Safe Gene Modification

What does it mean to say that a particular gene modification is safe or not safe? What harms, or potentially ethical or legal wrongs are involved?

It is clear that “safe” and “unsafe” should be read in a narrow technical sense in this discourse. Safety is conceptualized as being about biological harms that may befall the gene edited organism as a result of the editing, and not about wider social or ethico-legal harms. This restricted use of “safe” is often not made explicit, but it is implicit in distinguishing between safety and governance needs,^{Footnote 6} or in focusing on examples of biological lack of safety.^{Footnote 7} Such a restrictive account of safety may in itself be problematic, but a closer analysis of this issue is outside the scope of this paper where we will accept the narrow biomedical framing of safety.

If we perform gene editing in an organism, that organism can be harmed in various ways (let us call these harm₁, harm₂, harm₃, and harm₄):

1. 1) The gene editing may be intended to be harmful

2. 2) The gene editing is technically inefficient and leads to mosaicism

3. 3) The gene editing may be technically efficient (i.e., it makes the desired change in the genome), but may nevertheless turn out to be harmful either on its own, in combination with some other part of the organism’s (epi-)genome, or in combination with some infection or environmental exposure, either immediately or during the lifetime of the organism

4. 4) Off-target genetic changes may be harmful

In relation to clinical use in humans, including the research leading up to clinical use, we can discount harm₁ and concentrate on harms_2-4. No health care professional would perform gene editing with the express intention to harm. If we perform gene modification in humans it will always be in cases where we predict net benefit from the intervention, but where harm may nevertheless occur.

The likelihood and magnitude of these harms can be estimated from research evidence, and may be reducible by future research and development. It is, however, as with all technologies, unlikely that the risks of these harms occurring can be removed completely. A recent example of a harm falling into category 3 above is the emerging evidence that gene editing is more efficient in cells that are deficient in p53 function, and that this may lead to the selection of gene-edited cells with increased risk of tumor formation because p53 is an important tumor suppressor.^{Footnote 8,Footnote 9} It was only through research that this problem was identified, and it is only through research that we may find ways to overcome it.

If the gene modification is taking place in vitro—outside of the body in cultured cells—we may have technical procedures for eliminating the risks of harm₂ and harm₄, by modifying cells and then testing and selecting only cells with the correct gene modification, but in vivo gene modification is likely always to entail some risk of these harms eventuating.

When is a particular gene modification safe enough to introduce in clinical practice? The simple answer for adults and children, in terms of harms, seems to be: when we can be reasonably certain that

$${\rm{benefits}}\,{\rm{ - harms}} > 0,$$

i.e., when we are reasonably certain of obtaining net benefit from the intervention. There are many in principle and practical questions about how we estimate the benefits and harms quantitatively and about what we mean by “reasonably certain,” but resolving them is not relevant to the current analysis of the function of the “only when it is safe” rider.

This threshold must be the right one for approval from a regulator for routine clinical use. A regulator could not allow an intervention to be marketed if there were still significant concerns about its safety. But what if the condition that is targeted by the gene modification is very severe, and there are patients willing to take the gamble before we have sufficient evidence to determine the risk/benefit ratio? Should such patients be allowed to use the gene modification, and should researchers be allowed to offer it to patients as experimental treatment? This question is not in principle different from similar questions raised in relation to other therapeutic technologies and methods and discussed in the literature on “right to try” and expedited/expanded access,^{Footnote 10-Footnote 12} and will not be analyzed further here. It is, however, important to note that it raises important questions not only about the limits, if any, of personal autonomy, but also about the balance between personal liberty to choose and societal interests in having safety and efficacy questions definitively answered through well-planned research.

Reproductive Safety

Should we understand safety differently if the gene editing is either performed as part of the creation of a child by means of a reproductive technology, or affects the germ cells of the patient?

Bracketing the nonidentity problem and considerations of reproductive liberty, it seems that there are no relevant differences between the reproductive context and the nonreproductive context in relation to safety. If gene editing is an intended part of a reproductive project, “being safe” again simply means that we are reasonably certain that there is net benefit for the child that is the end goal of the reproductive project.

In the directly reproductive context, in which the entity being edited is either a gamete or an embryo, the decision of whether or not to use gene editing will be taken by the people involved in the reproductive project. They choose for someone else, i.e., their future child, not for themselves, and the child cannot consent. But all reproductive decisions by the reproducers are made without the consent of the children, but if they chose a nonsafe gene modification intervention (i.e., one where we are not reasonably certain that there is net benefit) they would unjustifiably put the child at risk of harm.

