Network Update for Programmers, Biostatisticians and clincal data enthusiasts #66

AI

Gowri Sivakumar A - AI’s Influence on SAS Programming. AI tools are transforming the role of SAS programmers, making them faster and more effective, but human expertise remains crucial in ensuring high-quality outcomes. Gowri shares how the future of #programming lies in a hybrid approach that leverages both human expertise and #AI-driven efficiencies. Link

Holger Fr?hlich - RADAR-AD: assessment of multiple remote monitoring technologies for early detection of Alzheimer’s disease. Alzheimer's Disease is the most prevalent neurodegenerative disease, which affects millions. However, it is often diagnosed too late. Remote Monitoring Technologies (RMTs), including digital medical devices and smartphone applications, offer a promising solution for early detection by tracking changes in behavioral and cognitive functions, such as memory, language, and problem-solving skills. However, before application in medical routine a careful evaluation is necessary. In that regard, the previous #IMI project RADAR-AD investigated a panel of 6 RMTs.

Our study with more than 200 participants showed how RMTs, tracking changes in activity, sleep, gait, and cognition, can identify early signs of Alzheimer’s Disease in the prodromal stage with 73% area under ROC curve using machine learning models. Link

Artificial Intelligence in Health and Health Care: Priorities for Action.

AI, driven by deep learning and generative models, presents vast opportunities, particularly in health care, where its safe, effective, and equitable use is critical. As part of the National Academy of Medicine’s Vital Directions for Health and Health Care initiative, this work outlines essential steps for responsible AI integration. Link

Artificial intelligence-driven pharmaceutical industry: A paradigm shift in drug discovery, formulation development, manufacturing, quality control, and post-market surveillance. This paper examines the transformative impact of AI on the pharmaceutical industry, detailing advancements across drug discovery, formulation development, manufacturing, quality control, and post-market surveillance. Health Affairs. 2025. Link

Alexander Schuhmacher Benchmarking R&D success rates of leading pharmaceutical companies: an empirical analysis of FDA approvals.

We don’t typically make big announcements about our scientific articles, but this time we’re particularly thrilled. Our latest publication, has just been published in Drug Discovery Today. We believe it offers valuable insights into pharmaceutical R&D and carries significant practical implications.

1. The average LoA (first approval) for leading pharmaceutical companies stands at 14.3%, significantly exceeding the previously reported 10% in the scientific literature.

2. Success rates vary significantly among the leading companies, ranging from 8% to 23%. These differences are largely attributable to the companies’ distinct strategies.

3. The data also shed light on R&D organizational structures, as reflected by the observed ratio between Phase 1 and Phase 3 success rates.

One emerging solution in recent years has been the application of AI in drug development to make the development process more efficient. Digital tools such as patient recruitment platforms, virtual trials, digital biomarkers, real-world evidence and?in silico?clinical trials provide the opportunity to significantly decrease dependency on costly?studies, enabling the industry to transform R&D into an economically sustainable value creation process. Link

Programming

Philip Bowsher - The open-source movement in pharma continues to thrive! ???? The Admiral package is revolutionizing ADaM dataset creation in R—a modular, collaborative, and standardized approach developed by 11 co-development partner companies as of 2025!

?? Why is this a game-changer? ? Harmonized ADaM creation across companies ? Standardized workflows & regulatory compliance ? Open collaboration for innovation & efficiency

Read more about this collaboration here - Link

Ross Farrugia and Michael Rimler - Collaborating Across Pharma: Open Source Highlights from the PHUSE US Connect 2024 Keynote. The PHUSE US Connect 2024 keynote, delivered by Michael Rimler (GSK) and Ross Farrugia (Roche), served as an inspiring call to action for the pharmaceutical industry to embrace cross-organizational collaboration in developing open source solutions. The speakers emphasized how open source, when embraced through industry-wide cooperation, can drive transformative change by accelerating innovation, eliminating redundancy, and uniting organizations with a shared purpose: improving patient outcomes. Link

Athenkosi Nkonyeni - Dataset-JSON: The Future of Data Exchange in Clinical Research.

