The SEQTEST Procedure

Information Level Adjustments at Future Stages

In a group sequential clinical trial, the information level for the observed test statistic at the current stage generally does not match the corresponding information level in the BOUNDARY= data set. By default (or equivalently if you specify INFOADJ=PROP), the SEQTEST procedure accommodates the observed information level by adjusting the information levels at future interim stages. The adjustment of information levels depends on the boundary key to be maintained in the boundary adjustments, which in turn is determined by the BOUNDARYKEY= option.

If you specify BOUNDARYKEY=ALPHA (which is the default) or BOUNDARYKEY=BETA, the maximum information level (the information level at the final stage) provided in the BOUNDARY= data set is maintained. In this case, if an observed information level at the current stage is different from the level provided in the BOUNDARY= data set, you can use the INFOADJ= option to determine whether the information levels at subsequent interim stages are to be adjusted. Specifying INFOADJ=NONE preserves the levels provided in the BOUNDARY= data set without adjustment. Specifying INFOADJ=PROP proportionally adjusts the levels provided in the BOUNDARY= data set as follows.

Denote the information level at stage k for the K-stage design that is stored in the BOUNDARY= data set by $I_{k}$, $k= 1, 2, \ldots , K$. Also denote the information level that corresponds to the test statistic at an interim stage $k_0$ by $I’_{k_0}$, $1 \leq k_0 \leq (K-1)$. Then for the updated design, the information level at stage k, $k= k_0+1, \ldots , (K-1)$, is computed as

\[ I’_{k} = I’_{k_0} + ( I_{K} - I’_{k_0} ) \, \frac{ I_{k} - I_{k_0}}{I_{K} - I_{k_0}} \]

Note that if $I’_{k_{0}} \geq I_{K}$, the information level at stage $k_{0}$ reaches the maximum information level in the design, the trial stops at stage $k_{0}$, and no future information levels are derived.

If you specify BOUNDARYKEY=BOTH, the maximum information level for the trial is not necessarily the same as the maximum information level saved in the BOUNDARY= data set. In this case, the INFOADJ=NONE option is not applicable, and the INFOADJ=PROP option is used to proportionally adjust the information levels at future interim stages with the updated maximum information $I’_{K}$. That is, with an updated $I’_{K}$, the information level at a future interim stage k is computed as

\[ I’_{k} = I’_{k_0} + ( I’_{K} - I’_{k_0} ) \, \frac{ I_{k} - I_{k_0}}{I_{K} - I_{k_0}} \]