|This work is available here free, so that those who cannot afford it can still have access to it, and so that no one has to pay before they read something that might not be what they really are seeking. But if you find it meaningful and helpful and would like to contribute whatever easily affordable amount you feel it is worth, please do do. I will appreciate it. The button to the right will take you to PayPal where you can make any size donation (of 25 cents or more) you wish, using either your PayPal account or a credit card without a PayPal account.|
A Perspective on the Ethics of Clinical Medical
Research on Human Subjects
The following is the essential initial line of reasoning involving the general ethics of clinical medical research involving experimental interventions on human subjects. While some of this paper may have application to more general research involving human beings, its focus is what is normally called “therapeutic research”, and by that we mean simply research involving interventions that test experimental therapies, particularly in regard to phase II and phase III types of studies that seek to find out whether an intervention is safe and effective. It is not about phase I research, nor about non-therapeutic research, such as participation in a retrospective study or a survey.
Proposition 1) Medical progress requires medical research.
Proposition 2) Medical progress benefits future patients with the condition for which progress occurs.
Proposition 3) Medical research sometimes requires experimentation with newly tried treatments (i.e., experimental treatments) for human subjects.
Proposition 4) Experimental treatments can be dangerous and result in serious harm or death.
Proposition 5) Benefitting patients is the justification for conducting research that is risky or potentially dangerous.
Proposition 6) Therefore (from 1, 3, and 4) medical research and medical progress can sometimes be risky and result in serious harm or death.
Proposition 7) àBecause of the risks involved, medical experimentation with human subjects requires the most rigorous scientific validity in order to be ethical, because to put people at risk for no good reason (i.e., to put people at risk in order to obtain results that are not necessarily accurate or meaningful) would be morally wrong.
Proposition 8) There are other conditions besides scientific invalidity that make putting innocent people at risk morally unjustifiable also, such as doing so without their consent, doing so in ways that the likelihood or quality of the potential benefit is not commensurate with the risk or magnitude of the potential harm for the subject.
Proposition 9) àTherefore (from 6, 7, and 8), the most rigorous scientific validity is a necessary, but not sufficient, condition in order for experimentation with human subjects to be ethical.
Proposition 10) àThe most rigorous scientific validity requires using random, double-blind, placebo-control trials in medical research.
Proposition 11) àScientifically rigorous and valid random control trials require strict adherence to treatments that cannot be individualized.
Proposition 12) Some subjects in random control trials receive placebo treatment or experimental treatment that is not tailored to their individual needs.
Proposition 13) Hence, (from 12) some subjects in random control trials do not receive the best treatment for their conditions.
Proposition 14) àIt is unethical for a physician not to use the best treatment available for a patient and it is unethical for a physician to recommend to a patient that the patient participate in a program of treatment, such as a research program, in which the patient will not receive, and/or will not be able to receive, the best treatment available, and which may also be dangerous, harmful, or fatal for the patient.
Proposition 15) àHence (from 13 and 14), it is unethical for physicians to utilize patients as research subjects or to recommend they participate in medical research.
Proposition 17) Thus, utilizing people as research subjects in ways against their will or in a manner that would be against their will, would be morally wrong.
Proposition 18) People who give valid consent to participation in a research project are not being utilized against their will or in a manner that would be against their will.
Proposition 19) Hence (from 17 and 18), valid consent to participate in a research program is a sufficient condition for not utilizing (or exploiting) people in medical experiments against their will or in a manner that would be against their will. And not utilizing people against their will or in a manner that would be against their will is a necessary but not sufficient condition for conducting research on human subjects in a morally right way.
(Note: this does not logically imply that valid consent is sufficient for conducting research on those who give it. And this also does not logically imply that valid consent is necessary for treating subjects experimentally either. For something to be a sufficient condition to meet a necessary condition is just to say it is one way to meet a requirement that itself is not sufficient. E.g., to obtain a college undergraduate degree, it is often sufficient to take an art history course to meet a humanities requirement, and it is often necessary to meet a humanities requirement to obtain an undergraduate degree, but that does not make taking an art history course either necessary or sufficient to obtain a college undergraduate degree.)
Proposition 20) If patients, no matter what they are told, still mistakenly believe that they will definitely receive treatment instead of a placebo, and if they mistakenly believe there is minimal or no risk to being a research subject, or that the potential benefits are commensurate with the risk, then their informed consent is not valid consent or is not rational consent.
Proposition 21) According to studies, many research subjects mistakenly believe that they will definitely receive treatment instead of a placebo and they mistakenly believe there is no or minimal risk to receiving the experimental treatment, or they believe the potential benefits are commensurate with the risk.
Proposition 22) Hence, many patients should not be research subjects (i.e., research on experimental treatments should not be conducted on these patients/subjects).
There is evidence against the premises preceded by arrows: propositions 7, 9, 10, 11, 14, and 15. We will argue that those statements, as they stand, are indeed false, and that some of the arguments to overcome the objections to them will not suffice. In some cases, however, those statements can be replaced by true statements which do not cause as many problems.
We will also argue that some of the arguments meant to refute the above arguments or false propositions are unsound or miss the point. In some cases other arguments will serve that purpose better.
Problem of Placebo Controls
It should be fairly clear from the above that placebo controls and the potential dangers of experimental treatments cause two different kinds of problems, and are in some ways the lynchpin of two significant ethical concerns in conducting research on human subjects – 1) causing actual harm to (or at least not helping benefit) research subjects or 2) exploiting people by conducting research on them that subjects them to the risk of harm (or lack of benefit in the case of placebo), even when harm does not actually occur.
Consider first the alleged problem of using placebo-controls:
“Randomized double-blind placebo-controlled trials have been argued to provide the strongest test of efficacy and, as such, are important tools for advancing the evidence base supporting [...] treatment. However, such trials present difficult ethical issues, because one group, by definition, receives no treatment for the condition being studied.”[i]
This may be put in the form of the dilemma that if one uses placebos in research, one will not be “treating” (i.e., administering to) patients (i.e., benefitting them through actual treatment) as one should be administering to them; and if one does not use placebos in research, one will not be doing proper research. Thus, one must choose between taking care of patients properly and doing research properly.
This dilemma is often characterized as the dilemma between a utilitarian consequentialist approach to ethics (the obligation to do what will bring about the greatest balance of good for the greatest number of people) and a “deontological” approach to ethics that says there are some things more important than just bringing about the greatest balance of good for everyone – that, for example, people should be treated with respect and not “used” at their expense to gain benefits for others, particularly they should not be used by those who have taken an oath to place their health and well-being first.
Consider also a parallel dilemma involving the risks and dangers for the subject of experimental research:
If one harms human subjects in research, one will be treating them maleficently and abrogating the obligation to “do no harm”thus not treating people as one should be treating them; and if one does not test treatments – even if the treatments might or do cause harm -- one will not be doing proper research. Thus, one must choose between taking care of patients properly and doing research properly.
Taken together, these two dilemmas are expressed in proposition 14 (and consequently in 15 also) above.
While there are many fine articles that have contributed to the discussion of individual elements of the arguments we are considering – articles which make excellent, insightful, reasonable, and sometimes passionately eloquent points -- this paper is intended to offer a more systematic and, in a sense fundamental, philosophical analysis of the underlying issues involved, showing their relationship to each other and to the overall topic.
Furthermore, the authors will contend that there can be some valid controlled studies that satisfy the underlying logic of randomized double-blind, placebo-control or “no treatment” control trials [RCT] even without withholding treatment from anyone.
With this as background, we would now like to examine the initial line of reasoning presented at the beginning of this paper and those answers and ensuing arguments generally given in response to it.
The conventionally accepted response to the dilemma is to deny the truth of proposition 14 (and subsequently proposition 15) by invoking the concept of equipoise, which comes in two forms, with the distinction and significance between them to be explained shortly: theoretical equipoise and clinical equipoise. The basic idea of equipoise is that it is not a breach of the duties of beneficence and non-maleficence if, without professional negligence, one truly does not know which act will bring about the greatest benefit or will cause more harm than good for a subject. “Equipoise” is the word that describes a non-negligent, lack of sufficient knowledge/evidence about which treatment option, if any, is best for the patient or subject, and therefore absolves the physician or researcher for conducting research on subjects or patients, and absolves a physician for recommending to patients that they join a research trial or for condoning a patient’s joining such a trial, even if the patient receives no treatment or a placebo (and thus is not given a real treatment) and is not benefitted by medical care as such, and even if the patient is harmed by the treatment.
What the principle of equipoise in research ethics does is to allow research to be conducted on proposed (or existing) therapies when the result is uncertain and may actually benefit the subject who has the condition for which the therapy is intended or used. This is because the researcher (whether physician or not) then does not already know the answer to the question as to which treatment (or whether non-treatment) is best for the patient or for similar patients and therefore is not intentionally (or negligently) harming or intentionally (or negligently) refraining from helping the subject by placing the subject in one arm of the study rather than another. And the physician is not intentionally harming or refraining from helping the subject by encouraging or allowing the subject to participate in the study.
In propositions 14 and, the term “unethical” and its implied antonym, “ethical” are ambiguous -- in some cases referring to what is right or wrong, and in other cases referring to what is morally responsible, conscientious, and reasonable, or what is morally irresponsible and culpable. What turns out to be wrong, for example, given full information (which normally appears only in hindsight or after the fact), is not always morally culpable or unreasonable. Putting or allowing a patient to be in a research project’s control arm, where the experimental arm yields benefits, is wrong (in the sense of outcome) but is not culpable or blameworthy since one could not know the outcome prior to the placement. Similarly, but opposite, when a treatment turns out to be harmful, the right thing to do would have been to put one’s patients in the control group instead of the treatment group, even though the above dilemmas might have made that seem to be a violation of the Hippocratic oath, when, prior to the trial, the experimental arm seemed to be the one most likely beneficial.
While ethical principles seek to accomplish what is actually right, they cannot hold one culpable for non-negligent ignorance that leads to err. While utilitarianism, for example, is the principle of doing the greatest good (i.e., bringing about the greatest benefit) for the greatest number, one is not normally morally culpable for falling short of that if one in fact does the act which all available evidence would lead any rational and knowledgeable person to believe would bring about the greatest good for the greatest number, but which ends up not doing so. Unfortunately, since people are not omniscient, and must instead use logic based on reasonably best available evidence, the intended goal, is not always achieved. The most reasonable act based on all available evidence is not always the act that turns out to be right. This is not just a problem for utilitarianism or other “consequentialist theories” (i.e., theories of right that involve reference to the best actual results or consequences), but it applies to all ethical theories, since what most rationally seems, for example, to be treating a person with respect may not be what actually respects that person. E.g., if you visit someone in the hospital who is seriously or terminally ill, that may not be what they want, even if it might be what most people might want. Or arguing with most people makes them feel you are not respecting them, whereas many people feel that arguing with them shows them you respect their opinions enough to challenge them and that you respect the person enough to want to show him/her what is best for them, even though that takes effort on your part and even though you risk being disliked by them for being thought to be arrogant, controlling, or argumentative.
