Workshop type: Discussion.

Objective: To debate criteria for deciding when it is reasonable to make recommendations about the effects of a treatment in the absence of RCTs.

Summary:
Two recent papers have proposed that there are circumstances in which requiring randomised controlled trials (RCTs) of the effects of treatments may be unreasonable.  Potts and colleagues argue that insisting on RCTs of interventions shown to be beneficial in observational studies is damaging health in less developed countries.(Potts et al, BMJ 2006).  Glasziou and colleagues proposed that the strength of association between a treatment and its effect is sometimes so large that biases can be ruled out as an explanation, making RCTs unnecessary (Glasziou et al, BMJ 2007).

On the one hand, insisting on RCTs when treatment effects appear dramatic may be seen as unethical, implying that effective treatments are being withheld, or that research resources are being used which might be better spent on other RCTs.  Holding out for RCTs may be seen as unreasonable by the public and damage the credibility of organisations such as The Cochrane Collaboration in its pursuit of high quality evidence.

On the other hand, having RCT evidence allows policies to be made with greater confidence and authority, and for the quantified benefits of treatments to be weighed against their potential harms and costs.  There is the danger of setting a precedent; if RCTs are deemed unnecessary in some circumstances will this be exploited as the ‘thin end of the wedge’?

Workshop participants will first try to identify factors relevant to decisions that RCTs are unnecessary, and then try to set thresholds beyond which it is unreasonable to demand RCTs.  Possible factors may include: (a) strength of association between treatment and effect; (b) the likelihood and severity of potential harms; and (c) the applicability of the implied associ ation in different health care settings.

Intended audience: Anyone interested in interventions that are difficult to evaluate by RCTs.

Technical level expected of participants: Intermediate.