The ‘do it yourself’ type 1 diabetes dilemma for medical practitioners

As highlighted in the recent article by Hng and Burren,1 motivated individuals with type 1 diabetes (and caregivers) have developed their own software algorithms, which when coupled with a continuous glucose monitor and insulin pump, allow the patient to ‘close the loop’ or automate insulin delivery. This development was driven by the latency in availability of commercial automated insulin delivery systems and stimulated further by their cost and perceived deficiencies (in particular individualisation) of commercial systems. There are three open source algorithms that are most commonly used; OpenAPS, Loop and AndroidAPS. All these algorithms are freely available online, as well as online tutorials on how to integrate the algorithm with existing technology and ‘build’ an automated insulin delivery system. This ‘do it yourself’ (DIY) approach is viewed with great caution and scepticism by healthcare professionals. However, with increasing use in the community, and in a bid to be more informed, leading diabetes conferences (American Diabetes Association Orlando 2018, Advanced Treatment and Therapeutics for Diabetes Berlin 2019) have recently included scientific sessions on this open source technology. When a patient presents to their healthcare professional having adopted the use of a DIY system, a dilemma ensues, which revolves around three primary issues: (i) What safety and efficacy data exist? (ii) What legal implications are there when providing care for a patient who uses an unregulated product? and (iii) What is the role of the healthcare professional with respect to adjusting/prescribing or education where there is none of the usual training or support infrastructure available? The purpose of this editorial is to update clinicians on these critical aspects. There is a growing body of published data available on the open source algorithms. Real-world data from an estimated global population of more than 1000 DIY users, and an estimated seven million hours of real-world experience with this technology,2 demonstrated remarkable glycaemic outcomes.3-6 The limitation to these data is that these individuals are highly motivated and are likely to do well; nevertheless, when compared with their own glycaemic data prior to using the DIY system, time in normal glycaemic range (4–10 mmol/L) improved by 16.4%.5 Further it could be argued that ‘real lived’ data are more robust than that gathered in the artificial environment of a carefully monitored clinical trial in pre-defined selected populations.7 Recently, the AndroidAPS algorithm (which shares the same core mathematical code as OpenAPS) has been shown to deliver excellent glycaemic control during a ski camp,8 and the same group have recently presented in-silico data demonstrating impressive results including simulations where no meal boluses are given.9 Nevertheless, large scale safety and efficacy trials are required to meet the highest level of evidence accepted by regulatory authorities (as well as to sway the intelligent caution most healthcare professionals possess). As all the open source algorithms have no regulatory approval, there are potential medico-legal implications for healthcare professionals consulting with patients who are using them. Diabetes Australia have recently released a position statement that identifies this issue and surmises that the therapeutic relationship between patient and clinician should take precedent, but also that clinicians cannot promote or educate patients on how to use such systems.10 A formal legal opinion from Maddock law firm (Sydney) has been commissioned by Australian Diabetes Educators Association, for which the final report is imminent. As yet, there have been no reported severe adverse outcomes for patients using an open source algorithm, and hence the legalities are yet to be officially challenged. Therefore, the medico-legal implications for healthcare professionals who see patients using a DIY system have not been resolved. The need for a position statement and legal opinion is in itself thought provoking, as the use of off label medications in diabetes has been common practice for many years. For example, glargine was prescribed off label for infants with type 1 diabetes commonly. In this case, the usual paradigm of diffusion of treatment was shown – initially, there were data in older individuals from large trials, then coupled with the relative comfort of prescribing a known entity (and associated safety profile) for some time in older patients, there was a natural progression to extend the use into populations where there was no approval – in this case, young infants. However, technology, especially that developed by consumers, is a different prospect altogether, with no safety data, no clinician experience and no training. As such, patients using a DIY system, embody the ultimate ‘patient-led care’ model, completely reversing the usual clinician-tutor/patient-pupil dynamic. Therapeutic innovations usually follow one of two distinct pathways: researcher-driven breakthroughs discovered through careful scientific interrogation or industry-led innovation (e.g. pharmaceutical companies) that commissions clinical trials to prove safety and efficacy. For any new treatments, if shown to be beneficial and safe, translation into routine clinical care can be slow due to the time taken to convince healthcare professionals that the new therapy is superior over their current strategy, and for the professional to feel proficient in the new model.11 However, the DIY closed loop initiative shows a third possible pathway – innovation driven by the patients or caregivers of patients themselves. Patient-driven innovation bypasses the usual scientific rigour, or regulatory pathway, and therefore leaves the healthcare professional in a challenging juxtaposition; between maintaining the therapeutic relationship with their patient and being an accomplice to a completely unregulated and ‘unproven’ (at least by large scale randomised controlled trials) therapy. The open source technology community patients and caregivers) is the source for education and support for new and existing users. This is a complete contrast to the usual situation, where patients seek help and advice from their healthcare professional. Currently, healthcare professionals have very little if any knowledge on the algorithms, or how to individualise settings, necessitating new users to turn to the open source community. This education model is interesting in its point of difference from routine care, where a patient is dependent largely on the expertise of their healthcare professional (n = 1). The patient has limited ability to assess this expertise, and seeking a second opinion is time consuming and can delay treatment. With a community able to offer expertise (n = many), users can rapidly seek advice and trouble shoot from experts all over the world, not even limited by time zones. Again, there are potential pitfalls, as there is no ‘credentialing’ of those giving advice and could potentially result in poor advice being given adding to the concerns facing healthcare professionals. However, the ‘open’ nature of such advice allows for monitoring from all experts within the community, thus limiting the potential for poor advice. In fact, the ‘open’ nature of such advice is far more transparent than the expertise offered behind a closed clinic room door. The future of DIY systems is dependent on regulatory approval, which in turn requires safety and efficacy data from large clinical trials. However, as the algorithms are open source, there has been no financial incentive from industry to undertake these trials, as conducted by Medtronic for their current commercially available system (Medtronic 670G),12 and those currently being undertaken by commercial competitors. However, in recognition of the significant demand for data and a pathway to possible regulatory approval, the Juvenile Diabetes Research Foundation in partnership with the Helmsley Charitable Trust, have announced funding (US$6m) to study the Loop algorithm and advance it through the regulatory requirements.13 However, there has been partnership with TidePool, a commercial entity, so it is yet to be known how this will affect the ‘open source’ concept with respect to the Loop algorithm. Investigator-led trials will likely follow for the OpenAPS and AndroidAPS algorithms. In summary, the DIY open source automated insulin delivery innovation has resulted in a unique challenge to the status quo of diabetes care. Translation into practice has come before clinical data, consumers have become experts before the healthcare professionals, and healthcare professionals have been forced into a medico-legal quandary at no fault of their own. This is epitome of patient centred, individualised care. However, it is anticipated that regulatory approval backed by safety and efficacy data is needed before considerably more patients adopt this innovative treatment. Until such time clinicians are likely to remain cautious both clinically and medico-legally.