It might be thought that given we can at best only be accountable for being non-negligently rational, that principles such as utilitarianism or the Hippocratic Oath should be worded as: “one should do what one rationally and non-negligently believes is best for all or for the patient.” Or that “one should do what one rationally and non-negligently believes shows the most respect for other people.” And while it is true in one sense that those are things one “should” do, it is also the case that doing that will not necessarily be the right thing to do – will not turn out to be correct. Ethical principles and obligations need to refer to what is objectively right, not just to what is rational, or otherwise everyone who is rational and who has best intentions would be automatically right even when what they do turns out to actually have terrible consequences or override someone else’s rights by mistake. It is important linguistically and conceptually that we keep the distinction between what is right and what is rational, so that when the two do not coincide, we have a way of pointing that out. Ethical principles need to refer to what is actually right, but moral character involves non-negligently doing what all available evidence would rationally say that is. So there is a distinction between being right and being reasonable. People should always (try to) do what is right, but they can only be held accountable for being (non-negligently) reasonable in the attempt. In this particular context of the above dilemmas, being “non-negligent” includes being current in the professional knowledge one should have.
Applying that idea to the concept of equipoise, as it is used to try to justify research on human beings, when such research may include no (beneficent) treatment (i.e., placebo) or actual harm, the idea is that if a physician or researcher does his/her best in trying to help an ill or injured patient while also seeking scientific knowledge, the physician or researcher, as long as s/he is reasonable and non-negligently knowledgeable is not morally culpable even if the subject turns out in actuality not to be helped, or even to be harmed. Hence, as long as a competent physician is not conducting or recommending a research study that will knowingly likely harm the patient as subject, s/he is not violating his/her oath or any reasonable ethical principle about treating people decently.
However, since Benjamin Freedman’s contribution to this subject (referenced below), “equipoise” relevant to research appears in two forms, 1) “theoretical” or individual equipoise, whereby the individual researcher does not already know what is in the subject’s best interest, and 2) “clinical” or professional equipoise, whereby the profession “as a whole” (however that might reasonably be understood or determined to mean) does not know, or cannot agree, on what is the best course of action for the subject with the condition. Since Freedman introduced the distinction, clinical equipoise is generally considered the important form necessary to justify research on human subjects, in order to prevent justifying research because of the mere hunches or beliefs of an individual researcher when that seems to ignore the professionally received view based on prior research evidence. It is the profession’s collective uncertainty about what is considered best treatment, not the individual’s uncertainty that is considered to be important..
We contend however, that each of these kinds of equipoise is important for a different reason, but each is necessary in order to justify doing research, because it would be wrong to subject people to harm or lack of help in order merely to formally demonstrate what is already clearly known. Clinical equipoise (i.e., equipoise within the profession as a whole) is necessary in order to morally justify a particular research study at all, regardless of who conducts it. And theoretical (i.e., individual) equipoise is necessary in order to allow a particular researcher to do a trial. Without the requirement for clinical equipoise, anyone could conduct research just by not knowing what s/he should professionally. Clinical equipoise is a safeguard against irrational research by individuals who do not have the prior knowledge they should. On the other hand, personal or theoretical equipoise is necessary for the moral character or moral nature of the individual researcher, who should not be conducting research s/he believes will be unnecessarily harmful to (or immorally neglectful and unhelpful) to the subjects in the study even if colleagues or the profession as a whole, or on balance, think the research is justified.
But one misunderstanding of theoretical equipoise is that it requires absolute or total uncertainty about which arm of a research study will best serve the subjects who have the condition for which a therapy is being studied. We contend it does not. We maintain that all that is required is some form of reasonable doubt or uncertainty about whether a treatment will be harmful or not and whether it will be helpful or not. A physician or researcher may have a belief about the safety and efficacy of treatment, whether standard or experimental, and yet still have sufficient doubt about the accuracy of that belief to warrant not considering it to be “knowledge” or “proven”. One can believe something and yet be aware the belief does not rise to the level of certainty or knowledge, or even much probability. That should be sufficient equipoise to justify a research study. When sufficient knowledge tips the balance to warrant believing with certainty that a particular treatment option (among those being considered) is or is not best for patients, then research using those options is not legitimate. This is sometimes referred to as reaching a point where equipoise is “disturbed”; and research is only said to be legitimate when equipoise exists or is “undisturbed.” These mean that the research is justified because the safety and efficacy of the treatment under study is not yet sufficiently known (i.e., “equipoise is undisturbed”, so there is still uncertainty about it), or that the research is unjustified because the treatment proposed or under study has been already shown to be safe and effective or because it has been already discredited either because it has been sufficiently demonstrated to be ineffective, dangerous, or is not worth the troubling side-effects even if they are not dangerous (i.e., “equipoise is disturbed” because there is now certainty about the treatment’s safety and effectiveness, or its lack thereof).
The same elements, whatever they might be, for determining whether sufficient equipoise exists to begin or conduct a research study, also determine the proper “endpoint” of a trial, the point at which there is sufficient evidence to end or discontinue the trial because there is no longer sufficient doubt whether the treatment is safe or effective. More will be said about this later.
The two different kinds of equipoise, however, are morally necessary, but for two different purposes. Clinical equipoise (uncertainty within the profession as to which arm of a trial will be best for subjects) is necessary in order to justify the research project being conducted at all, no matter who does it; and theoretical equipoise (uncertainty about the same thing by an individual researcher, or by a physician recommending a patient participate in a research trial) is necessary to justify the particular researcher’s conducting the study and to justify any physician’s allowing or recommending a patient’s participation in it. While lack of clinical equipoise precludes anyone’s doing the project, an individual physician may be obligated not to conduct that study or advise or condone his patient’s being in it, because s/he disagrees with the general professional opinion that there is insufficient evidence to warrant the trial. While colleagues may not believe there is sufficient evidence that a trial will harm or not benefit a patient as much as a different treatment, the physician who does would be morally inconsistent (perhaps approaching dereliction of duty) to conduct such a study and/or condone his/her patients’ participation in the study if someone else conducts it.
The Significance of the Duty to Minimize
While clinical and theoretical equipoise are both necessary to justify research involving human subjects, they are not sufficient.
As pointed out in proposition 17, it would be morally wrong to utilize people in research studies against their will or that would be against their will.
It is also normally morally necessary that a proposed treatment have sufficient evidential and/or theoretical basis for making it reasonable to believe it will be successful, before it is attempted on human beings. Normally one doesn’t morally just “try anything” on people to see what it does, though there may be some last ditch desperate efforts when otherwise all is lost anyway, that it is not unreasonable to at least try something, no matter how unlikely it might be. And it is normally morally necessary for the likely risk/benefit ratio – in terms of probability and magnitude, and in terms of significance for the patient – to be reasonable.
But we want to discuss here a different requirement necessary to morally justify research on human beings – the moral obligation not to subject them to unnecessary risk, where “risk” refers both to the risk of harm from the treatment and also the risk of harm (or unnecessary worsening of their condition) from withholding treatment (as in placebo-controls or “no-treatment” controls). In particular, we want to discuss the necessity of minimizing risk from not-treating a research subject, by withholding treatment in the form of giving no treatment or in the form of administering a placebo. (There is a difference because “placebo effects” are effects, though their cause is not always, or perhaps ever, understood or known.)
Much has been written in the controversy about use of placebo controls, particularly in regard to the controversy of their (moral) appropriateness as a control instead of standard treatment, and we do not wish to try to summarize or revisit all of it. Instead we wish to put it in a broader context.
First, [for later reference, we designate this as “Paragraph A”:] it is generally, though not universally, recognized that placebo controls in place of treatments which are known to prevent death or serious, particularly irreversible morbidity are not morally legitimate, at least not unless a case can be made for altruistic martyrdom in the name of science or unless there is an instance where the subjects will die very soon anyway from some other cause and dying from lack of effective treatment will not be any worse for the subject than the way s/he will otherwise die. We will return to this point because it is a specific corollary of a more general principle.
Second, we wish to deny proposition 10 at the beginning of this essay and point out that random control trials (RCTs), while extremely significant, are not the only way or necessarily, in certain circumstances, even the best way to gain scientific or medical knowledge. RCTs have certain logical limitations besides any practical or moral ones.
1) They cannot detect result differences when causal factors are unknowingly proportionally distributed among cohorts studied, which can and does happen when causally relevant factors are unknown. Unfortunately, it is logically possible, and sometimes actually the case, that large sample sizes, intended to cancel out incidental or coincidental factors, can also cancel out or mask actual causal factors, particularly when relevant, significant, actual causal factors are accidentally and unrecognized equally or proportionally divided between the control and test groups, or when they are accidentally and unrecognized distributed relatively equally among different cohorts. This can happen particularly when there are multiple causally influencing factors. E.g., in hormone replacement treatment for women (HRT), early evidence that HRT was substantially beneficial turned out not to be true when it was given to women a certain age beyond menopause, rather than the age of menopause. This harm was masked in that group by the gains made overall by women of different ages when combined in the same cohort. For randomized control tests to be effective, it is crucial that the causal factors be divided properly in the experimental group into cohorts that separate them from each other and that eliminate them from the control group cohorts. Unfortunately, when the causally influencing factors are not already known, and worse, when they are not even suspected, that cannot or will not likely be done. In the HRT study that showed age, or elapsed time past menopause, was a factor, it was fairly reasonable it would be detected in a controlled study, because grouping cohorts by age is one typically normal way to proceed. However, if the causal factors that made HRT dangerous for some women and beneficial to others were evenly distributed throughout all the cohorts, the overall gains might have masked the dangers for those with the unknown causal risk factors while simultaneously seeming to lower the beneficial effects of those without it, since they would be not be in (a) cohort(s) by themselves.