As the industry shifts toward open-source programming languages like R, Dataset-JSON eliminates the barriers imposed by SAS XPT, ensuring seamless data exchange across platforms. Its adoption will streamline regulatory submissions, enhance data transparency, and future-proof clinical data exchange. Read article from Athenkosi here. Link

Mike Stackhouse - datasetjson 0.3.0 has made its way to CRAN! In light of the release of Dataset JSON v1.1.0, which incorporates feedback from the FDA pilot and over a year of community feedback, Nicholas Masel of Johnson & Johnson pushed out these updates to provide support and improve the package as a whole. Link

? Ara Baghumyan - Turning RTF Tables into SAS Datasets. Manually extracting data from RTF tables and converting them into structured SAS datasets can be time-consuming and error-prone. But what if we could automate this process? Here’s a SAS program that reads an RTF table file and converts it into a SAS dataset with ease. Link

Michael Bretscher - Teal4Real - A reporting tool for Real-world Data based on Shiny & teal.modules.clinical. Adapted from Teal - an open-source, modular clinical trial reporting tool being developed at Roche. Link

Thomas Debray, PhD - a new R package, SimTOST, on CRAN! ?? SimTOST is designed to facilitate sample size estimation for bio-equivalence studies using a fast, C++-powered simulation-based approach. Detailed vignettes are available to get started, and a publication is on the way! Link

Kiran Venna - More than often data people do many-to-many joins in PROC SQL which could impact systems and also give undesirable results. This are often done in ignorance. Please check out my SAS video on this video and avoid this kind of joins. Link

Kiran Venna - By understanding concepts of Dictionary.tables, format=SIZEKMG. and &syslast, one can easily find size of latest dataset created and automate it using a macro. Please check out my SAS video on this topic. Link

Maciej Nasiński - Exciting news for the R community. Meet {tergo}, a new CRAN package that is here to transform how you format your R code. It is built with Rust under the hood, making it a powerful and efficient alternative to styler. Thanks to the author Konrad Pagacz

- It is up to 1,000 times faster than styler, as confirmed by benchmarks included in the package

- The speed boost means lower computational overhead, especially in CI pipelines

- By using fewer resources, you reduce your environmental impact. Performance and sustainability in one Link

Nan Xiao - Prevent RStudio 2024.12.0 from adding ProjectID to .Rproj files. Link

Biostatistics

Robert Rachford - Reviewing outputs is one of the most important jobs of a biostatistician.

Here’s a sneak peek into Robert's review checklist,

?? Step 1: Open the latest versions of key documents (Protocol, SAP, TLF Shells, etc.)

?? Step 2: Check if all outputs are present & timestamps are correct ??

?? Step 3: Ensure structure follows the shells—tables, titles, populations, footnotes ??

?? Step 4: Validate the BIG N, trends, calculations, and cross-check outputs! ??

these steps work if you are the CRO biostatistician delivering the outputs or if you are anyone on the sponsor team (not just the biostatistician) and need to review outputs. Read Robert's original post here - Link

Sofia S. Villar - Thompson, Ulam or Gauss? Multi-criteria recomendations for posterior probability computation methods in bayesian response-adaptive trials. How to know if approximating probabilities in a Bayesian RA(r) design is good enough? Exact calculations for binary endpoints allow us to suggest a benchmarking procedure that takes speed and accuracy into account. ?? MRC Biostatistics Unit joint work led by Daniel Kaddaj, Lukas Pin Yiu Nam Edwin TangStef Baas with David Robertson and Sofia S. Villar . Link

Jonathan Bartlett - Estimating hypothetical estimands with causal inference and missing data estimators in a diabetes trial case study. The ICH E9 addendum provides a framework for defining treatment effects in clinical trials but offers little guidance on estimation methods. This study analyzes a type 2 diabetes trial, estimating treatment effects using the hypothetical strategy to handle rescue treatment and treatment discontinuation. This analysis highlights the importance of choosing estimation methods based on practical and methodological considerations rather than a single predefined approach. Link

Cyrus Mehta et al. Testing One Primary and Two Secondary Endpoints in a Two-Stage Group Sequential Trial With Extensions.

While p-value-based tests are simpler, normal theory procedures provide greater power by incorporating endpoint correlations and gatekeeping effects, but they are computationally constrained to two secondary endpoints.

