Clinical trials of germline gene editing: The exploitation problem

The birth of the

These criticisms notwithstanding, the event had been preceded by a significant shift in attitude in the scientific and bioethical community. 3 For a long time, germline genetic interventions were widely considered ethically out of bounds. However, partly because CRISPR potentially renders such interventions more accurate than earlier gene transfer technologies, scientists, ethicists, and advisory bodies had increasingly come to consider them in principle permissible provided that certain requirements are met. 4 Reflecting this stance, several of He's critics not only highlighted ethical violations involved in the experiment, but also proposed conditions that, if fulfilled, would make clinical use of human germline gene editing (henceforth GGE) ethically justified. 5 The pursuit of this goal will therefore likely continue.
Supporters of pursuing GGE under certain conditions agree that clinical applications should, at least to begin with, be reserved for avoiding the transmission of devastating diseases in cases where these cannot otherwise be prevented. Crucially, they also agree that such applications must wait until more is known about their safety and effectiveness. 6 Progress in this area will therefore require research, first preclinical studies on animal models and in vitro embryos and then, if warranted by the preclinical evidence, clinical trials where genetically edited embryos are transferred to the uterus of a gestational mother and brought to term. Such research raises ethical issues distinct from those foregrounded in longstanding bioethical debates about germline genetic modification. These debates have mainly concerned how the relevant technologies should be used once safe and effective. 7 However, more immediate questions concern whether they can ethically be developed, i.e. rendered safe and effective, and, if so, how. 8 This paper contributes to the discussion of this issue by considering whether clinical trials of GGE would comply with the fundamental research ethical requirement of non-exploitation. 9 I argue that it is unclear that they would. This is because participants would face substantial risks and burdens while being in a situation where declining participation is difficult. This concern constitutes a significant ethical obstacle on the path towards clinical use of GGE.
I proceed as follows. Section 2 introduces the concept of exploitation along with two necessary and jointly sufficient conditions for transactions and relationships to qualify as wrongfully exploitative. Section 3 argues that these conditions would likely be fulfilled in GGE trials, indicating a clear potential for exploitation. Section 4 discusses how exploitation could nonetheless be avoided or mitigated in this context, highlighting difficulties with all approaches. Section 5 concludes.
Two notes on scope. First, I focus specifically on clinical trials of human GGE, setting aside preclinical germline editing studies and clinical trials involving somatic cell gene editing. These other kinds of research may also raise ethical concerns, including concerns about exploitation. However, since these concerns have already received scholarly attention, 10 I set them aside. Second, the potential victims of exploitation I have in mind are the would-be parents or reproducers that would be the main participants in such trials, rather than, e.g. the resulting children or funders supporting the research.

| E X PLO ITATI O N
Philosophers and bioethicists generally agree that exploitation involves taking advantage of people in some wrongful way. However, they disagree about which transactions and relationships qualify as exploitative and what, precisely, is morally wrong with them. My argument does not presuppose any particular view on these matters, but merely appeals to certain features that most exploitation theorists recognize. The following paradigmatic example of an exploitative exchange illustrates these features: Pit: B is alone in a pit in the desert and cannot get out, facing death from dehydration unless the situation changes. A comes along and proposes to extract B, but only if B agrees to transfer his life savings to A. B agrees. The rescue requires minimal effort of A. 11 Pit highlights four significant general features of wrongful exploitation. First, even mutually beneficial transactions and relationships can be wrongfully exploitative. In Pit, A stands to benefit considerably from B's money. But B benefits too, assuming (plausibly) that giving up his life savings is better for him than death.
Second, exploitation can occur even though the victim consents.
Of course, in Pit B has 'no choice' but to accept A's proposal given that the only alternative is death. But since B is fully informed and not coerced or manipulated by A, his agreement can plausibly be considered consensual. 12 Third, even when exploitation is mutually beneficial, there is always something objectionable about the level (or perhaps nature) of the benefits involved. The exploiter benefits excessively (or | 3 MALMQVIST inappropriately), whereas the victim benefits insufficiently (or 'inauthentically'). Differently put, the exploitation makes the victim better off than before and better off than he otherwise would be, but worse off than he ought to be, whereas the exploiter is made better off than she ought to be. 13

| G ENE ED ITING TRIAL S AND E X PLO ITATI O N
Establishing that GGE trials would raise exploitation concerns requires showing that they would likely fulfil the vulnerability clause and the advantage clause. Let us consider each in turn.