2) Even when they determine that a treatment might be safe under known or common conditions, they cannot determine whether an unexpected new environmental condition will make a treatment dangerous. I.e., some new diet may become trendy that turns out to put people at risk who are have taken a medication, or who have had a procedure, that was tested before such a diet was conceived and promoted.
3) The safety and efficacy shown by an RCT is only for the time period over which the test is conducted. Any longer time periods may show up problems that do not occur in the shorter run. That is why Phase 4 and also epidemiological and other sorts of studies need to be conducted.
Corollary to 3) Since evidence never comes clearly marked as evidence, and particularly does not come delineated as “sufficient evidence”, the “endpoint” of any RCT is always going to have to be judged by some kind of reasoning process that says when there is sufficient evidence to constitute knowledge about whether the intervention is safe and effective or not. Or to put it into the technical parlance, there needs to be a rational determination made as to when equipoise is sufficiently disturbed to require no longer giving the treatment to the experimental group -- because it is then known to be unsafe or ineffective-- or to require no longer withholding it from others -- because it is then known to be both safe and effective. Many of the problems involving what constitutes reasonable equipoise to begin a study also apply to when sufficient data is collected to end it because equipoise is no longer tenable; that is, short of obvious mortality and serious morbidity results during an RCT, of course.
These limitations of RCTs need to be taken into account when using placebos or no treatment in order to intentionally withhold treatments from people, because these limitations weaken the argument for using placebos in the name of research under certain circumstances. Insofar as an RCT will not likely yield sufficient valuable knowledge, there will be certain situations where the argument for using placebos (and thus withholding treatment from subjects) is sufficiently weakened to make it ethically unjustified.
It is notoriously difficult in human studies to totally control or even recognize the variety of factors that represent the (pre)conditions of the study. Study groups and treatments that seem identical in all but one way, may in fact involve differences that are unrecognized, and which are hidden by the size and cohort divisions of the study group. If we count the (pre)condition of the subject as part of the initial and testing conditions, it is virtually impossible to test identical subjects because no two people will likely have the same chemistry or previous histories, exposures, experiences, etc., let alone pairs of groups of individuals used as the test subjects and as the control group. Sufficient group size, along with controlling for the seemingly relevant factors is meant to try to cancel out or minimize those differences. Otherwise almost any study would need to be done only on identical twins, and even then there would be acquired and environmental, if not also genetic mutation, differences.
In many aspects of science where we are unable to find invariable patterns of concomitant or successive conditions, we still notice statistical patterns. Logically we could say either that there is an element of randomness involved that has no specific cause or invariably prior condition or we can say that we believe there is one or more causes or invariable sets of conditions that precede the effect (or that precede and determine the immediate cause of the effect) and we just do not yet know what that is or what they are. Given the variances between humans, and our still incomplete understanding of physiology, biochemistry, genetics, etc., much of modern medicine and research involves statistically correlated, rather than recognizably invariable, conditions. We tend to find apparently causally related factors, rather than definitive mechanisms or specific, determining invariable sets of isolable preconditions for states of health.
Given no possibility of perfectly matched control and experimental subjects, and given the potential variety of pathways and influencing factors, the idea is to come as close as seems reasonable with available evidence to having matched control and subject groups. That fact is sometimes lost on those who would overzealously require such impossibly matching control groups be found that no research can be attempted (particularly in regard to treatment for relatively rare or “orphan” medical conditions where it is difficult or impossible to get research and control sample sizes that meet the critical mass necessary to validate the research statistically). And it is lost on those who would have such lax standards of control that the research is invalid. Determining the criteria for a valid and relevantly matching control group is a matter of reasoning and reflection that takes into account available evidence and the knowledge and reasonableness of relevant assumptions. It is not a matter of slavish adherence to arbitrary or unreasonable rules, or rules that are not relevant to particular situations. Nor is it a matter of replacing one set of arbitrary, irrational, or irrelevant rules for another that is simply more lax.
Third, we wish to deny the strict truth of propositions 7 and 9, because it is not necessarily “the most rigorous scientific validity”, but the most reasonable evidence, whether strictly scientific or not, and whether strictly empirical or not (as opposed to theoretical and logical but based on, and consistent with, empirical evidence). Science certainly makes great contributions to knowledge, and clearly, empirical evidence is extremely important, but science is not the only form of attaining knowledge; and moreover, much of science is not strictly empirical but uses other forms of reasoning as well, particularly in research involving “theoretical constructs”, which may generally be described as suspected causes for observable phenomena, but which themselves cannot be directly observed. E.g., fossils give evidence for the prior existence of certain kinds of organisms which cannot now be directly observed to exist. Much of medicine involves indirect evidence of the existence and workings of theoretical constructs. And much of medicine involves probabilistic evidence which may imply but also mask specific causal relationships. Science also often employs mathematical models whose relationship with the physical world is not always known, which is why physicists often have to discuss ideal particles under ideal conditions. In short, science is not as strictly empirical as it is often supposed to be, and the empirical nature of science within the larger context of evidence and reasoning is not (always) fully understood and appreciated. Rather than limiting medicine to what can be demonstrated scientifically, especially perhaps when the limitations of science are not fully appreciated, it is more reasonable to amend propositions 7 and 9 to read something like:
Proposition 7’) Because of the risks involved, medical experimentation with human subjects requires the most rational methodology for gaining knowledge in order to be ethical, because to put people at risk for no good reason (i.e., to put people at risk in order to obtain results that are not necessarily rational or meaningful) would be morally wrong.
Proposition 9’) Therefore (from propositions 6, 7, and 8), the most rational methodology for gaining knowledge is a necessary, but not sufficient, condition in order for experimentation with human subjects to be ethical.
That means proposition 10 would have to be changed to something like the following (which will make it even less often and less apparently true than it is as it is stated above):
Proposition 10’) The most rational means of acquiring knowledge requires using random, double-blind, placebo-control trials in medical research.
For an example of the irrational, unnecessary, costly, mistaken, and egregious use of an RCT, see Robert Truog’s “Randomized Controlled Trials: Lessons from ECMO,” Clinical Research 40 (1992): 519 – 27. Also see “The Continuing Unethical Use of Placebo Controls” by Kenneth J. Rothman and Karin B. Michaels, New England Journal of Medicine 331 (1994): 394-98.
What the above signifies is that RCTs are not always ethically warranted or legitimate because they will not necessarily yield the most valuable or rational evidence that is being sought or because there will be far less Draconian ways to gain sufficient knowledge. Other methods may be more morally acceptable and yield as much, if not more, reasonable evidence for the safety and efficacy of a proposed treatment.
We wish to return now to the above designated “Paragraph A” to propose the more general principle:
Randomized Control Trials, because they withhold treatment from subjects, are not morally legitimate or scientifically necessary to use when there are other, less likely harmful, ways to know the information desired about the progress of the condition under study when no treatment is given. This particularly includes conditions whose progress without treatment is already well-known. And it applies even more strongly as the effects of withholding become proportionally more serious and harmful in regard to mortality and the significance of morbidity, especially irreversible morbidity.
Paragraph A derives its plausibility from the fact that medicine already knows what the progress of certain conditions will be if left untreated, or if treated in the conventional ways. That is the same rationale that gives “compassionate use” its plausibility in giving patients experimental treatments which are not likely to make them significantly worse off because you already know how bad they will be without a new intervention (where “worse off” includes aspects of quality and “significance” besides just life and death). Moreover, medicine often already knows in the cases of terminal illnesses the manner in which death will come; i.e., the physical deterioration that will precede it, the pain or agony likely to precede it, if any, etc. Similarly with regard to diseases or conditions which cause harm other than death, whether it is a long drawn-out process of loss of various physical or mental abilities or whether it is increase in pain or the effects of something such as arthritis, diabetes, paralysis from spinal cord or brain injury, etc.
Even in cases of less significant morbidity, it is unnecessary to test for placebo or “no treatment” when those effects are already known. The whole point of placebo controls is to see what happens that is different from the experimental treatment. And if you already know what happens without treatment, there is no need to withhold a treatment from people whom it might significantly help, simply for the purpose of finding out what will happen to them. There may be safety reasons not to give a new treatment to many subjects, of course, in order not to put any more people at risk than is necessary to get initial safety and effectiveness information but that is different from withholding treatment just to see what happens when treatment is not given.
This seems particularly true for physical (in the sense of non-psychiatric) illnesses or conditions where a placebo effect seems highly unlikely, as in surgical repairs of the tetralogy of Fallot for dying cyanotic babies. Some surgeries, of course, may have various components, not all of whose affects on the condition are fully understood. But if one has different trial arms that involve different forms of (the) surgery that are tried on different subjects, that is not a placebo nor a sham surgery. It involves numerous real surgeries of different sorts, with different total components, in order to see which one is more effective, if any. Any time one is “opened up” or drilled into by a physician for medical purposes and then closed, even without anything else’s being done, that is still surgery, even if it not ‘the’ full surgical procedure being tested as a treatment. And if something in numerous patients’ condition responds favorably to any form of the surgery, then that is an effect of the that surgery, regardless of the extent of the surgery or the unexplained surprise of the result.
Research into treatments for conditions involving more of a psychological component may benefit from knowledge of a placebo effect, but even then that may show the placebos to have some therapeutic properties (whether positive or negative).
Furthermore, in general, particularly for primarily non-psychological conditions, it seems to us unnecessary to use placebo control groups for a condition where placebo control has already been used in studying another treatment with the same kind of delivery system. E.g., if there are two different experimental treatments to be tested for arthritic knees in similar populations, where both treatments involve taking pills twice daily, it seems unnecessary that if the one trial is conducted first, with placebo controls, for the other trial to run another placebo arm instead of just using the same data from the first trial for its placebo-control test, and just run the experimental part of the test, to see whether the treatment is relatively safe and effective (i.e., relatively more safe and/or effective than placebo).
And it does no good to reply that a placebo control might turn out to have different results in one research study measured against treatment A than it does in another study measured against treatment B. If that were the case, as seems to happen sometimes, then it is really not clear what the placebo control even means, for research or pure science, let alone for clinical practice. Miller and Brody, in discussing one particular research trial to argue for the necessity of placebo controls (in this case, versus active controls) make the following two comments (not terribly far apart), the first being about the results of the trial in question, and the second describing the evidence available to the IRB before they approved the trial:
“Neither hypericum nor sertraline was found to be superior to placebo on the primary outcome measures.”
