An Argument for Fewer Clinical Trials

BY KIRSTIN BORGERSON

The quantity of published clinical research, much of it of poor quality, is out of hand. Physicians cannot keep up with it, and patients are likely suffering from it. Research ethics committees need to rein it in, and they can do so by drawing on the principle that clinical research must be justified by its social value.

he volume of clinical research is increasing trials, and there are no signs of this slowing down: exponentially—far beyond our ability to but there are still only 24 hours in a day,”2 or, more process and absorb the results. Given this sit- directly, “[T]he quantity of new data exceeds the T 3 uation, it may be beneficial to consider reducing the field’s ability to process it appropriately.” Accord- flow at its source. In what follows, I will motivate ing to an article recently published in BMJ, medi- and critically evaluate the following proposal: re- cal research output doubles every seven years.4 An searchers should conduct fewer clinical trials. More analysis by Ming-yueh Tsay and Yen-hsu Yang in- specifically, I consider whether researchers should be dicated that the publication rate of randomized permitted to conduct only clinical research of very control trials (RCTs) has not only grown since 1965 high quality and, in turn, whether research ethics but grown exponentially.5 And there is no plateau in committees (RECs) should prohibit all other, lower- sight.6 Tsay and Yang identified 4,600 medical jour- quality research, even when it might appear to meet nals worldwide, as of 2002, and this number is also some minimal ethical standard.1 Following a close increasing. In another study from 2005, An-Wen analysis of the social-value requirement of ethical Chan and Douglas Altman estimated the number of clinical research, I argue that this proposal is defen- human subjects in clinical research at more than two sible. million a year.7 Approximately one million papers from clinical trials have been published to date.8 And The Sorting Problem publications are longer, more detailed, and contain considerably more data than in the past.9 Dramatic Quantity. The problem identified in this paper statements about the “flood” of biomedical research has two parts. The first part has received consider- data seem entirely reasonable given these figures. able attention lately, with critics remarking, “Every As Tammy Hoffmann and colleagues point out, day there are now 11 systematic reviews and 75 physicians see the explosion of research as both a blessing and a curse: the potential benefits are “in- Kirstin Borgerson, “An Argument for Fewer Clinical Trials,” Hastings hibited by the information overload experienced Center Report 46 (2016): 25-35. DOI: 10.1002/hast.637 by clinicians struggling to keep abreast of new

November-December 2016 HASTINGS CENTER REPORT 25 research.”10 The authors point out to increasing patient demands: “It is more toward pragmatic trial design that while the pace of research has ac- unlikely, therefore, that the amount and the use of methods other than celerated, the time available to read of time physicians devote to continu- RCTs. Pragmatic trials aim to pro- the research has not seen a similar in- ing medical education will keep pace vide evidence that directly supports crease. Finding the information scat- with the rate at which medical knowl- clinical decision-making in “usual” tered across a variety of journals adds edge is growing.”17 care settings. Unlike explanatory tri- to the challenge: “To find even half Quality. The exponential growth als, which aim to abstract away from of the papers published in one year of medical research may not be, in particular settings and patients, in . . . a clinician would need to read and of itself, an insurmountable prob- the hopes of creating ideal condi- an impracticable number of journals lem. If every study was of the highest tions for the success of an interven- . . . for example, an estimated 39 quality, there may be some hope that tion, pragmatic trials deliberately journals for randomised trials on dia- systems of knowledge synthesis could pursue knowledge of high applicabil- betes and 23 journals for systematic be developed to process all the data ity, through the use of representative reviews on myocardial infarction.”11 (an approach discussed below). Un- subjects, clinically important ques- Information overload involves be- fortunately, quantity is not matched tions, flexible treatment protocols, ing “swamped with information, but by quality in the literature. In fact, patient-oriented outcome measures, starved of data.”12 the quality of much of the clinical and so on. Most physicians won’t need to stay research is judged to be poor.18 This One way of trying to achieve more on top of all of the medical research is a widely recognized problem in the pragmatic trials is to use a tool such literature, just the literature in their medical literature, with deep histori- as the pragmatic explanatory contin- area of specialization. Unfortunate- cal roots. Before James Lind published uum indicator summary (PRECIS), ly, this doesn’t help as much as one his review of the medical literature which is intended to evaluate the fit might hope in dealing with the prob- on scurvy in 1753, he wrote that “it between research intention and de- lem. In 1995, the estimated number was necessary to remove a great deal sign. PRECIS was originally proposed of articles a physician would have to of rubbish.”19 In 1994, Douglas Alt- by a group of international clinical re- read to stay on top of his or her field man described the problem in stark searchers in 2009 and has been modi- was seventeen per day.13 A 2004 study terms: “[H]uge sums of money are fied and adapted in the years since. estimated the volume of literature rel- spent annually on research that is In a 2015 paper, the group laid out evant to one area of medicine (prima- seriously flawed through the use of the improved and validated instru- ry care) at 7,287 articles per month. inappropriate designs, unrepresenta- ment, PRECIS-2, which provides Physicians trained in the methods of tive samples, small samples, incorrect nine domains on which to assess trial clinical epidemiology would require methods of analysis, and faulty inter- design: eligibility criteria, recruit- approximately twenty-nine hours per pretation.”20 Altman didn’t hold back ment, setting, organization, flexibil- workday to evaluate the evidence on in his assessment of the situation, ity (delivery), flexibility (adherence), their own.14 Given what we know stating unequivocally, “We should follow-up, primary outcome, and about the increasing rate of publi- be appalled. . . . This is surely a scan- primary analysis.24 In general, the cation, these figures would almost dal.”21 Hilda Bastian, Paul Glasziou, more these design elements match certainly be an underestimate of the and Iain Chalmers write eloquently usual care, the more pragmatic the time required today. of the modern version of this chal- trial. Although the designers of this As one would expect, these num- lenge: “[T]he problem of having to tool leave it open as to how it might bers map poorly onto the time avail- trawl through and sift vast amounts be used by researchers—a particular able to health care professionals for of data has grown . . . [M]ountains of researcher might, for instance, use it reading and critical analysis of the unsynthesized research evidence ac- to design a maximally explanatory literature. One study, which surveyed cumulate.”22 They cite the “overload trial—I have argued elsewhere in fa- the reading habits of physicians work- of unfiltered information” as one of vor of using tools such as this to de- ing in internal medicine, found that problems lingering (and worsening) sign more pragmatic trials.25 There is internists reported spending 4.4 hours in recent years.23 growing support for this position, for per week reading articles in medical Some clarification is needed at instance, in trends toward compara- journals.15 One survey respondent this point, because “quality” is a tive effectiveness and translational re- said, “It is unrealistic to expect that, highly contested term in the medi- search, research-practice integration, even if you have the skills, you will cal literature. When some scholars and quality-improvement studies. have time to critically review all the advocate for high-quality trials, they The shift to pragmatic trials may literature that is out there.”16 The au- mean large-scale, simple, explanatory seem at first like an extension of ef- thors of the study indicate that inter- RCTs. Others, including myself, have forts to include particular underrepre- nists face intense pressure both to stay defended a different characterization sented groups in research, as occurred on top of the literature and to attend of high-quality research that tends in the 1990s for women and as seems

