Variceal bleeding: does variable delay in trial entry invalidate comparison between trials?

Variation in time of patient admission to studies of prognosis after variceal haemorrhage has been proposed as a major factor in the wide range of reported results. A study of 144 unselected subjects with a low initial mortality (3% at two days) suggests that the effect has been overemphasised and studies in which time of entry is later than the date of bleeding may be usefully compared. Reanalysis of previous work suggests that even in populations with a high initial mortality limited comparisons between studies can still be usefully made. It is stressed that survival analysis must be started from the date of study entry and not from the date of the index bleed which involves incorporating a retrospective period of survivorship.