First year of the Software Engineering working group

Regulatory-Industry Statistics Workshop 2023

Daniel Sabanés Bové (Roche) and Ya Wang (Gilead) on behalf of the working group

2023-09-29

Introducing the WG

Software Engineering Working Group

Founded last year:

  • When: 19 August 2022 - just celebrated our 1 year birthday!
  • Where: American Statistical Association (ASA) Biopharmaceutical Section (BIOP)
  • Who: 11 statisticians from 7 pharma companies developing statistical software
  • New name: openstatsware

Why a new WG?

  • Started with specific R-package project (more below)
  • Makes sense to stay together as a group also for other package projects
  • New focus on good engineering practices and collaborative work
  • Importance of reliable software for statistical analysis can not be underestimated
  • Be rooted in the biostatistics community (rather than statistical programming)

But there are other WGs?

  • Comparing Analysis Method Implementations in Software (CAMIS) of PhUSE 🌎
  • Application and Implementation of Methodologies in Statistics (AIMS) Special Interest Group (SIG) of PSI 🌎
  • R Submission Working Group of R Consortium 🌎
  • R Tables for Regulatory Submissions Working Group of R Consortium 🌎
  • R Validation Hub 🌎
  • R Repositories WG of R Consortium 🌎

WG Objectives

  • Primary:
    • Engineer R packages that implement important statistical methods
    • … to fill in gaps in the open-source statistical software landscape
    • … focusing on what is needed for biopharmaceutical applications
  • Secondary:
    • Develop and disseminate best practices for engineering high-quality open-source statistical software
  • By actively doing the statistical engineering work together, we align on best practices and can communicate these to others

Relation to the validation topic

  • Validation is on every one’s mind when talking about open source in Pharma:
    • “Establishing documented evidence which provides a high degree of assurance (accuracy) that a specific process consistently (reproducibility) produces a product meeting its predetermined specifications (traceability) and quality attributes.”
  • The R Validation Hub is a collaboration to support the adoption of R within a biopharmaceutical regulatory setting
    • Develops tools that supports risk assessment of R packages

Relation to the validation topic (cont’d)

  • Base R and recommended packages are developed by the R Foundation, developed using best practices.
  • Our working group is spreading the word about best development practices such that contributed packages can achieve the same level of quality, hence aiding subsequent validation of them.
  • And we develop specific R packages with these best practices that can thus be easily validated.

Members and meetings

  • Currently 59 members
    • new members are welcome! (incl. academia/regulators/etc.)
  • Currently 38 organizations
    • Affamed, AstraZeneca, Bayer, Berry Consultants, BMS, Boehringer Ingelheim, CIMS Global, Cytel, Daiichi Sankyo, Denali, Edwards Lifesciences, Eli Lilly, EMD Serono, Erste Group, Genentech, Gilead Sciences, GSK, ICON, Independent, Johnson & Johnson, mainanalytics, MaxisIT, Merck, Merck KGaA, MSD, Novartis, Novo Nordisk, Pfizer, R&G US, RCONIS, Red Door Analytics, Regeneron, Roche, RPACT, RStudio & R Consortium, Sanofi, Transition Technologies Science, UCB
  • Meet every 2 weeks

Workstreams

  • Mixed Models for Repeated Measures (MMRM) 🌎
    • Develop mmrm (see below) to use frequentist inference in MMRM
  • Bayesian MMRM 🌎
    • Develop brms.mmrm (see below)
  • Health Technology Assessment 🌎
    • Develop open-source R tools to support HTA dossier submission across various countries, particularly the topics with unmet needs in R implementation and/or related to EUnetHTA
  • Note: Also “just” contributing to workstreams is great!

Achievements in the first year

New R packages released to CRAN

  • mmrm
    • R package for frequentist inference in MMRM, based on TMB (which provides automatic differentiation in C++ and R frontend)
    • See documentation 🌎
    • Easiest to install from CRAN 🌎
  • brms.mmrm
    • R package for Bayesian inference in MMRM, based on brms (as Stan frontend for HMC sampling)
    • See documentation 🌎
    • Easiest to install from CRAN 🌎

Why was the MMRM topic important?

