Evaluation of NONMEM 7.3.0 and Monolix 4.2.2 by Parametric Bootstrap
Nick Holford, University of Auckland
NONMEM 7.3.0 and Monolix 4.2.2 were evaluated in terms of parameter estimation bias and uncertainty coverage bias using a parametric bootstrap procedure. All calculations were performed using the NeSI PAN cluster.
Four problems of increasing difficulty were tested: warfarin pharmacokinetics, simple and complex tumour growth inhibition, viral load kinetics.
Uncertainty bias was based on the bootstrap standard deviation relative to the standard error which described 95% coverage of the bootstrap distribution.
Based on parameter and standard error bias the FOCE estimation method is better in some cases and SAEM is a better method in others. Both NONMEM and Monolix had biased uncertainty relative to bootstrap coverage (see Table 1 for viral growth kinetics [1)).
Table 1 Estimation Bias for Model Parameters and Uncertainty
| NONMEM SAEM | Monolix SAEM | |||||
| Parameter | TRUE | MDL | RSE | MDL | RSE | |
| POP_RR0 | 7.15 | -1.7% | -40% | 3.6% | -41% | |
| POP_P | 25.1 | -27% | -99% | 5% | -76% | |
| POP_C | 4.53 | -25% | -61% | 4.0% | -43% | |
| POP_DELTA | 0.192 | -0.5% | 19% | 0.6% | 15% | |
| BETA_DELTA | -0.1403 | 0.6% | 16% | 4.4% | 12% | |
| POP_ED50PEG | 1.19 | -12% | 10% | 5% | 11% | |
| BETA_ED50PEG | 1.245 | 3.6% | 1.5% | -2.6% | 8% | |
| POP_ED50RBV | 14.4 | -22% | -100% | 5% | -100% | |
| POP_K | 0.0238 | 10% | -100% | -2.8% | -100% | |
| POP_R | 0.00562 | 14% | -100% | -3.2% | -100% | |
| RUV_SDVL | 0.51 | 0.4% | 9% | 0.9% | -100% | |
| PPV_RR0 | 1.37 | 10% | 17% | -0.1% | 4.1% | |
| PPV_C | 1.2 | -2.5% | -5% | -0.1% | -6% | |
| PPV_DELTA | 0.58 | 0.2% | 1.8% | 0.4% | -3.4% | |
| PPV_ED50PEG | 2.81 | 5% | -3% | -1.4% | -5% | |
TRUE=Parameter used for simulation MDL=Model parameter bias, RSE=asymptotic standard error bias relative to bootstrap 95% coverage standard error