“Approximately twenty-five clinically available antidepressants, including sertraline, have been shown to be superior to placebo.” [ii]
Miller and Brody argue that the second statement shows that IRBs (Internal Review Boards) do not really consider equipoise as being necessary to do research on human subjects, because if they did, they would not have allowed the placebo control arm of the study, since the active control was already demonstrated to be superior to placebo and there was no “equipoise” any longer existing about that. There are other possible reasonable explanations for the potential rationale of the involved IRB, but the curious aspect of these two statements is that on the surface they are incompatible in a way that needs to be explained, for in general if S is superior to P, and P is equivalent to H, then logically, S must be superior to H. Moreover, S cannot logically be both superior to P and not be superior to P; but RCTs Miller and Brody describe supposedly show it to be both.
It is possible that the “primary outcome measures” for the study under discussion are different from what was studied in the previous studies, but, if so, that would obviate Miller and Brody’s point that equipoise does not exist (for the primary outcome measure under study). It is possible that the IRB did not trust the previous study results, but that seems to obviate the point of depending on research studies or of including placebo controls in them. Neither of these explanations will help Brody and Miller. But more importantly, if the outcome measures are the same in this study as were tested previously, then it means that placebo controls give different outcomes in different studies; and that should be troubling for anyone who holds that random placebo-control studies are a meaningful, essential part of research on human subjects.
It seems more likely to us that the whole point of placebo controls is to try to determine whether the results of the treatment arm of the trial really show safety and effectiveness of the treatment, or whether some significant part of the result can be attributed to just being administered to in a research study, even with a supposedly inactive, inert, or irrelevant agent. As explained previously, Freedman makes the strong case that placebos do not allow one to obtain a “net therapeutic effect” of an experimental treatment. And the Miller/Brody example seems to indicate that placebo control studies are not consistent, or not carefully enough described, understood, or interpreted to warrant their being considered absolutely necessary or sufficient for determining the net therapeutic effect of a treatment.
This is not to say there is no significance to the results of a placebo control. It is to say that placebo studies need to be understood better themselves, and their significance needs to be more carefully considered within the broader context of trying to determine how much effect (for benefit or harm) an experimental intervention really has. Placebo controls can likely contribute to that understanding, but not in as obvious or straightforward way as is usually assumed. There are numerous articles available that make good cases for when placebo controls are justified and when they are not, and why.
Plus it seems obvious to us that if the progress of a condition is already known when left untreated and is also known when treated with different previously used medications or procedures, then it should be reasonable to believe that any significant diversion from any of those paths, when the delivery method of the new treatment is the same as has been used previously (e.g., pills or injections in the same location, similar frequency, etc.) is to be attributed to the treatment, giving a net therapeutic effect. This would be a net therapeutic effect for clinical purposes even if some sort of (especially permanent) “placebo effect” is part of the cause of the treatment’s success or failure.
Placebo or non-treatment seems more important and reasonable when the progress and/or outcome of a condition is not already really known, and/or where the effect of standard accepted treatment is not already really known. Unfortunately the effects of many “standard treatments” are not necessarily known as well as they could and should be. In such cases, a placebo-treatment arm or, we think more importantly, a non-treatment arm can be justified if there is also no reason to believe significant additional harm will befall the subjects because they are in that arm.
But to repeat, the only point of placebo or non-treatment controls is to see what the net therapeutic effect of an experimental treatment is – and insofar as that can be determined in a rational way without withholding treatment from research subjects for which an experimental treatment for a significant condition seems promising, it would be wrong to put them in a non-treatment or placebo control group, because the results of that are either already known (in the case of non-treatment) or they are difficult to interpret (in the case of significant placebo effects that have never shown up before with the same kind of delivery system used previously in studying that condition).
Note, it is the purposeful “withholding” of treatment for which there is great evidence of safety and efficacy that causes the original ethical dilemma for researchers in general and for physicians in particular. It is not merely “not providing treatment” that is ethically problematic, but intentionally not providing available treatment which could be provided that is morally problematic. Physicians were not culpable for not successfully treating those with bacterial pneumonia before penicillin was discovered. Physicians are not responsible for not treating people they do not know are ill. They are not responsible for not treating those who suffer heart attacks miles from civilization while hiking alone in the mountains or desert. They are not responsible for not using equipment, procedures, or medications they do not have available to them even if they are in the desert or mountains with the person who suffers an illness or injury.
Even researchers are bound by ethics in general (as is everyone) not to harm innocent people without some justifiably overriding reason. And everyone is bound by ethics to help those they can when it does great good for the beneficiary without an incommensurately greater cost or risk to the person doing the act. Physicians do have an added obligation to help others medically beyond the minimally required effort that everyone should make, because they have taken an oath to do so, because they have the knowledge to do so, because they accept that additional obligation in offering to treat people who come to them, and because in many cases the public helps support their medical education in return for having physicians available to provide for them, and because physicians are often accorded certain financial benefits that incur obligations with them, just as any service for fee does. In short, there are a number of obligations over and above those everyone has, which people (including physicians) can and do incur, and which they are then responsible to honor. Life is full of such additional obligations, for example, keeping one’s word to do something simply because one gave it, even though one might not have had to do the act if one had not said s/he would and even though one did not have to say s/he would in the first place. But promising or committing to others that one will do something which they then depend on you to do because of your commitment, bestows a prima facie obligation then to do it. Most verbal commitments, promises, vows, oaths bestow such additional prima facie obligations that have to be honored unless there are morally overriding circumstances that justify or even require abrogating them. But even without voluntarily incurred additional obligations, everyone still has prima facie obligations to do no harm, and to be reasonably helpful to others.
Moreover, medical researchers, by virtue of their expertise, and by virtue of their seeking or accepting people into their programs, we believe, have an added obligation to look out for their subjects’ best medical interests as it pertains to the study. It is not physicians alone who have this additional obligation. Nor does merely acquiring consent allow one to mistreat another individual. This is particularly true, of course, if the consent is for the incurring of some reasonable risk, not just any risk, such as due to malevolence, insensitivity, negligence, etc. And even if someone were to be so irrational as to consent to being mistreated, that does not give anyone the actual right to mistreat them just because of that consent. If a patient trusts a physician or a subject trusts a researcher and gives carte blanche approval to do what they think is best, that does not give the physician or the researcher moral justification for mistreating them to serve his/her own ends, any more than entrusting one’s finances to a financial planner with the proviso to do what s/he thinks best, is an invitation or advanced pardon for them to embezzle the assets.
The Purpose of Informed Consent
There are a number of purposes for requiring the informed consent of research subjects:
1) It is supposed to signify that the research subject understands and agrees to what s/he is “getting into” or to what risks s/he is allowing him/herself to be exposed, and for what possible gain if any or what possible beneficial results there might be for medical science and future patients. Studies have shown this purpose is not often met, because research subjects often assume, no matter what they have been told, that the researcher or their physician will not allow them to be harmed, and they assume regardless of their intellectual understanding, that they will receive the treatment they need, and that they personally will not be given a placebo even if placebos are given out randomly.
2) It affords the opportunity for the potential research subject to let the researcher or the physician know which potential outcomes s/he might find acceptable and which ones might be unacceptable to him or her, in order to double check with the researcher or physician that this research meets or has a reasonably good chance of meeting those expectations and not likely causing significant unacceptable harm. Meaningful informed consent can meet that purpose, but formal procedures for obtaining informed consent might not, and probably won’t, particularly if they are written in legal and scientific or medical technical language that is unintelligible to the population which has to sign them. But even when the language is understandable to the subject population, that does not mean they will understand the significance of the words or statements even if they can restate the provisions of the agreement in their own words. One can imagine someone’s taking a drug for erectile dysfunction after consulting with their physician and after having seen a million ads for it on tv, still saying after an ocular blood vessel occlusion “Well, yes, I knew it could cause sudden ‘loss of vision’, but I didn’t know I would be blind! That I would end up like this! – unable to get around, go to work, drive, watch tv, read, and…..” It is our contention that the researcher has an obligation to make sure the research really is in the subject’s best potential interest and that it is not potentially devastating to the subject. And it is our contention that the researcher has an obligation to make sure the subject understands the risks, the probilities, the magnitudes, and the significance of the magnitudes! And those things are not accomplished by merely stating facts to the subject and then obtaining the subject’s signature on a form. So although a verbal statement of the facts and a signature on a form may be necessary, it is not sufficient for making sure that a researcher is not exploiting or taking advantage of the subject, whether purposefully or inadvertently.
3) It meets certain legal requirements to absolve the researcher from charges of battery, or from non-negligent iatrogenic injury, though not from negligence that causes harm. From a legal and somewhat cold or calculating standpoint, it also acts to make the volunteer subject essentially complicit in the result so that s/he “cannot complain” about it later. But morally that would not follow if the research procedure is not as good as it should be, even if it meets an approved protocol, or if the explanation to the subject prior to their consent was not as good as it should have been, even if it meets the legal requirements for the “informed” part of informed consent.
4) It signifies, if the consent is truly or validly informed, that the research subject is volunteering of his/her own volition for the research study. It may or may not achieve this purpose depending on the understanding of the subject, which we addressed above in terms of the subject and the researcher both understanding the signficance of the possible outcomes besides their probabilities and magnitudes. It perhaps is safer to say it shows the research subject is not being acted upon in a way that is against their will, but it does not show they are not being acted upon in a way that would be against their will (if they understood it better or after they see the outcome).
5) It allegedly proves the research subject is not being exploited or taken advantage of. It cannot do this, however, except under certain circumstances which do not always obtain, nor which can be demonstrated in any formal way to obtain.
In the context of “coerced” compliance to join a research study, Franz J. Ingelfinger argued:
“[I]t must be granted that natural contingencies (“acts of God,” things which come to pass naturally, those contingences which we cannot hold anyone responsible for) do not render a person unfree, nor do they render unfree the choices which a person makes in light of those contingencies.”
If one holds that uncoerced cooperation precludes exploitation, this might reasonably be interpreted to imply that no one is exploited who consents to reasonably potentially beneficial research studies because they have a condition resulting from a natural contingency (e.g., serious illness, accident, genetic condition, etc.) That would be untrue, however. Take the simple case of your car’s breaking down in a small town with one mechanic who, upon seeing your out of state license plates, and ascertaining that you are on an important business trip for which time is crucial to you, quotes you an exorbitant fee for making what is a fairly simple repair for him for which he would charge townspeople much less. Your predicament is due to a natural contingency in the above sense, but nevetheless because the mechanic has you “over a barrel”, he is exploiting you, or attempting to exploit you. Of course you are free to forfeit the business deal by turning him down, so his exploitation is not guaranteed. And you are not necessarily coerced into accepting it. But we think it is reasonable to consider this a case of coercion, and certainly of exploitation, if you feel you need to accept because you need the business deal and it will still be very profitable for you. You have a certain, miniscule amount of freedom in the matter, but it might be irrational in this case to reject his services, even though you know you are being taken advantage of.
