mercury (scenario 1), a net benefit of 49000 QALYs per year was predicted for the total population, primarily due to the cognitive benefits of DHA, which would contribute an average of 0.1 IQ point per child born. Scenario 2 was based on a study that reported a 17% decrease in overall fish consumption by pregnant women following the 2001 FDA advisory (Oken and others 2003). This dietary scenario was associated with a substantially smaller net benefit to health as compared with scenario 1, with + 0.02 IQ point per child and a net impact of 9700 QALYs per year. In scenario 3, where all members of the population decrease fish consumption by 17%, the net health impact was negative (−41000 QALYs per year), with the greatest losses experienced by elderly males (ages 75 to 84), whose annual CHD mortality risk would be increased by about 2 in 10000. On the other hand, the greatest net benefit to health (+120000 QALYs per year) occurred in scenario 4 when fish consumption was increased by 50% among all females not of childbearing age and all males, with a reduced risk for CHD mortality of 5 in 10000 among elderly men. In the case where all adult males and females increased fish consumption by 50%, the total benefits were offset slightly by the negative impact from MeHg on cognitive development (−0.07 IQ point/child), with a net health impact of +90000 QALYs per year. The effects of POPs were not considered in this risk-benefit assessment because they were not expected to be major contributors to the net health impact; for example, a sample calculation for scenario 4 based on the data reported by Hites and others (2004a) showed that exposure to organic contaminants was associated with a loss of about 600 QALYs per year, compared to the net benefits of 120000 QALYs per year. The overall results of this risk-benefit assessment suggest the importance of fish consumption in terms of the net health impact for the total population. When comparing the predicted results of scenarios 1 and 2, the importance of correctly following advisories for women of childbearing age to consume low-mercury fish, but not reduce total fish consumption is also apparent. The results of scenario 3, where the total population reduces fish consumption, indicate the potential for substantial reductions in public health and increased risks of CHD mortality when advisories targeted at a specific population inadvertently discourage fish consumption among other population groups.

Guevel and others (2008) utilized a QALY approach to assess the risks and benefits of high fish consumers in France. Consumption data for individuals consuming 2 or more servings of seafood per week were obtained from the CALIPSO study (Leblanc and others 2006; Bemrah-Aouachria and others 2008) and the most common types of fish and seafood were sampled locally and analyzed for fatty acids and MeHg. This information was combined to determine the exposure levels for these compounds among the target populations and then to compare risks and benefits for consumers with medium and high EPA + DHA intake. Consumers in the 1st quintile of the CALIPSO study (medium EPA + DHA intake) had an average fish consumption of 334 g/wk, an average EPA + DHA intake of 391 mg/d, and an estimated MeHg exposure of 0.8 μg/kg bw/wk. On the other hand, consumers in the 5th quintile (high EPA + DHA intake) consumed an average of 1104 g seafood/wk, with a daily EPA + DHA intake of 2700 mg and a weekly MeHg exposure of 2.6 μg/kg bw. The net QALYs associated with increasing EPA + DHA intake from medium to high levels were calculated for the adult population in France using dose-response curves developed previously linking fish consumption to cognitive and cardiovascular endpoints (Bouzan and others 2005; Cohen and others 2005a, b, c; König and others 2005). Additional dose-response curves linking EPA + DHA intake to the same cardiovascular endpoints were also developed by the authors based on previous studies (Dolecek and Grandits 1991; Iso and others 2001; He and others 2002; Hu and others 2002; Mozaffarian and others 2002). The net result of increasing EPA + DHA intake from 391 to 2700 mg/d was beneficial for both cognitive (+5949 QALYs) and cardiovascular endpoints (91229 to 114475 QALYs), regardless of the dose-response model. When all endpoints were combined, the total QALYs associated with increasing fish-derived EPA + DHA intake were 97248 using the EPA + DHA loglinear model developed by Guevel and others (2008); 116800 using the fish linear model from the Harvard Center for Risk Analysis (Cohen and others 2005a); and 285774 using the EPA + DHA exponential model developed by Guevel and others (2008). Despite the net benefits associated with increasing fish consumption and EPA + DHA intake, the potential effects of MeHg on neurodevelopment resulted in a negative lower bound of the 95% confidence intervals, ranging from −104380 QALYs for the EPA + DHA exponential model to −278665 for the EPA + DHA loglinear model. Overall, these results were in agreement with Cohen and others (2005a) that the benefits of fish consumption outweigh the risks; however, the magnitude of these effects is influenced by the type of dose-response curve utilized. Integration of additional risks and benefits into the QALY model was recommended for future studies on this topic.

