A warning: NOECs are inappropriate for regulatory use

The first two authors of this letter, independently and unknown to each other, were contracted to measure the toxicity of a pulp mill effluent. Specifically, two different commercial, certified bioassay laboratories independently tested the same effluent sample by measuring abnormal development to the prodissoconch I larval stage for both oysters and mussels [1,2]. They also used an echinoderm fertilization test [2]. Both 50% effective concentrations (EC50s) and no-observed-effect concentrations (NOECs) were determined; the latter, which are based on hypothesis testing, are roughly equivalent to the regulatory endpoint recently adopted for use in Washington State, where the pulp mill is located. The Washington whole effluent toxicity (WET) testing rules [3] define an acute WET limit as ‘‘. . . a demonstration of no acute toxicity in a concentration of effluent equal to the acute critical effluent concentration (ACEC)’’ and the chronic WET limit is defined as ‘‘. . . a demonstration of no chronic toxicity in a concentration of effluent equal to the chronic critical effluent concentration (CCEC)’’ [3]. The ACEC and CCEC are the expected maximum concentrations of effluent under critical conditions at the edge of acute and chronic mixing zones, respectively. The demonstration of no toxicity is based on hypothesis testing between a control and the ACEC and CCEC concentrations. The results of this otherwise minor work are important for two reasons related to the general problem of calculating measures of potency. First, the EC50 and NOEC comparison emphasizes critical differences between scientific usage and regulatory application of scientific information. In particular, we will demonstrate that the point estimate (EC50) is more reliable for regulatory use than a hypothesis test endpoint. Second, synoptic testing of a sample with all three tests by one laboratory, let alone two, has not been done previously (to our knowledge). Regulatory usage of two of these three different tests often assumes that they are interchangeable [1,2]. These results were mentioned at a European Symposium (August 1994). Our main conclusion, that NOECs are inappropriate for regulatory use, received strong support from some participants and puzzled looks from others. One of the supporters was the third author of this letter (P.F. Chapman), who was in the process of finalizing a presentation for SETAC ’94 questioning the suitability of the NOEC on statistical grounds [4]. Our subsequent decision to write this letter was made because we believe its relevance is far from limited to North America and Europe. Further, this letter supports recent European comment that the NOEC, because it is theoretically invalid, should be replaced by a more realistic measure of no effects [5,6]. The data presented here not only support statistical theory but should be much more persuasive to nonstatisticians.