In the indirectly reproductive context, in which germ cells are edited in vivo either deliberately or as a side effect of some type of in vivo somatic gene editing, similar considerations about safety and harm apply. We should not introduce genetic modifications into germ cells unless the modifications are unlikely to lead to net harm to any future child created using gametes generated by these sperm cells. Or, if the gene editing is beneficial for the person in whom it is done, but likely to be harmful to future children, then that harm and the reproductive lack of safety should be an important consideration in determining whether to bring the particular type of gene editing into clinical use.

So, the interim conclusion so far is that safety matters both ethically and in relation to regulation, and it matters because we are at risk of causing significant harm if we use unsafe gene editing. The “only when it is safe” rider therefore seems perfectly justified.

Reproductive Safety Reconsidered

The reader who is steeped in the Anglo-American bioethical literature will by now probably be screaming “but why, oh why have you ignored reproductive/procreative liberty and the nonidentity problem in the analysis. Because of those two considerations reproduction is different!”

And, it is true that reproduction is different, so different that there is considerably less agreement about our ethical obligations in the procreative sphere than there is about our general ethical obligations (see for instance the book-length exchange between David Benatar and David Wasserman^{Footnote 13}). So, let us take reproductive liberty and nonidentity seriously in our analysis of the meaning of harm in relation to gene editing and imagine a situation where a set of prospective parents wants to use gene editing as part of the process leading to the creation of their next child. Do they do something wrong if they use a gene modification intervention that is unsafe in one of the four ways outlined above? We will here bracket the issue that this can only be done as part of technically advanced assisted reproduction and therefore necessarily involves a number of third parties who may or may not have independent ethical importance.

The standard interpretation of the nonidentity problem is that if an action (or an inaction) I is necessary for the coming into being of person P, then P cannot have been wronged by I being performed, even if I leads to P being in a harmed state as long as P’s state is not so bad that it constitutes “a life not worth living.”^{Footnote 14} P has not been wronged, because you cannot be wronged if, without the action causing the putative wrong having been performed, you would not have existed. Some proponents of nonidentity also claim that P cannot be harmed by I, since harm should be understood counterfactually and because I is necessary for P coming into being there is no possible world in which P exists, but in a different unharmed state. Disentangling the many different possible interpretations of nonidentity and its implications, and the many and varied highly counterintuitive consequences that flow from it, is far beyond the scope of this paper. What is important in the present context is that if we accept something like nonidentity, then there is definitely no one who is wronged, and plausibly no one who is harmed, by the use of an unsafe gene modification intervention in reproduction, and it is important that this is accepted by some of those who put forward the “only when it is safe” rider,^{Footnote 15,Footnote 16} whereas others are more equivocal.^{Footnote 17} That there is no wrong done, and plausibly no harm created, is true of the instant child, and by recursion true of anyone who will ever carry the modification, given that the first act of modification is a necessary condition for their existence. If the parents are doing something ethically problematic, it is only in relation to the state of the world, if the world would, counterfactually, have been better without the existence of P.

It might be argued that not all genetic modifications introduced by gene editing are identity forming or identity changing. Some may, for instance correct a disposition to a condition with very late onset (e.g., Alzheimer’s disease) without changing the identity of the embryo that is edited or the person it becomes. And, as a thought experiment, we can conceive of gene editing of a part of the genome that has no known functions at all. So, gene editing of an embryo does not necessarily or automatically lead to nonidentity. There are many different conceptions of identity at play when considering nonidentity. As I have argued in a previous paper in CQ, nonidentity discussions can involve considerations of numerical, genetic, phenotypical/physiognomical, psychological, and narrative/social identity, and it is often important to be clear about what kind of identity we are discussing.^{Footnote 18} However, in the present context, almost all conceptions of identity lead to the conclusion that the gene editing of embryos affects identity. Not because the actual modification introduced necessarily affects identity, but because it is very unlikely that the same unedited embryo would ever have been created, implanted, gestated, and born.

In a context where gene editing is available to them, prospective parents will, with their clinicians, plan the IVF + gene editing, and it is very unlikely that this will happen at exactly the same time and in exactly the same way as it would have happened if gene editing had not been available, i.e., it is highly unlikely that exactly the same ova will be retrieved and fertilized by exactly the same spermatozoon. The resulting child will therefore be numerically, genetically, phenotypically, and psychologically different from the child which would have existed (if any) if gene editing had not been available. The only type of identity that might be preserved is an attenuated form of narrative or social identity. The child would still have the same number in the birth order and might still have the same sex, so could still be “the first son born to Jack and Jill”; but because of the many nonidentities it is likely that its narrative would quite quickly and quite substantially diverge from the narrative of the child brought into the world without gene editing being available.