6) Informed consent is alleged by some, mistakenly we have argued above, to absolve researchers or physicians from treating the subjects with a primary emphasis on beneficence and non-maleficence because they are then instead supposedly administering to the patient/subject with a primary emphasis on meeting his or her wishes and respecting his or her autonomy. Specifically it supposedly absolves physicians from either treating subjects, or recommending participation in research that treats subjects, in ways that are not necessarily in the patient’s best interest if that is what the patient wants or is willing to do. Even in medical practice, sometimes patient autonomy and medical beneficence/non-maleficence conflict, and the recent trends in medicine place a higher value in many cases on respecting patient autonomy than on a professional expertise, often considered “paternalistic”, approach of treating (or wanting to treat) the patient in a way that is in his/her best interest purely from the standpoint of medical health, whether the patient accepts that or not.
It is our contention that whether one exploits research subjects is not just about how one behaves toward them but also involves one’s state of mind and intentions toward the subjects. If one sees research subjects primarily as facilitators of his/her own research goals, no matter how laudable those goals might be, rather than as human beings who deserve to be attempted to be helped, one is exploiting them even if one is doing the very same thing as one who sees his/her mission as trying to help the person, and who is not exploiting or taking advantage of their conditions. Apart from conforming to certain necessary standards of disclosure, it is the researcher’s compassion, understanding of his/her subject, his/her sensitivity in general, and integrity in general that determine whether s/he is exploiting the subject or not. This is not to say that an observer cannot tell whether a researcher is exploiting his/her patients or not. It is to say it takes more work for an outside observer (such as an IRB) than merely collecting signed consent forms that have been approved by attorneys or even ethicists. It involves knowing the sensitivity and integrity of the researcher. More importantly, we believe that a researcher who really cares about a subject’s wishes in regard to his or her potential treatment outcomes will likely (though not necessarily) do more of an ethical nature to try to insure those wishes than will the researcher who merely views the subject as an opportunity to gain knowledge and personal benefit from discovery or publication. It is not that the one will be any more successful necessarily than the other in securing medical benefit for the patient, but that the former will simply more likely (but not necessarily) treat his/her subjects more humanely and ethically in the process.
Now Jay Katz has devised a list of minimally necessary disclosures to insure obtaining valid informed consent. Among these are information that the research subject’s therapeutic interests will be subordinated to scientific interests; that they may be harmed by the research instead of being helped; that other options besides participating in research may be more in their self-interest, etc. While such lists are important to prevent leaving out of the informed consent process something previously realized to be important, these lists do not preclude the investigator’s having an exploitive frame of mind, that might put the subject in more jeopardy that is unacceptable than necessary to him or her no matter how many provisions are formally met in terms of the letter of the law or considered ethical standard of the time.
We suggest that while such lists are important, it is equally important that the researcher genuinely cares about the well-being and best interest of the individual subject in order to more likely notice potential significant risks and effects not covered in the lists, and that the researcher makes sure that s/he and the subjects understand not only the probabilities and magnitudes of the possible outcomes, but their significances for the subject as well. And this is an ongoing process even after the treatment starts, because the subject is allowed, and should be encouraged, to withdraw from the study at any time the results appear to be causing or tending toward a negative significance. All this requires sensitive insight and effort by the researcher, not mere formality, but what is required is a reasonably normal moral and humane effort, not a superhuman, perfect one.
Implied and Substitutionary Consent
We pointed out earlier that obtaining valid or legitimate informed consent was a sufficient condition to meet the necessary condition of not treating people in ways against their will or in a manner that would be against their will. There are emergency conditions where it is impossible to get consent from the potential patient/subject, however, such as in emergency situations where an experimental intervention must be done within a limited time frame if it is to be effective at all. The patient may be in no condition to give a meaningful or valid consent (or possibly even a valid denial). And there may not be available a relative or legal representative who can give consent. If there is, they may not be able to weigh the pros and cons of accepting or rejecting the intervention and may want to rely on the physician’s best judgment. Emergency rooms normally are permitted, unless there is clear evidence to the contrary, to presume patients would want to be treated; and in some cases (e.g., where a patient is in pain and traumatic psychological despair) may even disregard pleas to be allowed to die untreated. So valid informed consent from the patient may not be necessary to enroll them in a research study for an intervention, say, following serious injury where the intervention has to be administered in a timely manner that precludes getting informed consent, particularly if there is good reason to believe it will be helpful and little reason to believe it will be harmful.
This is one area where IRBs might legitimately substitute their judgment for that of the patient or might allow the physician to substitute his judgment for that of the patient. But that is because no matter what is decided, it cannot be decided in a credible (or perhaps even possible) way by the subject/patient. So no matter what is done, the decision will, of necessity, be made by the IRB and/or the treating physician.
The more usual case however is that physicians, researchers, and IRBs guiding their practices in obtaining informed consent, act as filters that control the information given to recruits for research studies. The information that is given, and the manner in which it is given, particularly the likely reliance on the researcher’s and or physician’s advice by the patient, will influence the consent decision that is made when a patient or his/her health care proxy is able to be consulted. So, in part or to some degree, researchers, physicians, and IRBs cannot avoid substituting their judgments for the potential subject’s judgment, though it is not totally the same thing as deciding on behalf of the subject whether to participate in the trial or not. In some cases it may be tantamount to the same thing. Hence, there is a strong obligation for IRBs and researchers to understand the potential subjects’ perceived needs and significance to those potential subjects of the different options, so they know what information to provide, which is basically the information that even if the worst outcome occurs, the patient/subject or his/her health care proxy (or most sensitive, rational people) will not feel something significant was withheld in the consent process that should have been divulged and made crystal clear. Informally, this might be expressed as the consent process should be one that would be found to be acceptable and not likely to be embarrassing or disgraceful if it were to be shown on 60 Minutes after some tragedy. In regard to projects involving vulnerable patient populations, IRBs must have members who are either part of those populations or whose expertise involves them. But still, the patient has the opportunity to ask further questions, seek other counsel, and perhaps reject the advice of the physician or researcher.
However, Truog et al have suggested a model that includes what we call substitutionary consent for certain kinds of cases, whereby research subjects would not have to give specific (as opposed to general) consent for particular treatment arms or options.[iii] One of the requirements of their proposal is that “no reasonable person should have a preference for one treatment over any other, regardless of the differences between the treatments being compared.” They recognize, however, “[a]lthough the reasonable-person standard is widely used in the law, it is far from perfect. For example, there is always the possibility that a patient may be unusual in ways that cannot be anticipated and that would lead the patient to have a preference for one treatment over another.” But this is precisely why informed consent, when it can be given is necessary, because there is no reason to believe that a patient will be willing to be treated in a way that his physician, a researcher, or IRB members from the community who are presumably supposed to represent “everyman’s” thinking in regard to what they would want, perhaps ala the Golden Rule. It is one thing for the members of the community review board to heavily influence the decision a research recruit might make by how they structure the informed consent information required to be given; but it is quite another, we believe morally inappropriate act, to make that decision for the subject when the subject could reasonably make the decision him/herself, which is basically the review board’s saying “we didn’t think you’d mind.” But informed consent is meant precisely to find out whether the patient or subject would mind.
The Philosophical/Ethical Function and Justification for IRBs/ERCs
Notice that nothing in the propositions 1-22 mentions Institutional Review Boards or Ethics Review Committees in regard to conducting research ethically on human subjects.
Yet, Institutional Review Boards or Ethical Review Committees are required by the Common Rule[iv], by the International Ethical Guidelines for Biomedical Research Involving Human Subjects[v], and by The ICH Harmonised Tripartite Guideline – Guideline for Good Clinical Practice[vi], and their duties are listed in these documents. But the point of using them as a primary way of protecting human subjects, in accord with other provisions in these documents is apparently presumed obvious. We believe that leads to problems that need to be addressed, because the justification of IRBs and ERCs is not so obvious as it might appear; and the underlying functional role they thus (should) play in protecting human subjects is not as clear as it needs to be.
It is fairly clear that IRBs and ERCs are intended to be a check and balance for overly zealous, overly optimistic researchers, and to guard against biases, insensitivity, cruelty, reckless or selfish disregard, conflicts of interest, or anything else that would cause or allow the exploitation of human research subjects and/or that would cause or allow the misuse of data collected and interpreted. The question is how they are to do that, and why they are necessary if there are guidelines, rules, codes, standards, laws, etc. already in place. They are not merely intended to be police who enforce these rules, though they are to enforce the rules. But they also are meant to interpret the rules in order to apply them to specific situations; i.e., specific research projects.
But what insures that the IRBs or ERCs do this right or that they do it any more correctly than would individual researchers with insight and integrity who are the ones designing and proposing the projects in the first place?
In Nazi Germany, would not IRBs have simply been populated by like-minded people as those who did the heinous research? Would not a simple legal requirement simply of obtaining “informed consent” have been sufficient to prevent the atrocities done if the society had really cared to protect human subjects, particularly Jews and other non-Aryans? At the time at least that the Tuskegee study was started, would not an IRB have been as accepting of the beliefs of the time as the researchers were? What exactly, besides much labor and expense, does an IRB or ERC add to the process that ethics rules and laws, with suitable penalties do not?
And what if the research suspicion in the Tuskegee study had been right that syphilis in the African-American population was not harmful, as it is in the Caucasian population and that the risky treatments of the time (prior to penicillin) would have been unwarranted? Is it not just the hindsight result that syphilis acts the same in the black population as it does in the white population, that makes the idea seem so egregious and racist to us now? Or was it a racist idea? Is it racist to be concerned today about how hypertension or tuberculosis in a population of African descent might be different from the way it is in regard to a population from European descent? Or are there genetic and/or historically induced natural selection, adaptation factors of native diet, lifestyle, or acquired immunity from epidemics centuries ago in Europe, that make these two different populations different in regard to hypertension and tuberculosis? But more important for the issue at hand, why is an IRB more likely to decide these matters better than researchers or legislators or judges? And why, if an institutional review board is affiliated with an institution doing research, should they be less optimistic or zealous at wanting to see successful research projects than the individual researcher or department designing, promoting, and conducting the research?
