TABLE 5.
Type I error and power (%) comparison when using both Centiloid-11C-PiB and Centiloid-18F-florbetapir in 1000 simulated clinical trials
| No treatment effect (type I error comparison) | With observed treatment effect (power comparison) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Percent (%) of participants scanned with 18F-florbetapir | 100 | 75 | 50 | 25 | 0 | 100 | 75 | 50 | 25 | 0 |
| 20 participants per arm | 4.8 | 9.3 | 14.8 | 9.7 | 4.3 | 52.9 | 63.3 | 65.0 | 72.5 | 84.3 |
| 32 participants per arm | 4.9 | 8.2 | 14.6 | 9.4 | 5.7 | 74.4 | 81.1 | 83.7 | 88.4 | 98.0 |
| 40 participants per arm | 4.5 | 8.0 | 15.0 | 9.6 | 4.9 | 85.0 | 88.2 | 88.6 | 93.9 | 98.9 |
| With 5% annual dropout | ||||||||||
| 20 participants per arm | 4.6 | 9.8 | 12.2 | 8.5 | 4.1 | 48.7 | 54.8 | 60.7 | 67.2 | 76.5 |
| 32 participants per arm | 6.0 | 8.6 | 13.6 | 9.9 | 5.7 | 67.8 | 73.1 | 76.6 | 84.9 | 92.1 |
| 40 participants per arm | 4.7 | 8.0 | 11.9 | 9.4 | 5.0 | 78.5 | 82.2 | 85.5 | 94.4 | 97.6 |
Each simulated dataset was analyzed using linear mixed models and the treatment effect was defined as the difference between the slopes (annual rate of change in Centiloid). Two-sided t-tests were used with a nominal type I error of 5%.