Testing toxic compounds in human subjects: ethical standards and good science.

We read with interest the letter by Sass and Needleman (2004) and the responses it triggered (Charnley and Patterson 2004; Chart et al. 2004; McAllister 2004; Tobia et al. 2004). In our opinion, Sass and Needleman made two main points: a) human studies with low statistical power are not helpful in determining the presence and magnitude of adverse effects from environmental toxicants; and b) industry sponsorship may consciously or unconsciously bias study design, data analysis, or interpretation. We were thus surprised by the focus of the respondents, which was on issues of ethics in human research per se. One of us (J.M.L.) co-chairs an institutional review board (IRB). In general, IRB approval of a human research protocol hinges on two major factors: a) the relationship between risk and benefit, and b) the adequacy of the informed consent process.

In relation to risks and benefits, virtually all research studies convey direct risk to subjects. These risks fall into one of three categories: physical (e.g., adverse health effects from a study intervention), psychological (e.g., stress from uncomfortable interview questions), and social (e.g., breach of privacy). Benefits, on the other hand,

always accrue to society and only infrequently are directly conveyed to subjects. Thus, direct risks to subjects must be balanced by the potential benefits to society in many protocols; this is certainly the case for studies of environmental agents with intentional exposures to subjects. One of the major considerations of an IRB in deciding whether or not the risk-benefit balance is appropriate is the scientific design of the study; studies with poor design or inadequate statistical power cannot produce results of benefit to society and thus inappropriately expose subjects to risk. In this regard, the ethical issue that Sass and Needleman (2004) raise is not whether it is appropriate to intentionally expose volunteers to environmental toxicants per se, but whether it is appropriate to do so in small numbers that do not have adequate statistical power to answer the question in a meaningful way.

Sass and Needleman (2004) also took issue with the interpretation of the data in two studies, and support their concerns by citing work authored by one of us (Goldman et al. 1990a, 1990b). Both studies reported evidence that the pesticide aldicarb is toxic to humans at levels much lower than those predicted by a 1971 study of four Union Carbide employees (Haines 1971) and Rhone-Poulenc's 1992 human dosing study (Wyld et al. 1992). While the 1990 studies (Goldman et al. 1990a, 1990b) have been criticized because they did not use controlled dosage levels, it is important that we use information from outbreaks in order to test our assumptions regarding the applicability of limited premarket test data for regulatory standard setting for the general population. In this regard, protocols involving intentional exposures to limited numbers of human subjects often face a major challenge beyond low statistical power: restriction of volunteers to healthy male adults, who are not representative of the general population, especially its most vulnerable members. As a result, the interpretation of results from such human dosing studies, especially if the toxicant has a relatively high lowest observable effects level, may be misleading in the context of the general population.

In the context of Sass and Needleman's points (Sass and Needleman 2004), we propose that a) studies should only be approved by the IRB if the results will have scientific validity based on a priori considerations of design, sample size, and statistical power; and b) authors, reviewers, journal editors, and ultimately readers strive to ensure that interpretations of results conform to the data and acknowledge limitations in extrapolation to the general population.

One of the responses to Sass and Needleman (2004) was a statement by a representative of CropLife America, the pesticide industry trade association. In that statement, McAllister (2004) asserted that the U.S. Environmental Protection Agency (EPA) put a moratorium on human testing due to "an intense media campaign and public pressure." One of us (L.R.G.) was an official at the U.S. EPA responsible for the U.S. EPA's pesticide program at the time the first moratorium was adopted in 1998 (U.S. EPA 1998). While nearly all important matters at the U.S. EPA are accompanied by "an intense media campaign and public pressure" (McAllister 2004), the reason for the moratorium was that officials at the U.S. EPA at that time were stunned to find that the agency had not taken the necessary steps to ensure that its actions to generate and utilize human data met the standards for protection of human subjects. In 1998, we assembled a panel of expert scientists and ethicists to advise the U.S. EPA, under the agency's Science Advisory Committee and Scientific Advisory Panel. That committee concluded that human testing to determine adverse effect levels was not scientifically justified, but it could not concur about other types of human experimentation, and thus was not able to present to the U.S. EPA a complete consensus on the issue (U.S. EPA 2000).