What exactly is it about IRBs and ERCs that is supposed to make research on human subjects that they approve more ethical or better than if qualified researchers were left to conduct research on their own, and if researchers were punished in court or by those who license or employ them when they treated subjects immorally and/or violated codes of research?
We believe there are a number of important elements in the rationale for IRBs and ERCs; and we believe that understanding those elements helps explain how IRBs and ERCs should function.
It seems to us there are four underlying ways in which IRBs or ERCs are meant to provide protection for human research subjects:
1) They serve the idea that collective knowledge and wisdom is better than individual knowledge, particularly when members are from different backgrounds. IRBs and ERCs typically have to have not only members with scientific interests and medical backgrounds, but also at least one member outside of science and one member not part of the institution. Insofar as “vulnerable” populations are the subject of any research, members of that population or those familiar with their needs, are to be on these committees or boards. The point is to try to make it more reasonably likely, first, that the study under consideration is scientifically valid and meaningful so that subjects will not be exposed to risk or harm for no good reason or actually useful purpose, but that meaningful knowledge will likely come from the study. Second, inclusion of the non-scientific members is to try to make it more reasonably likely that the moral or “human needs” of the subject population are duly considered and met, and are not ignored or given a merely secondary importance to the scientific search for knowledge. In terms of the deontology/utilitarianism dichotomy, the point is to use collective wisdom of board/committee members to make it more likely that the most knowledge can be obtained without abrogating subjects’ rights.
2) As stated earlier, the collective wisdom is also supposed to balance and restrain any part of the potentially problematic part of the protocol that is the prompted by an overzealous or overly optimistic view the researcher has, whether it stems from intellectual enthusiasm for his/her own idea, from his/her intense curiosity about the topic, from career ambition, financial profit, or any other conflict of interest that blinds the researcher to scientific, ethical, pragmatic, or other problems that need to be addressed and remedied if they are found to exist.
3) They are supposed to make sure the rules and guidelines are met by researchers, thus serving a policing function.
4) They are meant to interpret rules in regard to particular application, and we contend, they should be allowed to add new rules or permit exceptions to existing ones under certain circumstances we will explain momentarily. We contend they should afford rational flexibility in order to make sure there is ethical compliance with what is right, not just legalistic, compliance with a set of previously prescribed rules that may have loopholes or that may be irrationally restrictive under particular circumstances.
IRBs also serve as a watchdog committee for the institution to prevent the institution’s liability in being associated with, conducting, or condoning unethical research, but that function is served by meeting the first four and might be considered a derivatory function; and it has less to do with protecting research subjects than with protecting the institution.
Meeting the first, second, and fourth functions requires having a board or committee whose members are knowledgeable, sensitive, rational, and logical and articulate enough to be able to teach and learn from each other. That does not always happen. And clearly, credentials (including membership in the subject population) do not always guarantee it.
Often researchers in a field, particularly perhaps a field studying “orphan” or less prevalent medical conditions feel that the scientists and physicians on the board or committee do not fully appreciate (the importance of) the research proposals brought before them.
In some cases members of a subject population, who serve on the committee, may be very atypical of the research population and thus not have the necessary insight about the subjects which they need to be able to share with the committee or board.
And one of the criticisms often of “ethicists” on review boards or committees is that they are of no help because either they cannot explain principles in a way the other members can understand them or see their applicability, or the ethicist him/herself cannot actually apply the principles to the cases at hand. Some ethicists even suffer “paralysis by analysis” and cannot seem to take a meaningful position that could help crystallize agreement or disagreement with it.
Hence, we propose that there be some sort of appellate or review provisions or system in place for researchers who wish to appeal a decision of their particular review board, and for members of a board or committee who believe the majority has made an error in approving a project or approving it without additional security measures. Also, organizations for practitioners in those medical fields not likely to be represented on review boards or ethics committees might want to form their own ethics advisory committees whose prior analyses of proposed research projects might have helpful influence on IRB or ERC deliberations.
Codes, Rules, Laws, Policies, Guidelines, and Judgment
In regard to function 4, the reason that IRBs and ERCs should have flexibility is that ethics, like science, is something of a “bootstrap” enterprise in that rules in ethics, like laws in science, are codifications and collections of the state of best ideas based on the logic of the evidence known at the time. As new phenomena are encountered or discovered and as new perspectives about known phenomena are conceived, rationality demands that the new evidence be taken into account, and if necessary, scientific laws and methods, and ethical principles, be modified. Ethical principles and scientific laws are meant to be collections and codifications of what is most reasonable to believe, given previously known evidence. They are not meant to be calcifications of rules that deny the significance of new evidence, perspectives, or arguments. Unfortunately, while that sounds simple in theory, it is difficult in practice to decide or agree when new evidence or perspective warrants new rules or laws, and whether the new rules or laws accurately capture in their wording the essence of what is discovered. Rui Wang, et al express this well in an article in The New England Journal of Medicine that proposes and explains guidelines for reporting statistical subgroup analyses in The Journal: “As always, these are guidelines and not rules; additions and exemptions can be made as long as there is a clear case for such action."
It is very difficult, particularly in ethics, as in law, to capture in words the exact insight one has in mind. It would be most helpful for principles to be explained as fully as possible when they are formulated not only so that others can understand what is intended but so that one can more likely see oneself whether the formulation is flawed. But modern society tends to have a penchant for stipulating rules rather than explaining and justifying them, which turns out too often to be not very helpful as a guideline for understanding what the rule or law really is intended to mean or convey. In law, the courts often have to spend considerable time, effort, money, and resources to figure out what probably should have simply been explained in the first place – the intention and “spirit” of the law-- alongside the law when it was passed, in order to prevent the colossal waste of time and resources later.
In formal systems, such as law, sports or other institutions with rules, or written codes of professional ethics (as opposed to actual ethics or moral philosophy), the outcome is considered to be correct when it results from following the rules and procedures accurately. But morally that is insufficient when the rules or procedures are flawed and/or inappropriate in the first place. It is just as wrong to follow a bad rule as it is to follow an immoral order, and as it is to ignore a just rule. The one possible exception to not following a bad rule is if not following it will lead to (significantly) more harm or abrogation of rights than would following it. But when following a rule leads to morally wrong results by causing harm, abusing rights, or by preventing good, then following it just because it exists is not the morally right thing to do. “We were just following the rules” is no more moral than “We were just following orders.”
The Misconception About “Therapeutic Misconception”
In the paper “False Hopes and Best Data”, Appelbaum et al describe, define, or discuss at least two different meanings of “therapeutic misconception”. The first concerns a patient who has consented to become a research subject after it is carefully explained to him that he may receive a placebo and other aspects of the way research is conducted:
“Yet when the patient is asked why he agreed to be in the study, he offers some disquieting information. The medication that he will receive, he believes, will be the one most likely to help him. He ruled out the possibility that he will receive the placebo […]. In short, this man, now both a patient and a subject, has interpreted, even distorted, the information he received to maintain the view […] that every aspect of the research project to which he had consented was designed benefit him directly.”
The second description simply states the factual elements about research protocols that show the difference between them and clinical care a patient would receive from a physician dedicated to treat him/her as best the physician can: randomization, inflexible medication quantities and delivery times in spite of indications more or less medication might better serve the patient, researchers not being able to monitor the effects of the treatment as the research is in progress, etc. The misconception then is supposedly that the research subject believes that he will receive the same kind of medical care in the study he would receive when being administered any valid treatment option by a qualified physician, even if the subject realizes that not every step of the research trial is designed specifically to benefit him.
We agree that the first example is a serious misconception that needs to try to be avoided insofar as possible. Research literature and clinical experience present various better and worse ways to do that. We wish however to discuss the second description, because that presents more of an ethical policy or moral philosophy issue than does trying to figure out and guarantee the most practical methods to try to prevent misunderstanding by research subjects.
It is our contention, in light of what we have argued so far, that from an ethical standpoint, the subject’s belief that s/he will be treated as a patient, even though this is an experimental treatment, is a reasonable one, and that when research deviates from that concept, it is the research that is unethical; and it is the research that is (scientifically or ethically) unreasonable and mistaken, not the patient.
First, Applebaum and his colleagues do point out that “There are at least two reports in the literature of physicians’ reluctance to refer patients to randomized trials because of the possible decrement in the level of personal care.” But surely this is a common occurrence. Anecdotal information from physicians in conversation about this generally quickly points out cases in which physicians have advised patients not to participate in trials when they think it will not serve the patient well. They consider that their ethical duty as physicians, not much different from advising their patients who ask them, not to try unproven and likely sham alternative medicine approaches, particularly ones that may be expensive and harmful, let alone disappointing. Physicians also routinely advise patients against medications they might inquire about that they have seen advertised on tv or that the patients’ friends have recommended. In some cases they give the reasons; in others, they simply say it would not be a good thing for the patient.
Moreover, researchers and physicians will often remove subjects from a study when they see a serious decline or signs that it is about to happen. They do not need to know which treatment arm the subject is in at the time. It is not that research subjects are not being sufficiently monitored (ideally); in fact, in a research study, they may be monitored more closely than they would be as patients in a physician’s care (apart from being hospitalized). It is only that the monitoring does not include knowing what the causal factor that is precipitating the decline might be, whether treatment or placebo, or something coincidental and unrelated to the study at all. So it is not that a patient’s best interests are being ignored in an ethical research study – particularly with regard to harm being intentionally and knowingly caused or permitted (to continue).
Furthermore, while some, such as Miller and Brody, argue (in our view, mistakenly) that people consider the phrase “therapeutic research” to mean research that is therapeutic, it pretty clearly means research into potential therapies, which is a significant point. The “therapeutic” in “therapeutic research” means that therapies are the subject of the study, not a quality of the study. The distinction between therapeutic research and non-therapeutic research is that the latter is not research that fails its goal (to be therapeutic) but is research intended to find out information perhaps unrelated to therapies, such as surveys intended to elicit causal commonalities among people who may have a disease or condition. It may also be that “therapeutic research” is a phrase used to describe not only research (such as a retrospective study) about a past experimental treatment, but is normally applied to a new study in which an experimental treatment is being tried as the central element of the study. At any rate, therapeutic research is research into potential therapies, not research that is by its nature meant to be a therapeutic activity, in the way that “therapeutic massage”, “therapeutic gardening” or a vacation or afternoon off might be by those who find those activities restorative.
