Multi-arm multi-stage trials can improve the efficiency of finding effective treatments for stroke: a case study

Table 2 Properties of MAMS design and running two separate trials

Design	Treatments ineffective		One treatment effective
Design	Total type-I error rate	Expected sample size	Power to recommend effective treatment	Expected sample size
MAMS, futility only, two-stage	0.033	623	0.761	834
Separate trials	0.034	992	0.644	992
Separate trials, group-sequential	0.036	608	0.625	764
Multi-arm, no interim analyses	0.036	990	0.782	990

As described further in the manuscript, the ‘Treatments ineffective’ scenario uses a success probability of 0.304 for the control treatment, 0.302 for the first experimental treatment and 0.294 for the last arm. The ‘one treatment effective’ scenario uses 0.404 for one experimental treatment while the other two arms use 0.304. Statistical properties are found by approximating the log-odds ratio as normally distributed

ISSN: 1471-2261