Read the paper here

Yuliya Leontyeva - Bayesian Spatial Relative Survival Model to Estimate the Loss in Life Expectancy and Crude Probability of Death for Cancer Patients. The article focuses on using Bayesian approach, Flexible Parametric Survival models, and spatial models to quantify life expectancy for cancer patients. Link

Miguel Pereira MD PhD - How to choose a method for a phase 1 trial? Phase 1 trials may not be as flashy as later phases, but selecting the right dose-escalation method is critical for success. Read this post by Miguel to understand how to pick the best methology for your trial. Link

Miguel Hernán and Mats Stensrud write that

• Hazard ratios are biased – Due to the depletion of susceptibles, hazard ratios have inherent selection bias and should not be the sole focus of analysis. Adjusted absolute risks are needed for better decision-making.
• Proportional hazards rarely hold – Hazard ratios are usually not constant over time due to changing causal effects or selection bias. The only exception is when treatment has no effect.
• Alternative methods are better – Proportional hazards-based methods are unnecessary; analysts should use approaches that provide absolute risks and do not rely on unrealistic assumptions.

Check Miguel's post and discussion that it created. Link

Integrating Randomized Controlled Trial and External Control Data Using Balancing Weights: A Comparison of Estimands and Estimators. Randomized controlled trials (RCTs) have limitations due to ethical and resource constraints, leading to small sample sizes. To overcome this, researchers integrate external control (EC) data, such as historical trial or real-world data, with RCT data for treatment effect evaluation. This approach raises questions about target population specification, causal estimands, and optimal pooled estimators. Balancing weights help ensure comparability between patient groups, but their application with ECs is still underexplored. This study defines key estimands, proposes balancing-weight-based estimators, and evaluates their bias and efficiency through simulations, highlighting the impact of RCT-EC similarity on performance. Link

Peijin Wang, Hwanhee Hong, Kyungeun Jeon, Laine Thomas

Bayesian Borrowing With Multiple Heterogeneous Historical Studies Using Order Restricted Normalized Power Prior. The FDA's recent guidance on complex innovative trial designs supports Bayesian approaches for incorporating historical data. While informative priors can leverage past studies based on data compatibility, external factors like patient enrollment year may impact their relevance, creating a natural ordering among trials—something many priors fail to address. To tackle this, authors introduce the ordered normalized power prior, which enforces order restrictions while allowing data-adaptive borrowing. Authors compare two normalization strategies, detail computational methods, and validate our approach through clinical datasets and simulations. An efficient implementation is available in the updated NPP package on CRAN. Link

Zifei H., Qiang Zhang, Ram Tiwari, Ph.D., FASA, Tianyu Bai Statistics in Medicine

Designing and evaluating advanced adaptive randomised clinical trials: a practical guide.

Advanced adaptive randomized clinical trials offer greater flexibility than conventional trials, potentially increasing efficiency and the likelihood of conclusive results with smaller sample sizes. These trials also enhance the chances of participants being assigned to more promising interventions. However, there is limited guidance on their design and performance evaluation. This guide outlines the key considerations for planning and assessing such trials using techniques like adaptive stopping, arm dropping, and response-adaptive randomization within a Bayesian framework. It provides practical advice on topics such as intervention types, outcome measures, analysis timing, allocation rules, adaptation strategies, and performance evaluation, using realistic examples and simulation tools. This resource aims to assist clinical trialists, methodologists, and biostatisticians in designing and evaluating advanced adaptive trials. Link

Anders Granholm, Aksel Karl Georg Jensen, Theis Lange, Anders Perner, Morten Hylander M?ller, Benjamin Skov Kaas-Hansen

Example of a study with control arm augmentation using propensity score-based frequentist methods and Bayesian dynamic borrowing.

Tiragolumab in combination with atezolizumab and bevacizumab in patients with unresectable, locally advanced or metastatic hepatocellular carcinoma (MORPHEUS-Liver): a randomised, open-label, phase 1b–2, study. TIGIT is a novel inhibitory immune checkpoint present on activated and exhausted T cells, natural killer cells, and regulatory T cells. Tiragolumab plus atezolizumab resulted in clinically meaningful improvements in objective response rate and progression-free survival compared with placebo plus atezolizumab. Link

Events/Webinars

Haitao Pan - ASA Western Tennessee Chapter (Workshop): Causal Inference and AI/ML in Pharmaceutical Statistics

