Thomas Frank, Ahmed A. Suleiman, Rüdiger Siek, Andreas Kovar
Sanofi
Objectives: Pharmacometric analyses are used to streamline clinical drug development, and their reports are required as part of submission packages. Pharmacometrics reports are typically complex as they include many figures, tables, text files, in-line text, equations, and numerical values. Standards of these reports are defined by different regulatory agencies (FDA, EMA and PMDA). We have previously described a reproducible LaTeX document preparation system that can automate and standardize the creation of these reports [1]. Pros and cons in using R Markdown for writing pharmacometrics reports have been detailed recently [2]. The objective of this poster is to share our experience on combining the ease of use of R Markdown with the typographical quality of LaTeX to produce reports that meet all regulatory requirements while ensuring reproducibility, automation, and standardization.
Methods: The computational environment used to setup the reporting system consisted of RStudio Server (version 2023.03.1 Build 446), Perl 5 (version 5.26.1), R (version 4.2.0) with ‘bookdown’ (version 0.3.4), ‘rmarkdown’ (version 2.25), ‘knitr’ (version 1.44) packages and Tex Live 2023 running under openSUSE Leap 15.3. Established report templates based on ‘scrartcl’ document class of KOMA-Script [1] were used to receive the LaTeX code generated by the pandoc converter.
Results:
A modularized report was setup using several R Markdown (*.Rmd) files. The master file included the YAML header, defining report properties such as title, author, reviewer, approver, project code, and R chunks to process child documents. It also defines the LaTeX engines and template to be used. Child *.Rmd files represent report sections, containing code and narrative text describing analyses and appendices. R chunks / LaTeX commands in the master file also control LaTeX packages like ‘biblatex/biber’ and ‘glossaries’ for creation of reference lists and term/abbreviation definitions.
For simplification, users can use R code chunk options and ‘kable’ function instead of complex LaTeX figure and table environments. A Perl script streamlines coding, gathering user information for the YAML header and automating analyses environment creation. It places templates in the correct location to prevent time-consuming tasks like copying/pasting or renaming files.
Templates and the Perl script reside in a corporate GitHub repository for shared and controlled revision.
Conclusions: Applying our reproducible workflow based on R Markdown supports an uninterrupted workflow along “data-analysis-report” chain. Experience gained so far shows that the integration of “analysis & results” and “report” phases resulted in savings of ~20 – 50% of standard reporting time. Also, with LaTeX in the background, the PDF reports generated by R Markdown are completely submission-compliant and caused little and no follow-up steps for dossier managers. By combining narrative text and code, the traceability and reproducibility of pharmacometrics analysis is ensured enabling the preservation of institutional memory. Users positively noted the ability to stay in the familiar work environment (RStudio), take its advantage of the Word-like visual editor features, and minimizing the need to write LaTeX code. The RStudio Server provides a convenient interface to all required software components and does not require the user to interact with a separate editing environment for LaTeX.
References:
[1] Frank T et al. LaTeX tutorial for the standardization and automation of population analysis reports. CPT Pharmacometrics Syst Pharmacol (2021) 10 (11), 1310–22
[2] Moroso V, Magnusson MO, Jonsson NE. Writing reports of modelling and simulation analysis: Our experience in the field of pharmacometrics. Med Writing (2023) 32 (3), 56-63
Reference: PAGE 32 (2024) Abstr 10798 [www.page-meeting.org/?abstract=10798]
Poster: Methodology - Other topics