Moussa M A
Faculty of Medicine, Kuwait University, Safat.
Control Clin Trials. 1989 Dec;10(4):378-85. doi: 10.1016/0197-2456(89)90003-2.
It is shown how the predictive power, which takes account of prior clinical opinion about treatment differences, can be extended to group sequential designs based on a range rather than point of equivalence. It is found that such design when averaged over the whole prior distribution seems to be most conservative in terms of probability of firm conclusion followed by the nonzero null hypothesis design. Planning a trial based on calculations of the "exact" power or zero null hypothesis design may result in a practically unacceptable probability of firm conclusion.
研究表明，考虑到关于治疗差异的先验临床观点的预测能力，如何能够扩展到基于等效区间而非等效点的成组序贯设计。研究发现，就得出确切结论的概率而言，这种在整个先验分布上进行平均的设计似乎最为保守，其次是非零原假设设计。基于“精确”检验效能计算或零原假设设计来规划试验，可能会导致得出确切结论的概率在实际中无法接受。