This brings us to reproductive liberty/freedom/autonomy. Reproductive liberty can be understood either as a jurisprudential or an ethical principle or both. The core of reproductive liberty is the claim that our reproductive choices are strongly protected from outside interference, and that this strong protection is justified by the central importance of reproduction to the life plans and personal identity of people. In his seminal first paper arguing for the importance of procreative liberty, John Robertson writes:

A strong account of reproductive liberty entails that procreative acts are protected from outside interference even in cases where the resulting child is harmed and where the people reproducing are acting in ways that are recognized even by them as ethically wrong and ordinarily blameworthy.

Taken together, nonidentity and reproductive liberty lead to the conclusion that almost no uses of genetic modification in reproduction can be deemed as unsafe (because of nonidentity considerations), and if there are any unsafe uses, they can nevertheless not be prohibited or interfered with in other ways if the procreating parents want to use them (because of reproductive liberty). The only exceptions are gene modifications that are so harmful that they lead to a “life not worth living,” since creating such a life is a wrong in all circumstances and also a sufficient justification for overriding even a strong right to reproductive liberty.

Procreative Beneficence and the Ethics of Reproduction

As noted above, the ethics of reproduction is immensely complex and the range of positions on the duties of prospective parents contemplating reproduction go from a strong duty not to reproduce (and a fortiori not to reproduce in ways involving gene editing) to support for the Biblical injunction to be fruitful and multiply. The part of this wide ranging field that is of relevance here is the question of whether prospective parents have duties to consider the welfare of the child to be in their reproductive decision-making. It has been argued that parents have such duties, either a duty not to bring a child into the world that is in a harmed state,^{Footnote 20} or a stronger duty to follow a principle of procreative beneficence and bring into the world the best child they can.^{Footnote 21} There has been considerable criticism of these posited duties, both in terms of their justification and in terms of their wider implications.^{Footnote 22-Footnote 24}

Some of the writers in the gene editing debate are committed to a trinity or triple of positions encompassing nonidentity, reproductive (legal) liberty, and procreative beneficence. This commitment to the triple is rarely expressed explicitly, but accepting two of the positions simultaneously in argument is quite comment. Julian Savulescu and Guy Kahane, for instance, discuss the interplay between reproductive liberty and procreative beneficence in their seminal paper on the principle of procreative beneficence:

Can the “only when it is safe” rider be justified from the triple of positions? There is a question about whether the triple itself is internally consistent, since there is a possible conflict between nonidentity and a duty of procreative beneficence, and a possible conflict between reproductive liberty and reproductive beneficence if reproductive liberty is understood as a moral principle. If the triple of positions is internally inconsistent, it follows trivially that the rider is justified, since according to the rules of first order predicate logic, any conclusion can be validly inferred from a set of inconsistent premises. But, this is of course not real justification but pseudo justification, because soundness is undermined by the problem with the premises. In the following, we will therefore assume that the triple is internally consistent and take reproductive liberty to be a legal and not a moral principle.

What can be justified is then the claim that prospective reproducers do something which is seriously morally wrong if they use an unsafe mode of gene editing (from procreative beneficence), that this wrong is not directly based on any harm caused to the child (from nonidentity), and that although they are doing something morally wrong, they should not be legally prevented from doing it (from a legal understanding of reproductive liberty). This initially looks very much like the “only when it is safe” rider, but this initial appearance is deceptive because the “only when it is safe” injunction is now a purely moral injunction with force only for the prospective reproducers. It is not an injunction aimed at regulators telling them how to regulate or at health care professionals telling them when to refuse to assist in a reproductive project involving unsafe gene editing. Health care professionals can express their disapproval of the reproductive choice (as per Savulescu and Kahane^{Footnote 27}) or they can remonstrate, reason, persuade or entreat (as per John Stuart Mill just after the famous enunciation of the so-called “harm principle” in On Liberty^{Footnote 28}), but they cannot refuse to participate unless they give up their neutrality as health care professionals.

The Function of Safety Arguments

The argument seems to have arrived at a point of perplexity. Many of the interlocutors in the debate who seem to be committed to the “only when it is safe” rider are also committed to the view that only extreme lack of safety can matter when regulating the use of gene editing in a reproductive context. The nonidentity problem entails that any future child whose coming into existence has involved gene editing cannot have been wronged by that use of gene editing, and the invocation of reproductive liberty mean that prospective parents have a strong right to reproduce in any way they choose, even if the child that is being brought into existence is in a harmed state. And, even if we commit ourselves to procreative beneficence this has no bearing on when we should allow gene editing to be used in reproductive contexts. So what is the real function of the “only when it is safe” discursive move?

