Why Informed Consent Is Impossible
##Why informed consent regarding therapeutic decisions is impossible (or at least very difficult)
Background
Just about everyone agrees that a patient being prescribed a therapy should be given enough information regarding the likely benefits and harms of the therapy such that the patient can make an informed decision about whether they wish to take it.
While much of this debate has occurred in relation to informed consent for clinical trials the ethics is readily applicable to day-to-day therapeutic decision-making by healthcare professionals and patients. If anything the problem is more acute in this setting because presumably we are supposed to have enough information in order to justify the use of the therapy for the patient.
{[green … whereas in the clinical trial setting we only need to claim equipoise, which can be claimed on the basis of ignorance. Jason ]}
The problem
The key problem has been the question of how to express to patient the risks and benefits of the therapy in an objective as possible manner, i.e. without substantial framing from the health professional, while still communicating the inherent uncertainty present within all available information.
{[pink Hm. Maybe. But I don’t think there is such a thing as saying anything without substantial framing. Maybe there’s a better statement of that phrase. Or maybe there isn’t, and the problem can only be expressed in a way that makes it impossible to solve it. Jason ]}
It seems that the bulk of the literature sees this as a problem of communicating complex information. {[orange Yes. What’s more, I think the literature takes it as obvious that that’s what the problem is. Jason ]} There is growing literature on how patients may be provided with information based on findings from RCTs such that something like informed consent can be achieved. These approaches typically involve putting the results of trials, in particular NNT and NNH, into a format that aids communication. One frequent approach is to display such numbers pictorially, e.g. for a RCT with NNT of 25 and NNH 50 you might display 100 faces and show four faces as happy (outcome prevented) and two faces as unhappy (adverse event experienced) and 94 faces a neutral. {[red Need examples, of course. Alison might know some particularly good ones if you don’t. Jason ]}
Redefining the problem
It seems that the problem could be rephrased as a problematic consequence of the metaphysical assumptions of classical statistics. Essentially, classical analysis of RCTs is unable to provide the right kind of information for achieving informed consent. I would argue that the kind of information required is probabilistic (in a different sense to what is provided by classical methods).
In short, I would argue that what decision-makers need is probabilistic information regarding therapeutic hypotheses i.e. rough conditional probabilities as well as ways to adjust these rough conditional probabilities for patients with differing clinical situations to those patients included in the trials. Classical statistics does not provide this information.
Classical statistics rejects probabilism. It does not permit the assignment of probability functions to hypotheses about the world. {[olive Not even for hypotheses about populations with the SAME characteristics as the sample studied in the trial, and doubly not for the patient whose consent is being sought. Jason ]} Within the classical statistical model a given hypothesis is either true or false. Probability measures are used in classical statistics in only a limited way. The most noteworthy example is p-values: Under the assumption that the null hypothesis is true, p-values provide the probability that the observed result, or a result more extreme, would occur should the experiment be repeated an infinite number of times. Classical statistics then provides an arbitrary cut-off for when to accept or reject the null hypothesis based on the p-value of the result. {[purple I always start with p-values too, but in this particular case it might be better to use confidence intervals throughout, because (a) this is an applied paper, so it should speak to current practice in the most advanced medical fields, which is conf. intervals; and (b) because there won’t be any of that tricky maths in this paper, presumably, so there’s no need to go for the extra simplicity of p-values. I can help with conf. interval examples and/or maths if you like. Jason ]}
The hypothesis is either accepted or rejected. There is no natural measure of uncertainty about the hypothesis. {[blue People THINK there is, if they’re using conf. intervals, but we can show that they’re wrong, using only uncontentious frequentist theory. Jason ]} If the hypothesis is accepted then the observed magnitude of the effect is also typically accepted (I need to look further into the classical problem of estimation ‘— but as far as I can tell this is an accurate statement {[pink I think so, yes. Jason ]} ). Current approaches to the problem of adequately informing the patient about the treatment look to ’probabilise’ this magnitude information (i.e. NNT and NNH) ’— all other uncertainty is ignored or framed in whatever way the health professional feels appropriate — this approach seems wrong-headed.
I would argue that it is this — the fact that classical statistics does not provide the right kind of information for informed consent — that makes risk communication within therapeutic decisions so difficult.
I don’t think that the problem has been discussed in the literature in this manner (I need to check). {[olive Hard to check. Can ask a couple of prominent Bayesians. Doesn’t matter too much even if it has. Jason ]} I think it is possible to argue (i) what people need to make decisions is probabilistic information (ii) what is currently provided is inadequate (iii) the key source of this inadequacy is the outputs of classical statistics (in particular the lack of probabilism) (iv) probabilistic statistical analysis as a output of RCTs would be a step forward in achieving the ethical requirement of informed consent in therapeutic decision making.