26 HASTINGS CENTER REPORT November-December 2016 to be occurring today for pregnant trials can support, such as adequate literature, and an extensive body of women. While it is consistent in some sample size, trial designs appropriate research tracks their widespread use.30 ways with these developments, there to the question being asked, appro- Any measure of quality that takes ob- are at least two key differences worth priate methods of analysis, minimal jectivity seriously will find fault with noting. First, the eligibility criteria bias, and so on. I will also return to many contemporary clinical trials. for enrollment in clinical research discussions of quality at the end of In a 2012 Nature Medicine ar- constitute only one of nine factors the paper. ticle, Peter Humaidan and Nikolaos that make a trial more (or less) prag- Returning to the current situa- Polyzos report that the problem ex- matic. A PRECIS-2–based approach tion, then, one undisputed indication tends to meta-analyses. Because me- goes far beyond narrow representa- of the low quality of research is the ta-analyses are, among other things, tion of particular groups in making relatively small sample size of many increasingly seen as providing “an trials more generally representative RCTs. Small trials may be appropri- easy way to get published,” the num- (although it should improve repre- ate in some cases—early-phase re- ber of such publications has increased sentation in this narrow sense). Con- search or research on rare diseases, for fivefold (from 849 in 2000 to 4,720 cern extends to the setting of care, the flexibility of treatment regimens, and The problem faced by those wanting to use the results so on, as described above. Second, the approach I favor of research in practice today is one of sorting good shifts the burden of justification to evidence from bad evidence and, given the overload of researchers who want to add ideal- izations of any kind to their trial de- research studies, doing so efficiently. signs. I have argued that the context in which research is meant to be ap- instance—but are generally thought in 2011).31 Many of these meta- plied should be the context in which to be undesirable once researchers analyses are unnecessary (reporting new interventions are evaluated. In are investigating effectiveness (and no new research since previous meta- other words, I see applicability as a comparative effectiveness) of new analyses or based on few or no stud- marker of high-quality research. As treatments in phase III trials. Robert ies in the area) or are of poor general Kirsty Loudon and colleagues put Califf and colleagues found that, be- quality (for instance, because they it, “Applicability (the ability for a tween 2007 and 2010, 62 percent of were performed on the basis of only trial result to be applied or used in a trials registered on ClinicalTrials.gov a few small trials). And guidelines do particular situation) is the outcome enrolled one hundred or fewer par- not fare any better. Two recent studies of these choices, which affect the ticipants;28 50 percent enrolled fewer indicate that guideline recommenda- ease with which the trial results can than seventy participants. Many of tions in clinical practice are based on be applied to and by the usual com- the trials were originally designed to high-quality evidence less than 15 munity of users of the intervention in enroll more participants but fell short percent of the time.32 The pace of re- the settings in which the trial design- at recruitment, leading the authors to search is so quick that some commen- ers envisioned it being used.”26 I do speculate that there may be a wide- tators wonder whether it will ever be not argue that certain groups need spread problem with underpowered possible for critical syntheses of the to be “added” to trials or targeted for trials. In addition to these problems, research to catch up. Bastian, Glaszi- inclusion, which would require that many trials failed to report details of ou, and Chalmers express this skepti- we name and identify these groups. randomization, blinding, and use of cal view: “There is nevertheless a risk Steven Epstein’s excellent book, In- data-monitoring committees. that the increasing burden placed on clusion: The Politics of Difference in Another way in which the quality the methods of - Medical Research, has carefully docu- of research is compromised is the sig- ing could make the goal of keeping mented the challenges of this sort of nificant presence of biases in research. up-to-date with the knowledge won “niche-standardization.”27 Rather, I At this point, the list of known tac- from trials recede ever more quickly suggest that the default should be the tics used to manipulate clinical trials into the distance.”33 standard-care context and all exclu- would number in the dozens and in- To sum up, whether one’s mea- sions and idealizations should be care- clude extensive and unjustified use of sure of research quality tracks appli- fully justified across all nine domains exclusion criteria (about 40 percent of cability, as mine does, or sticks with above. This understanding of quality potential subjects were excluded from more widely shared markers, such as shapes some of what follows, in that the 280 most influential trials from objectivity and sample size, we can I regard trials with limited applicabil- 2002 to 2010), suboptimal dosing of agree that there is much low-quality ity as lower quality, but I have also comparison treatments, and manipu- research being produced. The prob- attempted to highlight elements of lated data analysis.29 Many of these lem faced by those wanting to use the quality that defenders of explanatory tactics are prevalent in the medical results of research in practice today

November-December 2016 HASTINGS CENTER REPORT 27 is one of sorting good evidence from drugs shown to be inferior to others acting on the basis of heavily biased bad evidence and, given the overload in published but not yet recognized research results and industry-spon- of research studies, doing so efficient- trials (in other words, “suboptimal sored meta-analyses designed to ly. I will refer to this as the “sorting prescribing”). Positive results in trials ignore or exclude evidence not in problem.” of new treatments have been shown support of a drug. It isn’t always clear to influence practice only very slow- which evidence synthesis to trust, The Harm ly, and sometimes negligibly. Lack and there are many such syntheses of awareness of the results is one of competing for physicians’ allegiance. t seems likely that the overproduc- the reasons offered in explanation Trusting the wrong source can be Ition of low-quality research would for this phenomenon.36 This lack of harmful to patients, who once again be harmful to patients. The harms awareness can easily arise when high- are exposed to treatments that are not of research may be either direct or quality trials must first be sorted and best for