2. Ethical tensions

In its narrower guise, precision medicine aims to provide treatment that will be most effective and bear the fewest undesirable side effects, while hopefully also improving economic outcomes for individual patients (Chen et al., Reference Chen, Guzauskas and Gu2016). Broadening the use of such a ‘revolutionary’ approach, precision health platforms tend to emphasize the importance of healthy lifestyles, disease prevention and risk awareness (Akdis & Ballas, Reference Akdis and Ballas2016). Yet, here it may seem that the precision model is not an entirely newfound approach. In their recent work on precision in critical care, Shihab Sugeir and Stephen Naylor note that Hippocrates, well over two millennia ago, ‘believed that disease was a product of environmental forces, diet and lifestyle habits, and that treatment should focus on patient care (prevention) and prognosis (prediction)’ (Sugeir & Naylor, Reference Sugeir and Naylor2018). As Sugeir and Naylor submit, personalized patient care has eroded into a protocol-based practice. What we may be seeing then, with the rise of precision models, is a return to a very classic form of medicine, albeit with the use of up-to-date tools.

Granted, the precision approach will make use of previously unheard-of social and behavioural information, and it stands to usher in a reclassification of diseases based upon one's molecular profile (Mirnezami et al., Reference Mirnezami, Nicholson and Darzi2012). Determining how to properly harness the trove of highly sensitive information – how to ensure privacy and security of the data – is naturally among the key hurdles (Mega et al., Reference Mega, Sabatine and Antman2014; Jameson & Longo, Reference Jameson and Longo2015). Many commentators simply acknowledge the most apparent concern, namely that we will need to balance public safety with maximized productivity (e.g., Ashley, Reference Ashley2015). Others have suggested that the public health benefits are unclear, that gene-based drugs will cost more due to targeting smaller populations and accordingly that we should temper the narrative of ‘transformative change’ (Joyner & Paneth, Reference Joyner and Paneth2015). Nonetheless, as things stand, the US National Institutes of Health (2019a) All of Us research program and its awardees have been encouraging patients to contribute medical records and biological samples in an effort to amass a database of at least one million individuals.

As of August 2019, the number of participants in the National Institutes of Health research programme has reached approximately 245,000 (National Institutes of Health, 2019b). By contrast, the now well-known private direct-to-consumer genetic testing company 23andMe reported the genotyping of their one millionth customer in June 2015, months before the US Precision Medicine Initiative was officially launched (Wojcicki, Reference Wojcicki2015; Stoeklé et al., Reference Stoeklé, Mamzer-Bruneel, Vogt and Hervé2016). According to one report, 23andMe is ‘amassing one of the largest genetic bio/databanks in the world’ as its ‘back-end business model’ (Spector-Bagdady, Reference Spector-Bagdady2016). Unsurprisingly, the company has faced setbacks but also approvals on certain genetic testing from the US Food and Drug Administration (Check Hayden, Reference Check Hayden2017). And although private ownership of sensitive personal data might be less alarming under the pretext of governmental regulation, concerns are being raised over vulnerabilities from foreign access and commercial interests (e.g., Berger & Schneck, Reference Berger and Schneck2019).

Along with the concerns over data privacy, security, ownership and use, questions of how to justify non-therapeutic research on human subjects appear equally challenging. Here, we see one of the most fundamental tensions in research ethics: that is, determining the extent to which using a selection of individuals for the purpose of benefiting others can be morally permissible. We might suppose there simply is no problem where researchers have obtained the most robust sort of informed consent available. However, as numerous authors have made clear, the traditional standards of obtaining informed consent do not map onto secondary, future uses of subjects’ data. Indeed, consent procedures are often highly questionable – for example, due to obscurities or omissions in terms and conditions documents (Stoeklé et al., Reference Stoeklé, Mamzer-Bruneel, Vogt and Hervé2016). For these reasons, newly proposed models of informed consent attempt to articulate a continuous, open-ended agreement to studies for which subjects’ data were not intended and which were likely unknown at the time of initial data-gathering (e.g., Hansson, Reference Hansson, Dillner, Bartram, Carlson and Helgesson2006; Steinsbekk, Reference Steinsbekk, Kare Myskja and Solberg2013; Ploug & Holm, Reference Ploug and Holm2016). Undoubtedly, precision medicine, with its reliance upon future uses of enormous quantities of data, aptly raises such concerns for informed consent. But even if we determine an appropriate recourse for traditional consent models, a procedurally prior enquiry remains. How, if at all, can we conscript subjects in the first place? To what extent should we incentivize the voluntary donations of sensitive data?