| Gene editing trials and vulnerability
The vulnerability clause applies when one party to a transaction lacks decent alternatives to transacting on the other's terms. 19 If, in Participants in GGE trials would plausibly face a sufficiently constrained set of options for the vulnerability clause to apply. To see this, note that GGE is (at least in the short-to-medium term) proposed for a very small group of would-be reproducers. 21 Such interventions are mainly considered for severe incurable singlegene disorders (e.g. Tay-Sachs disease) since the genetic basis of most more complex conditions (e.g. cardiovascular disease, diabetes) is insufficiently understood to allow reliable editing. Most prospective parents are not at increased risk of transmitting such a disorder, and those at risk usually have other ways of avoiding having afflicted children. Adoption is one alternative, but is not available to everyone and cannot produce the genetic tie that many people strongly desire. However, at-risk couples who would be satisfied with a child that is genetically related to one of them may, in many jurisdictions, reproduce using donor gametes. And those who lack access to or reject this option may create, select, and implant embryos without the condition in question using preimplantation genetic diagnosis (PGD). 22 However, in rare cases, e.g.
when one prospective parent is homozygous for a dominant condition or when both are homozygous for a recessive one, these procedures fail to produce any unaffected embryo. GGE has clear utility only in this kind of case, i.e. for prospective parents who are at high risk of transmitting a severe incurable single-gene disorder, who strongly desire a genetic child, and who lack all alternative options just described. 23 These are also considered the primary candidates for participation in GGE trials. 24 Consider the choice situation of these participants. They strongly desire a healthy genetic child. They are likely to see the fulfilment of this desire as central to their well-being. They could try to reproduce without using gene editing. However, assuming that pregnancy and birth are achieved, this would likely mean caring for a child with a short and agonizing life. Alternatively, they could decide not to become parents at all or seek to adopt.
However, either option would leave the desire for genetic relatedness unfulfilled. It is therefore plausible that, when presented with the option of enrolling in the trial-which does offer a chance of a healthy genetic child-this is the only option they do not find 13  prohibitively burdensome. 25 And if this is the case, the vulnerability clause applies.
It might be objected that this exaggerates the importance of genetic relatedness. Cannot the goods of parenting be realized by caring for and raising a non-genetic, e.g. adopted, child? There are indeed strong reasons to think so. 26 However, my argument does not assume that the preference for genetic children is defensible, only that it would be deeply held by participants in GGE trials. And this does seem plausible. Absent a strong such preference it is unclear what would motivate these participants to enrol, given the alternative reproductive options surveyed above and the burdens of participation discussed in Section 3.2 below. Indeed, absent this preference, the rationale for pursuing GGE, whose main advantage over other means for avoiding transmitting severe genetic conditions lies in establishing a genetic parent-child link, would significantly weaken. 27 But even granted that participants in GGE trials would strongly prefer a genetic child, it might be argued that a lack of options for fulfilling this preference is not a sufficiently serious vulnerability to raise exploitation concerns. Remaining childless or adopting does not seem quite as bad as dying of thirst in a pit, after all, even for somebody who deeply desires a genetic child. In response, vulnerability is admittedly scalar, and it can be debated just how restricted people's choices must be for them to be vulnerable to exploitation. 28 However, it does not seem that all alternative options have to be truly catastrophic-like, say, imminent death or loss of a life partner.
By analogy, people are plausibly considered vulnerable to exploitation when they enrol in research to access otherwise unavailable medical treatment, even if the treatment is not life-saving but only relieves debilitating symptoms. 29 Similarly, workers are plausibly considered vulnerable to exploitation when they take precarious jobs to escape unemployment, even when unemployment would not mean starvation but rather feeding themselves and their families less than adequately. 30 The alternative option in these cases (experiencing debilitating symptoms, inadequately feeding oneself and one's family) is not catastrophic, but unattractive enough to make one vulnerable to exploitative arrangements. The alternative option in the GGE case (foregoing the chance of a much-desired genetic child) may well be equally unattractive.
These analogies could be challenged on the grounds that the alternative option is objectively bad in the other cases, but merely subjectively unattractive in the GGE case. Thus, it could be argued that while participants strongly prefer a genetic child, leaving this preference unfulfilled may not ultimately make them worse off. 31 In response, we should use (or at least allow for) a subjective standard when assessing the burdensomeness of options for the purpose of determining whether somebody is vulnerable to exploitation. 32 Clearly, whether people are willing to agree to exploitative arrangements depends on the subjective value they attach to their options at the time of choosing between them, not (directly) on the eventual objective outcome of the options. 33 In conclusion, it is plausible to think that GGE trials would fulfil the vulnerability clause. transaction has not yet occurred). 37 Third, their respective gains should be assessed from an ex ante perspective, focusing on the expected outcome at the time of entering the transaction rather than the outcome that eventually results. 38 Otherwise, many clearly exploitative transactions involving uncertainty-e.g. extremely risky human subject research-would be misclassified as non-exploitative merely because of chance.

