A review of the Food and Drug Administration risk analysis for polychlorinated biphenyls in fish

Abstract

This paper reviews the Food and Drug Administration (FDA) analysis of the risk to humans from consuming fish contaminated with polychlorinated biphenyls (PCBs). In brief, the FDA methodology employed “high” dose experiments on animals and extrapolated the observed rates of certain types of cancer at these elevated doses to the low doses found in human diets. These extrapolations were then used to define a recommended tolerance level of 5 ppm, and proposed reduction to 2 ppm, for fish sold in interstate commerce. Unfortunately, as is shown here, such a procedure is extremely sensitive to the basis for extrapolation. Important elements of the FDA analysis include the following: (i) FDA assumed a particular form of the dose-response model: the one-hit model. Many other models have been proposed and, on balance, appear equally plausible. These models estimate lower risks than does the one-hit model. (ii) FDA calculated 99% upper confidence bounds on these risks and, moreover, emphasized cases of fish eaters who consume greater amounts of PCB-contaminated fish than do 98.5% of the U. S. population. (iii) FDA based PCB ingestion computations on consumption of raw fish, whereas most fish are cooked before eating, and it is known that PCB levels in cooked fish are lower than PCB levels in raw fish. (iv) FDA based estimates of cancer risk on the assumption that PCB levels in fish would be constant over the nominal 70-year human life span used in the FDA “lifetime risk” computation. Recent data suggest that PCB levels have been declining in fish (particularly in sport fish) and humans as well. Such trends imply significantly lower cumulative lifetime PCB doses than were assumed in the FDA analysis. (v) FDA assumed that humans and test animals are equally sensitive to PCB ingestion when measured on a parts per million in diet basis. Extrapolations on an equivalent consumption per unit of body weight, thought appropriate by most researchers, result in much lower health risks. In short, when confronted with methodological choices, the FDA consistently selected “worst case” or conservative assumptions over other alternatives of at least equal plausibility. This philosophy of choice was explicitly acknowledged by the FDA. What was omitted from the FDA analysis, however, was the possible degree of overstatement of these risks. The results of replicate risk computations using alternative assumptions to examine the possible magnitude of overstatement of health risk are summarized in Table 12. As can be seen, this overstatement could easily account for a discrepancy of several orders of magnitude between actual and calculated risks. As well, the FDA analysis overstated the effectiveness of the imposition of tolerance limits, considering the sampling procedures currently in use for testing commercial fish. Taken together, these points challenge the analytical foundation underlying an FDA proposal to reduce the tolerance level from 5 to 2 ppm and, in addition, lend necessary perspective to the debate on the health risks of eating PCB-contaminated fish.