Mike K Smith (1), Phylinda Chan (1), Luke Fostvedt (2), Ana Ruiz (2), Francois Gaudreault (3,5), Camille Vong (3), Jae-Eun Ahn (4).
(1) Pfizer, Sandwich, UK; (2) Pfizer, La Jolla, USA; (3) Pfizer, Cambridge, USA; (4) Pfizer, Groton, USA; (5) Biogen, Cambridge, USA
Objectives:
Capturing best practices, communicating and sharing these is a first step on the path to industrialising population modelling and increasing efficiency within organisations. Recent manuscripts [1-5] have disseminated “best practice” around various aspects of modelling, model diagnostics and reporting. These are invaluable for the new modeller, and useful for the experienced modeller in calibrating their work against expectations from the community, academia and from regulatory agencies. But few of these discussions of what is expected describe how this should be achieved [4, 5]. After all, each analyst will have their own preferred tools and even when analysts use a common tool, their experience and depth of knowledge in that tool may bring heterogeneity in how they implement even standard models and workflows. Providing “standardised” code, code snippets and templates for commonly used population modelling tasks allows analysts to focus on the pharmacology and statistics of a problem and less on the computer science and coding. We present how Pfizer are attempting to harmonise Pop PK modelling within our organisation, implementing the “best practices” expressed in our guidance into code that analysts can use as a skeleton for refining according to their needs.
Methods:
By using the R package `bookdown` we are able to incorporate the advice from the guidance alongside reproducible code and code snippets that illustrate how to implement the guidance. We can easily keep the “book” up to date, refining code that becomes out of date or where a better solution emerges, and extending the book by adding chapters on new topics as required. Code and output can be fully explained and annotated in the text, and output from the code is kept alongside so that the analysts can see both inputs and outputs from each step. Snippets of code can be presented that allow the analysts to pick and choose the code that applies to their situation. Topics covered include exploratory data analysis using the `tidyverse` R packages, structural model building for Pop PK models in NONMEM, model diagnostics using the R package `xpose`, covariate model building using PsN FREM and SCM implementations, model qualification using PsN routines and creating run record and parameter tables ready for reporting. The book can be compiled to GitBook (HTML), PDF or EPUB formats.
Results:
Example datasets and model code are presented to provide working examples. Example code is presented within the book and is executed at the time of compilation which ensures reproducibility. Current best practice methods using NONMEM, R and PsN are presented. The material covered aims to support the analyst in going from data checkout and exploratory data analysis to a qualified model ready for inference, and to provide standardised code for implementing the steps along that path. The book has been shared within Pfizer and analysts are providing feedback on the content and methods presented.
Conclusions:
While the standardisation of tools and coding practices within a company help efficiency in delivering model informed decision making, sharing these practices more widely would help refine the code and truly standardise these simpler aspects of Pop PK workflow, forming the basis of training material, and allowing analysts to focus on science rather than coding.
References:
[1] Byon, W., Smith, M., Chan, P., Tortorici, M., Riley, S., Dai, H., Dong, J., Ruiz-Garcia, A., Sweeney, K. and Cronenberger, C. (2013), Establishing Best Practices and Guidance in Population Modeling: An Experience With an Internal Population Pharmacokinetic Analysis Guidance. CPT: Pharmacometrics & Systems Pharmacology, 2: 1–8, 51. doi:10.1038/psp.2013.26
[2] Dykstra, K., Mehrotra, N., Tornøe, C.W. et al. (2015), Reporting guidelines for population pharmacokinetic analyses. J Pharmacokinet Pharmacodyn 42: 301. https://doi.org/10.1007/s10928-015-9417-1
[3] Karlsson, M. O. and Savic, R. M. (2007), Diagnosing Model Diagnostics. Clinical Pharmacology & Therapeutics, 82: 17–20. doi:10.1038/sj.clpt.6100241
[4] Nguyen, T. H. T., Mouksassi, M.-S., Holford, N., Al-Huniti, N., Freedman, I., Hooker, A. C., John, J., Karlsson, M. O., Mould, D. R., Pérez Ruixo, J. J., Plan, E. L., Savic, R., van Hasselt, J. G. C., Weber, B., Zhou, C., Comets, E., Mentré, F. and for the Model Evaluation Group of the International Society of Pharmacometrics (ISoP) Best Practice Committee (2017), Model Evaluation of Continuous Data Pharmacometric Models: Metrics and Graphics. CPT Pharmacometrics Syst. Pharmacol., 6: 87–109. doi:10.1002/psp4.12161
[5] Keizer, R., Karlsson, M. and Hooker, A. (2013), Modeling and Simulation Workbench for NONMEM: Tutorial on Pirana, PsN, and Xpose. CPT: Pharmacometrics & Systems Pharmacology, 2: 1–9, 50. doi:10.1038/psp.2013.24
Reference: PAGE 27 (2018) Abstr 8606 [www.page-meeting.org/?abstract=8606]
Poster: Methodology - Other topics