The significance of that is that the whole point of ethical therapeutic research is to find something helpful for the patients with that condition, some of whom will be the subjects in the trial and who will be the first to be helped if the experimental therapy is successful. Proposition 2 at the beginning of this paper “Medical progress benefits future patients with the condition for which progress occurs” is sometimes construed to mean that only future patients benefit or are intended to benefit from research, not the patients who are in the trials. That seems simply false in most ethically right research; and when true, is normally unethical – as when vulnerable populations (e.g., the poor, the imprisoned, the desperate) are the subjects of experimentation with no plan to continue to help them or members of the group to which they belong if the therapy proves to be beneficial, but to use it to help others instead, particularly others who will make its development highly financially profitable.
There is at least one kind of situation where a person undergoing research as a research subject is not the one who will likely benefit from the experimental procedure or therapy, but that subject’s participation then is still ethically right if they understand that and still consent, and if other conditions are also met. That is the case in something like a pioneering radical surgery such as a heart transplant in someone who is about to die soon and who is suffering with no prospect of relief or worthwhile quality of life with standard treatment, but where it the experimental procedure is pretty sure to fail in some unknown way that is necessary to discover in order to try to promote progress for the benefit of future patients. Such a trial is ethical only insofar at least as the subject understands this and its significance for him/her, and insofar as the amount and quality of life potentially forfeited is not of any real significance to the patient volunteering to be a subject. This is the kind of case where the likelihood of failure in the first attempts is considered to be greater than usual for experimental therapies, but the benefits for future patients holds some promise. But this is not the usual situation in therapeutic research. The usual situation is research into a therapy that is expected to benefit the research subject as well as future patients with the condition, though future patients may also benefit more as additional information is learned as time goes on.
Normally, there must be some reason to believe an experimental intervention will be helpful and not unduly harmful before it is tried on a human subject. There may be prior observations of success (as when a valid intervention for one purpose is seen to have beneficial “side effects” for other conditions that the patients for whom it is approved happen to have along with the condition for which the intervention is approved. There may be good theoretical reasons to expect success, based on, say biochemistry or genetic principles, and/or there may be prior animal study successes. There may be some other kinds of evidence that justify trials on human subjects as being potentially successful.
All this being said, we wish to raise, but not try to fully analyze, one particular issue of concern to us in regard to research on human beings – conducting research whose risk seems to outweigh its beneficial value, but for which there is a “demand” or “market value” for the product if it is successful. There are a number of conditions where the potential harm of the therapy (whether experimental or even approved) is significantly worse than the condition on any objective level, but where people will want or use or be willing to try the therapy anyway. This is a case where (they believe) the condition they have is significant enough to warrant risking some other harm. We have presented “significance” as being personal and subjective, but it seems to us that there is a limit to that subjectivity, in the same way that any act has limited justification simply on the grounds of autonomy. We have written previously[vii] about such limitations in regard to autonomy, and two salient examples of limiting autonomy would be not letting friends drive drunk and not letting someone commit suicide over a temporary despair that seems permanent to them at the time. There are objective features of such cases that make overriding autonomy not just mere paternalism or merely imposing one’s will on others. One such feature we will mention here is strong reason to believe that if the negative effect occurred, the person would then seriously regret having had the treatment, even though they beforehand cannot imagine having such regret.
One way to state the problem is that it arises from a conflict between the two qualifying conditions in proposition 16, because what someone wills (or believes s/he wants) prior to treatment may not be what they “would will” or be willing to accept after the treatment. The agent making the treatment available then has to decide which option is best because either they will have to honor the person’s current will or the person’s likely future will (based on the evidence of resulting past cases where the current will was honored), so they will violate proposition 16 as it is stated no matter what they do. This is not a matter of overriding, ignoring, or disrespecting the person’s autonomy because the question is which person’s autonomy to honor – the “before” person or the “after” person, when there is no way to honor both because the “before” will and the likely “after” will are contradictory. In such cases, professional benevolence will likely opt for the “after” person’s likely will but will still be considered arrogant, condescending, paternalism by the “before” person or even by the “after” person if the physician or researcher is mistaken. On the other hand, if the “before” person’s will is honored, many will deem the act to be one of self-serving, pandering, exploitation of the subject, rather than honoring autonomy or merely “meeting a demand by providing a desired service.” Caught in this dilemma, it seems that a physician’s oath (and common decency) requires his/her choosing the option most beneficial or least harmful for the patient because the will of the before person and the likely will of the after person cancel each other out.
An explanation will be required even if it is not convincing to the before person, but sometimes a demonstration of one sort or another can be given to them that helps make the explanation more persuasive, whether it is testimonials of prior patients, video of their reactions, or in some cases a physical temporary demonstration to the person of what it would be like for him or her if things went awry, allowing them to see at least some semblance of the consequences for themselves in order to try to heighten their awareness of the significance to them of those consequences.
The Inflexibility of Research Protocols
We also wish to take issue with proposition 11 at the beginning of this article, because it is not clear why a Procrustean protocol is required to give evidence of clinical safety and effectiveness when clinical medicine is not practiced in a Procrustean way.
“One of the problems presented by much research designed to determine the safety and efficacy of drugs is that this activity is much less experimental than the practice of medicine. It must, in general, conform to the specifications of a protocol. Thus, the individualized dosage adjustments and changes in therapeutic modalities are less likely to occur in the context of a clinical trial than they are in the practice of medicine. This deprivation of the experimentation ordinarily done to enhance the well-being of a patient is one of the burdens imposed on the patient-subject [in] a clinical trial.”
It is not clear, for example, why a pre-determined
dosage, if excess seems pretty clearly to be the cause of unpleasant
side-effects, yields valid results about what would happen if that would not be
the dosage that would be used in practice.
And it is not clear why a pre-determined dosage, if it gives some
benefit that would be more helpful if increased, is proscribed when the
increased dosage level might actually be the effective dose. Even if having patient-centered dosages
requires more work to analyze the data, it still seems to be the more reasonable
way to conduct a trial. It is certainly not clear that saving time, labor, and
money to analyze irrelevant data is preferable to spending more time and effort
analyzing the relevant data. Doing the
wrong thing more easily and quickly should not be preferable to doing the right
thing more arduously.
 This is often stated as a principle of respect for persons or a person’s right to autonomy, but we believe the way we have stated it is more basic and more intelligible. Moreover, the proposition we have stated is not explained or justified by appeals to autonomy or respect for persons because those usually just mean in that context that one ought not to control people or make them do something against their will. Thus saying one should not control people because they ought to have autonomy is merely to say one should not control people because one ought not to control people.
 The “would be against their will” part is essential because one is exploiting someone else if one knows one can talk them into something one knows they will later regret and one knowingly and intentionally does it anyway. Similarly if one does something wrong to someone who is asleep or unconscious or otherwise unaware you are doing it. The point is not that they do not disapprove now, but that they would disapprove if they knew what has been done to them, or that they will disapprove when they discover what has been done to them.
 Some authors have advanced the position that informed, valid, or educated consent absolves physician researchers from having to treat their research subjects (or patients to whom they recommend participating in research) in a way that their oath requires them to treat patients, in particular the requirement by oath not to cause or allow harm and to treat to the best of their ability. The rationale is based on respect for the patient’s autonomy and wishes (to participate or continue participating in a research trial). The above, along with explanations and arguments yet to be presented here, are meant to rebut that view.
 It is sometimes thought this ethical dilemma of having to choose withholding treatment from some individuals in order to benefit others for the greater general good is peculiar to physicians as researchers because of these lines in their professional oath:
Hippocrates version: I will apply dietetic measures for the benefit of the sick according to my ability and judgment; I will keep them from harm and injustice.
Modern version 1: I will apply, for the benefit of the sick, all measures [that] are required.... I will prevent disease whenever I can....
But everyone is bound by ethical maxims not to harm others unjustly and to help others where they reasonably can. For all intents and purposes any researcher who is qualified to administer treatments, or who utilizes the services in the study of those who can, would have as much moral obligation to do so as a physician would. The moral dilemma would thus apply to them as much as it would apply to a physician as researcher.
 The consequentialist/deontological dichotomy is an unnecessary one brought about by trying to reduce ethics to some single or simplistic principle of what is right and what is good. In everyday life, there are plenty of examples of utilitarianism’s legitimately taking precedence over deontological principles, and there are plenty of examples where utilitarianism falls short and fails to legitimately override deontological principles. For those interested in a full account of that, and a more reasonable principle that incorporates the best of both theories see “An Introduction to Ethics” at www.akat.com/Ethics.html
 There is an interesting trap that is easy to fall into: when focusing on the potential benefits of an experimental intervention, it is easy to think that placebo arms of a treatment do a disservice to the subject/patient and violate the physician’s oath to be beneficent. So it is easy to oppose placebo controls when concentrating on the potential benefits rather than the potential harms. But surely in some trials, such as the hormone replacement treatment (HRT) clinical trials, where the treatment proved fatal to too many women in the treatment arm, it is better to have been in the control group.
 Fred Gifford has pointed out what he considers to be other “types of equipoise” [Gifford, Fred (2000) 'Freedman's 'Clinical Equipoise' and Sliding-Scale All-Dimensions-Considered Equipoise'', Journal of Medicine and Philosophy, 25:4, 399 — 426]. While he raises important points, we do not think that subsuming them under “types of equipoise” is the clearest or best way to address them, and we will do it differently here where some of the concepts we discuss happen to coincide with his.
 “Theoretical equipoise exists when, overall, the evidence on behalf of two alternative treatment regimens is exactly balanced. […] Theoretical equipoise is overwhelmingly fragile; that is, it is disturbed by a slight accretion of evidence favoring one arm of the trial. In Chalmers’ view, equipoise is disturbed when the odds that [treatment, or trial arm] A will be more successful than [treatment, or trial arm] B are anything other than 50%. […] We may say that theoretical equipoise is balanced on a knife’s edge.” (Benjamin Freedman, “Equipoise and the Ethics of Clinical Research,” New England Journal of Medicine 317 (1987): 141-45.)
Samuel Hellman and Deborah S. Hellman basically reinforce this view when they argue that physician’s beliefs, including hunches, based on uncontrolled evidence or experience are ethically sufficient to proscribe the physician’s putting research subjects (or acquiescing to patients becoming research subjects) in arms of trials that do not act on those beliefs. (“Of Mice but Not Men: Problems of Randomized Clinical Trail,” New England Journal of Medicine 324 (1991): 1585-89.) While that is true in some cases, it is not true in all. Some beliefs are stronger than others and are based on better evidence than others. All that is necessary for morally relevant equipoise to occur is that the physician has some reasonable doubt about the belief. Hellman and Hellman are correct that RCTs are not the only method of preventing such doubt, but that does not mean that every belief based on anecdotal or personal experience outside of clinical trials is unaccompanied by reasonable doubt. It is simply reasonable uncertainty that matters morally in determining theoretical equipoise.
