Lay conceptions of the ethical and scientific justifications for random allocation in clinical trials

Abstract

Randomised controlled trials (RCTs) play a central role in modern medical advance, and they require participants who understand and accept the procedures involved. Published evidence suggests that RCT participants often fail to understand that treatments are allocated at random and that clinicians are in equipoise about which treatment is best. We examine background assumptions that members of the public might draw upon if invited to take part in a RCT. Four studies (N=82; 67; 67; 128), in the UK, identified whether members of the public (i) accept that an individual clinician might be genuinely unsure which of two treatments was better; (ii) judge that when there is uncertainty it is acceptable to suggest deciding at random; (iii) recognise scientific benefits of random allocation to treatment conditions in a trial. Around half the participants were loathe to accept that a clinician could be completely uncertain, and this was no different whether the context was one of individual treatment or research. Most participants found it unacceptable to suggest allocating treatment at random, though there was weak evidence that a research context may reduce the unacceptability. Participants did not judge that more certain knowledge would be gained about which treatment was best when treatments were allocated at random rather than by patient/doctor choice: scientific benefits of randomisation were apparently not recognised. Judgements were no different in non-medical contexts. Results suggest a large mismatch between the assumptions underlying the trial design, and the assumptions that lay participants can bring to bear when they try to make sense of descriptive information about randomisation and equipoise. Previous attempts to improve understanding by improving the clarity or salience of trial information, or of making explicit the research context, while helpful, may need to be supplemented with accessible explanations for random allocation.

Introduction

Randomised controlled trials (RCTs) are widely considered to be the best method of advancing knowledge about the effectiveness of medical treatments, and their incidence has increased from 2116 trials listed in 1960 to 348,740 in 2002 (Cochrane Controlled Trials Register, 2002). These trials rely on millions of patients’ consent to participate. Consent or refusal to participate in such trials is adequately informed only if patients grasp two key features in addition to the details of what will happen to them in their particular trial: that participants will be allocated randomly to treatment arms, and that at the start of the trial there are no convincing grounds for supposing that any patient would be advantaged or disadvantaged if allocated to one treatment arm rather than another. This consensus state of uncertainty amongst clinicians has been labelled clinical equipoise (Freedman, 1987) or collective equipoise (Johnson, Lilford, & Brazier, 1991).

There is ample evidence in the published literature that trial participants often show signs of misunderstanding the basis of their treatment allocation, and of wrongly assuming that one treatment is already known to be better than the other or others. Many researchers find a relatively high incidence of trial participants’ failure to report that their treatment was allocated at random (e.g. Glogowska, Roulstone, Enderby, Peters, & Campbell 2001; Hietanen, Aro, Holli, & Absetz, 2000; van Stuijvenberg et al., 1998; review by Edwards et al., 1998). Several studies report failures to acknowledge equipoise (e.g. Ellis & Butow, 1998; Edwards et al., 1998).

Importantly though, being able to report accurately that treatments were allocated at random does not necessarily indicate coherent understanding. Some participants state that treatments were allocated at random, but nevertheless reveal other inconsistent beliefs by reporting that they were assigned treatment according to their particular symptoms (Appelbaum, Grisso, Frank, O’Donnell, & Kupfer, 1999; Appelbaum, Roth, Lidz, Benson, & Winslade, 1987; Featherstone & Donovan (1998), Featherstone & Donovan, 1998 (2002); Snowdon, Garcia, & Elbourne, 1997). Some participants construct their own incorrect explanation for random allocation, such as that it removes responsibility from individual doctors or parents, or that it provides a way of rationing a scarce or costly resource (Snowdon et al., 1997). Both these kinds of misconception—that doctors assigned treatment according to individuals’ symptoms, or that random allocation is used for non-scientific reasons—imply a lack of recognition of equipoise. The interviews in Appelbaum, Roth, Lidz, Benson & Winslade (1987), Appelbaum, Grisso, Frank, O’Donnell & Kupfer (1999) took place immediately after trial information was given. By contrast, interviews took place up to 2 years after consent in the case of Snowdon et al. (1997). Featherstone & Donovan (1998), Featherstone & Donovan, 1998 (2002) interviewed some participants after 6 months, after treatment was complete. Although the anger and upset which some of these participants revealed in their interviews do give cause for serious concern, we cannot infer from these particular studies that participants had an inadequate grasp of random allocation and equipoise at the time they consented to participate.

Failure to understand about random allocation and equipoise could occur if trial information is just too complex for the patient to comprehend, or if the patient is not given sufficient time or opportunity to take it in. More accessible written information, or follow-up telephone conversations with a research nurse, might help with problems of this kind (e.g. Davis, Holcombe, Berkel, Pramanik & Divers, 1998; Edwards et al., 1998).

