On the sequential design of the Deutsche Nikotinamid Interventionsstudie--DENIS

The "Deutsche Nikotinamid Interventionsstudie" (DENIS) is a sequentially designed double blind clinical trial. This paper describes the design and monitoring of the trial. In contrast to the classic fixed-sample design the chosen sequential triangular test allows the trial to be terminated and the final result to be concluded, keeping the predetermined error probabilities of the test after an interim analysis. The main reason for this sequential test was the desire for early results because of the expected large diabetes risk reduction on the one hand and the risk of side effects on the other. Furthermore, all probands of the trial may change therapy and thus also potentially benefit from an earlier outcome of the study. The trial is designed in such a way that, assuming a 3 year diabetes incidence of 30% in the placebo group, a reduction to 6% in the nicotinamide group can be ascertained with a power of 90% using a significance level of 5%. This results in the following monitoring procedure: Probands are randomly assigned to the placebo or the nicotinamide group until termination of the study or reaching a sample size of 130. Interim analyses are performed whenever 5 new diabetic cases are manifested but at the earliest after 18 months. At each interim analysis it is decided whether to continue or to end the trial. A final result will be reached at the latest after the occurrence of 22 diabetes manifestations. But as shown, probably less than 16 diabetes manifestations will be needed, which is the number required using a fixed sample design.(ABSTRACT TRUNCATED AT 250 WORDS)