their conditions. But when indirect. Direct harms are those ex- identified among the vast number of the research is too overwhelming to perienced by research subjects as a re- low-quality trials. Further, the aston- assess as an individual with far fewer sult of their participation in research, ishingly high rates of adverse events than twenty-nine hours per day to for example, an infection acquired in clinical care may be partly a result devote to the task, someone has to be as a result of IV placement. Indirect of a mismatch between the research trusted to perform independent criti- harms are those experienced by fu- evidence produced and the realities cal analyses. The frustration and un- ture patients whose care is affected by of clinical care. When physicians certainty generated by this situation the results of that research.34 Those don’t have the evidence they need are also likely to have harmful effects harms may be experienced differently to make individualized patient-care on physicians. by members of different groups, for decisions—because, for instance, There may also be more abstract instance, vulnerable groups, because their patients don’t resemble those harms associated with physicians’ harms can have a multiplicative effect, on whom the therapies were tested— failure to discharge their professional and any harm calculation would have those patients may be harmed. responsibility to provide evidence- to be responsive to these differences. Similarly, indirect harms could based care. These include moral Similarly, direct benefits are those arise when new evidence emerges harms to physicians themselves, who that affect the health and well-being about the side effects of a popular face a duty to provide treatments sup- of individual research subjects, while drug but this evidence takes years to ported by the best available evidence indirect benefits accrue to future pa- reach physicians and affect prescrib- but are impeded in discharging this tients as a result of greater knowledge ing habits, subjecting patients to sig- duty by the overproduction of vari- of “what works” in medicine. nificant harms in the meantime. We able-quality research. This failure may Of course, tracking the harm aris- might use the Vioxx scandal as an also have implications for the trust- ing as a result of the sorting problem example.37 Of course, in the Vioxx based physician-patient relationship is extremely difficult. There may be case, the delay was the result of a de- and, by extension, the well-being of historically controlled or observation- cision to withhold information from patients. If a patient stumbles on a al studies capable of doing so, but I the public, whereas the time lag we recent excellent study on a new treat- have yet to encounter them.35 In the are imagining would occur as a re- ment for her condition and brings it absence of clear data, we can make sult of the time needed to perform to the clinical encounter, the physi- some progress by identifying harms an adequate research synthesis. But cian’s failure to stay up to date may that could be reasonably anticipated. the case does help us to see the harms well come to light. Perhaps this can Perhaps most obviously, there are possible with delayed action: it was be managed judiciously, but it could direct harms associated with partici- estimated that tens of thousands of also lead to a (justified!) lack of confi- pation in low-quality trials, includ- patients were harmed by the delay in dence in the physician, with predict- ing any of the net risks to subjects releasing data about the harms of the able depreciation in trust. of research arising from blood draws drug. When drugs are prescribed to A related systemic problem is that and other monitoring tests, as well millions of patients, as many block- poor-quality research can damage the as inconveniences and wasted time, buster drugs are today, any delay in social trust required by the research which are not balanced by benefits to changing prescriptions in light of enterprise as a whole.38 This can make society. Research subjects may also be dangerous side effects can lead to recruitment of research subjects even deprived of the standard of care and tremendous harm to patients. These more difficult and lead to even more thus be harmed by being deprived of harms often have a ripple effect on underpowered studies. In sum, there an effective treatment. the families and communities of the are many possible, and even likely, In clinical care, reasonably antici- people affected. harms of overproducing low-quality pated indirect harms would also in- Further harms could reasonably research evidence. clude the continued prescription of be expected to arise from physicians

28 HASTINGS CENTER REPORT November-December 2016 The Evidence Synthesis without appreciation for the neces- used in selecting ‘similar’ treatments Solution and Its Shortcomings sary limits of such systems; undue causes a problem.”45 Abstraction may conservatism about what counts as make it easier to access research re- he problem outlined above has evidence (the exclusion of obser- sults, but this is often at the cost of Tattracted considerable attention vational research, case studies, and making the results far more difficult, in the medical literature. Almost all other nonrandomized research from if not impossible, to implement in commentators identify some range of higher-level synopses); the division of practice. synthesis solutions as if this exhausts communities into separate groups of This study provides support for the options available for addressing synthesizers, researchers, and practic- my claim that the problems of evi- this problem.39 I will provide a brief ing physicians, resulting in decreased dence synthesis are not easily or de- description of one such representative communication across these groups; cisively remedied, since it highlights solution and summarize the problems contamination of the evidence by an the tension between any attempt to identified with the proposal. Because endless supply of loopholes in trial provide clear, decisive, simple answers my interest is not in replicating this methodologies; de-emphasis on criti- to clinical questions and the complex debate over knowledge synthesis but in advancing a new solution to the Poor-quality research can damage the social trust problem it was designed to address, this section will be relatively brief. required by the research enterprise as a whole, The evolution of what I will refer to leading to even more underpowered studies. as the “S hierarchy” approach—“4S” (from 2001), “5S” (2006), and “6S” (2009)—to evidence-based clinical cal thinking; cultivating a false sense and detailed evidence required from decision-making effectively illustrates of certainty (and the challenges this physicians on, for instance, variable this shift to knowledge synthesis.40 creates for full informed consent); dosage, the time required for a drug The original hierarchy proposed by and a lack of appreciation for detail to take effect, interactions between evidence-based medicine (EBM) and context.42 drugs, side effects, and so on. With identified small-scale and observa- This last criticism has been bol- respect to the evidence synthesis ser- tional trials as lower-quality meth- stered by empirical evidence from vices available in 2012, Hoffman and odologies and prized randomized Paul Glasziou and colleagues, suggest- colleagues conclude that “few current controlled trials and meta-analyses of ing that syntheses of clinical research systems seem adequate.”46 Perhaps it is RCTs as the most reliable sources of rarely contain the sorts of informa- time to consider other options, how- research evidence. Each addition to tion required in clinical decision- ever unattractive they may appear at the original hierarchy by the S hier- making.43 Syntheses may include first. Of course, these solutions aren’t archy movement (syntheses, synopses