Sir––My colleague Martijn B Katan (Sept 7, p 806) argues that the term prospective trial is a pleonasm because trials are prospective by definition. However, I know of at least one published retrospective randomised controlled trial that proves him wrong. In this study, done in 2000 in the intensive-care unit of an Israeli hospital, patients admitted between 1990 and 1996 with a severe disorder (blood stream infection) were randomly allocated either no intervention or a retroactive prayer said for their well-being and full recovery. The author justified the design as follows: "As we cannot assume a priori that time is linear, as we perceive it, or that God is limited by a linear time, as we are, the inter vention was carried out 4–10 years after the patients' infection and hospitalisation." In my experience, discussion of this short report provides an excellent opportunity for students to learn about publication bias and to consider several of the principles proposed by Hill for inferring causality in epidemiological studies. In addition to support obtained from experimental evidence, these principles include the necessity that the cause precedes the effect, and that the hypothesis under investigation is biologically plausible. Hans Verhoef works in the same division as Martijn Katan, and has received a bottle of red wine following a bet about whether or not it is possible to do a retrospective randomised controlled trial.