In Reply: Authors of all 4 letters acknowledge that truncated randomized controlled trials (tRCTs) overestimate treatment effects; controversy remains as to by how much and what should be done about it. A pivotal issue is what real-world data should be used to quantify overestimates. Ideally, trials with fixed sample size and no provision for early stopping would be available. Such trials are rare and typically not identifiable through reading their methods. Feasible alternatives are therefore all trials not stopped early (our primary comparison, with which Dr Berry and colleagues and Dr Goodman and colleagues disagree, describing it as a logical error) or all trials including tRCTs (our secondary comparison).
Guyatt GH, Bassler D, Montori VM. Bias and Trials Stopped Early for Benefit—Reply. JAMA. 2010;304(2):156-159. doi:10.1001/jama.2010.934