|
a) Review: approach taken to multiplicity
|
| |
Formal adjustment
|
Hierarchical testing
|
Other approach
|
None
|
|
Interim analyses
|
26/41a (63%)
|
0/41 (0%)
|
2/41b (5%)
|
13/41c (32%)
|
|
b) Survey: responses to posed scenarios
|
| |
Always
|
Sometimes
|
Never
|
Unsure
|
|
Would you adjust for multiplicity if interim analysis(es) were pre-specified in the study protocol?
|
8/27 (30%)
|
12/27 (44%)
|
3/27 (11%)
|
4/27 (15%)
|
- Notes: a Eight trials used the Haybittle-Peto procedure, eight used O’Brien-Fleming, seven partitioned the significance level between final and interim analyses (with no further details given), one used Pocock, one used Lan DeMets and one did not give details.
- bOne trial used a group sequential design and one used a conditional rejection probability approach.
- cOf these, three trials stated a pre-specified significance level for stopping the trial.
