Least-Squares_Regression

Least-Squares Regression models include the following. Many of the models can be run using the NLME engine (even if you do not have an NLME license). This is done by setting up a Maximum Likelihood Models object for individual modeling and using the Set WNL Model button to select the model. See the “PK model options” section in the Phoenix NLME documentation for more information. Refer to “An example of individual modeling with Maximum Likelihood Model object” for an illustration of how a dataset can be fitted to a two-compartment model with first-order absorption in the pharmacokinetic model library using either the Least-Squares Regression PK Model or a Maximum Likelihood Model object GUI.

Dissolution Models

Choose from Hill, Weibull, Double Weibull, or Makoid-Banakar dissolution models.

Indirect Pharmacodynamic Response Models

Four basic models have been developed for characterizing indirect pharmacodynamic responses after drug administration. These models are based on the effects (inhibition or stimulation) that drugs have on the factors controlling either the input or the dissipation of drug response. See the “Indirect Response models” section for more details.

Linear Models

Phoenix includes a selection of models that are linear in the parameters. See “Linear models” for more details on available models. Refer to the “Linear Mixed Effects Object” section for more sophisticated linear models.

Michaelis-Menten Models

Phoenix’s Michaelis-Menten models are one-compartment models with intravenous or 1st order absorption, and can be used with or without a lag time to the start of absorption. For more on Phoenix’s Michaelis-Menten models, see the “Michaelis-Menten models” section. Information on required constants is available in the “Dosing constants for the Michaelis-Menten model” section.

Pharmacodynamic Models

Phoenix includes a library of eight pharmacodynamic (PD) models. The PD models include simple and sigmoidal Emax models, and inhibitory effect models. For more on Phoenix’s PD models, see the “Pharmacodynamic models” section.

Pharmacokinetic Models

Phoenix includes a library of nineteen pharmacokinetic (PK) models. The PK models are one to three compartment models with intravenous or first-order absorption, and can be used with or without a lag time to the start of absorption. For more on Phoenix’s PK models, see the “Pharmacokinetic models” section. See also the “PK model examples”.

PK/PD Linked Models

When pharmacological effects are seen immediately and are directly related to the drug concentration, a pharmacodynamic model is applied to characterize the relationship between drug concentrations and effect. When the pharmacologic response takes time to develop and the observed response is not directly related to plasma concentrations of the drug a linked model is usually applied to relate the pharmacokinetics of the drug to its pharmacodynamics.

The PK/PD linked models can use any combination of Phoenix’s Pharmacokinetic models and Pharmacodynamic models. The PK model is used to predict concentrations, and these concentrations are then used as input to the PD model. This means that the PK data are not modeled, so the linked PK/PD models treat the pharmacokinetic parameters as fixed, and generate concentrations at the effect site to be used by the PD model. Model parameter information is required for the PK model in order to simulate the concentration data. Refer to the “PD output parameters in a PK-PD model” section for parameter details.

User-Defined ASCII Models

Phoenix does not support the creation of ASCII models. ASCII models have been deprecated in favor of the Pharmacometrics Modeling Language (PML). However, legacy WinNonlin ASCII models can still be imported and run. Refer to the “ASCII Model dosing constants” section for details on required constants. For more on PML, see “Pharmacometrics Modeling Language” documentation.

Note: There can be a loss of accuracy in Least-Squares Regression Modeling univariate confidence intervals for small sample sizes (NDF < 5). The Univariate CIs in use an approximation for the t-value, which is very accurate when the degrees of freedom is at least five but loses accuracy as the degrees of freedom approaches one. The degrees of freedom are the number of observations minus the number of parameters being estimated (not counting parameters that are within the singularity tolerance, i.e., nearly completely correlated).

In extremely rare instances, the nonlinear modeling core computational engine may get into an infinite loop during the minimization process. This infinite looping will cause Phoenix to “hang” and the application must be shutdown using the Task Manager. The process wnlpk32.exe may also need to be shutdown. The problem typically occurs when the parameter space in which the program is working within is very flat. To work around the problem, it is first suggested that the minimization algorithm found on the Engine Settings tab be changed to Nelder-Mead and then retry the problem. If this fails to correct the problem, varying the initial estimates and/or using bounds on the parameters may allow processing to complete as expected.