Unlike statistics which has been around in some form for hundreds of years, pharmacometrics is, by comparison, a relatively new discipline and only entered the clinical development world in the last 30 years. Situated at the intersection of mathematical modeling, simulation, and big data, pharmacometrics leverages the best practices of translational research to generate clinical development strategy.
To conduct their complex work, pharmacometricians will tend to work with a variety of software applications, including data analysis workhorses such as SAS and R. However, specialist non-linear regression population modeling software is also usually required.
With a number of such options on the market, what are the important factors to consider when considering which non-linear regression population modeling software to use? We asked Cytel pharmacometrics expert Cecilia Fosser for her thoughts:
1) Analytical Power
This is of primary importance. How accurate are the models which are generated by the software? This should be evaluated across different types of models of different levels of complexity. A paper ( Kiang et al ) conducting a detailed comparison on these aspects is linked below:
2) Interface/ usability
One important question is what is the learning curve/ investment required for the pharmacometrician to be able to use the software- how user-friendly is it? One bottleneck in pharmacometric modeling is identifying the initial conditions for the model and a great deal of time can be expended trying to identify these at the outset . Therefore it’s important to evaluate whether the software can help with this activity. Another important aspect is to ascertain how easy it is to qualify the model (i.e., run extensive statistical tests to validate the model as being representative of the data.). A helpful feature of a modeling software tool is the automatic generation of diagnostic tables and goodness-of-fit plots, in addition to easily running bootstrap runs and posterior (or visual) predictive checks on model estimates.
Software may have functionality ‘built-in’ to allow for quick exploration of various different model components and faster and easier computations. This is helpful and essential. However, in the complex world of pharmacometrics modeling there are always going to be occasions when it’s necessary to do non-standard work. In these cases, it’s important that the software is flexible enough to allow for easy customization.
4) Industry usage
What are the trends in usage across the industry? What community support and thought-sharing is available for modeling with your specific software? Obviously, the regulatory perspective is also important to consider. Pharmacometricians at the FDA will be running models on your data, so the software they use is relevant to you.
Cytel's Quantitative Pharmacology and Pharmacometrics group have experience in a variety of different software tools and are able to tailor a solution to meet our clients’ varied needs.
To find out more about our services in this space, click the button below to view our Quantitative Pharmacology and Pharmacometrics brochure.
(1) 'Fundamentals of Population Pharmacokinetic Modelling, Modelling and Software’ by authors: Tony K.L. Kiang, Catherine M. T. Sherwin, Michael G. Spigarelli and Mary HH Ensom, Clin Pharmakokinet 2012: 51 (8) pp515-525
Read other Cytel blogs on Pharmacometrics and Quantitiative Pharmacology:
Liked this blog? Join our audience of biopharma innovators and sign up for biostatistics and biometrics insights, news, best practices and industry trends direct to your inbox . Click the button below to subscribe to the Cytel blog.