Health-based monetary impacts. The use of monetary values to quantify risks and benefits of seafood consumption was explored in a study that considered the effects of mercury advisories on dietary patterns (Shimshack and Ward 2010). The dietary changes following the 2001 U.S. FDA advisory on mercury in seafood were examined using household-level seafood consumption data obtained from the Information Resources, Inc.'s InfoScan Consumer Network database. The data used by Shimshack and Ward were collected from close to 15000 consumers that were asked to scan the universal product codes for all products purchased from all stores upon returning home over a 3-y period (2000 to 2002). Changes in seafood types and amounts were investigated following the 2001 advisory for “at-risk” households with pregnant women, nursing women, or children under 6. Similar to the results reported by Oken and others (2003), a 17% decrease in fish consumption by pregnant women was observed following this advisory (Shimshack and Ward 2010). Of those who decreased their fish consumption, “at-risk” households decreased fish consumption 21.4% and there was a 60% increase in the number of consumers with no significant fish and shellfish consumption. Overall, there was no evidence of differential avoidance of high mercury fish, with at-risk groups reducing consumption of low-mercury seafood like salmon (27.9% reduction) and shrimp (17.5% reduction). Consumption of white tuna and light tuna fell by 14.0% and 19.5%, respectively. However, when education level was considered, households with a college degree did show selective avoidance of high-mercury fish, with an overall decrease in MeHg exposure of 27.9% (compared to 0.003% for less educated households), but there was also a substantial decrease of about 20% in the n-3 fatty acid intake among both education groups. The health impacts of these declines in seafood consumption were calculated using the cognitive and cardiovascular dose-response models developed previously (Bouzan and others 2005; Cohen and others 2005a, b, c; König and others 2005). However, rather than using a QALY approach, the net health impacts were expressed in terms of monetary values using U.S. EPA benefit transfer figures of $13084 per IQ point and $7.52 million for the value of statistical life, based on 2007 U.S. dollar amounts. Overall, the 21.4% reduction in seafood consumption following the 2001 U.S. FDA advisory was estimated to have a net health impact of U.S. −$30. Declines in MeHg exposure were associated with a benefit of +0.012 IQ points/child, while declines in the EPA + DHA intake were associated with −0.008 IQ points/child, resulting in a net benefit of 0.004 IQ points/child following the advisory (equivalent to U.S. $61). On the other hand, the decline in EPA + DHA intake was also associated with a net increase in CHD and stroke mortality of +0.63 deaths per 100000 adults (equivalent to U.S. −$91). The effects of an idealized scenario presented by Cohen and others (2005a; scenario 1), in which at-risk households maintain overall fish consumption amounts but only eat low-mercury fish, were also examined by Shimshack and Ward (2010) in terms of risks and benefits to neurodevelopment. This scenario resulted in a net benefit of 0.039 IQ points/child (U.S. $587). These results indicate the importance of targeted strategies for reducing MeHg exposure without concomitantly reducing n-3 fatty acid intake in order to receive the greatest health benefits from seafood.