This short course introduces the basic concepts and fundamental methods of causal inference relevant to pharmaceutical statistics. Starting with the central questions in drug development and the roles of causal inference and AI/ML in answering them, the short course consists of three parts: (1) estimand framework, (2) efficient estimators, and (3) targeted learning. The short course covers causal thinking for different types of commonly used study designs in the pharmaceutical industry, including but not limited to randomized controlled clinical trials, single-arm clinical trials with external controls, and real-world evidence studies. The short course concludes with a case study along with a roadmap to conduct causal inference in clinical studies. The materials covered in this short course are extracted from the instructor’s book, Causal Inference in Pharmaceutical Statistics, published by Chapman & Hall/CRC in 2024. Link

???March 12 from 1:00 to 4:00 PM CST

The Pharmaceutical Data Science Conference (PharmaDS) is back for its second year, bigger and better than before!

Three days filled with cutting-edge insights, valuable networking opportunities, and unparalleled learning via our short courses.

PharmaDS 2025 will focus on the transformative impact of data science in revolutionizing the pharmaceutical industry. Engage with renowned experts, participate in thought-provoking discussions, and connect with a dynamic community of professionals shaping the future of pharma.

???Crowne Plaza in Edison, NJ, from April 7th to 9th, 2025, Link

Pavel Mozgunov - Workshop on Adaptive and Bayesian designs in real trials: clinicians', patients' and statisticians' perspectives.

I am happy to share that I will be organizing a two-day workshop in Cambridge on the application of adaptive and Bayesian designs in practice!

We are going to have talks from clinicians, statisticians, and a patient representative.The first day will be primarily dedicated the talks by clinicians and a patient representative on their views and experience with innovative design in various therapeutic areas. The second day will be primarily dedicated to a series of case studies of the implementation of innovative designs in practice.

The topics will cover: ???Dose-finding trials (monotherapy, combinations, dose-schedule studies)???Basket trials???Biomarker trials???Bayesian Dynamic Borrowing Trials

???Thursday, 5 June, 2025 - 09:30 to Friday, 6 June, 2025 - 17:00

Cancer Research UK Cambridge Institute, Li Ka Shing Centre, Robinson Way, Cambridge CB2 0RE Link

Thomas Debray, PhD I’m thrilled to be co-hosting the pre-conference course "Unlocking Insights: Advanced Pooled Analyses Techniques for Clinical Trial Statisticians" with Prof. Tim Friede at PSI: Statisticians in the Pharmaceutical Industry conference 2025 in London.

This course will run on June 8th and is designed for clinical trial statisticians looking to deepen their expertise in pooled analyses of randomized controlled trial (RCT) data.

? Advanced Meta-Analysis Techniques.

? Applications in Drug Development.

? Real-World Case Studies.

???Sunday 8 June from 1pm -? 5pm. Novotel London Wembley Link

PSI: Statisticians in the Pharmaceutical Industry Estimand Online Training 4 Day Workshop

Delegates will gain experience of applying the estimand framework to real case studies. The training will give an opportunity to discuss estimands with experts from the estimand implementation working group (EIWG) and gain the skills to discuss and apply the estimand framework to their own studies. Link

???Mar 10 - 28, 2025, Online

Join Philip Bowsher (Posit PBC) at PharmaSUG 2025 for a hands-on workshop on end-to-end submissions in R! This session will provide practical, cutting-edge insights into how R is transforming clinical data analysis and regulatory submissions. Link

? Overview of Pharmaverse, GT, Quarto, WebAssembly, and Shinylive

? Hands-on prepping ADAM data and generating TLGs (Tables, Listings and Graphs) in R

? The latest pilot project incorporating Shiny into FDA submissionsSign up today!

Sunil Gupta will be also presenting with the following topics:

?Does SDTM Validation Really Require Double Programming? in the Metadata Management section and

?Reach for R Low Hanging Fruit for Faster Results in the R, Python, and Open Source Technologies section! I hope to see you there!

BBS - Basel Biometric Society seminar on Novartis Campus, Basel with Stephen Senn, Dominic Magirr, Jack Kuipers and Tim Morris.

Covariate Adjustment in Clinical Trials. Covariate adjustment is well known and common in statistical modelling, but its use in the analysis of clinical trials is still rather limited. This half-day meeting is dedicated to covariate adjustment in clinical trials and will provide a theoretical as well as practical view on the topic. Case studies will be presented and their alignment with FDA and EMA guidance on adjusting for covariates will be discussed.

