Home
East
|
|
|
East: Example 1
|
|
Weight Control Trial of Orlisat in Obese Subjects Eighteen U.S. research centers participated in a clinical trial where obese adults were randomized to receive either Orlisat or placebo plus dietary intervention (JAMA, 1999). A one-sided test with 5% significance and 90% power was designed to detect a mean weight reduction of 3 kilos and the common standard deviation of the observations was assumed to be 8 kilos.
The standardized difference of interest is thus 0.375.
The East screen-shot displayed below presents three plans; fixed-sample, 5-look sequential with early stopping only if Orlisat works, 5-look sequential with early stopping either if Orlisat works or does not work.
The O'Brien-Fleming boundary shape was used for plans 2 and 3.
|
|
The Design Phase
|
|
With Plan 1 you must commit up-front to a study involving 244 subjects with no possibility of early stopping.
With Plan 2 you must commit up-front to a study involving up to 253 subjects. If, however, H1 is true (Orlisat works), you can expect to cross the upper stopping boundary after accruing only 178 subjects on average.
With Plan 3 you must make an even greater up-front commitment of 271 subjects. But now, you have the possibility of a swift exit either if H1 is true (Orlisat works) or H0 is true (Orlisat does not work). Under H1 you expect to cross the upper stopping boundary after 175 subjects have accrued on average. Under H0 the average number is 156.
|
|
| The Interim Monitoring Phase |
|
The interim monitoring table displayed on the left presents the actual schedule of interim monitoring that was performed on the study. Only three looks were taken and the test statistic crossed the upper stopping boundary at the third look with a total accrual of 195 subjects, demonstrating that Orlisat is statistically superior to placebo plus dietary intervention at the 5% significance level. |
|
|
|
|
|
|
