Do the appropriately titrated administration of opioids, and palliative sedation, shorten survival time?

Schofield et al 2 argue that my literature review excludes evidence regarding the safety of opioid and sedative use outside of the palliative care setting, and that this has introduced a cherry-picking bias and skewed the results. I agree with the authors that in other settings the safety of opioids and sedatives has been clearly established, for instance the large quoted range of daily oral morphine equivalent which is sometimes required and well tolerated (although note that the 21 600 mg was an outlier and the next highest dose in the study described was 720 mg).6 Despite this, the lack of good quality evidence in the specific palliative care setting is highly problematic. It is often intuitive that a drug with an effect in one setting will have that same effect in another setting; however, this intuition is often subsequently proven incorrect. Recognising this is a fundamental tenet that drives much of the empirical research conducted in medicine. A good example demonstrating this is the blood-thinning drug dabigatran, which was found to be an effective alternative to warfarin for preventing strokes in the setting of an abnormal heart rhythm called atrial fibrillation, although it was subsequently (and surprisingly) found in a landmark study not to be an effective alternative to warfarin in preventing strokes in the different setting of patients with mechanical heart valves.7 By the same token, although opioids and sedatives are safe in the setting of an elective total hip replacement, they may not be safe in the entirely different context of a frail patient with metastatic lung cancer in a hospice, and we simply do not have good evidence to know this one way or the other. The study referred to by the authors regarding the paradoxically increased survival among patients receiving sedatives after withdrawal from mechanical ventilators is interesting,8 although it is not particularly relevant to the vast majority of the broader palliative care population, and it was a retrospective case series of five patients without a control group; hardly a rigorous level of evidence. Recognising all this, defining the population group of interest (ie, the palliative care population) is an essential step in generating the initial research question being asked, and my search strategy reflects this and was carefully developed to prevent any such cherry-picking bias. The authors do make the good point that my research was not prospectively registered or designed according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines; however, my literature review simply searched for pre-existing systematic reviews and meta-analyses and then provided an appraisal of these, rather than conducting a systematic review or meta-anlysis of primary studies de novo (nor did it claim to), so this concern is not applicable in this context. The systematic reviews appraised each had their own methodologies detailed, and these can be directly and transparently reviewed by our readership. I therefore reject the authors’ repeated claims that the rigour of my research process has been substandard, incomplete and misleading, and I maintain my original conclusion that agnosticism is the most appropriate position to take on whether palliative opioid and sedative use shorten survival time.

Schofield et al also argue that continuous deep palliative sedation (CDPS) falls beyond the scope of routine palliative care and is not considered best practice, and that discussion of this practice therefore obscures the question of whether other forms of palliative sedation shorten survival time. The authors correctly acknowledge that palliative sedation is an umbrella term describing a heterogeneous group of practices, utilising a variety of drugs and drug classes at different doses and dosing intervals, in order to achieve any combination of light/deep and intermittent/continuous sedation (or anywhere on these spectrums in between). The authors assert that CDPS ‘remains controversial globally’ and ‘should not be grouped with routine palliative prescribing’, particularly in their own experience working in the UK. They refer to the European Association for Palliative Care’s (EAPC) ‘recommended framework for sedation in palliative care’ to substantiate this claim,9 and Materstvedt3 and Symons4 also refer to this framework. In my experience working in Australia, CDPS is occasionally used as a measure of last resort—mainly in the setting of refractory terminal agitation in the final hours or days of life, when less intensive forms of sedation have failed—and is considered an accepted component of routine palliative care. Despite the authors’ claim, this seems consistent with the EAPC framework, which describes specific intolerable symptoms including ‘dyspnoea, pain and convulsions, (and) emergency situations (such as) massive haemorrhage, asphyxiation, severe terminal dyspnoea or overwhelming pain crisis’ as accepted indications for CDPS. It acknowledges that ‘continuous deep sedation should only be considered if the patient is in the very terminal phases of their illness with an expected prognosis of hours or days at most’, and states that ‘deeper sedation should be adopted when mild sedation has been ineffective’. It even describes a set of criteria in which CDPS may be selected as first-line therapy and situations in which CDPS may be considered for refractory psychological (not physical) suffering. Interestingly, the framework also refers in multiple places to the risk of hastened death with palliative sedation, and also the occasional need for flumazenil (the benzodiazepine antidote) for cases in which respiratory depression occurs despite the judicious use of sedatives even in specialist palliative care centres (although I have not encountered this in my experience). The Australian and New Zealand Society of Palliative Medicine (ANZSPM) also aligns itself with the EAPC’s framework.10 In light of this, I reject the authors’ claim that CDPS falls beyond the scope of routine palliative care.