  • MMRM is a popular analysis method for longitudinal continuous outcomes in randomized clinical trials
  • No tailored R package with sufficient capabilities/reliability
  • Also used as backbone for more recent methods such as multiple imputation
  • We have run comparison analyses with other R packages, namely nlme, glmmTMB and lme4 as well as SAS PROC GLIMMIX and have found that mmrm is the fastest of them and provides closest results to PROC GLIMMIX

Best practices

  • Includes version control, git workflow, code review, unit/integration testing, continuous integration/delivery (ci/cd), reproducibility, change log, documentation, package design, user experience, maintainability, publication, etc.
  • Workshop “Good Software Engineering Practice for R Packages” on world tour
    • Basel, Shanghai, San José, Rockville (last Tuesday), Montreal 🌎
  • Start of video series “Statistical Software Engineering 101” 🌎
    • currently 2 videos, hopefully we can still produce more content

Conference contributions and Publications

  • Dedicated sessions with discussions at ISCB, CEN, ASA/FDA workshop (now)
  • Presentations at PSI, JSM, Pharma RUG, BBS, etc. 🌎
  • BIOP Report 🌎
  • Blog 🌎

Ingredients for successful and sustainable collaboration

Human factors

  • Mutual interest and mutual trust
  • Prerequisite is getting to know each other
    • Although mostly just online, biweekly calls help a lot with this
  • Reciprocity mindset
    • “Reciprocity means that in response to friendly actions, people are frequently much nicer and much more cooperative than predicted by the self-interest model”
    • Personal experience: If you first give away something, more will come back to you.

Development process

  • Important to go public as soon as possible
    • don’t wait for the product to be finished
    • you never know who else might be interested/could help
  • Version control with git
    • cornerstone of effective collaboration
  • Building software together works better than alone
    • Different perspectives in discussions and code review help to optimize the user interface and thus experience

Coding standards

  • Consistent and readable code style simplifies joint work
  • Written (!) contribution guidelines help
  • Lowering the entry hurdle using developer calls is important

Robust test suite

  • Unit and integration tests are essential for preventing regression and assuring quality
  • Especially with compiled code critical to see if package works correctly
  • Use continuous integration during development to make sure nothing breaks along the way

Documentation

  • Lots of work but extremely important
    • start with writing up the methods details
    • think about the code structure first in a “design doc”
    • only then put the code in the package
  • Needs to be kept up-to-date
  • Need to have examples & vignettes
    • Testing alone is not sufficient
    • Builds trust with users
    • Reference for developers over time

Long term vision

Vision: Statisticians have software engineering skills

  • These skills are taught at university
  • Statisticians can use basic practices in their daily work …
  • … to ensure reproducibility of statistical analyses and research results

Vision: Innovation does not stop with publication

  • Methods research does not end when the first methods paper is published
  • Initial prototype code as paper supplement is not sufficient
  • Continue to developing open source and reliably tested software packages …
  • … to enable users to easily use the new methodology in their own applications

Vision: Industry develops common code base

  • Increasingly companies work in the open source as much as possible
  • Rather than repeating similar developments internally …
  • … to become more cost-effective and transparent towards society

Next steps

R Packages

  • New workstream on covariate adjustment is starting up
  • Think more strategically about identifying gaps in the statistical software landscape
  • Help maintaining the CRAN Task View on Clinical Trials

Branding and Collaboration

  • Publicize our new short name openstatsware
  • Proposed to associate also with EFSPI to emphasize global nature of the group
  • Ensure a strong connection to the new pan-pharma methodology group

Communication and Outreach

  • Add more content to our video series
  • Start a chat channel to start informal discussions within larger community
  • Organize hackathons working together on workstream packages e.g.

Q&A