Environmental pollutants: downgrading the fish food stock affects chronic disease risk
Determining the benefits and risks associated with fish intake is a major public health issue. On the one hand, fish contain docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA), long-chain omega-3 fatty acids which have been considered beneficial for human health. Consistent with this concept, a meta-analysis showed moderate, inverse associations of fish consumption with cerebrovascular risk . On the other hand, fish may contain various substances such as persistent organic pollutants (POPs), including dioxins, polychlorinated biphenyls (PCBs), organochlorine pesticides, and polybrominated diphenyl ethers and heavy metals, including methylmercury, lead, and cadmium, that can negatively affect human health. Consistent with this point, circulating POPs predicted incident stroke in an elderly cohort in Uppsala, Sweden .
A study by Bergkvist et al.  in this issue of the Journal of Internal Medicine complements these studies [1, 2]. Bergkvist et al.  present original findings about associations of dietary PCB exposure, fish intake and the risk of stroke. In this 12-year prospective study of middle-aged and elderly Swedish women, neither fish nor intake of PCBs was related to stroke risk. However, with adjustment for fish intake, dietary PCBs were associated with an increased risk of total stroke. In contrast, with adjustment for dietary PCBs, consumption of fatty fish, EPA and DHA was associated with lower risk . One interpretation of this epidemiologic phenomenon is that PCB risk masks fish benefit. Thus, by separating fish from PCB effects, Bergkvist et al.  provided evidence that PCBs importantly downgrade the fish food stock, at the same time supporting the concept that fish in its unpolluted state might protect against stroke. It is of concern that the median daily dose of PCBs in the upper quartile of dietary intake, which was associated with excess risk of stroke, was only 288 ng per day. We compared this level of exposure to acceptable exposure for the PCB mixture Arochlor . The reference dose, which is ‘an estimate (with uncertainty spanning perhaps an order of magnitude) of a daily exposure to the human population (including sensitive subgroups) that is likely to be without an appreciable risk of deleterious effects during a lifetime’ , is 0.00002 mg kg−1 day−1, which translates to 1200 ng per day for a 60 kg person. This American recommendation for a safe dose of PCBs is substantially higher than the dietary dose estimated to have an adverse association with future stroke in Swedish women.
Strengths and weaknesses of study of dietary PCBs
Unexpected lack of benefit of fish has been found in other conditions, for example, in type 2 diabetes  and in low birth weight . Our research group has previously commented on the possibility that contamination of fish could be responsible for the unexpected lack of benefit . Chemical pollutants pose difficulties for nutritional science; we have recommended that POPs including PCBs should be considered when studying diet .
Bergkvist et al.  have directly addressed our concern by estimating dietary PCBs. Their analyses in relation to stroke risk came out just as expected under the hypothesis that fish without contaminants should protect against stroke, whilst PCB exposure should enhance risk of stroke. Thus, study of dietary intake of PCBs is a real strength. However, interpretation of estimates of dietary intake of PCBs should consider inherent methodologic cautions and limitations.
Although Bergkvist et al.  rely on validity coefficients of 0.30–0.58 , these values may be high. Bergkvist et al. [3, 9] estimated correlation coefficients between body weight-adjusted food frequency questionnaire (FFQ)-based PCB estimates and age- and blood lipid-adjusted serum concentrations of six serum PCB congeners. Age is highly correlated with duration of exposure and is the strongest determinant of serum concentrations of PCBs. Calibration of dietary intake of PCBs to age-adjusted serum PCB values therefore introduces error. Their correction for within-person variability in the FFQ  improved the estimate of the true underlying association between dietary intake and serum PCBs, but the estimates used in the study of stroke risk were without such correction . The corresponding validation correlations without this correction are 0.21–0.40 . Absolute exposure levels to pollutants may be more important than level relative to total intake, bringing into question adjustment for total energy intake. Additionally, content of PCBs or other pollutants in food vary greatly by calendar time, geography and human activities. For example, the lipid content (and consequently PCBs) of wild-caught mackerel varies according to the season. Some years ago, the aquaculture feed was essentially based on marine ingredients, thereby bringing dioxins, PCBs and mercury to the farmed fish and mimicking the risk profile of wild fish. However, the aquaculture industry is now relying heavily on vegetable ingredients , which bring new and uncontrolled pollutants, at the same time causing a dramatic nutritional change in farmed fish. Farmed Atlantic salmon now has as much EPA and DHA as n-6 polyunsaturated fatty acids, whereas a wild salmon or mackerel has between 5 and 9 times more EPA and DHA than n-6 polyunsaturated fatty acids . As the extent of industrialization or regulation directly affects the quality of food, validity of dietary estimates of pollutants should be addressed in each specific human study (which Bergkvist et al.  did). Even within one country, whether the PCB content of food is sufficiently regular to study using a table of PCB content per unit of specific food items is a major issue. Thus, a food table that ‘works’ at one time and place might not ‘work’ in another.
