Application of restricted sequential design in a clinical protocol.

Restricted sequential design is an alternative to fixed-sample analysis as a statistical tool in clinical trials. This paper presents a specific example of the rationale which led to the choice of sequential design in a clinical protocol evaluating granulocyte transfusions in children with leukemia. The main advantage of sequential design is that fewer patients may be necessary to declare statistical significance. Its application is limited to trials where (a) the result of therapy is easily defined, (b) the result of therapy is discernible in a short time interval, and (c) one randomization is being tested, although patients may be stratified. The main disadvantage of sequential design is that if the therapies being tested are similar in efficacy, the trial may require more patients than fixed-sample analysis. This potential disadvantage may be acceptable when the concern in designing a trial is to evaluate as quickly as possible a reputedly superior therapy.