| Gene editing trials and excessive advantage
While these points are fairly uncontroversial, the tricky part concerns (c) above, i.e. what normative standard should be used for assessing A's and/or B's gain. Some propose a comparative standard: A's gain is excessive when it is too large compared to B's gain. 39 Others defend a non-comparative standard: A's gain is excessive when B's gain fails to reach some threshold, such as the fulfilment of needs. 40 However, specifying the standard in greater detail has proven elusive. Indeed, Wertheimer-the leading exploitation theorist in research ethics-concluded a thorough 2011 examination of extant proposed candidates (including one earlier defended by himself) by noting that none was ultimately plausible. 41 Absent a compelling standard of excessive gain, the most promising way of assessing whether GGE trials would fulfil the advantage clause is arguably to proceed casuistically, comparing this case to other relevantly similar cases where the clause plausibly is fulfilled.
One such case is pharmaceutical trials in low-and middle-income countries (LMICs) in cases where participants enrol to gain temporary and/or uncertain access to needed but otherwise unavailable medicines or healthcare, and the investigational drug will be marketed in affluent countries but will be unaffordable in LMICs. 42 Another is phase 1 trials involving paid healthy volunteers in cases where participants depend economically on trial income, but are paid modestly, lack insurance, and risk exclusion from the study without pay, while assuming health risks and enduring unpleasant or intrusive research procedures, strict diet and resting regimes, and weeks of confinement. 43 Both these cases are plausibly thought to involve excessive gain to researchers and sponsors and insufficient gain to participants (and, since the latter are vulnerable, to be exploitative). The following features support this judgment: the potential gain to the powerful party-the sponsors and researchers-is substantial, yet not, in a single case, necessary to secure an adequate level of well-being. On the other hand, the potential gain to the vulnerable party-the participants-is necessary to their well-being; this is what makes them vulnerable. However, this gain is uncertain or insufficient to fully meet that need, and the burdens incurred in pursuing it are quite substantial. Crucially, these features suggest that researchers and sponsors benefit excessively even though their net gain may be considerably smaller than that of the participants, who, after all, stand to benefit from much-needed medicine or income. As in many other cases (including Pit), the victims of exploitation, due to their vulnerability, have more to gain than their exploiters from the exploitative transaction. 44 It is plausible to expect GGE trials to display similar features, which provides reason for thinking that they would indeed fulfil the advantage clause. To begin with, such a trial would be a prestigious However, even an unsuccessful trial would likely provide some such benefits, as long as it is perceived as scientifically and ethically legitimate. Moreover, recall that benefits should be assessed ex ante.
Despite the risk of failure, the expected gain to these parties seems substantial, assuming (again) that the trial is seen as legitimate.
Turning now to the participants, the experiment would expose them to significant risks and burdens. This is partly because producing a genetically edited child would require the prospective mother to undergo IVF, pregnancy, and delivery. 45 IVF using her eggs, which would be needed to create a genetic link to the child, would require hormonal stimulation, which commonly causes mild abdominal pain, bloating, nausea and vomiting, and mood changes.
A more serious complication is moderate or severe ovarian hyperstimulation syndrome, 46 which is associated with severe abdominal pain, severe nausea and vomiting, rapid weight gain, respiratory difficulties, and, exceptionally, death. Since IVF has a modest success rate, the woman may face these IVF-related risks and burdens repeatedly. Once pregnant, she would incur several additional costs, including discomforts like nausea, fatigue, and pain; libertyrestrictions related to diet, recreational drug use, and physical activity; temporary and/or permanent undesired changes in bodily appearance and/or function; risks of pregnancy-related illnesses like pre-eclampsia and gestational diabetes; and a risk of miscarriage. 47 Finally, delivery is at minimum painful and exhausting, often causes injuries requiring surgery or follow-up care, is not uncommonly followed by postpartum depression, and may, in rare cases, prove fatal. The fact that this particular pregnancy and delivery would likely be exceptionally closely monitored might somewhat mitigate the health risks involved but would also add a further cost: interference with the woman's privacy.
It might be argued that the risks and burdens of IVF, pregnancy, and delivery should not be considered risks and burdens of GGE trials, either because they do not result directly from the GGE intervention or because people willingly (and often unobjectionably) assume them when reproducing outside of the research context.
However, as argued earlier in this section, when analysing whether some transaction involves exploitation, we should take into account all benefits and burdens involved and assess these against the alternative of not transacting. As argued in Section 3.1 above, at the time of deciding whether to enrol in a GGE trial, these participants would likely have no other reproductive option they consider viable.
Thus, at this time, their no-transaction alternative does not involve the risks and burdens of IVF, pregnancy, and delivery. This justifies including these risks and burdens in the analysis.
Moreover, many uncertainties remain about the effects of altering the genome, making such interventions risky to the resulting child (and, possibly, to future generations). Should the child experience serious harm due to off-target or unwanted on-target effects, this would severely set back the participants' interest (qua parents) too. 48 However, even if no such harm occurs, the risk of harm will likely engender significant anxiety. On the inclusive approach to benefits and burdens defended above, these burdens, too, should be included in the analysis.
On the other hand, the trial would represent an opportunity for the participants to fulfil their desire for a healthy genetic child.
Since (as argued in Section 3.1 above) they can be assumed to have a powerful such desire while lacking other ways of fulfilling it, this is clearly an important potential benefit-one that is plausibly considered central to their well-being. However, given present knowledge gaps about the effects of altering genes and the modest success rate of IVF the experiment may very well fail to yield a healthy child, or indeed any child. Thus, unlike some of the key burdens involved, the main benefit to participants would be highly uncertain.
This analysis suggests that GGE trials, much like better known cases of exploitative research, would involve excessive gain to researchers, sponsors, and employers and insufficient gain to participants, i.e. that the advantage clause would apply.