reference to “behavioral interven- exclusive: we can continue to try to [of studies and of syntheses], sum- tions,” “salt reduction,” or an “exer- improve our system of knowledge maries, and systems) was proposed cise program” without providing any synthesis (if we think there is value to in order to make clinical research further information about the inter- this enterprise) while exploring other evidence more digestible and easier vention, and while at the same time possibilities. If the synthesis solution to use. Developers of the approach obscuring differences in the terms as is less than complete, as it surely is, seem interested in getting to a point used by the trials being analyzed. The we have reason to look for other so- where all of the information needed challenges of implementing any such lutions to the problem (particularly by clinical decision-makers would be intervention, however positive the when they are complementary). contained in the titles of short synop- results, should be immediately obvi- In spite of extensive commentary ses or summaries.41 This is done with ous. For evidence on pharmaceutical on how best to synthesize knowledge an awareness of the information over- treatments to be usable by physicians, and translate it to practicing clini- load present in the literature and an a description should include “the cians, there is near silence on the al- appreciation for the relatively limited dose, titration, route, timing, dura- ternative considered below. The only time health care professionals have to tion, and any monitoring used.”44 researchers bold enough to mention, search through and critically appraise Yet many syntheses fail to provide if briefly, the possibility of conduct- research data. precisely these pieces of information. ing fewer research trials write, “First, The S hierarchy approach has Systematic reviews fared especially we need to prioritise effectively and been criticized on a number of poorly when it came to providing reduce avoidable waste in the pro- fronts, including its continuity with details essential to the implementa- duction and reporting of research evi- the original—now widely discred- tion of research results. The authors dence.”47 Of course, they are careful ited—EBM ; acknowledge that “[i]n systematic to say, “although funding for evalu- overreliance on computer systems reviews, the high level of abstraction ative clinical research internationally

November-December 2016 HASTINGS CENTER REPORT 29 remains a priority,” before offering sary to protect and promote people’s things, ensuring that finite social re- any such suggestions.48 I will proceed health.”50 What does it mean for re- sources are used responsibly in pur- with a related disclaimer: my account search to be socially valuable? Among suit of that goal. One of the resources is neutral on matters of funding. other things, “[t]he social value of shared by all clinical research, wheth- Fewer large-scale trials, if that is what this research is ultimately grounded er publicly or privately funded, is hu- is required, may well turn out to be in the quality of the information man subjects. And, as the frequency the same cost as a series of smaller tri- that it produces.”51 This is consistent of trial delays at recruitment suggests, als. This is an empirical matter, and it with the discussion above, in which this resource is in short supply. Given turns on the debates over quality in I suggested that one key marker of how few people are both willing and clinical research. the quality of clinical research is the able to participate in (often quite bur- Moreover, my suggestion that we relevance or applicability of results. densome) clinical research, it is not ought to pursue fewer clinical trials The beneficiaries of clinical research surprising that there are tremendous should be understood as defending are future patients within the same challenges with recruitment.56 The the position that many of the low- health care context.52 In what follows social-value requirement exists partly quality trials being proposed right I take a closer look at the social-value to ensure that we are doing the best now should not be conducted. In requirement, consider why it has not we can with the limited collective other words, if we take a snapshot prevented the sorting problem out- resources at our disposal, given our of the trials being reviewed by RECs lined above, and evaluate whether it shared interest in advancing human today, around the world, I am argu- might play a more active role in re- health.57 When collective resources ing that fewer of them should be stricting the overflow of low-quality are limited, some form of principled approved than are currently being ap- research. rationing is surely defensible. The proved. If, over time, researchers ad- There are two standard arguments social-value requirement can serve as justed to higher standards by turning supporting the social-value require- a kind of rationing tool. This is one all low-quality trials into high-quality ment. These are the “responsible use of many topics within bioethics that trials, it is possible we would not end of finite resources” or efficiency argu- would benefit from closer attention up with fewer trials in the long run; ment and the avoidance of exploita- from political philosophers. however, I think there are efficiency- tion argument.53 In a recent paper, An even more problematic feature related reasons (discussed below) to Alan Wertheimer devoted consider- of the requirement as it is understood doubt that this would be a likely out- able effort to debunking the exploita- today concerns the risk-benefit calcu- come. In any case, since it is probably tion argument, drawing on his own lation. For most proposed clinical tri- best to avoid speculating about the account of mutually advantageous als, the risks to subjects outweigh the future, I will state as clearly as I can: exploitation to make the case that benefits to those same subjects. This I argue for fewer trials relative to the social value is not as robust or univer- shouldn’t be surprising: the enterprise status quo in research. sal a requirement as it might seem.54 of clinical research involves ventur- In sum, then, given the overload Though I think that alternative ac- ing into the unknown; in this sort of of low-quality biomedical research counts of exploitation would likely situation all kinds of risks will present data published every day and the fact provide a better defense for the social- themselves. Researchers try to con- that there are serious problems with value requirement, for the purposes tain those risks and learn from what knowledge synthesis as a full solution of this paper I am more interested in doesn’t work by systematically track- to this predicament, how might re- examining a particular version of the ing what takes place over the course search be responsibly pursued? efficiency argument, as well as draw- of research and comparing outcomes ing attention to (and correcting) an across groups, but the context of re- The Social-Value Solution oddly blinkered, or one-sided, read- search does not, on its own, eliminate ing of the requirement within the risks. The residual risks to subjects, linical trials should be socially bioethics literature.55 then, demand attention. The social- Cvaluable. In the list of seven ethi- One reason social value is identi- value requirement exists in part to cal requirements of clinical research fied as an ethical requirement of re- enable us to find a measure of benefit identified by Ezekiel Emanuel, David search is because clinical trials should that will balance out the risks taken Wendler, and Christine Grady, social fairly and efficiently evaluate new by human research subjects. This has or scientific value is first.49 The most medical interventions. The social- led to a focus on the potential posi- recent draft of the CIOMS guidelines value requirement connects the con- tive effects of prospective trials. After mentions social value in its opening duct of research to the ultimate goal all, the task of evaluating the ethics sentence: “The ethical justification of clinical research: better health for of a proposed trial appears to require of health-related research is its social humans. It does so by requiring that that researchers seek out potential so- value: the prospect of generating the research aim to benefit society, in the cial benefits in order to deal with the knowledge and/or the means neces- long run. It can do so by, among other

