17 Aug, 14 | by Iain Brassington
Guest Post by Ruth Stirton and Lindsay Stirton, University of Sheffield
One of us – Ruth – was on Newsnight on Wednesday the 13th August talking about the PARAMEDIC2 trial. The trial is a double blind, individually randomised, placebo controlled trial of adrenaline v. normal saline injections in cardiac arrest patients treated outside hospital. In simpler terms, if a person were to have a cardiac arrest and was treated by paramedics, they would usually get an injection of adrenaline prior to shocks to start the heart. If that same person was enrolled in this study they would still receive an injection but neither the person nor the paramedic giving the injection would know whether it was adrenaline or normal saline. The research team is proposing to consent only the survivors for the collection of additional information after recovery from the cardiac arrest. This study is responding to evidence coming from other jurisdictions that indicates that there might be some significant long term damage caused by adrenaline – specifically that adrenaline saves the heart at the expense of the brain. It is seeking to challenge the accepted practice of giving adrenaline to cardiac arrest patients.
Our starting position is that we do not disagree with the research team. These sorts of questions need to be asked and investigated. The development of healthcare depends on building an evidence base for accepted interventions, and where that evidence base is not forthcoming from the research, the treatment protocols need changing. This going to be tricky in the context of emergency healthcare, but that must not be a barrier to research.
There are two major ethical concerns that could bring this project to a grinding halt. One is the opt-out consent arrangements, and the other is the choice of methodology.
Consent, then. There are two aspects of a consent process: information provision and decision-making. It’s worth saying that it’s absolutely standard in emergency research not to seek consent from potential participants if doing so will delay the start of treatment. It would be abhorrent to delay treatment where time was of the essence in order for a participant to read a 30 page information sheet and sign a consent form. It’s equally problematic to try to seek consent from a relative, or legal representative, as required by Schedule 1, part 5 of the Medicines for Human Use (Clinical Trials) Regulations 2004, since they would have to read the same form. Delaying the onset of treatment for any reason and reducing the chances of survival is unethical. So, explicit opt-in consent is not possible in this sort of emergency research; and this leaves us with opt-out consent. Now, opt-out consent is ethically controversial: it should be a last resort, and if it is resorted to then the details have to be absolutely nailed down. In addition, it has to be a real opt out system.
According to Jerry Nolan (Newsnight, 13/8/14) the process is still being worked out. But the details we have are that this will be done by
including information about the trial on Ambulance Trust and University Websites, Ambulance Service Public Newsletters, Posters / information leaflets, discussion at public meetings, Annual reports etc.
and people will opt out in some way. Now, we don’t know about you, but neither of us sit there every evening and think, “Ooh, I might have a cardiac arrest this week, I’ll just check the ambulance service website to see if there’s some research happening that I might need to opt out of.” If people are not going to think proactively enough to seek the information then it is a fiction to say that this is an opt out consent process. In order for opt out consent to be real, potential participants must be properly informed and there must be a real opportunity to opt out. At the very least this requires active information provision. This trial needs to be in the public consciousness in the area in which it is happening. Every household needs to be leafleted or mailshot; there needs to be posters in all GP surgeries, pharmacies, community centres, churches, local shopping centres… in fact, anywhere where people congregate; there should be advertisements on the local radio, and discussions in local TV news and in the local print media. What we are saying is that there needs to be an information storm, and depending on the ethnic make-up of the area, this information needs to be available in a variety of languages. People in this part of the country should not be able to go anywhere without seeing some information about the trial. Part of that information should be about how to make a decision. If the information provision is right and is exhaustive in this way then it is legitimate to say that anyone who has not opted out has consented to participate. Anything less than this, and it’s a complete fiction to say that there has been consent.