| M ITI G ATI N G E X PLO ITATI O N
I have argued that participants in GGE trials would be hard put to decline participation, while also facing risks and burdens that are substantial and not obviously offset by compensating benefits. Therefore, they would be at danger of exploitation. However, this does not mean that they would necessarily be exploited. There could be ways of preventing or mitigating the exploitation. I now consider how this might be achieved. My assessment assumes that compelling mitigation approaches must reflect plausible general moral principles, but also be feasible, which inter alia requires sufficient fit with established research ethical standards. 49

| Participant selection
One approach could be to enrol participants who are not vulnerable in the relevant sense, either because they do not strongly desire a healthy genetic child or because they could fulfil that desire in some other way. However, while this would defuse exploitation concerns, it would also make it difficult to recruit sufficiently many participants. It is not clear that people would agree to undergo IVF, pregnancy, and delivery without a strong interest in parenting the resulting child, nor that they would opt for experimental gene editing if they could secure healthy offspring more easily by means of PGD or donor gametes.
Moreover, this move would make the risk-benefit profile of the experiment hard to justify. As argued in Section 3.2 above, assessing whether research meets the non-exploitation requirement is distinct from making a risk-benefit assessment. Nonetheless, a defensible risk-benefit profile is a necessary ethical requirement on all human subject research. 50 Selecting non-vulnerable participants would arguably conflict with this requirement, since these participants would face all risks and burdens of IVF, pregnancy, and delivery without enjoying the main compensating benefit: a unique chance of fulfilling a powerful reproductive desire. 47 Pregnancy may involve significant benefits too, including increased attention and care from others, and it might be argued that these outweigh the costs. However, since many of the benefits specifically aim at easing the burdens involved, it is unclear that