30 HASTINGS CENTER REPORT November-December 2016 pesky problem of residual risk (even harder to meet. Researchers will need Canada: Tri-Council Policy State- when risks are minimized). to work to ensure that the benefits ment. Two passages in the most re- There seems to be some general brought about by a particular trial cent version of the TCPS (TCPS-2) support for the view that it is ac- overcome the higher risk and harm provide general guidance on social- ceptable to include only the social threshold arising out of a more bal- value assessments. The first provides effects that “do the work” we want anced calculation. The current prolif- support for my suggestion that so- them to do in balancing risks of eration of low-quality trials would be cial value—which adds benefit to harm to individuals (in other words, less rampant according to this more the risk-benefit calculation—be en- social benefits).58 An otherwise ex- complete understanding of the social- hanced where possible for each pro- cellent forty-page paper on risk-ben- value requirement. At the very least, posed trial: “Researchers and REBs efit calculations, for instance, relies if researchers are pressed by RECs to must attempt to minimize the risks throughout on discussion of social explain why their trials aren’t better associated with answering any given benefit: “Whether a given level of than they are—aren’t perhaps even research question. They should at- research risk is excessive depends on the best they could be under the cir- tempt to achieve the most favourable the magnitude of the risks, the level of corresponding potential benefits The idea that it is acceptable to include only the social for the participant, if any, and the level of potential social benefit from effects that “do the work” we want them to do in performing the intervention and the balancing risks of harm to individuals seems to be study” (emphasis added).59 But this common position is transparently commonly held. But this is transparently unprincipled. unprincipled. After all, we don’t gen- erally make important decisions by cumstances—this might be a much- balance of risks and potential benefits considering only the benefits of par- needed nudge. If this is paired with in a research proposal” (emphasis ticular choices while ignoring any a robust understanding of research added).62 I take this to support my harm.60 The requirement is an assess- quality, of the sort outlined earlier, proposal (and perhaps even take it ment of the importance of research— neither researchers nor RECs will one step further).63 A second passage of social value, not social benefits. Full have an excuse for inaction. makes it clear that the welfare of soci- consideration of social value includes ety—where that includes assessment both social harms and social benefits. Support for the Social-Value of social benefitand social harm—is Benjamin Freedman’s classic analy- Solution of considerable importance in the sis of social value makes this point: design of research: “Groups may ben- “Exogenous factors may be relevant he status quo in research eth- efit from the knowledge gained from . . . . On [the] positive side, high pri- Tics is the view that researchers research, but they may also suffer ority research; on [the] negative side, and research ethics committees need from stigmatization, discrimination potential for abuse of knowledge not consider—and are perhaps even or damage to reputation.”64 In other gained.”61 There will be a measure prohibited from considering—the words, social harms matter to the eth- of speculation required, unquestion- potential indirect harms of research. ical assessment of research. Together, ably, in assessing social harm, as there In what follows, I demonstrate that, these passages support my proposal always is in assessing social benefit, contrary to popular interpretations as outlined above. In fact, as with the but refusing to venture an estimate of current regulations, the Canadian common rule, I take it to be the case of effects on grounds of difficulty just Tri-Council Policy Statement and that the TCPS already endorses my sidesteps the issue, and negligently the American Common Rule permit, position, and members of ethics com- so. So, whatever argument we offer and in the Canadian case may even mittees and researchers who fail to in support of the social-value require- be said to encourage, consideration of consider indirect harms are failing in ment, it should support the assess- social harms in the ethical assessment their existing ethical responsibilities. ment of both potential benefits and of research. In what follows, I outline United States: Common Rule. risks of harm, for any investigation. my interpretation of these guidance In response to a recent proposal by If indirect harms have a place in the documents in the hopes of convinc- Alan Fleischman and colleagues that social-value assessment performed by ing skeptics that the position I have called for a national advisory group researchers and RECs, as I argue they defended is more plausible than it to assess the potential social harm do, then the harms outlined above might seem at first. Of course, even if of particular research projects, legal should tip the balance of favor against all regulatory documents took a dif- scholar John Lunstroth clarified the otherwise marginally acceptable clin- ferent position on the matter, I would status of claims made in the Com- ical trials. In effect, this means that argue that my argument as presented mon Rule regarding the obligations the social-value requirement will be above is still sound. of institutional review boards in

November-December 2016 HASTINGS CENTER REPORT 31 the assessment of the potential so- According to Lunstroth, it is “indis- based on its failure on the second cial value of proposed research tri- putable [that] the general mandatory listed requirement from Emanuel, als.65 Lunstroth draws attention, in rule requires consideration of all as- Wendler, and Grady: scientific va- particular, to a mandatory rule and pects of the outcome.”68 lidity.72 There is no need to appeal subrule expressed in 46.111, “Cri- Lunstroth asks what would hap- to social value. This is an important teria for IRB approval of research”: pen if the results of research were objection because it reminds us of the “(2) Risks to subjects are reasonable highly socially beneficial. Surely the complex relationship between the re- in relation to anticipated benefits, if committee would take this favorable quirements of social value and scien- any, to subjects, and the importance outcome into consideration. Since the tific validity. of knowledge that may reasonably full range of potential risks and ben- I am sympathetic to an argument be expected to result” (the manda- efits is rightly within the purview of against low-quality trials on grounds tory rule), and “[t]he IRB should not what is reviewed by IRBs, these com- of scientific validity. Such an argu- consider possible long-range effects mittees fail to fulfill their responsibili- ment might well be complementary of applying knowledge gained in the ties when they don’t take such factors to the one offered here. The problem research (for example, the possible ef- into account. Further, he chastises as I see it is that the requirement of fects of research on public policy) as those who resist the suggestion