Randomised studies of emergency therapy for variceal bleeding are difficult to design, conduct and evaluate. Despite international consensus definitions covering descriptive characteristics of patients, evaluation of end points and checklists for presentation of results, many trials are difficult to interpret.1 The multicentre English study by Jenkins et al(Gut1997;41:526–33) is a case in point. In this randomised study injection sclerotherapy (n=77) was compared with octreotide 50 μg/h (n=73) over a 48 hour period, after which sclerotherapy was used in the octreotide group. By its nature this comparison is unblinded for both the clinician and the patient. Although it could be argued that end points for control of bleeding are fairly “hard end points”, this is not the case in practice. I highlight some examples. Balloon tamponade was added for 12 hours in both trial groups: in the injection group the indication was if there was persistent oozing at endoscopy and in the octreotide group, the indication was “if the bleeding persisted after starting octreotide”, presumably after the endoscopy had been completed. Although the numbers needing tamponade were similar (sclerotherapy n=20, octreotide n=13), I suspect the severity of bleeding was different as the parameters of assessment were not the same (endoscopic oozing versus clinical evidence of persistent bleeding). Neither use of balloon tamponade was counted as a failure to control bleeding.  Further need of balloon tamponade and/or haematemesis or melaena accompanied by changes in pulse or blood pressure (not specified) or decrease in haemoglobin (not specified) within 48 hours constituted failure to control bleeding (n=14, injection sclerotherapy; n=11, octreotide). This was the same in both groups. However, if transfusion requirements are evaluated to corroborate the purported similar efficacy then again the data are not presented in a standard way. Transfusion need is described, but only from hospital …