While the application of risk-benefit analysis to specific foods is a relatively new field of study, it represents a powerful tool that allows decision makers to set policy that can have a universal beneficial effect. Although science-based, risk-benefit analysis remains complex but will continually improve with increased analyses, new methodologies, and larger studies. It also provides new research opportunities for investigating the multiple components of food as well as their physiological and nutritional interactions in populations. There are continuing efforts to maximize benefits and minimize risks through optimization of food consumption patterns. Consumers are informed about what foods to avoid, while being encouraged to consume others that will, hopefully, promote better health. Seafood is somewhat unique as few other foods present so many deleterious and beneficial components in one packet. Many of the studies highlighted in this review show the importance of maximizing the benefit: risk ratio by consuming fish that are high in EPA + DHA but low in contaminants. Although the benefits of EPA + DHA consumption can be negated by excessive consumption of high-MeHg seafood by sensitive populations, such as pregnant women and young children, under-consumption of EPA + DHA across both the at-risk and general population groups is of equal concern. Studies also demonstrate that attempts to “protect” an at-risk population can have negative unintended consequences on that group and on the much larger overall population. It is therefore prudent that agencies consider the impact of unintended consequences that messaging can create. Regardless of the risk-benefit method utilized, this review indicates that results from studies overwhelmingly support the idea that messaging regarding seafood consumption risks requires a targeted strategy that does not discount the benefits. Sensitive populations will likely benefit by following risk advisories and reducing consumption of high-mercury fish; however, there is a need to be careful about how these advisories are worded and released to the public. Studies have shown that messaging reduces fish consumption in general, in both the target and nontarget populations, resulting in an overall reduction in the potential health benefits derived from EPA + DHA. This is a cause for concern as most populations are not currently meeting the recommended intake levels for EPA + DHA. While a majority of research has shown that seafood consumption greatly outweighs the risks, it is important to keep in mind that this field of science is just at the incipient stages of determining how to accurately assess the everyday choices we make in our diet and how these ultimately affect our lives.


This work was partially supported by a grant from the National Food Safety Initiative (Grant Nr 2007–5110-03815) of the National Inst. of Food and Agriculture, U.S. Dept. of Agriculture.


Appendix A: Summary of Risk-Benefit Assessment Studies

Table 3.  Table A1–Risk-benefit assessments for seafood consumption using common metrics.
Target population Seafood type Risk factors Risk endpoints Beneficial factors Benefit endpoints Method of analysis Results/conclusions Reference
American adults and subpopulation of subsistence fishersUnspecified types of fish containing six cancer-causing compounds at maximum allowable levels according to FDASix organic pollutantsRisk of cancerFish consumptionReduced risk of death from CHDRisk tradeoff analysis—probabilities of each health endpoint were calculated and comparedCHD-related benefits of eating fish far outweigh cancer risks for American adults and for subsistence fishers. Consumption of 20 g fish per day was estimated to reduce CHD risk by 35%, while increasing cancer risk by 1% (1 × 10−2) if fish are contaminated at maximum levels, and a more realistic increased cancer risk of 0.0075% (7.5 × 10−5) based on actual contaminant levels in fish. Anderson and Wiener (1995)
