Those who care for the health of scientific research in this country have watched with growing alarm as more and more news stories report incidences of fraud, mistreatment of patients during clinical research and abuse of animals in laboratory experiments. The stories, as far as can be told, are legitimate. The strength of the reaction to them suggests that some important nerves are being touched.
Some of the responses from the scientific community have been overly defensive. Fraud, we are told, is an aberration so infrequent as to be unworthy of serious attention; ethical objections to experiments involving human or animal subjects are answered by describing the benefits of the research, which amounts to changing the subject. Science is not now a sacred cow, if it ever was one; and scientists ought to stop behaving as though it were. The existence of abuses, however, does not reveal a fundamental, structural flaw in the conduct of science, as some have suggested. The real question is how to establish ethical standards and make choices about how much violation of the standards we can accept. This challenges us to decide how much freedom from abuse we wish to purchase, and at what cost to the scientific enterprise and to society itself.
Let me begin with fraud. Scientists are rewarded for their work not, in general, by large salaries or royalties, but by a measure of personal security and, more important, by such things as the joy of discovery and the regard of their peers. Scientists care a great deal about such things, and anything that many will work hard for, a few will cheat for.
There has always been some fraud around, and it has, for better or worse, been the academic tradition to handle it quietly. The literary plagiarist or the biochemist who cooks his data are permitted to pass quietly out of the system (or, more unfortunately, to continue to work about its edges). With the heavier involvement of public dollars and the greater insistence on accountability for them, more cases are being exposed. Has the actual incidence of fraud, as a fraction of the total work, increased? No one knows, and furthermore there are no data upon which to risk even an educated guess.
Frauds are uncovered in two ways: by oversight (utilizing an auditing process or some other compliance strategy), and by the processes of peer review. Research findings in science are confirmed, or overthrown, by replication, or its failure. That is not a test of fraud--one can reach a wrong result by innocent error, or a correct one by dishonesty-- but it is an effective way to uncover and discard the incorrect. Mechanisms of this kind have uncovered the vast majority of scientific frauds that have come to light, while the catch of auditing programs has been meager.
Given that an effective system of error detection exists, should we invest heavily in a system of fraud detection? I do not believe that would be practical. To detect a small number of events we would have to audit a large amount of work. The expense, in manpower and in chilling effect upon the research enterprise, would be extraordinary, and the social gain would be limited.
The critical problem in the use of human subjects in biomedical research is quite different. The rules to be followed under the current system of protection are clear, and there is widespread sympathy for them in the scientific community. Institutional review boards, which must approve the plans for all experiments involving human subjects, operate, over nearly their entire range of responsibility, in an atmosphere of approval and cooperation.
The Washington Post's series of articles on cancer research by Ted Gup and Jonathan Neumann ("The War on Cancer," Oct. 18-21) illustrates what happens in those rare instances when there is a special class of subjects who stand in a unique position with respect to the costs and benefits of treatment. Advanced cancer patients have the same rights as other subjects. But the prospective benefits of relief from a debilitating, life- threatening disease justify additional risks; the only argument is how much additional risk. The Post's series was for the most part anecdotal; but where it did attempt a more global analysis, it relied heavily on views from inside the Food and Drug Administration about the National Cancer Institute program. It would have been fairer had the authors researched and reported the long and often highly personal history of that disagreement. At the time I served as FDA commissioner, some of the same FDA officials quoted inThe Post's series were claiming that the NCI program had been placed "off bounds" to them, or that it received favorable treatment. I could find no basis for such claims then, and I doubt there is any now.
Had The Post's series attempted to grapple effectively with the risk/benefit question, it could have performed a valuable service. Instead, that function was left to a later editorial which, despite a certain obscurity of intent, at least recognized that "an experimental drug's chance of success and high risk of causing side effects must be balanced against the patient's high risk of death." Just so; that is the essence of the problem with trials of cancer drugs. It is a poignant one indeed, but it does not suggest a global failure of our system of protection for the human subjects of biomedical research.
No equivalent system exists for the protection of animal subjects. There are strong principles, but their enforcement rests mainly within the profession. Scholarly journals refuse papers whose authors fail to provide evidence of proper treatment of animals. This is supplemented by a system of government inspections. But these apply to conditions under which animals are kept and not to the experimental protocols in which they eventually participate.
Our discomfort about animal experimentation is based upon a complex array of feelings. Taxonomic distance surely has a lot to do with it: in the United Kingdom, antivivisection laws apply stringently to warm- blooded vertebrates, but not at all to cold- blooded ones. In this country the Animal Welfare Act exempts rats and mice, suggesting perhaps that suitability as a pet may be involved in the degree of protection we accord a species here. These features do not add up to a very clear and consistent ethical position.
Nevertheless, there is solid agreement among scientists and advocates of animal rights that animals of whatever sort should be well cared for and subjected to minimum pain in the course of an experiment. Where real differences arise is over assessments of the need for animal experimentation. It is important that such judgments rest on a realistic understanding of what animal experimentation means to biomedical research; and unfortunately, the air is full of nonsense.
The only sure method we have for assessing the health risk of chemicals is to expose laboratory mammals at relatively high doses, and to estimate by extrapolation the risks of lower doses. No other method now available has the sensitivity or reliability to displace such tests.
It is silly to argue that if we merely throw money in the direction of developing alternatives to animal testing these alternatives will magically appear. Most animal experiments take advantage of basic physiological or biochemical similarities between the animals and people. They thus employ highly complex, delicate living systems, most of which cannot be modeled by in vitro systems, nor, as is sometimes suggested, by computer models.
The idea that one could usefully spend 30 percent of the NIH budget, or anything like that amount, on the development of alternatives to animal experimentation strikes most biologists as ludicrous. If we believe that we should no longer purchase improvements to human health through the experimental use of nonhuman vertebrates, we should make that decision in the clear knowledge that we are trading something away. In this area as in others, there is no free lunch.
Science is not magic, not different from other human activities. It is an enterprise we pursue for social gain, and it has been, by and large, an effective one. If we don't like the present arrangements for its conduct, we can change them: we could probably reduce the incidence of fraud a little, reduce the risks to human subjects somewhat and reduce or even eliminate the use of animals. But we cannot do any one of these things without effects--profound ones, in my judgment--on the productivity and vigor of the research enterprise and on its capacity to improve the human condition. We can have what we want, in short, but we shouldn't kid ourselves about the price.