Written trial information will normally be supplemented with oral information, but it is the leaflet that is scrutinised by research ethics committees to check that it offers the necessary information for informed consent or refusal to participate in the trial. In the UK, guidelines have recently been established for approval of trial information leaflets by Multicentre Research Ethics Committees. Leaflets are expected to make a statement about equipoise, and a statement that treatments will be allocated at random. The particular wording suggested is “Sometimes because we do not know which way of treating patients is best, we need to make comparisons. People will be put into groups and then compared. The groups are selected by a computer that has no information about the individual, i.e. by chance. Patients in each group then have a different treatment and these are compared.” There follows information about the chance of receiving each of the treatments (Committees, Ethics Committees (COREC), 2001).

Appelbaum, Roth, Lidz, Benson & Winslade (1987), Appelbaum, Grisso, Frank, O’Donnell & Kupfer (1999); Appelbaum & Grisso, 2001) argue, however, that simply providing clear factual information is not sufficient to ensure patient understanding. They argue that patients hold a very strong assumption that their doctor acts in their best interests and offers what he or she considers is best for them, and patients are inclined to maintain or fall back on that assumption despite being given clear information that this will not happen in a RCT. According to this view, the solution is to ensure that patients realise they are entering a research study which does not follow the procedures of standard clinician-patient consultations.

Appelbaum, Roth, Lidz, Benson & Winslade (1987), Appelbaum, Grisso, Frank, O’Donnell & Kupfer (1999) assumption that patients tend to interpret novel trial experiences in terms of their familiar framework of standard clinician-patient interactions is entirely in line with the longstanding view in cognitive and social psychology, according to which people are not passive recipients of information but rather are active interpreters who try to make sense of new input by drawing on their background knowledge and beliefs (e.g. DiSessa, 1996; Shank & Abelson, 1977). But Appelbaum and colleagues’ approach apparently rests on the assumption that so long as patients abandon their inappropriate assumptions about treatment allocation in a trial setting, they are in a strong position to grasp and hold on to the information they are given about random allocation. Disabusing patients of incorrect assumptions may not however be sufficient to ensure they understand and accept the true situation. Evidence from the qualitative studies mentioned above (Featherstone & Donovan (1998), Featherstone & Donovan, 1998 (2002); Snowdon et al., 1997), suggests that at least some patients actively try to make sense of why randomisation is used. Their attempts may in some cases produce incorrect justifications and this may lead them to lose sight of the fact of initial equipoise. Some patients may fail to come up with any plausible reason for randomising, and may then lose sight of the fact of randomisation.

From the trialists’ perspective, there are two intertwined justifications for randomisation, one ethical and the other scientific. The state of initial equipoise which motivates setting up the trial also provides the ethical justification for randomisation: it is acceptable to allocate participants to treatment arms at random only if relevant experts as a group have no convincing grounds for believing that one treatment arm has overall benefits compared with another. Note that the initial state of collective equipoise does not mean that potential trial recruits need themselves be in equipoise. A patient might legitimately have a preference for one treatment arm over another, based on his or her particular values in combination with accurate understanding of the current state of knowledge about the treatments under comparison (Ashcroft, 1999; Lilford & Jackson, 1995). Nor does the state of collective equipoise mean that individual clinicians need themselves be in personal equipoise; we come back to this point below.

The state of collective equipoise does not provide a sufficient justification for offering random allocation to treatment arms. The further necessary justification is the scientific one, that by minimising selection bias, a randomised trial is considered most likely to yield results that genuinely increase our knowledge (Schulz, Chalmers, Hayes, & Altman, 1995).

On the assumption that the UK recommended wording for approval by Multicentre Research Ethics Committees (given above) provides an example of current best practice, explanations of the scientific or ethical justifications for randomising in trial information leaflets may be either missing or obscure. The guidelines’ recommended statement about equipoise is not explicitly presented as a justification for randomising, though it could be interpreted as a justification in the absence of any other, and it could be that at least some people find it an adequate justification. We examine this possibility in Studies 1 and 4 reported below.

Trial information leaflets, which provide only bald statements about randomisation and equipoise, may be adequate to permit a patient who has the relevant background knowledge to construct a true justification for randomisation. They may not be adequate, though, for many members of the lay public. The aim of the studies reported below was to investigate the background knowledge and assumptions that members of the public could bring to bear should they be invited to participate in a trial. We use hypothetical scenarios involving individual treatment and clinical and non-clinical research to examine the extent to which lay people consider randomisation to be justified given a state of equipoise (Studies 1 and 4), and the extent to which they consider random allocation to have scientific benefits over other methods of allocation (Studies 2–4). Since the ethical justification for randomisation relies on acceptance of the possibility of equipoise, we examined lay views about the possibility of equipoise in Studies 1 and 4.

Background knowledge may be accessed differently in different contexts: medical vs. non-medical, and medical treatment vs. research. We included additional comparisons to examine such context effects. The rationale for these is given in the introduction to each study.