| Payment
Another, perhaps related, approach could be to pay people to enrol in GGE trials. This would increase the level of net gain to participants, addressing the worry that they benefit insufficiently.
Moreover, if high enough, payment would presumably help incentivize participation, as it does in other research, 51 potentially overcoming recruitment problems.
However, this approach would potentially clash with established research ethical norms in two ways. First, since the risks and burdens would be high, the payment would also need to be high to effectively incentivize participation. This would raise familiar worries about 'undue inducement', i.e. worries that the prospect of large sums of money may entice people to participate against their better judgment. Second, while offering payment to incentivize enrolment or reimburse costs that participants have incurred is considered acceptable, there is widespread resistance against the idea of paying to offset risk. 52 Now, this resistance concerns risk-benefit assessment, not the evaluation of whether the research is exploitative, which (again) is a distinct issue. However, it does not seem unlikely that the reluctance of ethics committees to consider payment a benefit for the purposes of the former assessment may affect their consideration also of the latter issue.
Of course, both the notion of undue inducement and the rejection of payment as a benefit may be criticized. 53 But whatever the merits of these criticisms, both notions are parts of research ethical thinking and practice. So regardless of whether paying people to enrol in GGE trials is permissible in principle, it may well be unfeasible in current circumstances.

| The long road
The most promising mitigation approaches are likely more long-term than those considered so far. As noted in the Introduction, proponents of developing human GGE agree that clinical trials must be preceded by rigorous preclinical studies on animal models and human embryos. This requirement limits the burden on participants that stems from the uncertain effects of gene editing and associated risks to the resulting child, and more so the higher the bar for necessary preclinical evidence is set. However, preclinical research can only reduce this uncertainty, not eliminate it. 54 Moreover, such research will not reduce the risks and burdens of IVF, pregnancy, and delivery, which I have argued are substantial. Thus, some concern about excessive advantage will remain no matter the quality of preclinical research.
Given these limitations, it is worth asking whether exploitation could instead be mitigated by reducing participants' vulnerability, which arises from a powerful desire for healthy genetic offspring and a lack of alternative options for fulfilling that desire. While addressing this vulnerability by participant selection seems difficult (see Section 4.1 above), there might be another way. People's desire for genetic offspring is partly produced by social norms that attach great importance to genetic relatedness. Social norms change, however, and these particular ones are arguably being challenged by new kinds of family constellations and increasing reproductive options among groups previously lacking them. Should the societal preference for genetic offspring significantly weaken, people may be less inclined to enrol in risky and burdensome experiments to satisfy it, and hence less vulnerable to exploitation in such experiments.
Perhaps GGE trials will not raise exploitation concerns in a future where the societal preference for genetic relatedness has significantly weakened. However, in such a future, the attraction of GGE will also be smaller since the main rationale for the procedure is to provide an opportunity of securing genetic offspring to those who otherwise lack it.

| CON CLUS ION
Many scientists and ethicists now advocate the pursuit of clinical uses of human GGE. However, before this technology can be rolled out its safety and efficacy must be established in clinical trials involving human subjects. I have argued that such trials would potentially conflict with the fundamental research ethical norm of non-exploitation. This is because they would expose people who are in a vulnerable situation to quite significant risks and burdens.
Moreover, this problem seems hard to mitigate without undermining the main rationale for GGE. Some may disagree with my analysis of participants' vulnerability or of the risks and burdens they face.
However, I hope to at least have shown that GGE trials raise a potential exploitation concern that warrants serious attention if (or when) such trials are considered.

ACK N OWLED G EM ENTS
This paper was first presented at the workshop 'Gene Editing in Human Reproduction: Experimenting on Future Children', which was held at the Brocher Foundation, Geneva, January 14-16, 2020. I wish to thank the organizers for the opportunity to present my work, the Foundation for hosting the event, and other participants for useful feedback. I also thank audiences at Linköping University and the University of Gothenburg, as well as two anonymous reviewers for Bioethics, for helpful comments.

CO N FLI C T O F I NTE R E S T
The author declares no conflict of interest.