Not specifiedFarmed and wild salmonOrganic contaminants (PCBs, dieldrin, and so on)Cancer and noncancer risks from lifetime exposureEPA + DHAProbability of reducing sudden death from CHDU.S. EPA risk-based assessments for organic contaminants compared to results of clinical trials with n-3sConsumption of 1 g/d of EPA+DHA from salmon would save about 300 times the number of lives that would be lost to cancer from exposure to organic compounds. Individuals can maximize benefit while reducing contaminant-associated risk by choosing salmon with lower contaminant concentrations. Foran and others (2005b)
French consumers with high fish consumption (n= 1000)82 types of fishes, mollusks, crustaceans, and seafood-based dishesContaminants (6 trace elements and 3 categories of POPs)Probability of exceeding PTWI for each contaminantOmega-3 fatty acidsProbability of obtaining RDA of 500 mg/dCombined data obtained through food frequency questionnaires with levels of contaminants in various fish types to determine probability of each population group to meet the RDA for omega-3 fatty acids and to exceed the PTWI for contaminants.84% of study participants were meeting the RDA for omega-3 fatty acids and 34% of participants were exceeding the PTWI for MeHg. Participants had a 39% probability of exceeding the PTWI for dioxins and dl PCBs and a 74% probability of exceeding the PTWI for iPCBs. Blood levels of MeHg were well below the no adverse effect level, even in individuals consuming 4.5 kg seafood/wk. The authors recommended that the general population should consume at least 2 servings of fish, especially oily fish, per week and pregnant and nursing women should only consume predatory fish once per week. Leblanc and others (2006)
Women in North America aged 20 to 39 y with an average body weight of 60 kgFarmed Atlantic salmon and rainbow troutMeHg, PCBs, and dioxinsReproductive and developmental risksEPA + DHAIntake of 500 mg EPA + DHA/dThe EPA + DHA intake with two, 180-g servings of Atlantic salmon per week calculated to be 440 mg/d. MeHg, PCBs and dioxins was determined and compared to the PTWIs for these contaminants.Based on the results, two 180-g meals per week of farmed salmon or trout available in North American markets would be expected to provide sufficient levels of EPA + DHA for the target population without concern over exceeding the PTWIs for MeHg or PCBs and dioxins. Dewailly and others (2007)
Dutch consumers (n= 6250)11 types of seafood and 7 other foodsDioxins (PCDDs, PCDFs, and DL-PCBs)Probability that dioxin exposure is below TDI of 2 pg WHO-TEQ/kg bw/dayEPA + DHAProbability that EPA+DHA intake is above AI: 7 mg EPA+DHA/day/kg bw (based on AI of 450 mg/day and average body wt of 65 kg)Calculated exposure levels based on food diaries and levels of compounds in each food type and compared to dietary guidance valuesFor most consumers, exposure to dioxins and EPA+DHA was below the TDI and AI, respectively. A scenario where 10–25% of beef and pork was replaced with fatty fish allowed for the AI of EPA+DHA while remaining below the TDI. Higher percentages of substitution (50–100%) led to exceeding the TDI. van der Voet and others (2007)
70-kg adult consumer in Spain14 most consumed fish species in SpainMeHg, Pb, Cd, and 7 categories of organic pollutantsExposure as related to PTWI, TDI, and cancer and noncancer risks from lifetime exposureEPA + DHADaily intake of EPA + DHACalculated daily intakes of risk and benefit factors based on dietary data for top 14 fish species and compared to dietary guidance valuesEstimated daily intakes for 70-kg adult consumer provided 244 mg EPA + DHA and were below risk levels for all categories, except that PAHs exposure resulted in an increased cancer risk of 2.4 × 10−5 Domingo and others (2007)
Adult consumers (n= 401) in a coastal region of France with high fish consumption38 categories of fish and seafoodDioxins and PCBsExposure as related to PTWI for dioxins and dioxin-like PCBs of 14 pg TEQ/kg body wt/wkLC n-3 PUFAsIntake of LC n-3 PUFAs as compared to recommended level of 500 mg/dCalculated exposure and intake levels based on 1-mo dietary notebook records and compared to dietary guidance values41% of the subjects achieved the recommended level of LC n-3 PUFAs and remained below the tolerance threshold for dioxins; 19% of subjects met the nutritional recommendation but exceeded the tolerance threshold; 38% of subjects were below both the tolerance threshold and the nutritional recommendations Verger and others (2008)