Finally, Schofield et al correctly identify opiophobia as a significant barrier to the provision of palliative care globally. They express their concern that my research will add to this opiophobia and may be harmful by virtue of creating ‘significant concern and fear for current and future patients, their families and healthcare professionals’. I want to emphasise that I have not been arguing against palliative opioid and sedative use; on the contrary, I know that they are essential and underutilised. My arguments regarding agnosticism on whether they shorten survival time are instead aimed at advancing the case for euthanasia, which can occur in parallel with palliative opioid and sedative use as complementary rather than mutually exclusive practices. I therefore do not think that my research ought to have harmful real-world ramifications.

Is there an ontological distinction between relieving and ending suffering, and if so, is it ethically significant?

Materstvedt’s3 main argument is that euthanasia ‘ends’ rather than ‘relieves’ suffering, and is therefore not a ‘treatment’ per se, and that by virtue of this ‘category mistake’ the arguments presented in my essay comparing euthanasia with palliative sedation are a non-starter. He correctly states that euthanasia causes death, and that following death the patient is unable to experience any further suffering, whereas in deep palliative sedation the patient is sedated to the level of unconsciousness so that s/he cannot experience suffering (and in the case of CDPS, s/he will never again experience suffering). Owing to this, Materstvedt then states that palliative sedation ‘relieves’ and euthanasia ‘ends’ suffering. This is an interesting proposition although it does not seem immediately obvious to me, and requires an argument to substantiate it which I do not think has been provided. It could equally be held that this is just splitting hairs and that the linguistic difference here does not represent a true ontological difference. The two words are almost synonymous in this context and both describe a state of affairs in which one bad thing (eg, suffering) was initially present and is now absent following some intervention. If a relatively healthy person experienced pain following an elective hip replacement and that pain subsequently went away (ceased to exist) with opioids, then it would not strictly be incorrect to say that pain was ‘ended’ although it would admittedly be an odd choice of words. Similarly, if a dying patient had pain that subsequently went away (ceased to exist) after euthanasia was performed, then it does not seem incorrect (to me at least) to say that the pain was ‘relieved’. I do not have an argument to support my competing intuition, and I may be wrong and do not feel strongly about the matter either way; however, I doubt that this linguistic difference is a real substantive issue that lends itself to meaningful debate. Furthermore, as argued in my essay, we should be agnostic regarding whether palliative sedation shortens survival time (and thereby hastens death); if it does, then it at least in part ‘ends’ suffering in the same way that euthanasia does on Materstvedt’s account. Materstvedt’s description of CDPS resulting in a state of being ‘socially dead’ as opposed to ‘biologically dead’ is also interesting, and may therefore also entail ‘ending’ rather than ‘relieving’ suffering on his account. It is also worth noting that if the ‘higher centres’ definition of death rather than the more traditional ‘whole brain’ definition of death is true,11 then CDPS would actually constitute death and would therefore ‘end’ rather than ‘relieve’ suffering on Materstvedt’s account.

Even if it were the case that there is an ontological difference between ‘relieving’ and ‘ending’ suffering, it is even more doubtful whether this distinction is ethically significant. It is unlikely that this small difference should in itself render palliative sedation ethical and euthanasia unethical. The qualitative research that Materstvedt describes regarding the personal guilt experienced by some medical practitioners providing euthanasia in the Netherlands and USA is interesting, but is anecdotal and not overly relevant to the ontological and ethical questions of whether euthanasia ‘relieves’ or ‘ends’ suffering.

Furthermore, even if this difference exists and is ethically significant, it does not follow that comparing palliative sedation and euthanasia is not feasible or that the arguments offered in my essay are a non-starter. I recognise that CDPS and euthanasia are two separate practices, with different intentions as Materstvedt correctly identifies (Symons4 also correctly highlights this); however, I think that the ‘procedure’ of administering sedative drugs is similar, and the ‘outcome’ of relieving (or ending) suffering is similar. The purpose of my original essay was to compare euthanasia with palliative opioid and sedative use and to suggest that their differences may not be ethically significant, rather than to argue that the two practices are ontologically identical.

Should the Doctrine of Double Effect (DDE) be interpreted from a Thomistic rather than from a Kantian/Millian perspective, and is it sound when interpreted Thomistically?