Let’s turn to information provision. In the NIHR summary the researchers consider the arguments about information provision and conclude that “the burden of actively informing [the relatives of non-survivors] will outweigh the potential benefit”and so they will rely on the passive information provision arrangements. Now, we think that anything less than an information storm is inadequate to ensure that potential participants are informed. The arrangements in place, insofar as they are known, are significantly less than an information storm. Most people will not know that the trial is happening. The survival rate from a cardiac arrest is less than 1 in 10. Most of the participants will not survive. Therefore, most of the relatives of those who do not survive will never know that their relative was in a clinical trial. This is a problem – a significant problem. When it comes down to it, people understand the need to do research into these questions. But if we learned nothing else from Alder Hey, and the Bristol Heart Surgery investigations, it was that people want to know what is happening. They wanted to know that bits of their children were being retained for research purposes. They were overwhelmingly supportive of the idea of doing research into these problems. But they insisted on being informed.
Things have moved on since then; but information is still considered essential. The Care.Data project has suffered significantly because of failings in the information provision arrangements. There is an increasing public awareness of just how much can be done to us without our knowledge and consequently where the public could have been told there is outcry when they are not. There is no doubt at all that the cardiac arrest of a relative is a very difficult thing to experience. But trust in the health service is essential, and there is nothing that breaks trust more than learning after the fact that someone could have informed you about something, but they chose not to. Why not give ambulances a stack of short (one page) summary information sheets with contacts for further details to hand out to the relatives of any participant in the trial? The sheet should include a very brief explanation of the purpose of the study, what was done to the participants, who reviewed the study and contact details, which could include a website with a fuller information sheet, for further information. Yes, it will be painful and difficult, and it will cause distress. But it would be nothing like as much distress as discovering weeks after that a relative was experimented on without anyone’s knowledge.
Finally: methodology. There are good reasons why double blind, placebo controlled, randomised controlled trials (RCTs) have become the gold standard methodology for medical research. RCTs randomise out extraneous factors related to individual patient’s likely outcome. They even factor out potential placebo effects, and the unconscious biases of medical personnel. These are considerable benefits, and if we are going to undertake research involving human subjects, there is an ethical obligation to undertake that research in a way that is methodologically sound.
But in evaluating the ethics of a RCT into adrenaline, to compare it with an unsound research design, or with a decision not to undertake the research at all, is to make the wrong ethical comparison. Given the obvious difficulties, the case for a RCT needs to be justified in terms of the incremental benefits of double-blind randomisation over the next-best research design. When Ruth appeared on Newsnight to discuss this trial, she suggested doing a mass observation study in the UK to get some reliable evidence that adrenaline is harmful to cardiac arrest patients. Such a study might proceed by informing those present with cardiac arrest patients of the known risks of adrenaline and asking whether they wanted adrenaline to be given. Or alternatively, given that there is concern about the harm that might arise from adrenaline, there must be a body of paramedics who are not supportive of giving adrenaline, as well as a body of paramedics who are. It would be possible to create the comparator groups on this basis, rather than rather than using randomisation.
How much would be lost by an approach such as this, or by whatever other non-randomised research design the researchers thought would best capture the potential neurological effects of adrenaline? Certainly something is lost: the research might be biased because (for example) adrenaline was opted for in the most serious cases, or because medical personnel (again, perhaps subconsciously) acted differently towards patients in other ways, or because patients who received adrenaline believed they had better (or worse) treatment and this affected their recovery.
Then again, this is the sort of problem that social scientists (who rarely have the luxury of the level of control over the research context to conduct RCTs) face all the time. And as a result they have developed techniques to deal with just these sort of problems. James Heckman won a Nobel Prize in Economics for “his development of theory and methods for analyzing selective samples”. So-called “Heckman models” have become an accepted way of handling the problem of selection bias in economics and throughout the social sciences. Propensity score matching is another common technique that social scientists use to control for the effect of factors which might confound an estimate of a treatment effect.
Are these methods as effective as a properly conducted RCT? Probably not, but perhaps they come close. Properly used, we could think of them as a “silver standard”, providing pretty reliable means for the analysis of non-randomised data. Ultimately, there is a balancing act to be done. The reliability of the research must be balanced against the ethical concerns associated with collecting the data in a particular way. In the case of adrenaline for the treatment of cardiac arrest, the silver standard – a mass observation study, properly analysed – seems to be more than adequate. Unless it can be shown that the incremental (and not the total) benefits of using the gold standard – the RCT – are substantial, then the significant ethical issues highlighted above point towards the silver standard as being the more appropriate methodology.