???March 25th 2025, ? 9.30 -12.00, F2F at ??Novartis Campus, Basel

2nd Contemporary Clinical Trials Methodology Meeting. Do you work in clinical trials and want to find out about the latest in the design and delivery of contemporary trials? Join for sessions that will dive into practical insights, including real-world examples and in-depth discussions. Day 1: For trial statisticians and statistically minded trialists. Day 2: For trial managers, data managers and staff interested in trial operations, data systems and data management. Link

???2-3 April 2025, London, South Kensington Campus , London

Introduction to Linear Mixed Models using R. The course will give you the skills to formulate, fit and interpret linear mixed models for a range of practical situations, as well as an appreciation of some of the benefits of mixed modelling. Link

???24-25 June 2025, Online

Advanced Survival Analysis using R. The most commonly used methods of dealing with survival and other time-to-event data are based on the assumption of proportional hazards. But often this assumption may not be tenable, or the data structure may be more complex. This course is concerned with models for different types of data structure, or with different underlying assumptions.Examples used will be drawn from a variety of applications in medicine and health.Practical work will be based around the statistical software R

???14-15 May 2025, Online, Link

BAYES2025:

Invited Plenary Speakers- Juan José Abellan, EMA, NL Towards (long overdue) regulatory guidance on Bayesian statistics in clinical trials.

Andy Grieve, King’s College London, London, UK Predictive and Pre-Posterior Distributions in the Planning of Clinical Trials

Harrison Quick, University of Minnesota, USA The Intersection of Informative Priors and Differential Privacy in Bayesian Spatial Biostatistics

Virgilio Gómez Rubio, University of Castilla-La Mancha, SP Approximate Bayesian inference for the analysis of population health data

???22-24 October 2025, Leiden, NL, Link

Free webinar: Can simulation guided trial design really improve strategic decision making in drug development? . A conversation where Michael Krams will debate with Andy Grieve, Tobias Mielke and Kert Viele.

???28 February 2025, Online, Link

10th?International Meeting on Statistical Methods in Biopharmacy - Advancing drug development through innovative designs and efficient data use".??Multiple topics will be concerned such as challenges in small population trials, novel endpoints, use of external data, data visualisation, causal inference and bayesian approaches.

???Paris, 8-10 October 2025, Link

Sample size calculations in randomised clinical trials: beyond the basics by Babak Choodari-Oskooei, Ian White, Andrew Copas, Matteo Quartagno

This one-day course consists of five lectures. The first lecture provides the statistical theory for sample size calculations and covers the methods for continuous and binary outcomes. The second session extends the theory to trials with time-to-event outcomes and presents methods for the prediction of power and trial timelines in complex and realistic scenarios with non-uniform recruitment and follow-up patterns. The third and fourth lectures present methods for sample size calculations with more complex designs, including factorial and cluster, and group sequential trial designs. The final lecture covers sample size calculation methods for non-inferiority designs and presents methods on how to deal with uncertainties in design parameters when calculating the sample size as well as the application of simulations in calculating sample size for more complex designs.

???4 March 2025, MRC Clinical Trials Unit at UCL, London. Link

Slides from the presentations at the 2024 Boston Pharmaceutical Statistics Symposium (31 October - 1 November 2024)

- SHORT COURSE: Mark Chang, AI/ML for Clinical Trials and Humanized AI for Future Healthcare.

- Margaret Gamalo, PhD, FASA, Statistical Boom: Opportunities for a Statistician in an Era of Quantitative Medicines Development.

- Joseph Ibrahim, Optimal Priors for the Discounting Parameter of the Normalized Power Prior.- Veronica Bunn, Predicting Probability of Success for Phase III Trials via Propensity-Score-Based External Data Borrowing.

- Kyle Wathen, AI-Generated R Functions Integrated with Proprietary Software for Clinical Trial Simulation.

- Cynthia Hau, Pragmatic Clinical Trials: Future Directions and Design Considerations.

- James Rogers, Practical Steps Toward Model-informed Drug Development Using Causal Diagrams.

Link

Webinar: Navigating FDA's Expectations for Drug Approval - Adaptive designs for dose-optimization and accelerated approval by IDDI - Regulatory Statistics & Clinical Data Science Experts (12 December 2024). Link