Both Symons4 and Giebel5 correctly acknowledge that Thomas Aquinas lived centuries prior to both Immanuel Kant and John Stuart Mill, and that my use of a Kantian and Millian perspective to interpret Aquinas’ DDE is anachronistic. They both then argue that a Thomistic interpretation of the DDE should be preferred and that on this interpretation the DDE is sound. I think that this first claim is questionable and that the second is incorrect. Firstly, it is not necessarily that case that a contemporaneous perspective is superior to an anachronistic one, and each should be appraised according to their own merits. We are fortunate to have Kant and Mill to borrow from and do have the philosophical liberty to use them to interpret the DDE, in hope of achieving fruitful insights. The intrinsic and proportionality criteria do lend themselves to deontological and consequentialist interpretations, respectively. By the same token, I agree with Symons and Giebel that a contemporaneous Thomistic interpretation is more historically appropriate and is valuable in this respect. It is unclear to me whether one can be deemed superior to the other, and I think that they can be considered in parallel.

Symons and Giebel both subsequently argue that the DDE is fundamentally different on a Thomistic rather than a Kantian/Millian account, and that on the former account the DDE is sound. Conversely, I think that the two accounts are actually very similar, and that the Thomistic account is unsound for one of the same reasons that the Kantian/Millian account is. Symons describes the first component of the DDE as pertaining to the ‘right-making’ or ‘wrong-making’ intrinsic features of an action; the ‘right-making’ features promote some basic good, and the ‘wrong-making’ features constitute a direct attack on some basic good. This seems to be defining intrinsically right or wrong actions on the basis of their outcomes, which to me sounds consequentialist and circular with the proportionality criterion. Contrastingly, on Giebel’s account, the intrinsic criterion refers to any ethically relevant factor other than the outcome of the action, providing the example of a virtue ethics interpretation to this effect. I believe that virtue ethics accounts (eg, saying that the action is intrinsically right because it is just or brave, rather than unjust or cowardly) are circular and not do any explanatory work, and do overlap with deontological accounts when deconstructed; however, this lies beyond the scope of this response.

Symons then describes the circumstances of an action—who performed the act, when and where it was performed, the intention of the act, the means used to perform the act, and the manner in which the act was performed—as the next component of the DDE. These seem to closely reflect the intention and means criteria as described in my essay. Giebel lists the second and third criteria of the DDE in essentially the same way, although challenges my use of the term ‘coincidental’. I am using the term ‘coincidental’ literally, to mean that there is a ‘co-incident’ or second incident occurring in conjunction with the first incident, but this second incident is not necessarily unexpected. Some coincidents are unexpected (eg, realising that my estranged brother and I happen to be in the same movie cinema at the same time), whereas others are expected (eg, washing dishes and one’s fingers simultaneously becoming pruney, as Giebel describes). The coincident described in the means criterion does not have to be unexpected; conversely, it is usually expected, and this is what makes ethical deliberation ahead of time so difficult in the first place. Interestingly, I note that Aquinas’ original description12 does not explicitly reference the means criterion; however, it can be implied, and in most subsequent iterations of the DDE, including in Catholic theology, the means criterion is explicitly emphasised.13

Symons then describes the final component of the DDE as saying that the bad outcomes of an action may outweigh an otherwise right or ethically-neutral action. He provides the example of hiking in the mountains as being an intrinsically ethically-neutral action, but it becoming unethical if one neglects one’s children in doing so. This ‘outweighing on the basis of outcomes’ sounds essentially the same as the consequentialist interpretation of the DDE provided in my essay. Giebel’s formulation is that ‘the agent must have a proportionately serious moral reason for performing the act’. I think this is vague, being ambiguous as to what constitutes a ‘proportionately serious moral reason’; however, if this is determined by the outcomes of the action, then it is also essentially the same as the consequentialist interpretation provided. Giebel argues that this is not just as simple as one outcome ‘trumping’ another, although it seems to me that this is indeed precisely what is occurring. Giebel’s example concerning the marginal utility of feeding one’s own children or donating that money to charity is interesting, and pertains to Singer’s thesis in ‘Famine, Affluence and Morality’.14 Whether one chooses to allocate limited resources to immediate or distant others is an ethical question lying beyond the scope of this response, but it is fundamentally a question comparing the outcomes of two actions, so I do not think this is removed from a consequentialist (or Millian) reading at all, contrary to Giebel’s claim.

Separate to this discussion of whether there really is a significant difference between the Thomistic and Kantian/Millian interpretations of the DDE, I believe that the DDE is unsound even on a Thomistic interpretation owing to the Loop Variant thought experiment devised by Thomson15 and described in my essay. On Symons’ and Giebel’s accounts of the Thomistic interpretation, I believe that both the intention and means criteria are both still violated, and therefore that pulling the switch is deemed unethical, which is proof via contradiction that the DDE is unsound. Giebel does acknowledge that pulling the switch would be forbidden on the Thomistic account. Even if the Thomistic interpretation was devised without the means criterion being included (noting that it is not explicitly stated by Aquinas himself), the intention criterion would still be violated as I argue in my essay. So even on the Thomistic interpretation, the DDE is a useful ethical tool and is a good example of a pluralistic ethical system, but it is ultimately unsound by virtue of setting the bar too high and erroneously deeming some ethical actions unethical.