that scientific validity has been interpreted among those research risks that fall ethics review is the right place to con- so narrowly that it only captures our within the purview of its responsi- sider the sociopolitical consequences interest in methodological rigor.73 But bility”66 (the subrule). Because the of research.69 How might we think it may well be the case that method- subrule seems to suggest that IRBs of providing any assessment of the ologically rigorous trials are the least must not consider the possible long- importance of research without con- socially valuable trials. Benjamin term effects of the results of research, sidering these factors? The proposal Freedman called attention to this most scholars have believed that con- he is critiquing, which puts forward when he wrote that “a useless study sideration of the long-term harms to a national advisory group as the ap- is more likely to be valid than a use- society of particular research trials is propriate body for such assessments, ful study.”74 At the end of the day, the beyond the purview of ethics review. would weaken the current system of social-value requirement, which pre- What Lunstroth meticulously review by removing one of its central supposes scientific validity, goes fur- points out is that the mandatory rule, responsibilities, and “the beneficiaries ther in attending to concerns beyond which refers to the importance of the of this weakening would be scientists those narrowly in the methodological knowledge resulting from research, and firms who engage in controver- domain. It is not only scientific va- could not properly be assessed if half sial research.”70 I would add that they lidity, narrowly construed, but also the information (the potential harms might also be the researchers con- clinical relevance or applicability that to society) were ignored. Further, ducting low-quality trials. matters to the assessment. It requires proper legal analysis demands a more While the revisions proposed in that research fairly and efficiently ad- open reading of the “should not” 2015 to the Common Rule offer vance human health. As noted, it also claim in the subrule than might seem helpful clarifications of many of the provides an opportunity to include to be required by a common language original rules, no such illumination social harms in the assessment of pro- analysis of the passage. I quote him is provided for the rule cited above.71 posed research. at length: It is hard to say whether other devel- A second objection concerns the opments in the latest version, such as breadth of indirect harm as a cat- There are no limitations on the calibrating the level of review to the egory. If RECs are encouraged to kinds of importance that the IRB level of risk to individual subjects, consider potential indirect harms in must consider, and therefore if will enhance or diminish discussion their evaluations, they may end up there is the possibility the knowl- of indirect harms by IRBs. Only time rejecting too many research studies. edge may be politically or socially will tell. This is because many important and important, the IRB is required to fruitful scientific discoveries have re- consider those possibilities. Fur- Objections and Replies sulted in technologies that caused sig- thermore, if the IRB were restricted nificant harm to humans. Consider, from considering the risks of socio- ne anticipated objection to my for instance, research in nuclear phys- political effects, then the advisory Oposition runs as follows. If the ics in the 1930s and 1940s, which rule would have the effect of putting problem is that there is too much permitted both the development of a metaphorical thumb on the ben- low-quality research, shouldn’t we life-saving medical treatments and efit side of the scale, inasmuch as the raise our standards of scientific va- the construction of deadly nuclear risk side of the scale would be artifi- lidity rather than our standards of bombs. It may be better for scientists cially lightened by removal of a risk social value? If current research is of simply to pursue knowledge for its factor.67 such low quality, it should be rejected own sake, without limitation. Key to

32 HASTINGS CENTER REPORT November-December 2016 this objection is an assumption about questions are prioritized, they can new obligation, as, for instance, was the limits of foreknowledge: we never shape the direction of research much arguably the case with the addition of really know, in advance, how knowl- more easily than members of RECs the “responsiveness” criterion follow- edge will be put to use. can by rejecting individual proposals. ing an increase in ethically dubious It will be helpful to clarify my po- In reply, I would return to the international research. Rather, I am sition, by way of initial response: I am widely accepted view that social helping to develop our understand- not arguing that we should stop or benefits are necessarily a part of the ing of an existing obligation. Second, prohibit any research with the poten- assessment done by RECs. This is resources exist to assist researchers tial to lead to harm. That would be necessarily of concern to RECs be- and RECs with assessments of qual- to tip the balance too far in the direc- cause many trials have net risks to ity; the PRECIS-2 tool outlined ear- tion of weighing harm, thus ignoring subjects and would not otherwise lier is one such resource. Third, the the need to balance benefit. This sort achieve a favorable risk-benefit ratio. fact that there is a crisis in a system of engineered imbalance, whichever If RECs are assessing social benefit, of regulation is not in itself reason way it tips, is precisely what I want there is no principled reason for ig- to reject efforts to make that regula- to avoid. If we keep in mind that as- noring social harm. tory system more rigorous. Ethical sessments include social benefits as well as social harms, it isn’t clear that Contrary to popular interpretations, the Canadian most research will fail this test, since there will be potentially beneficial Tri-Council Policy Statement and the American applications to weigh in the balance. Common Rule permit, and in the Canadian case may To go a bit further, though, it is not always the case that we are deeply un- even be said to encourage, consideration of social certain about the harmful effects or harms in the ethical assessment of research. applications of research. For instance, some research will be aimed primar- A fourth, related objection con- obligations weigh on us whether we ily at bad ends: think here of research cerns the limited capacity of RECs like it or not, and they don’t go away done to enhance the transmissibility to carry out the sorts of robust so- simply because they seem to be too or virulence of a pathogen. While cial-value assessments required. It is demanding. people may argue about the permissi- no secret that RECs in most inter- In sum, harms associated with bility of research that is ambiguous in national jurisdictions have limited the overproduction of low-quality its applications (and I am in favor of resources—whether time, expertise, research evidence are rarely included precisely this sort of deliberation by or training—to discharge their many in the social-value calculations con- RECs), most people also agree that responsibilities. We have evidence of ducted during the ethical review of research with only or primarily harm- this from many sources, including an proposed clinical trials.77 This hap- ful effects or applications should be IOM report in 2003.75 More recently, pens for (at least) two reasons: first, restricted on ethical grounds. I argue in a survey of NIH-funded research- because of a failure to recognize the in this paper that low-quality research ers, “many researchers viewed the need to preserve limited research re- cannot produce useful knowledge IRB as cumbersome and slow; some sources, for instance, human