McAllister (2004) and others are tightly outraged that the U.S. EPA still has not taken action to resolve this issue. However, we may disagree about what actions are appropriate. In his letter, McAllister (2004) asserted that human testing of pesticides is "equivalent" to Food and Drug Administration (FDA) phase I investigations of pharmaceuticals. However, important ethical distinctions can be made between the benefits of pharmaceuticals to both individuals and society versus the benefits of pest control agents, and there is at least potential benefit to subjects in many clinical trials of new drugs. Further, the U.S. EPA still does not have in place any mechanism to safeguard the use of human subjects in their approval process. In contrast, in the case of phase I clinical trials and other human studies, the FDA does have such a mechanism. Specifically, the FDA adopted regulations in 1980, 1981, and 1996 (FDA 2003a) that provide enforceable requirements for informed consent of human subjects in any studies that are submitted to the FDA for regulatory approval of products. Also, in 1981 and in 1991, the FDA adopted regulations requiring IRB approval for such studies (FDA 2003b). These regulations are consistent with the Common Rule cited by McAllister (2004) but go well beyond it, as appropriate given the financial interest of third parties. Regulations do not assure compliance, but the absence of safeguards at the U.S. EPA is a dangerous situation that needs to be rectified.

This, indeed, is what recently was concluded by the National Research Council (NRC) in its report Use of Third Party Toxicity Research with Human Research Participants (NRC 2003). Among the dozens of recommendations to the U.S. EPA were a number of specific recommendations regarding the scientific validity of such studies. In recognition of the increased potential for bias when there is much at stake for study sponsors, the NRC (2003) recommended that the U.S. EPA put in place a number of safeguards to assure that industry-sponsored studies performed in support of regulatory standards receive especially careful scrutiny. In many respects, these recommendations go beyond current practice at the FDA. The NRC report (NRC 2003) provided a blueprint for action--actions that need to be taken as soon as possible to dispel the uncertainties that have been created by administrative policies and court cases revolving around these issues. Whether studies such as those critiqued by Sass and Needleman (2004) would have been allowed under the stricter standards recommended by the NRC (2003) is a matter of debate, but we doubt it. Of greatest importance in our view, the U.S. EPA needs to adopt all of the reforms recommended by the NRC (2003) in order to assure, to the extent possible, that both their own research and the research they incorporate into regulations meets the highest ethical standards.

The authors declare they have no competing financial interests.

REFERENCES

Charnley G, Patterson J. 2004. Ethical standards of studies involving human subjects [Letter], Environ Health Perspect 112:A152-153.

Chart IS, Manley A, Youngren SH. 2004. Study criticisms unjustified [Letter]. Environ Health Perspect 112:A151-152.

FDA (Food and Drug Administration). 2003a. Protection of Human Subjects, 21CFR50.

FDA (Food and Drug Administration). 2003b. Institutional Review Boards, 21CFR56.

Goldman LR, Belier M, Jackson RJ. 1990a. Aldicarb food poisonings in California, 1985-1988: toxicity estimates for humans. Arch Environ Health 45:141-147.

Goldman LR, Smith DF, Neutra RR, Saunders LD, Pond EM, Stratton J, et al. 1990b. Pesticide food poisoning from contaminated watermelons in California, 1985. Arch Environ Health 45:229-236.

Haines RG. 1971. Ingestion of Aldicarb by Human Volunteers: A Controlled Study of the Effect of Aldicarb on Man. Union Carbide Corporation Study No. ALD-03-77-2215. MRID No. 00101911. HED Doc. Nos. 007001, 010450. Washington, DC:U.S. Environmental Protection Agency.

McAllister RS. 2004. Statement of CropLife America on pesticide testing involving human subjects [Letter]. Environ Health Perspect 112:A154-155.

National Research Council. 2003. Use of Third Party Toxicity Research with Human Research Participants. Project No. STLP-Q-02-02-A, Washington, DC:National Academies Press.

Sass JB, Needleman HL. 2004, Industry testing of toxic pesticides on human subjects concluded "no effect," despite the evidence [Letter], Environ Health Perspect 112:A150-151.

Tobia A, Ayers A, Blacker A, Hodges L, Carmichael N. 2004. Aldicarb study misrepresented in human testing debate [Letter]. Environ Health Perspect 112:A155-156.

U.S. EPA (U.S. Environmental Protection Agency), 1998. EPA Statement on Human Testing, Available: http://www. epa.gov/scipoly/sa p/1998/december/epastmt.htm [accessed 3 May 2004].

U.S. EPA (U.S. Environmental Protection Agency). 2000. EPA Comments on the Use of Data from the Testing of Human Subjects. EPA-SAB-EC-00-017. Available: http://www. epa.gov/sab/pdf/ec0017.pdf [accessed 3 May 2004].

Wyld PJ, Watson CE, Nimmo WS, Watson N. 1992. A Safety and Tolerability Study of Aldicarb at Various Dose Levels in Healthy Male and Female Volunteers. Rhone-Poulenc, Lyon, France. Inveresk Clinical Research Report No. 7786. MRID No. 42373-01. HED Doc No 010459. Washington, DC:U.S, Environmental Protection Agency.

Lynn R. Goldman

Jonathan M. Links

Johns Hopkins Bloomberg School of

Public Health

Baltimore, Maryland 

E-mail: [e-mail]; [e-mail]