Belgian adult fish consumers (n= 821)10 types of fishPBDEsExposure levels for different dietary scenariosEPA + DHAIntake levels for different dietary scenarios as compared to reference intake of 9.7 mg/kg bw/d (681 mg/d)Calculated probable dietary exposure and intake levels based on food frequency questionnaire and compared fish consumption scenariosCurrent consumption levels showed mean PBDE and EPA + DHA intakes of 0.85 ng/kg bw/d and 3.5 mg/kg bw/d, respectively. Consumption of two 150-g portions of salmon per week allowed for 11.9 mg/kg bw/d EPA + DHA without major increases in PBDE exposure (1.3 ng/kg bw/d). Sioen and others (2008)
13 regional clusters of the global adult population7 categories of fish and seafoodMeHg, iPCBs, dioxin-like PCBs, PCDDs, and total TEQExposure related to PTWIEPA + DHA, iodine, and vitamin DIntake levels for different regional clusters as compared to dietary guidance values (500 mg/d EPA + DHA; 5 μg/d vitamin D; 150 mg/d iodine)Estimated exposure and intake levels worldwide and compared to dietary guidance valuesMost regional clusters were below the PTWIs for contaminants, but did not achieve the recommended dietary guidance values. The countries with the highest seafood intake exceeded the PTWIs for MeHg or dioxin-like compounds. The mean intake values for these countries did not exceed the reference values. proposed by SACN/COT. Sioen and others (2009)
U.S. consumersDifferent fish species consumed in the U.S.MeHgFetal neurodevelopmentSeafood consumptionFetal neurodevelopment, risk of fatal CHD, risk of fatal strokeMaternal info based on Iraq and Seychelles Islands and UK cognitive development study; CHD based on meta-analysis and pooled analysisCurrent fish consumption is contributing a net benefit to neurodevelopment, and if women of child-bearing age increased fish consumption to 340 g/wk, neurodevelopment benefits would also increase (0.225 IQ point); commercial fish baseline consumption is averting a central estimate of over 30000 deaths per year from coronary heart disease and over 20000 deaths per year from stroke.FDA (2009a)
Adult population (cardiovascular risk group) and subpopulation of pregnant women (neurodevelopmental risk group)16 types of fish commonly available in Connecticut marketsMeHgIncrease in risk of adult MI; decrease in infant VRM scoreEPA + DHADecrease in risk for adult CHD mortality; increase in infant VRM scoreCalculation of net changes in cardiovascular risk and infant VRM scores for each fish species based on dose-response relationshipsSpecies-specific consumption recommendations varied depending on risk group. Both groups were recommended to avoid shark and swordfish; there were 7 to 9 fish species that could be eaten more than twice per week and 5 to 7 fish that should be limited to one or two 6-oz meals per week. Ginsberg and Toal (2009)
French pregnant women (n= 161)48 types of fish and seafood itemsMeHgExposure as related to PTWI for MeHgLC n-3 PUFAsIntake of LC n-3 PUFAs as compared to recommended level of 500 mg/dCalculated dietary exposure and intake levels based on responses to food frequency questionnaires and compared to dietary guidance valuesAbout 50% of the subjects achieved the recommended nutritional intake and about 5% of subjects exceeded the PTWI for MeHg Pouzaud and others (2009)
European consumers from 8 countriesTop 5 consumed species for each countryMeHgProbability of exceeding the PTWI for MeHgEPA + DHAProbability of exceeding a daily intake of 500 mg/d of EPA + DHALog-normal distribution for seafood consumption patterns and plug-in (PI) or tail-estimation (TE)-based estimates to calculate probabilities of exceeding dietary guidance valuesProbability of exceeding 500 mg/d of EPA + DHA in each country was 0.32% to 66.05%; probability of exceeding the PTWI for MeHg was 0.04% to 9.61%. Cardoso and others (2010)
Table 4.  Table A2–Risk-benefit assessments for seafood consumption using composite metrics.