 “Significance for the patient” has to do with how important the potential outcomes might be for the patient, which can vary from one individual to another even though the quantitative magnitudes might be the same, or which can vary even in the same individual at different times in his/her life. E.g., those with a high tolerance for pain might eschew pain killers that put their thinking ability into a drug “fog” if they do not like drug “fog”. Someone with a low tolerance for pain who does not mind having his or her thinking “dulled” might prefer the pain killer. Or a quadriplegic with a possibly deteriorating condition might not want to risk surgery that could end up with a worse outcome than the outcome of deterioration, if the latter is acceptable and the former is not, particularly if the outcome of successful surgery merely leaves him/her at the baseline condition where s/he is now. The baseline condition may or may not be significantly preferable to the worst outcome. And at the same time, the best outcome might not be sufficiently more personally valuable to the patient than the outcome of the deterioration. In regard to a research study, while the probabilities and magnitudes of potential burdens and benefits may be the same for two different subjects, the value or significance of the different possibilities may be strikingly different for them – justifying participation of the one and not of the other. A common American example of the difference between magnitude and significance is seen in the difference in American football between trying a two-point scoring conversion early in the game versus at the end of the game with no time left on the clock and the team that has just scored a touchdown still trailing by two points. In all situations except where point spreads are important (for determining playoff or future draft choice opportunities, for example) or where an individual record may be more important than team victories (e.g., most point-after conversions by a kicker), while the probabilities and magnitudes of the different attempts are the same early and late in the game in the case described, the significance of the magnitude – the two points, as opposed to the one point-- is very different. Similarly, if someone who is likely to die before a certain important event they would like to see (e.g., the birth of a grandchild), an experimental treatment that might prolong their life by a few days or shorten it by a few days may have different significances. If they are expected to die before the date that is important to them, the few days of extra life the treatment might afford them could be very significant, and the time the treatment would cut off if it fails might not be significant to them. On the other hand if they are expected to die after the date that is important to them, there is more negative significance attached to the treatment’s failure than positive significance attached to its success. The magnitude in terms of days lost or gained is the same, but the significance of the magnitude is different. The concept we are referring to as “significance” is very important in decision-making, and we will show it is important also in trying to prevent exploitation of research subjects.
 Benjamin Freedman, for example, points out that you cannot merely “subtract” the placebo data from the experimental data to arrive at a supposed “net therapeutic advantage” of the treatment. Moreover, he contends that even if you could, that might have significance for pure research, but not for clinical practice. See his “Placebo-Controlled Trials and the Logic of Clinical Purpose,” IRB: A Review of Human Subjects Research 12, no. 6 (1990): 1-6.
 Examples of this in educational research occur in those studies that have “shown” class size and length of school day to be the most causally relevant factors related to student learning. When controlled for other variables thought to have been important (e.g., parents’ level of education, social/economic status, gender, ethnicity, etc.) in large scale research groups, all other factors cancel out. But in none of these studies is quality of teaching a factor, and it is presumably proportionally equally distributed between control and experimental groups (or between different experimental groups, depending on how the study is conceived). If, as would seem reasonable, that quality of teaching matters, that fact is hidden by the aggregate results. But it is difficult to imagine that students who have good teachers for half a day will learn less than students who have bad teachers for a full day, or that those students in large classes with excellent teachers will learn less than students in smaller classes with bad teachers. Even as anyone knows who has ever been individually tutored at length by a bad teacher, the personal attention and duration of time is of little avail. Large scale studies, where the actual causal factors are not known or suspected and not properly “tested for,” can obscure those causal factors.
 E.g., in some cases, a 20% success rate may not mean that a treatment is 80% ineffective but that it is 100% effective under conditions we don’t yet know and 100% ineffective absent those conditions. If Einstein was right to believe that metaphorically “God does not play dice with the universe”, then probabilities are more a sign of lack of knowledge of specific causes than they are a sign that something is causally effective only a certain percentage of the time.
 E.g., see Paul S. Appelbaum, Loren H. Roth, Charles W. Lidz, Paul Benson, and William Winslade, “False Hopes and Best Data: Consent to Research and the Therapeutic Misconception,” Hastings Center Report 17, no. 2 (1987): 20-24.
 Franz J. Ingelfinger, “Informed (But Uneducated) Consent,” New England Journal of Medicine 287 (1972): 465-66.
 Jay Katz, “Human Experimentation and Human Rights,” St. Louis University Law Journal 38 (1993): 7-54,
 They write “We believe that as with clinical care, in the case of many randomized, controlled trials, the patient’s participation can and should be considered to be authorized by his or her general consent for treatment and that specific consent should not be required.”
 Rui Wang, Stephen W. Lagakos, James H. Ware, David J. Hunter, Jeffrey M. Drazen; “Statistics in Medicine – Reporting of Subgroup Analyses in Clinical Trials,” The New England Journal of Medicine 357;21, Nov. 22, 2007: p. 2194.
 What happens in any formal system is that rules and regulations are utilized or developed which seem to be fair, which seem to be necessary and useful, and which seem to capture the spirit and purpose of the endeavor being established. Unfortunately not all circumstances or applications of the rules can be anticipated, and three different kinds of problems will typically arise in any formal, procedural system. (1) The loophole problem -- instances will arise that meet the letter of the rules but which will appear at least to some people to violate their spirit or the purpose of the enterprise. In sports, an example of that was Dean Smith’s University of North Carolina invention of the “Four Corners Offense” which essentially that year turned their college basketball games into games of sophisticated “keep away”. (2) Instances will arise where even sincere and faithful adherence to the letter and the spirit of the rules leads to an unanticipated, undesirable result because the rules will be incomplete, contradictory, or because one or more of them will not have accurately captured what was intended or what was unconsciously understood. (3) Unanticipated circumstances will arise for which the rules are inadequate or antiquated, yielding undesirable results. An example of the this last case arose in American football when kickers became so strong that kicking the ball off from the 40 yard line virtually eliminated "kick-off returns" and their potential excitement, because kick-offs routinely went beyond the field of play, for a touchback.
The American collegiate and professional football experience with "instant replay" checking of disputed referee calls is a good example of the last part of the second kind of flaw. Instant video replay analysis of the accuracy of on-field officials' calls was instituted because from time to time during football games officials made mistaken calls that were obvious and important. But the procedures instituted to detect and correct the errors did not address the real problem. This is true of the collegiate experience with rules governing instant replay official review and with both of the, so far two, NFL attempts. Coaches were allowed to request replay reviews, and play was held up while the video was analyzed. This slowed down the game so much and did not make sufficient difference most of the time to warrant the time and trouble, so "instant replay" officiating review was dropped. But, of course, this left the original problem that "instant replay" analysis was intended to prevent. It is my view that the way the original instant replay was conceived and formalized is where they went wrong. The original, and current problem, is that some calls in football are not only wrong and not only important about the outcome of the game, but they are also seen that way by all the fans watching the game on tv, and those who see the replay in the stadium on a large screen. Those calls are the only ones that need to be reversed -- the obvious and egregiously wrong calls that everyone watching tv sees. A referee who could stop the game and reverse such calls could simply be stationed in a booth with a tv, acting when, and only when, an obviously wrong call was made. That would not slow the game beyond what is both necessary and acceptable to fans. Fans do not expect perfect referee judgment about every difficult or close call; what they do not want is for really terrible calls to be made that are obvious to everyone but the referees. Understanding that would help football administrators develop better policies for instant replay review than they have so far. Instead the current instant replay rule in professional football requires that coaches make a challenge, and give up a time-out if the challenged call is upheld. Plus, play has to be stopped and the referee who made the call has to do the review -- sometimes allowing him to repeat the error if it involves misinterpretation or misunderstanding of a rule. And coaches only get two challenges a period, so if there are four bad calls during a period, at least two of them will be ignored. "Instant replay challenges" then, as they are formally instantiated, do not reflect the intention and point of allowing all and only obviously egregious calls to be overridden more or less automatically whenever they occur.
 Levine has pointed out different instances of medical research or practice that benefits others instead of the research subject, such as organ donor research, quarantine, etc. But even in those cases, the subject rightly expects minimal harm. And in some of those cases there is benefit to the subject in the form of his or her helping a loved one whose life and well-being is dear to him or her. There is also, we think, benefit in contributing to one’s society’s being one in which people help each other out when they can. That is an indirect personal benefit of helping others and setting an example; and, of course it does not always “come back” to one, but it at least might help, and it makes one deserving of future benefit in time of need, even if one does not receive it.
 Robert J. Levine, Ethics and Regulation of Clinical Research, 2nd edition (Baltimore: Urban and Schwarzenberg, 1986), 3-10.
[i] Whyte J: Treatments to enhance recovery from the vegetative and minimally conscious states: ethical issues surrounding efficacy studies. AM J Phys Med Rehabil 2007;86:86-92
[ii] Franklin G. Miller and Howard Brody, “A Critique of Clinical Equipoise: Therapeutic Misconception in the Ethics of Clinical Trials,” Hastings Center Report 33, no. 3 (2003): 19-28.
[iii] Robert D. Truog, Walter Robinson, Adrienne Randolph, and Alan Morris, “Is Informed Consent Always Necessary for Randomized, Controlled Trials?” New England Journal of Medicine 340 (1999): 804-7.
[iv] U.S. Department of Health and Human Services, Nation Institutes of Health, and Office for Human Research Protections, The Common Rule, Title 45 (Public Welfare), Code of Federal Regulations, Part 46 (Protection of Human Subjects).
[v] The Council for International Organizations of Medical Sciences (CIOMS) in collaboration with the World Health Organization (WHO) (Geneva: 2002).
[vi] International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use, ICH Harmonised Tripartite Guideline – Guideline for Good Clinical Practice (ICH-GCP Guideline) (Geneva: 1996).
[vii] Garlikov, R. and Jackson, A, “Introduction to Perspectives on Ethical Issues and Dilemmas in the Treatment of Patients with Spinal Cord Injury,” Topics in Spinal Cord Injury Rehabilitation (2008); 13 (3):1-17.