We also draw attention to the distinction between dietary PCBs and serum PCBs. Dietary analysis does not capture the full effect of PCB exposure on health. Although diet is a primary source of PCB exposure, duration and extent of prior exposure plus other factors contribute to serum PCB concentrations. Indeed, Kvalem et al.  in Norway estimated PCBs based on dietary intake complemented by sex, parity, age, residence, smoking status, energy intake and education; they found even higher correlations with serum PCB levels than did Berkvist et al. . Therefore, estimated dietary intake of PCBs should not be used to evaluate health effects of PCBs generally in humans; direct measurement of these chemicals in blood is preferred for this purpose. Where blood assays are not available, prediction model-based estimation including many POPs determinants  would be preferred to prediction based solely on dietary intake. This point may explain why the strength of association between serum concentrations of PCBs and stroke was much stronger  than that of dietary intake estimates of PCBs .
Other evidence that pollutants modulate the benefits of fish
The evidence that environmental pollutants can affect the health benefits of fish is supported by previous experimental studies. In basic research, feeding high-fat diet to both rats and mice is a classical model to induce insulin resistance, obesity and other disorders linked to type 2 diabetes. Adding decontaminated farmed salmon oil in a high-fat diet was found to protect rats from metabolic disorders . However, when the same oil, coming from farmed Atlantic salmon, was not decontaminated, the presence of persistent organic pollutants in this commercially available food fully counteracted the beneficial effects of the decontaminated very long chain omega-3 fatty acids and, indeed, accelerated the development of insulin resistance-related disorders in rats. In another study , the health effects associated with the consumption of a commercial farmed Atlantic salmon fillet with usual levels of POPs were compared with Atlantic farmed salmon fillet specially bred to have low levels of POPs (−50%). Consonant with the study of oil , mice-fed farmed salmon with low POPs showed better metabolic profile than animals exposed to higher POPs commonly present in farmed salmon .
How should the risk and benefit of fish be evaluated?
The study of Bergkvist et al.  may have important consequences for recommendations on fish consumption, and further studies are warranted. Determination of the health effects of fish requires awareness of chemical pollutants. Giving advice on fish intake based on scientific findings obtained from the EPA and DHA present in fish oil or omega-3 capsules, as is common practice , is inadequate in the presence of PCBs. Turning to fish oil capsules could make sense, because the fish oil used to make fish oil supplements has been decontaminated and contains much lower pollutants. Unfortunately, fish oil supplements did not predict stroke , and the levels of pollutants in fish oil supplements may also vary significantly because some decontamination processes are more efficient than others . Future epidemiological and clinical studies should therefore assess the levels of pollutants present in the fish or the fish oil supplements used to avoid any confounding factors. Finally, although risk assessment of fish should focus on dioxins, PCBs and mercury because these pollutants are highly present in wild fish, farmed fish present another risk profile. It will be therefore crucial for future risk and benefit assessments to provide separate evaluation of farmed fish and of pollutants typically found in current day plant ingredients, such as endosulfan and others.
Conflict of interest statement
No conflict of interest to declare.