subjects, with positive applications because, viewed their IRB as not competent to for high-quality trials and, second, for instance, it is too biased or it is review the research, described as not because social-value calculations— statistically incapable of answering understanding the protocols or ana- when they are conducted—focus on the question it poses. Useless research lytic methods; IRBs were criticized positive outcomes of potential tri- will have no benefit and some poten- for applying regulations inconsistent- als. But the overproduction of low- tial harm because it contributes to the ly or for ‘over-protecting’ subjects.”76 quality clinical research is very likely sorting problem. My proposal seems to further burden to be harmful to patients. On ethi- A third anticipated objection goes an already overburdened regulatory cal grounds there are persuasive rea- as follows: why does this responsi- system. sons to endorse the position that we bility to assess the full set of harms I want to acknowledge the force should conduct fewer clinical trials. and benefits fall primarily to research of this practical concern. There is a Researchers and research ethics com- ethics committees rather than, for serious capacity shortage on RECs mittees should work together to en- instance, funding agencies? After in most jurisdictions. I don’t dispute sure that trials truly benefit society, as all, funders—particularly public this. But let us keep in mind three they are meant to do. funders—are in a unique position things. First, the obligation I have Acknowledgments to assess social benefit and harm. If identified and argued for in this pa- they set out a research agenda, in per is one that is already existing. I Special thanks to the reviewers of which certain topics, methods, and am not suggesting we add an entirely the Hastings Center Report for helpful

November-December 2016 HASTINGS CENTER REPORT 33 comments and to colleagues at Dal- Identify the Information They Need,” Brit- no. 11 (2013): 1029-31, doi:10.1001/ housie University and members of ish Medical Journal 310, no. 6987 (1995): jamainternmed.2013.496. the New Scholarship in Bioethics re- 1085. 30. See K. Borgerson, “Why Reading the 14. B. S. Alper et al., “How Much Effort Title Isn’t Good Enough: An Evaluation of search network for ongoing scholarly Is Needed to Keep Up with the Literature the 4S Approach to Evidence-Based Medi- support. Thanks also to my mentors Relevant for Primary Care?,” Journal of the cine,” International Journal of Feminist Ap- in the American Society for Bioeth- Medical Library Association 92, no.4 (2004): proaches to Bioethics 2 no. 2 (2009): 152-75. ics and Humanities Mentor-Pairing 429-37. 31. P. Humaidan and N. Polyzos, Workshop—Rebecca Dresser, Scott 15. S. Sanjay et al., “Journal Reading “(Meta)analyze This: Systematic Reviews Habits of Internists,” Journal of General In- May Lose Credibility,” Nature Medicine 18, Kim, and David Wendler—for offering ternal Medicine 15 (2000): 881-84. 1321 (2012), doi:10.1038/nm0912-1321. detailed feedback on the first draft of 16. Ibid., 883. 32. D. H. Lee and O. Vielemeyer, “Anal- this paper. 17. Ibid., 883. ysis of Overall Level of Evidence behind In- 18. R. M. Califf et al., “Characteristics fectious Disease Society of America Practice Notes of Clinical Trials Registered in ClinicalTri- Guidelines,” Archives of Internal Medicine 171, no. 1 (2011): 18-22; P. Tricoci et al., 1. Where possible, I use the neutral term als.gov, 2007-2010,” Journal of the Ameri- “Scientific Evidence Underlying the ACC/ “research ethics committees” or “RECs” to can Medical Association 307 no. 17 (2012): AHA Clinical Practice Guidelines,” cover such variations as, for instance, “in- 1838-47; see also J. P. A. Ioannidis, “Why Journal 301, no. stitutional review boards” (“IRBs”), used Most Published Research Findings Are of the American Medical Association 8 (2009): 831-41. in the United States, and “research eth- False,” PLoS Medicine 2 no. 8 (2005): e124, 33. Bastian, Glasziou, and Chalmers, ics boards” (“REBs”), used in Canada. By doi:10.1371/journal.pmed.0020124. “Seventy-Five Trials and Eleven Systematic “clinical research,” I mean all human-sub- 19. J. Lind, “A Treatise of the Scurvy” Reviews a Day.” jects research that aims to produce general- [1753], cited in Bastian, Glasziou, and 34. A familiar account of direct and in- izable knowledge that will advance human Chalmers, “Seventy-Five Trials and Eleven direct benefit is as follows: “A health and well-being. Systematic Reviews a Day.” direct benefit is the benefit obtained by those who need 2. H. Bastian, P. Glasziou, and I. Chalm- 20. D. Altman, “The Scandal of Poor and receive a resource (as a result of their ers, “Seventy-Five Trials and Eleven Sys- Medical Research,” BMJ 208 (1994): 283- needing and receiving that resource), and tematic Reviews a Day: How Will We 84, at 283. 21. Ibid., 283. an indirect benefit is a non-direct benefit Ever Keep Up?,” PLoS Medicine 7 no. 9 obtained by a third party as a result of the (2010): e10000326, doi:10.1371/journal. 22. Bastian, Glasziou, and Chalmers, fact that the resource is given to a direct pmed/10000326.s001. “Seventy-Five Trials and Eleven Systematic beneficiary” (J. Du Toit and J. Millum, “Are 3. S. Sieber, L. M. Machesky, and R. H. Reviews a Day.” Indirect Benefits Relevant to Health Care Insall, “Overflow in Science and Its Impli- 23. Ibid. 24. K. Loudon et al., “The PRE- Allocation Decisions?,” Journal of Medicine cations for Trust,” eLife 4 (2015): e10825, 41, no. 5 (2016): 540-57. doi:10.7554/eLife.10825. CIS-2 Tool: Designing Trials That Are Fit and Philosophy 35. Any such empirical evidence would 4. T. Hoffman et al., “The Scatter of for Purpose,” BMJ 350 (2015): h2147, be most welcome since it is directly relevant Research: Cross Sectional Comparison of doi:10.1136/bmj.h2147. to the assessments made in this paper. A Randomised Trials and Systematic Reviews 25. K. Borgerson, “Are Explanatory Tri- als Ethical? Shifting the Burden of Justifi- speculative list similar to my own appears across Specialties,” BMJ 344 (2012): e3223, (in a somewhat different context) in A. J. doi:10.1136/bmj.e3223. cation in Design,” Theoretical London, J. Kimmelman, and B. Carlisle, 5. M. Tsay and Y. Yang, “Bibliometric Medicine and Bioethics 34, no. 4 (2013): “Rethinking Research Ethics: The Case Analysis of the Literature of Randomized 293-308. 26. Loudon et al., “The PRECIS-2 of Postmarketing Trials,” Science 336, no. Controlled Trials,” Journal of the Medical Li- Tool.” 6081 (2012): 544-45. brary Association 93, no. 4 (2005): 450-58. 36. P. Glasziou et al., “What Is Missing 6. Bastian, Glasziou, and Chalmers, 27. S. Epstein, Inclusion: The Politics of from Descriptions of Treatment in Trials “Seventy-Five Trials and Eleven Systematic Difference in Medical Research (Chicago: and Reviews?,” 336 (2008): 1472-74. Reviews a Day.” University of Chicago Press, 2007). Niche BMJ 37. A. Berenson et al., “Despite Warnings, 7. A. W. Chan and D. G. Altman, “Epi- standardization is “a general way of trans- Drug Giant Took Long Path to Vioxx Re- demiology and Reporting of Randomized forming human populations into standard- ized objects available for scientific scrutiny, call,” New York Times, November 14, 2004, Trials Published in PubMed Journals,” Lan- political administration, marketing, or oth- http://www.nytimes.com/2004/11/14/ cet 365 (2005): 1159-62. See also Bastian, business/14merck.html. Glasziou, and Chalmers, “Seventy-Five Tri- er purposes that eschews both universalism 38. A. J. London, “A Non-Paternalistic als and Eleven Systematic Reviews a Day.” and individualism and instead standardizes Model of Research Ethics and Oversight: 8. J. Ioannidis, “Why Most Clinical Re- at the level of the social group” (p. 135). Certain vulnerable populations have been Assessing the Benefits of Prospective Re- search Is Not Useful,” PLOS Medicine 13, view,” no. 6 (2016): e1002049, doi:10.1371/jour- both overincluded and underincluded in Journal of Law, Medicine and Ethics (2012): 930-44. nal.pmed.1002049. medical research, and niche standardization, 39. For instance, in a 2015 paper, Sa- 9. Sieber, Machesky, and Insall, “Over- combined with explicit targeting of such bina Sieber and colleagues say, “Reducing flow in Science.” groups for inclusion, has been one (fraught) overflow is hardly a solution. It is widely 10. Hoffman et al., “The Scatter of approach to ensuring representation. accepted that funding scientific research Research.” 