Back to the Top
The following message was posted to: PharmPK
Kinetica incorporates the EM algorithm that was originally in P-Pharm
for the population pharmacokinetic analysis. The population estimation
algorithm in Kinetica is an iterative process that computes the maximum
likelihood estimates. During the E (expectation) step, the individual
parameters using a Bayesian estimate are determined from the population
parameters and the individual data. And for the M (maximization) step,
the population parameters, using maximum likelihood, are estimated from
linearization about those Bayesian estimates from the E-step. The two
steps are iterated back and forth until the changes in the fixed, random
and residual error variances between two consecutive iterations is lower
than 0.01.
Kinetica outputs allow the user to evaluate the population model through
(1) statistical criteria using objective function, Akaike and Schwartz
criteria (AIC and BIC), (2) visual examination of the goodness-of-fit
curves (population fitted model and individual fittings); (3) visual
examination of scatterplot of the observed versus predicted
concentrations of the drugs; (4) the distribution of the normalized
residuals and individual pharmacokinetic parameter estimates. Kinetica
also allows the user to incorporate covariable model and to check for
the decrease in interindividual variability of the parameter estimates
after taking the covariable model into account. The user can also
compare the models by evaluating the statistical criteria discussed
previously.
Kinetica also allows the user to perform both external and internal
validation. The user can apply the existing population model to a new
dataset from another study (external validation) or the user can split
the data into either the test set or the validation data set (internal
validation). Kinetica uses two methods to validate population models,
namely (1) Prediction errors on concentration method (or concentration
method) and (2) validation through parameter method (or parameter
method). In the parameter method, model parameters are predicted from
the validation data set with or without covariates; bias and precision
are calculated for the predictions. The results from the population
model are used to run the Bayesian fit (E-step) on the validation
datasets. The concentration method uses the parameter values from the
test dataset to predict the concentration for individuals in the
validation dataset group. The 95% confidence intervals about each
concentration data are constructed.
Kinetica is a powerful, industry-standard
pharmacokinetic-pharmacodynamic (PK/PD) application for population
PK/PD, non-compartmental analysis, standard fitting-simulation and
reporting. With an intuitive point-and-click graphical interface,
Kinetica facilitates data analysis and reporting in a structured yet
flexible, easy-to-automate environment. Kinetica offers fast
high-throughput data analysis for clinical, preclinical, discovery, drug
metabolism and drug delivery settings.
Contact:
Sherwin Sy
Thermo Electron Corporation
Informatics
E-mail: sherwin.sy.-a-.thermo.com
http://www.thermo.com/informatics
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)