AUC Ratio Prediction Errora | Simcyp Time-Based Approach | Static Model: kdeg = 0.03 h−1 | |
---|---|---|---|
kdeg = 0.0077 h−1 | kdeg = 0.03 h−1 | ||
All trials (n = 54) | |||
≤0.5-fold | 11 (20%) | 29 (54%) | 19 (35%) |
≤1-fold | 17 (31%) | 39 (72%) | 28 (52%) |
≤2-fold | 31 (57%) | 47 (87%) | 41 (76%) |
>2-fold | 23 (43%) | 7 (13%) | 13 (24%) |
Average deviation | 2.03 | 0.72 | 1.03 |
RMSE | 5.59 | 2.30 | 2.79 |
GMFE | 1.84 | 1.31 | 1.51 |
Trials with observed AUC ratio ≥2 (n = 40) | |||
≤0.5-fold | 3 (7.5%) | 19 (48%) | 11 (28%) |
≤1-fold | 6 (15%) | 27 (68%) | 17 (43%) |
≤2-fold | 18 (45%) | 34 (85%) | 29 (73%) |
>2-fold | 22 (55%) | 6 (15%) | 11 (27%) |
Average deviation | 2.56 | 0.88 | 1.09 |
RMSE | 6.47 | 2.67 | 3.17 |
GMFE | 2.02 | 1.33 | 1.49 |
Trials with observed AUC ratio <2 (n = 14) | |||
≤0.5-fold | 8 (57%) | 10 (79%) | 6 (43%) |
≤1-fold | 11 (79%) | 12 (86%) | 11 (79%) |
≤2-fold | 13 (93%) | 13 (93%) | 12 (86%) |
>2-fold | 1 (7%) | 1 (7%) | 2 (14%) |
Average deviation | 0.54 | 0.27 | 0.72 |
RMSE | 0.85 | 0.41 | 1.04 |
GMFE | 1.09 | 1.06 | 1.55 |
Trials using midazolam as probe substrate (n = 27) | |||
≤0.5-fold | 4 (15%) | 17 (63%) | 10 (37%) |
≤1-fold | 8 (30%) | 22 (81%) | 14 (52%) |
≤2-fold | 16 (59%) | 25 (93%) | 22 (81%) |
>2-fold | 11 (41%) | 2 (7%) | 5 (19%) |
Average deviation | 2.13 | 0.13 | 0.91 |
RMSE | 3.69 | 1.02 | 1.86 |
GMFE | 1.79 | 1.20 | 1.33 |
Trials with moderate CYP3A inhibitors (n = 28) | |||
≤0.5-fold | 4 (14%) | 17 (61%) | 11 (39%) |
≤1-fold | 7 (25%) | 23 (82%) | 16 (57%) |
≤2-fold | 17 (61%) | 26 (93%) | 27 (96%) |
>2-fold | 11 (41%) | 2 (7%) | 1 (4%) |
Average deviation | 2.19 | 0.16 | 0.17 |
RMSE | 3.15 | 1.81 | 1.15 |
GMFE | 1.63 | 1.25 | 1.33 |
Trials with strong CYP3A inhibitors (n = 21) | |||
≤0.5-fold | 2 (10%) | 9 (43%) | 5 (24%) |
≤1-fold | 6 (29%) | 13 (62%) | 9 (43%) |
≤2-fold | 11 (52%) | 16 (76%) | 11 (52%) |
>2-fold | 10 (48%) | 5 (24%) | 10 (48%) |
Average deviation | 1.65 | 1.43 | 2.06 |
RMSE | 7.48 | 2.97 | 4.20 |
GMFE | 2.30 | 1.36 | 1.67 |
↵a Prediction error = |observed AUC ratio − predicted AUC ratio|.