Sequential Analysis: Reply to the Presentations

Abstract

At its best, sequential analysis is an excellent means of obtaining dependable information quickly without prolonging a study more than is absolutely necessary; quite apart from a saving in time and expense this enables one to avoid exposing patients to an ineffective or second-rate form of treatment once it has become evident that another therapy is superior. The essential problem is how much further one can go along this road without leaving a degree of residual uncertainty as to the validity of the result obtained. The “look and see” or “not-very-sequential” approach, which enables the investigator to break the code as soon as a trend appears to be developing in one direction or the other, must clearly be used with some reticence where there are no urgent reasons to employ it, e.g. where the condition under treatment is not dangerous or seriously disabling and the importance of a reliable verdict (which will have repercussions for large patient populations) must weigh more heavily than the inconvenience suffered by those participants receiving the less-effective treatment.