It could of course be inadvertent ignorance of the underlying inconsistency, i.e., the ethicists in question are not aware or have not yet realized that they are really committed to the view that safety does not matter very much, if at all, in reproductive contexts. Or, that if it matters it matters only morally. I think that is unlikely, since we are talking about highly intelligent people with excellent philosophical skills, but cannot prove that it is not the case. I will, however proceed on the assumption that they know the content of their own prior writings and their stated philosophical commitments in relation to reproductive ethics sufficiently well to be able to see, almost immediately, that there is an inconsistency here. So, something discursively more interesting must be going on. In the following, I will analyze five possible explanations for the use of the “only when it is safe” rider. These five options are not mutually exclusive; two or more may be combined in the explanation of any particular instance of this discursive phenomenon:

1. 1) simplification of argument

2. 2) ritualized hand-waving

3. 3) “consensus building” with nonphilosophers (in casu scientists)

4. 4) blurring of the line between philosophical analysis and policy advice

5. 5) deliberate obfuscation

The simplest and least contentious function of the “only when it is safe” rider is that it allows the argument about the ethics of eventual uses of gene editing to proceed, while we are bracketing safety issues. So, we can, for instance, isolate and analyze questions of gene editing for enhancement purposes without having to think about any possible interactions between considerations of purpose and considerations about safety. This is standard philosophical technique, tackling the issues one by one, and is unproblematic, as long as we remember that we have bracketed safety and that all of our conclusions are therefore qualified by safety. That is, if we find that X is ethically unproblematic when safety is bracketed, the conclusion we can draw and state is not that “X is ethically unproblematic,” but that “X is ethically unproblematic, if gene editing is safe.”

The second function, “ritualized hand waving,” conceptualizes the “only when it is safe” rider as an almost automatic invocation, on a par with the Muslim’s “inshallah” finishing of sentences signifying future intentions or “please” in British English. It has little actual meaning apart from being something that has to be said every time we discuss a new technology. This can be linked to the observation that it is one of those propositions which it is very difficult to negate and still be taken seriously by polite society. Claiming that we should use gene editing in general, or for a specific purpose “long before it is safe” simply does not sound like a good idea. We live in a “risk society” where it is a widely shared value that risks have to be identified and minimised.^{Footnote 29} On this account, any inconsistency with other philosophical commitments is therefore not a real inconsistency, because uttering or writing “only when it is safe” does not, despite surface appearances, show any real commitment to the content of the proposition.

The third function of “consensus building with nonphilosophers,” in casu researchers using gene editing and clinicians wanting to use it, is a plausible explanation for the frequent occurrence of the “only when it is safe” rider in reports and papers produced by multidisciplinary groups. As explained above, negating this rider is pragmatically and discursively difficult, within the general frame of discourses around new technologies. The precautionary principle and similar cautionary approaches are part of that general frame, whatever philosophers and regulation theorists might think about them. The philosopher may thus agree to “only when it is safe” to stay within the standard discursive frame and seem reasonable and sane, with the tactical goal of, for instance, getting a more receptive audience in the group for arguments indicating the ethical acceptability of a wide scope of use of gene editing, e.g., not just restricting it use in humans to core therapeutic use.

The fourth function of the rider, “blurring of the line between philosophical analysis and policy advice,” has received some attention in the bioethics literature, but usually the problem that is discussed is the reverse of the problem we are analyzing here. That is, the discussion is usually about how philosophers overlook or elide the difference between putting forward a radical philosophical conclusion and advocating for a radical public policy change in the direction of that conclusion.^{Footnote 30,Footnote 31} What we have here are philosophers explicitly abjuring themselves of potential public policy implications of their philosophical conclusions. They could say “safety does not matter very much” for the regulation of reproductive uses of gene editing and advocate that we develop our public regulatory policies accordingly, but instead they say that our policy should be only to use the technology “when it is safe.” Why this reticence to follow the philosophical conclusions where they lead? Now, it could in an optimistic mode be argued that since gene editing is a relatively new technology, what has happened is that philosophical bioethicists have read the papers referenced just above and the many similar papers and books on the gap between philosophical analysis and public policy prescription and have finally realized the significant difference between reaching a philosophical conclusion and advocating for that conclusion as directly implementable public policy. Having had this “road to Damascus” experience, they have therefore started to take account of the specific challenges in public policy development and have stopped immediately transforming philosophical conclusions into policy advice. I leave it to the reader to decide whether this is a plausible account.

We therefore have to consider the fifth and final possible function of the “only when it is safe” rider, i.e., deliberate obfuscation. By deliberate obfuscation I mean that the rider is inserted primarily in order to hide the full implications of the philosophical positions that underpin the analyses of the ethics of reproductive gene editing, i.e., it is inserted deliberately in order to placate and mislead the reader. Here its function is to signal something like “don’t worry too much about gene editing, or the fairly radical things we say it is OK to do with it; they are all far in the future and we will only do them when it is safe to do them.” But this signal elides the difference between biological safety and wider conceptions of safety, and it obscures the wider role of nonidentity and reproductive freedom considerations in reaching conclusions on the ethical acceptability of particular reproductive uses of gene editing.