3. Historical precedents?

In a recent contribution to this journal, McGonigle notes that the success of precision medicine depends upon the voluntary participation of healthy individuals, which, in turn, calls for changing the mindset of patients, healthcare providers and the wider public. Specifically, we will ‘need to embrace the “idea” that genetic information is an important part of medical treatment’ (McGonigle, Reference McGonigle2016). In order to promote this idea, we must foster collaborations – for example, between scientific and clinical communities – and increase public engagement. Yet, as McGonigle readily points out, the ethical conundrums are substantial, including questions of genetic data ownership, risks of sharing family data and uncertainties over future withdrawals of one's data.

As a way to ‘tackle’ the ethical concerns, McGonigle suggests making comparisons with historical precedents. The first analogy on offer is a system of voluntary blood donation in Israel, wherein donors ‘receive a government identity card assigning them priority to receive future emergency blood donations’ (McGonigle, Reference McGonigle2016). Similarly, on Israel's organ donation plan, those who sign the ‘Adi’ card agree to ‘donate their organs after death’ to help save patients awaiting transplants, and in return, donors and their relatives are granted priority for transplant wait-lists (McGonigle, Reference McGonigle2016). Undoubtedly, there is an intuitive appeal to such schemas, reinforced with reflection upon desert-based justice and the allocation of scare resources to the contributors and their family members. According to some research, this type of incentivized biobanking is rather effective in promoting an increase in voluntary contributions (Stoler et al., Reference Stoler, Kessler, Ashkenazi, Roth and Lavee2016).

However, as a precedent for operationalizing precision medicine, the blood and organ donation models understate the extensive change in mindset required. In short, incentivized blood and organ donation systems, such as those seen in Israel, are unlike the data donations needed for precision medicine. At least three important differences can be promptly clarified. First, human blood and vital organs are properly considered scarce and non-renewable resources. Patients’ medical records and genetic information, on the other hand, may be presently unavailable – particularly for legitimate use in large-scale shared databases – but they are surely not scarce in the same sense or for the same reasons. Unlike blood and organs, personal data are very easily duplicated, transmitted and shared with anyone, anywhere. Once data are consumed, they do not diminish in value (Hand, Reference Hand2018). For this reason, data can be used again and again, well into the future, bringing a much greater risk of harm to the ‘data donor’. Second, then, it seems that the further into the future one's data are used, the less likely one knows exactly how they are being used, for what purpose and so on. In this way, the risks of data donation – regardless of novel forms of consent that may be obtained – are simply incommensurable to the risks associated with the donation of blood or organs.

Third, and perhaps most glaringly, are the incongruous incentives attached to biobanking models that offer distinct and tangible rewards. As McGonigle showed, on Israel's systems, donors clearly benefit by being prioritized for potential reception of blood or vital organs. Donors benefit their families by allowing relatives to be prioritized. These sorts of immediate rewards simply cannot be translated to incentivizing data donation, nor should they be. It would be far less clear how to be prioritized as a beneficiary of precision medicine. McGonigle notes that data donors can benefit in terms of ‘access to personal health assessment, and … by helping the wider community become healthier’ (McGonigle, Reference McGonigle2016). However, healthcare is not a commodity with which we should reward those who submit sensitive personal data, and those inclined to support public health surely have clearer opportunities – for example, through immunization efforts or drug and nutrition awareness (Joyner & Paneth, Reference Joyner and Paneth2015). Further, even if such incentive schemas are translated to precision medicine, incentivizing data donation appears morally wrong in ways that Israel's biobanking systems are not. Personal data and genetic information are, of course, highly sensitive. In the wrong hands, one's data can disrupt insurance premiums, employment and much more (Stiles & Appelbaum, Reference Stiles and Appelbaum2019). With the donation of blood or donating organs after death, one no longer needs these resources. By contrast, unless we agree to donate data only upon death, we remain in need of these delicate resources even after donation.

It might be thought that incentivizing the contribution of health records and genetic data is morally permissible where subjects can identify with the goals of research or, at least, are not directly coaxed into contributing by the researchers themselves (e.g., Jonas, Reference Jonas1969; Macklin, Reference Macklin1981). In a recent report on the Genes for Good (2019) programme, for example, it is emphasized that altruism is among the primary incentives and that recruitment occurred ‘organically, with participants publicizing the study through their own networks’ – namely, Facebook (Brieger et al., Reference Brieger, Zajac and Pandit2019). Of course, questions should still be raised concerning the protection of human subjects and the use of social media as a tool for research and provision of healthcare (e.g., Pedersen & Kurz, Reference Pedersen and Kurz2016; Omaggio et al., Reference Omaggio, Baker and Conway2018). It is with good reason that we see recent institutional efforts to implement professional guidelines on the use of social media, such as the American Medical Association's policy (Kind, Reference Kind2015). Still, the technical, legal and ethical hurdles of big data in healthcare are far from resolved. Given the extraordinary technological advancements and the extent of the risk involved, we must take care not to mislead potential data donors into thinking the challenges of precision medicine are anything like what we have seen before.