Randomized controlled trials (RCTs) are pervasive in clinical medical research, which stands in stark contrast to other sciences such as physics, chemistry and biology. Most clinical researchers that use RCTs regard them as uncovering causal connections. R. A. Fisher best articulated the rationale for this position in 1935. According to Fisher, if randomization, blocking and replication demonstrated a connection between an intervention and an outcome, that connection is causal. This chapter argues that RCTs in clinical medicine do not reveal causal connections. Causal claims in clinical medicine, as in the rest of science, are justified by reference to a robust theory, not RCTs. Part of the argument rests on crucial differences between Fisher's use of RCTs in agriculture and the current use of RCTs in clinical medicine. Two key differences are: the different role of randomization and the legitimacy of assuming homogeneity of the intervention and control entities. A more significant part rests on the integrative power of robust theories; causal attributions are justified by demonstrating that they are, or can be, embedded in a large well-confirm framework. RCTs, by contrast, at best provide isolated input-output connections. A secondary thesis of the paper is that robust theories also allow causal claims to be well-confirmed.
Keywords: randomised controlled trials; theory structure; evidence; causality; experimental design
Chapter. 9211 words. Illustrated.
Full text: subscription required