Target population Seafood type Hazards Risk endpoints Beneficial factors Benefit endpoints Method of analysis Results/conclusions Reference
Total population and subpopulation of women of childbearing age and their childrenUnspecified types of fish with average MeHg levels of 0.5 to 2.0 ppmMeHgIncrease in neurodevelopmental delay (that is, talking after 24 mo of age) from prenatal exposureFish consumptionDecrease in myocardial infarction mortalityQALY-adjusted dose-response modeling examining consumption of 0 to 300 g fish/d with MeHg levels of 0.5, 1.0, and 2.0 ppmTotal population showed net benefits for fish consumption (gain of approximately 5000 QALYs for 2 servings* fish per week). Subpopulation of women of childbearing age showed negative effects of consumption of fish with 0.5 to 2.0 ppm MeHg (loss of approximately 250 to 2000 QALYs for 2 servings per week; benefits of DHA on neurodevelopment were not considered). Ponce and others (2000)
Total U.S. adult population and subpopulation of women of childbearing age42 types of fish consumed in the U.S.MeHg exposure through fish consumptionChange in IQ pointsFish consumption; prenatal DHA intake through fish consumptionChanges in the relative risks of CHD mortality and stroke; change in IQ pointsQALY-adjusted dose response modeling examining 5 different scenarios of fish consumptionTotal population showed a net negative impact of reducing fish consumption by 17% (−41000 QALYs) and a net benefit of increasing fish consumption by 50% (+90000 QALYs). Subpopulation of women of childbearing age showed substantial benefits of replacing fish high in MeHg with fish low in MeHg (+49000 QALYs/+0.1 IQ points). Cohen and others (2005a)
French adults consuming high amounts of fish and seafood (average 11 servings/wk)Most consumed types of fish and seafood among target populationsMeHgImpact on IQ pointsEPA + DHACHD mortality and stroke mortality and morbility; impact on IQ pointsQALY-adjusted dose-response modeling examining effect of increasing EPA + DHA intake from 391 to 2700 mg/d through fish consumptionBeneficial effects of increasing EPA + DHA intake through fish consumption, with QALYs totaling 97248 to 285774, depending on the dose-response model. Confidence interval indicates possible negative effects from MeHg for some individuals. Guevel and others (2008)
U.S. households with pregnant women or young childrenAll seafood purchased from 2000 to 2002MercuryImpact on IQ points/child and corresponding monetary valueEPA + DHA, total fish consumptionImpact on IQ points/child, reduction of CHD-related deaths and corresponding monetary valuesCalculation of health impacts based on seafood purchased in 14800 householdsFDA's 2011 mercury in fish advisory resulted in reduced fish consumption and an overall negative impact on health, valued at negative $30 USD 2007 Shimshack and Ward (2010)
Table 5.  Table A3–Qualitative risk-benefit assessments.
Target population Seafood type Hazards Risk endpoints Beneficial factors Benefit endpoints Method of analysis Results/conclusions Reference
AllGeneralMeHg, other metals, POPs, microbs, seafood allergens and naturally occurring toxinsMeHg was the greatest concern, POPs remain uncertainEPA + DHABenefits to women during pregnancy, duration of gestation and birth weight, infant and child development, cardiovascular disease, cardiovascular mortality, and all-cause morbidity and mortality.4-part qualitative protocol involving: identification of the magni-tude of R/B; identify risk; evaluate different consumption patterns; balancing to arrive at specific guidelinesBenefits of seafood consumption outweighed the risks in healthy adults and may reduce their risk for cardiovascular disease by consuming seafood regularly Adults as risk for cardiovascular disease may benefit from seafood high in EPA + DHA pregnant/nursing women and children may benefit from consuming seafood (up to 12 oz per week) IOM 2005 IOM (2005b)
Adult populationTotal fish consumptionMeHg, PCBs, dioxinsNeurologic, carcinogenic, and other health risksFish consumption and EPA + DHA intakeNeurologic developmental benefits, reduction in CHD, and sudden deathEvaluation of studies and meta-analyses evaluating human health effects related to fish intake, contaminants, and beneficial compounds.Among adults, the benefits of fish intake exceed the potential risks. For women of childbearing age, the benefits also outweigh risks, with the exception of a few selected species. Moderate fish consumption reduces coronary death by 36%. Mozaffarian and Rimm (2006)
Adult population and special groupsAll seafoodMeHg, dioxins, dl PCBs, furansNeurodevelopment, cardiovascular disease and cancerFish consumption; EPA + DHANeurodevelopment and prevention of cardiovascular diseaseEvaluation of specific nutrients and contaminants in seafood as well as scientific literature regarding the risks and benefits of seafood consumptionOverall: good source of nutrients, lowers CHD. No evidence of CHD from MeHg or cancer from dl PCBs were below CHD benefits. Omega-3 intake lowers risk of suboptimal neurodevelopment. Maternal dioxin/dl PCB intake is below PTMI neurodevelopmental risk is negligible. Recommendations: benefit of seafood on CHD mortality and improving neurodevelopment JECFA FAO/ WHO (2010)
All AmericansAll seafoodPOPs, MeHg EPA + DHAEPA + DHA and risk of CHD; EPA + DHA and health of infantsEvaluation of experts in the fields and systematic reviews in the areasSeafood intake reduces risk of CHD mortality or sudden death; omega-3s improved infant health visual acuity and cognitive development. Overall, health benefits from 2 servings per week outweigh the risks from MeHg and POPsDGAC (2010)
Table 6.  Appendix B: Abbreviations/Acronyms.