28. Califf et al., “Characteristics of Clini- leads to many social and economic benefits, 11. Ibid. cal Trials Registered in ClinicalTrials.gov, and calls to limit science participation are 12. Sieber, Machesky, and Insall, “Over- 2007-2010,” 1838. rarely supported by either governments or flow in Science.” 29. K. Humphreys, “Extent and Re- most scientists” (S. Sieber, L. M. Machesky, 13. F. Davidoff et al., “Evidence-Based porting of Patient Nonenrollment in In- and R. H. Insall, “Overflow in Science and Medicine: A New Journal to Help Doctors fluential Randomized Clinical Trials, 2002 to 2010,” JAMA Internal Medicine 173 Its Implications for Trust”). They go on to

34 HASTINGS CENTER REPORT November-December 2016 recommend changes to peer-review and 51. Ibid. 62. Government of Canada, “Tri-Coun- journal acceptance policies as means of ad- 52. The responsiveness criterion attempts cil Policy Statement: Ethical Conduct dressing the problem. to spell this out more explicitly for inter- for Research Involving Humans,” 2014, 40. R. B. Haynes, “Of Studies, Summa- national contexts, but the social value cri- p. 8, http://www.pre.ethics.gc.ca/eng/ ries, Synopses, and Systems: The 4S Evolu- terion is best understood as requiring that policy-politique/initiatives/tcps2-eptc2/ tion of Services for Finding Current Best the benefits of research are made available Default/. Evidence” [editorial], Evidence-Based Medi- to the communities in which research is 63. I do not defend a requirement of cine 6 (2001): 36-38; R. B. Haynes, “Of conducted. This is important for inner-city maximizing social value in this paper, al- Studies, Syntheses, Synopses, Summaries, Canadian emergency rooms in much the though I believe I would endorse a quali- and Systems: The ‘5S’ Evolution of Infor- same way that it is important for rural com- fied version of the position, if pressed on the mation Services for Evidence-Based Health- munities in Thailand. matter. care Decisions,” Evidence-Based Medicine 53. Emanuel, Wendler, and Grady, 64. Government of Canada, Tri-Council 11 (2006): 162-64; A. DicCenso, M. L. S. “What Makes Clinical Research Ethical?,” Policy Statement, 10. Bayley, and R. B. Haynes, “Accessing Pre- 2703. 65. U.S. Department of Health and Hu- appraised Evidence: Fine-Tuning the 5S 54. A. Wertheimer, “The Social Value man Services, Human Subjects Research 45 Model into a 6S Model,” Evidence-Based Requirement Reconsidered,” Bioethics 29, CFR 46, http://www.hhs.gov/ohrp/human- Nursing 12 no. 4 (2009): 99-101. no. 5 (2015): 301-08. subjects/guidance/45cfr46.html. 41. Sharon Straus and colleagues, for in- 55. A. Rid and D. Wendler, “A Frame- 66. Ibid. stance, claim that “in some circumstances, work for Risk Benefit Calculations in Bio- 67. J. Lunstroth, “The Role of Contro- the title [of a review] provides enough in- medical Research,” Kennedy Institute of versial Research in the IRB’s Risk/Benefit formation”; see S. E. Straus et al., Evidence- Ethics Journal 21, no. 2 (2011): 141-79. Analysis,” The American Journal of Bioethics Based Medicine: How to Practice and Teach 56. R. Foy et al., “How Evidence Based 11, no. 5 (2011): 14-16, at 16 (emphasis EBM (Toronto: Elsevier, 2005), 38. Are Recruitment Strategies to Randomized added). 42. Borgerson, “Why Reading the Title Controlled Trials in Primary Care? Experi- 68. Ibid., 16. Isn’t Good Enough,” 152-75. ence from Seven Studies,” Family Practice 69. One reason for resistance might be 43. Glasziou et al., “What Is Missing 20 (2003): 83-92. that members of RECs don’t have the skill from Descriptions of Treatment in Trials 57. Wertheimer considers this argument set or expertise to make this assessment. In and Reviews?” but dismisses it quickly, suggesting that it that case, perhaps there is a need to rethink 44. Ibid., 1472. extends the idea of public resources too far, the expertise required on RECs. 45. Ibid., 1474. and argues that there must be “demonstra- 70. Lunstroth, “The Role of Contro- 46. Hoffman et al., “The Scatter of ble social harms” in order to justify requir- versial Research in the IRB’s Risk/Benefit Research.” ing that privately funded research adhere to Analysis,” 14-16. 47. Bastian, Glasziou, and Chalmers, a strong social-value requirement. The first 71. Federal Policy for the Protection of “Seventy-Five Trials and Eleven Systematic sections of this paper may be read as an at- Human Subjects: Proposed Rule, Fed. Reg. Reviews a Day”; see also I. Chalmers and P. tempt to provide provisional grounds of this (Sept. 8, 2015). Glasziou, “Avoidable Waste in the Produc- sort: the abundance of low-quality research, 72. Emanuel, Wendler, and Grady, tion and Reporting of Research Evidence,” leading to the sorting problem, harms pa- “What Makes Clinical Research Ethical?,” Lancet 374 (2009): 86-89. tients, places research subjects at undue 2701-11. 48. Bastian, Glasziou, and Chalmers, risk, and compromises public trust in the 73. Borgerson, “Redundant, Secretive, “Seventy-Five Trials and Eleven Systematic research enterprise. and Isolated,” 385-411. Reviews a Day.” 58. The one exception to this, though 74. Freedman, “Scientific Value and 49. E. J. Emanuel, D. Wendler, and C. still on the fringes of bioethics, is work on Validity as Ethical Requirements for Grady, “What Makes Clinical Research malevolent uses of biomedical research; see Research.” Ethical?,” Journal of the American Medi- S. K. Green et al., “Guidelines to Prevent 75. The first problem identified in the cal Association 283 (2000): 2701-11; B. Malevolent Use of Biomedical Research,” report was “significant doubt about the Freedman, “Scientific Value and Validity as Cambridge Quarterly of Healthcare Ethics 15 capacity of the current system to meet its Ethical Requirements for Research: A Pro- (2006): 432-47. core objectives (dissatisfaction with the cur- posed Explication,” IRB: Ethics and Human 59. Rid and Wendler, “A Framework for rent system is widespread)” (Institute of Research 9, no. 6 (1987): 7-10. I set aside Risk Benefit Calculations in Biomedical Medicine, Responsible Research–A Systems further, more contentious, requirements at Research,” 148. The authors do make one Approach to Protecting Research Participants this time, simply for ease of argument. cryptic reference to social risks in their dis- [Washington, DC: National Academies 50. Revision of CIOMS 2002 Interna- cussion of “open questions” at the end of Press, 2003], 5). tional Ethical Guidelines for Biomedical the paper: “However, the framework also 76. T. Straight, “Clinical Research Regu- Research Involving Human Subjects. The makes clear that truly comprehensive risk- lations: Challenges to the Institutional Re- latest draft of the guidelines was retrieved benefit evaluations must consider social view Board System,” Clinics in Dermatology here: http://www.cioms.ch/publications/ risks from the research” (p. 168). 27 (2009): 375-83. guidelines/guidelines_nov_2002_blurb. 60. Well, perhaps we do. But we ought 77. A. Fleischman et al., “Dealing with htm. The language of “social value” has not to do so if we are aiming to act rationally. the Long-Term Social Implications of Re- been added to the first guideline of this 61. Freedman, “Scientific Value and search,” American Journal of Bioethics 11, most recent draft, accessed September 29, Validity as Ethical Requirements for no. 4 (2011): 5-9. 2015. Research.”

November-December 2016 HASTINGS CENTER REPORT 35