Do the appropriately titrated administration of opioids, and palliative sedation, meet the Doctrine of Double Effect’s four criteria when interpreted from a Thomistic perspective?

Finally, Giebel5 argues that palliative opioid and sedative use do meet the DDE’s criteria when correctly interpreted in a Thomistic way, and that they can therefore be ethically distinguished from euthanasia by the DDE. I will now address these points. Firstly, opioid and sedative use may represent pro tanto (‘to an extent’) bads, but their routine and unproblematic use in general medical care does not have the bad effect of hastening death, so they are not relevant to (and therefore not prohibited by) the DDE, as Giebel challenges. It therefore does not follow that pro tanto bads cannot violate the DDE when they are actually associated with a bad effect, such as may (or may not) be the case with palliative opioid and sedative use. As discussed in my original essay, the case for whether palliative opioid and sedative use are pro tanto bads, and whether pro tanto bads violate the intrinsic criterion, is not completely clear; however, we do have more compelling reasons to believe that CDPS is a tout court (‘without qualification’) bad, which would clearly violate the intrinsic criterion.

Secondly, Giebel’s discussion of ‘indirect’ and ‘direct’ means of palliative opioid and sedative use achieving the relief of suffering conflates the means criterion (relevant here) with the intention criterion (irrelevant here). Both palliative opioid and sedative use can relieve suffering directly (by means of the analgesia and level of sedation achieved), and perhaps indirectly (by hastening death, and death itself resulting in the ‘relief’ or ‘ending’ of suffering, noting Materstvedt’s3 critique above). I have argued that this results in the good effect (the relief of suffering) being ‘overdetermined’; both the direct and indirect means are individually sufficient but not individually necessary to relieve suffering, and it is unclear whether the means criterion is satisfied or violated in this instance. The context of refractory psychological (rather than physical) suffering is more complex, in which case palliative opioid and sedative use (save for CDPS) may not be able to relieve that suffering directly, but only indirectly via death. In this case, the direct means is neither necessary nor sufficient, and the indirect means is both necessary and sufficient, so overdetermination does not apply here and the means criterion is violated. CDPS could also achieve the required relief of suffering directly in this instance, but as discussed earlier it is similar to death (or actually a form of death on some accounts), so the means criterion is arguably still violated in this case. Of course, this does then have a bearing on the intention criterion. I agree with Giebel that a coincidental bad outcome does not have to be intended; however, an instrumental bad outcome (ie, a bad outcome functioning as a means to achieving the desired good outcome) does need to be intended, as was argued with reference to the discourse between Foot16 and Thomson15 17 and the Fat Man thought experiment. I therefore disagree with Giebel that ‘in palliative care not every causal contributor need be intended as an agent’s means’, apart from in the case of overdetermination which is unclear.

Thirdly, Giebel claims that the Doctrine of the Sanctity of Human Life (DSHL) states that human life must never be intentionally violated, although in some circumstances it may be permissible for human life to be unintentionally shortened. She therefore argues that the DSHL does not contradict the proportionality criterion of the DDE. Although not explicitly using the phrase ‘Doctrine of the Sanctity of Human Life’, the Evangelium Vitae of Pope John Paul II details this principle at length,18 describing at various points the ‘absolute’, ‘sacred’ and ‘incomparable and inviolable worth of every human life’, being present ‘at every moment of existence’ and ‘at every stage and in every situation’. It also refers to the ethical significance of intentionally ending human life, at least in part consistent with Giebel’s claim. It is contentious how far this principle extends, that is, whether it requires thoroughgoing pacifism or whether it permits some instances of killing such as in wars or self-defence.19 If Giebel is correct and the DSHL only applies to the intentional shortening of life, then the DSHL is only relevant to the DDE in the setting of palliative opioid and sedative use if the intention criterion is violated, and this is arguable as just discussed. On the other hand, if the DSHL at least sometimes also applies to the unintentional shortening of life, then it is clearly relevant to the DDE. In the setting of palliative opioid and sedative use there is a choice between not giving opioids/sedatives, or giving opioids/sedatives and (possibly) shortening life. If the latter option is chosen, when the former could have been chosen instead, then it is difficult to reasonably maintain that individual human life was ‘inviolable’. Either relieving suffering is enough to warrant (‘is proportionate to’) the shortening of life, in which case the proportionality criterion is satisfied but the DSHL is contradicted, or relieving suffering is not enough to warrant the shortening of life, in which case the proportionality criterion is violated but the DSHL can be upheld.