AHA: American Heart AssociationMeHg: methylmercury
AI: adequate intake (The Netherlands)MI: myocardial infarction
AHRQ: Agency for Health Research and QualityML: maximum level
AL: action levelMRL: minimal risk level
ALA: alpha-linolenic acidmt: metric ton
ASTDR: Agency for Toxic Substances of Disease RegistryNHANES: National Health and Nutrition Survey
BMD: benchmark doseNMFS: National Marine Fisheries Service
BMDL: benchmark dose lower boundNRC: National Research Council
Cd: cadmiumNOAA: National Oceanic and Atmospheric Administration
CHD: coronary heart diseaseNOEL: no observed effect level
CPF: cancer potency factorPAH: polycyclic aromatic hydrocarbon
CSF: cancer slope factorPBDE: polybrominated diphenyl ether
CVD: cardiovascular diseasePCB: polychlorinated biphenyl
CFSII: USDA Continuing Survey of Food Intake by IndividualsPCDD: polychlorinated dibenzo dioxin
CO: chronic oralPCDE: polychlorinated diphenyl ether
DALY: disability-adjusted life yearsPCDF: polychlorinated dibenzo furan
DDT: dichlorodiphenyltrichloroethanePCN: polychlorinated naphthalene
DGAC: Dietary Guidelines Advisory CommitteePMTDI: provisional maximum tolerable daily intake
DHA: docosahexaenoic acidPOP: persistent organic pollutant
dl PCB: dioxin-like polychlorinated biphenylsPTDI: provisional tolerable daily intake
DRI: dietary reference intakePTMI: provisional tolerable monthly intake
EC: European CommissionPTWI: provisional tolerable weekly intake
EFSA: European Food Safety AuthorityPUFA: polyunsaturated fatty acid
EPA: eicosapentaenoic acidQALY: quality-adjusted life years
EPA: U.S. Environmental Protection AgencyRCT: randomized controlled trail
EU: European UnionRDA: recommended daily allowance
EUML: European Union maximum levelRDI: reference daily intake
EVT: extreme value theoryRfD: reference dose
FAO: Food & Agriculture Organization of the United NationsSACN/COT: Scientific Advisory Committee on Nutrition/Committee on Toxicity (UK)
FDA: U.S. Food and Drug AdministrationSCF: Scientific Committee on Food (France)
GL: guidance levelTCDD: 2,3,7,8-tetrachlorodibenzo-p-dioxin
HC: Health CanadaTDI: tolerable daily intake
HCB: hexachlorobenzeneTEF: toxic equivalent factor
HCN: Health Council of The NetherlandsTEQ: total dioxin-like equivalency
HHS: U.S. Department of Health and Human ServicesTL: tolerance level
IOM: Institute of MedicineUSDA: U.S. Department of Agriculture
INRA: Institute for Agronomy Research (France)VRM: visual recognition memory
ISSFAL: International Society for the Study of Fatty Acids and LipidsWAPM: World Association of Perinatal Medicine
JECFA: Joint FAO/WHO Expert Committee on Food AdditivesWHO: World Health Organization