Skip to main content
Home / Resources / Blog / Why Phoenix NLME is the Best in Class Solution for Pop PK/PD

Why Phoenix NLME is the Best in Class Solution for Pop PK/PD

Over the past 20 years, I have seen a number of significant changes in the pharmaceutical industry’s approach to the study of pharmacokinetics and pharmacodynamics (PK/PD). With the high risk and large expense associated with drug development, it is imperative to have the best PK/PD analytical tools to aid in understanding the safety/efficacy profile of investigational drugs. I’m glad to have been a part of the development of Phoenix NLME. And I’m even more excited about the recent improvements to Phoenix NLME that are part of the broader Phoenix 1.4 release, and the direction that Certara will be taking Phoenix NLME in the future.

A short history of Phoenix NLME

WinNonlin, and its predecessors, have been an industry standard for more than 30 years. In 2009, Certara performed a major expansion of WinNonlin by developing the Phoenix platform, wherein all the tools—Phoenix WinNonlin, Phoenix NLME, Phoenix Connect, the IVIVC Toolkit for Phoenix WinNonlin—work seamlessly together. Phoenix WinNonlin is the tool for NCA, individual compartmental PK/PD modeling and simulation, and other types of analysis including bioequivalence. Phoenix NLME is a companion tool for the pre-processing, analysis, and post-processing of population PK/PD modeling and simulation projects. Because Phoenix NLME is part of the larger Phoenix platform, it’s very easy for Phoenix WinNonlin users to learn population modeling using Phoenix NLME.

Phoenix NLME has some great enhancements

While the entire Phoenix platform has undergone a number of updates and improvements, I believe that our users who perform population PK/PD modeling will be particularly enthusiastic about Phoenix NLME’s newest features. Phoenix NLME uses a wide set of optimization engines including the Quasi-Random Parametric Expectation Maximization engine (QRPEM). The QRPEM engine has been upgraded to be much faster for “no mu modeling”—that is a model that has fixed effects, but no random effects. We’ve also added the option for defensive importance sampling to the QRPEM engine. This makes it much easier for users to select a distribution that would give a higher probability of model convergence, ie, the best possible solution for your data. The QRPEM engine has also been upgraded to handle inter-occasion variability.

In addition, we’ve improved the start-up procedure for initial estimates for population models. To help initiate the destination for all engines, we implemented a new algorithm, Map NP (naïve pooled). Previously, users had to come up with initial estimates for parameters, which could be difficult if there are a lot of parameters or the dimensions are large. This could result in the model failing to converge. With the new Map NP option, users can perform an initial series of naïve pooled runs prior to starting the selected engines. This improvement will greatly increase chance of successful model convergence.

Finally, the newest version of Phoenix NLME uses dynamic memory allocation. Thus, there are no static limits on parameters or covariates (previously, there had been a limit of 32 covariates). This improvement provides users with a broader scope of models which can be successfully performed.

The future of Phoenix NLME

The world of pop PK/PD is always changing, and we will continue to enhance Phoenix NLME to serve our users’ needs. I can envision implementing several changes to Phoenix NLME in the next couple of years. First, we will increase the analytical power of Phoenix NLME by providing additional engines, Bayesian and SAEM (stochastic approximation of EM). Second, we will increase the usability of the program by fine tuning the existing engines. Finally, we will help our users save even more time by providing the ability to execute Phoenix NLME runs in the cloud to take advantage of parallelized processors.

Learn more about how Phoenix NLME is a powerful, easy to use tool for pop PK/PD!

It’s been a great experience to see how Phoenix NLME has grown over the years, and how it’s enabled so many scientists to easily become proficient in pop PK/PD. Recently, I’ve been working with pharmacometricians at the University of Maryland to expand Phoenix NLME’s capabilities for advanced users.

As part of this collaboration, we hosted a webinar with Dr. Tim Cacek, a recent graduate of their Master’s degree program in pharmacometrics. Watch it to learn about his journey through his introduction to Phoenix NLME, and how it has become his primary tool for population modeling.

About the author

By: Thomas Henry